1 ----------------------------------------------------------------------------
3 SimGrid (3.26.1) NOT RELEASED YET (v3.27 expected March 20. 2021, 09:37 UTC)
6 - Removed some option aliases that were deprecated since v3.23 (June 2019)
7 - 'verbose-exit' is now 'debug/verbose-exit'
8 - 'simix/breakpoint' is now 'debug/breakpoint'
9 - 'clean-atexit' is now 'debug/clean-atexit'
12 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
13 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
14 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
15 or Mailbox::get_async<void>() if you really want to play with void*.
17 ----------------------------------------------------------------------------
19 SimGrid (3.26) December 15. 2020
21 The Release release (the French lockdown was eased today).
23 Important user-visible changes:
24 - SimGrid now requires a compiler with C++14 support.
25 Sibling projects should upgrade their FindSimgrid.cmake
26 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
27 several users had difficulties to understand issues when using high bandwidth or
28 small latency events. The new value was already the default for SMPI and
29 should not cause too much performance hit. It can change some simulation
30 timings if simulation was skipping a lot of small events (which was actually
31 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
32 - On some network models (L07, CM02, LV08), an implicit loopback link is created
33 for local communication. Default bandwidth was 500MBps and latency was 15µs,
34 which was rather bad. Default values have been changed to 10GBps and 0 latency.
35 Options network/loopback-lat and network/loopback-bw have been added to adapt
36 these values if needed. Network and routing models already providing separate
37 loopbacks (clusters and such) are not impacted by this change.
38 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
39 metadata that are added to each MPI message by MPI libraries (tag,
40 src/dest, and communicator information). This may change simulation times
41 for very small message sizes, but should be more realistic.
44 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
45 - New plugins: wifi_energy and link_load. See the associated examples.
46 - New examples: wifi networks, and communication suspend.
49 - LTO is now enabled for Intel/clang compilers.
50 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
51 Setting it to "auto" will use all available cores, while setting it to n will
52 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
53 - Remove obsolete runtime option 'contexts/parallel-threshold'.
54 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
55 has been kept for compatibility.
56 - Finally remove obsolete snake_case() aliases for runtime options.
57 - Further improve the documentation.
60 - the on_exit() of each actor is also executed when the simulation deadlocks.
61 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
62 An example is provided for s4u::Comm but it should work with Exec and Io.
65 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
66 https://framagit.org/simgrid/SMPI-proxy-apps
67 - MPI/IO: supports names starting with "./"
68 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
70 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
71 building shared libraries, this may cause issues, so environment variable
72 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
73 - most temporary files should now be created in /tmp dir (or equivalent).
74 If this one does not allow execution of code (noexec flag), this may cause issues.
75 Please use another tmp directory (using TMPDIR or equivalent system variable)
79 - A large refactoring is ongoing, but this should be transparent for users so far.
82 - Drop xbt_str_split().
84 C binding and interface:
85 - The return type of the sg_actor_on_exit() callbacks is now 'void'
86 (int value was previously ignored)
87 - Many C functions were renamed to follow a common naming style
88 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
89 names are kept until version 3.30.
90 - Many MSG tests were converted to the new S4U's interface in C, that
91 was extended for that.
93 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
94 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
95 - FG#41: Add sg_actor_create C interface
96 - FG#43: xbt::random needs some care
97 - FG#48: The Impossible Did Happen (yet again)
98 - FG#50: Suspending an actor executed at the current timestamp fails
99 - FG#52: Zero-seconds timeout : "That's in the past already"
100 - FG#53: Crash while using ns-3 network model
101 - FG#54: How to suspend a comm?
102 - FG!22: Stochastic Profiles
103 - FG!24: Documentation and fix for xbt/random
104 - FG!35: Add a modeling hint for parallel links in doc
105 - FG!36: [xbt/random] Read/Write the state of the RNG
106 - FG!42: Link cumulated load plugin
107 - FG!44: Test condition variable wait_until+notify_all
108 - GF#18137: Allow different stack sizes?
109 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
110 - GH#139: Allow pthread creation in SMPI
111 - GH#336: Packet-level simulation using SMPI?
112 - GH#345: Error in the LMM while migrating a VM
113 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
114 - GH#352: pip install / python setup.py install fail to find pybind11
115 - GH!337: Fix link_energy plugin for wifi platforms
116 - GH!339: Add Mailbox set_receiver method to python binding
117 - GH!344: Cast hugepages macros parameters to int64
119 ----------------------------------------------------------------------------
121 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
123 The Palindrom Day Release.
125 Important user-visible changes:
126 - Improve the Python usability (stability and documentation).
127 - A nasty synchronization bug was ironed out, see also below.
128 - Python's doc was organized as a separate tree, now integrated with C++.
129 - C bindings of S4U were not part of the doc.
130 - The C++ doc was also improved as methods are now split by theme.
131 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
132 - This is mandatory to use the Java bindings.
133 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
134 - SimDAG++: Automatic dependencies on S4U activities (experimental)
135 - Some features are already implemented but not all of them
136 - Cannot block an activity until it's scheduled on a resource
137 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
138 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
141 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
142 - Drop support for Viva/Triva (old visualization tools), hence removed
143 TRACE_get_node_types() and TRACE_get_edge_types() functions.
146 - Require pybind11 version 2.4.
147 - Greatly improve locking during context switches with Python.
148 - Declaring Python callbacks to this_actor::on_exit is now working.
149 - No more tests are failing currently, even on win32.
152 - Actor: Merge signals on_migration_start/end into on_host_change
153 - Actor: Rename migrate() into set_host()
154 - Disk: Allow users to get the read and write nominal bandwidth values
155 - Exec: Implement wait_for(timeout)
156 - Io: Implement wait_for(timeout)
159 - Parse errors now raise a simgrid::ParseError that you may want to catch.
162 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
163 been renamed to fix a spelling error. As usual, the previous definitions are
164 kept with a deprecation warning. However, to avoid runtime errors, it is now
165 forbidden to override these deprecated methods in a derived class. Please use
166 the new names immediately if you need to override them.
169 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
170 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
171 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
172 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
173 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
174 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
175 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
176 - New module: random, providing classical random numbers generators.
179 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
180 allocations' size above which they are considered "shared" by default (as if
181 they were performed through SMPI_SHARED_MALLOC macros).
182 Default = 0 = disabled feature.
183 Note : malloc, calloc and free are now overriden by smpicc/cxx by default.
184 This can cause some troubles if codes are already overriding these. If this
185 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
186 help, but will make this feature unusable.
187 - Input check is now performed much more consistently and outputs precise
188 warnings for each error found
189 - New supported calls : MPI_File_set_view, MPI_File_get_view
190 - MPI I/O now only supports disks, as storage is discontinued.
191 - MPI I/O allows opening files without specifying absolute path
192 - Note : SMPI configuration options are now initialized by smpi_init_options()
193 call, which should be called if SMPI is not being used through smpirun.
195 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
196 - FG#9: Python bindings crashing
197 - FG#39: Missing s4u::Comm::wait_any_for example
198 - FG#42: Add support for ThreadSanitizer (TSan)
199 - FG!19: Removing RngStream
200 - FG!20: A module for RNG calls
201 - FG!21: Choice between ad-hoc and standard distributions implementations
203 - FG!25: Fix link in Mutex doc
204 - FG!26: Fix links in SMPI interface doc
205 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
206 - GH#305: Unscheduled tasks are still excuted
207 - GH#313: smpirun: manual outdated w.r.t. --help
208 - GH#321: [S4U] Get task remaining work ratio
209 - GH#323: Crash when an actor turn off his physical host
210 - GH#335: Missing links on dragonfly example svg
211 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
212 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
214 ----------------------------------------------------------------------------
216 SimGrid (3.24) October 9. 2019
218 The Clean Disk Release.
221 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
222 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
223 thus becomes the default resource to execute I/O operations. Examples
224 and the file system plugin have been modified to use disks instead of
225 storage elements. Storage can still be used but is doomed to disappear.
226 Users are thus adviced to update their simulators accordingly.
227 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
228 for consistency with pthread_barrier_wait()
229 - Host::get_englobing_zone() returns the englobing netzone
230 - Actor::on_destruction is now called in the destructor
231 Actor::on_termination new signal called when the actor terminates
233 - Global signals are now part of the Engine:
234 - on_platform_creation: after config settings, before the XML parsing
235 - on_platform_created: right after the XML parsing
236 - on_time_advance: each time the clock advances
237 - on_simulation_end: after simulation, before cleanups
238 - on_deadlock: as the name implies.
240 - sg_{actor,host,link}_{data,data_set}() now all exist.
241 Use them to attach user data to the object and retrieve it.
244 - Introduce an experimental Wifi model. It sounds reasonable
245 according to the state of the art, but it still has to be properly
246 validated, at least against ns-3 if not against reality.
247 - Improved the usability of ns-3. Several bugs were ironed out.
248 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
249 It is now expressed as 'idle:epsilon:allCores' to properly model the
250 consumption of non-whole tasks on mono-core hosts. More info in the doc.
253 - convert a new set of functions to the S4U C interface and move the old MSG
254 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
257 - Fortran bindings for DVFS have been removed.
258 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
259 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
260 - SMPI can now be selected by cmake's find_module(MPI) with
261 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
262 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
263 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
264 - trace-call-location can be used with TI traces, and replayed, and is compatible
265 with smpi/comp-adjustment-file.
266 - sleep events are now correctly traced/replayed.
267 - Default for trace-call-location is now to use file names and not full paths.
268 To revert to previous behavior (in case of collision of filenames), option
269 "smpi/trace-call-use-absolute-path" can be set to yes.
272 - Use the included xxHash as an hash implem when C++14 is usable.
273 - Option model-checker/hash was removed. This is always activated now.
274 - New option smpi/buffering controls the MPI buffering in MC mode.
275 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
276 This is useful to check for MPI compliance.
279 - New section on plugins: how to define a new one, and existing ones.
282 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
283 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
286 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
287 and <mount> tags. The rationale is finish to decouple what should be handle
288 by SimGrid kernel (i.e., resources and activities) from plugins built in
289 user space (e.g., the file system plugin). This new tag currently accepts
290 only three attributes (id, read_bw, and write_bw). All the other information
291 that was declared with the storage related tags now has to be expressed as
292 properties. An example of platform using this new tag is available at
293 examples/platforms/hosts_with_disks.xml
296 - 'expect signal' can now accept more than one potential signal.
298 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
299 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
300 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
301 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
302 - FG#31: per-actor data
303 - FG#34: SG_BARRIER_SERIAL_THREAD?
304 - FG#35: model-checker does not like buster-produced binaries
305 - FG!13: MC: complete workaround in the error msg seen on modern systems
306 - FG!15: execute_flops now logs compute
307 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
308 - FG!17: ns-3: unblock the right number of communications + others issues
309 - FG!18: Improving the performance of the ns-3 bindings
310 - GH#189: Energy consumption of parallel tasks
311 - GH#219: Error in the throughput of TCP transfer
312 - GH!330: Minor polishing of English: Configuring_Simgrid
313 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
314 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
315 - GH!334: Rework energy plugin, again
317 ----------------------------------------------------------------------------
319 SimGrid (3.23.2) July 8. 2019
322 - Nicer introduction page.
323 - Migrate the "Deploy your application" page to the new doc.
324 - Move Java as a subtree of MSG.
327 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
330 - Drop xbt_dynar_sort_strings().
333 - Really fix FG#26: Turning off a link should raise NetworkFailureException
334 - FG#27: Wrong exception thrown to wait_any when link is turned off
335 - GH#328: Java: Canceling multiple tasks in a single vm/host
337 ----------------------------------------------------------------------------
339 SimGrid (3.23) June 25. 2019
341 The Exotic Solstice Release.
344 - SunOS and Haiku OS support. Because exotic platforms are fun.
345 - Stop setting random seed with srand48() at initialization.
346 - Use addr2line as a fallback for stacktraces when backtrace is not available.
347 - Build option -Denable_documentation is now OFF by default.
348 - Network model 'NS3' was renamed into 'ns-3'.
351 - Simgrid can now hopefully be installed with pip.
354 - wait_any can now be used for asynchronous executions too.
357 - New log appenders: stdout and stderr. Use stdout for xbt_help.
358 - Drop xbt_dict_dump.
361 - 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
362 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
363 - MPI interface is now const correct for input parameters
364 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
365 - Fortran bindings for async collectives.
366 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
369 - Remove option 'model-check/record': Paths are recorded in any cases now.
370 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
371 always sparse. That's an immense gain in memory, and can even be
372 faster because of cache effects. So there is no need to clutter the
373 code to allow the user to go for the unefficient mode.
376 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
377 models proved to be more accurate than these old experiments.
379 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
380 - FG#1: Broken link in error messages
381 - FG#2: missing installation documentation
382 - FG#3: missing documentation in smpirun
383 - FG#6: Python bindings not available on PyPI
384 - FG#7: simple cmake call requires doxygen
385 - FG#8: make python bindings an optional dependency
386 - FG#10: Can not use MSG_process_set_data from SMPI any more
387 - FG#11: Auto-restart actors forget their on_exit behavior
388 - FG#12: -Denable_lto=OFF doesn't disable LTO
389 - FG#13: Installs unstripped file 'bin/graphicator'
390 - FG#14: Installs the empty directory 'doc/simgrid/html'
391 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
392 - FG#17: Dead link in doc (pls_ns3)
393 - FG#20: 'tesh --help' should return 0
394 - FG#21: Documentation link on http://simgrid.org/ broken
395 - FG#22: Debian installation instruction are broken
396 - FG#26: Turning off a link should raise NetworkFailureException exceptions
397 - GH#133: Java: a process can run on a VM even if its host is off
398 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
399 - GH#326: Valgrind-detected error for join() when energy plugin is activated
401 ----------------------------------------------------------------------------
403 SimGrid (3.22) April 1. 2019
405 The Easter Christmas Release (this one is somewhat late).
408 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
409 - Not complete yet: asynchronous activities (amongst others) are still missing
410 - Still ongoing: the interface may change in the future. We need more
411 testers! Please report any glitches.
412 - No new project using Java should start now. Please switch to Python.
415 - Some of the internal cleanups may lead to speed improvements:
416 - The hard limitation on the amount of simulated actors with
417 Java+Mac was removed. Now, the available memory is the only limit.
418 - Our refcounting was tidyied, leading to 10% speedups in some cases.
419 - We are still working on making our code robust to the actor kills
420 and hosts' churn. Things are improving, but it's not perfect yet.
421 - Replaced our own code to display a backtrace (that was forking addr2line)
422 with the Boost.Stacktrace library.
423 You won't see your backtraces without this optional dependency.
424 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
425 - Stop setting random seed with srand() at initialization.
428 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
429 XML file version remains 4.2 since old files are still compatible.
432 - Process termination which was broken at version 3.21 has been repaired.
433 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
434 - Hide the examples into examples/deprecated. New users should use Python.
437 - Drop MSG_process_create_from_stdfunc() from the API.
438 This C++ function was a pimple in the C API, made necessary at some
439 point by the Java bindings. This is fixed now.
440 - Hide the examples into examples/deprecated. New users should use S4U.
441 - MSG_process_create and MSG_process_attach now crash if the host used
442 to run this process is off.
443 - Fix the protype of MSG_process_on_exit()
444 Now use (int,void*) callbacks instead of (void*,void*) ones.
445 The implementation was ways too messy to actually work, I guess.
448 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
449 too often and provide more accurate timings.
450 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
451 - MPI_Alltoallw support
452 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
453 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
454 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
455 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
457 - Basic implementation of generalized requests (SMPI doesn't
458 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
461 - Drop sg_cmdline. Please use xbt_cmdline instead.
462 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
463 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
464 OsSemaphore is implemented in a portable way with C++11 threads.
465 This should allow much more threads to be created at the same time,
466 allowing Mac and Java users to simulate many more actors.
467 - Implement the 'thread' factory with std::thread instead of xbt ones.
468 It is not possible to set the stack size with threads anymore, but
469 -fsplit-stack is the way to go nowadays when using threads.
470 - Drop the xbt_os_thread_t module (now unused)
471 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
472 - Drop xbt_str_join_array().
473 - Drop cunit, use Catch2 instead.
476 - Many cleanups in the kernel::activity namespace. This was long
477 overdue, and shall open the path to many future endeavors.
480 - #132: Java: a process can not shut down its own host
481 - #220: S4U: detached send causes exception if sender terminates soon after sending
482 - #261: Document the parameters of parallel execution's constructor
483 - #300: [s4u] BarrierPtr is missing
484 - #314: SMPI args internal cleanup
485 - #316: Fix a bug related to the CPU utilization of multi-core VM
486 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
487 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
488 - #325: Turning off a host has different behavior on sleeping actors and computing actors
490 ----------------------------------------------------------------------------
492 SimGrid (3.21) October 3. 2018
494 The Restarting Documentation (TRD) Release.
497 - Convert the user manual to Sphinx for improved usability.
498 Unstable version now hosted on https://simgrid.frama.io/simgrid
499 Many glitches remain; stay tuned.
500 - Restrict the Doxygen reference API to the installed header files.
501 (documenting all internals this way is overkill + doxygen scales badly)
502 - New tutorial on S4U (highly inspired from the old MSG one, but with
503 a git repository to fork as a starting point, and a docker image)
504 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
505 - The Developper manual is still to be converted (not compiled until then)
506 - Some parts are still missing in this conversion (such as the
507 platform and deployment sections) while others were blindly converted
508 and would need more love (such as the configuration flags).
509 Things will certainly further improve in the future releases.
512 - Fully reimplement auto_restart mechanism that was utterly broken
513 (fix #22, #131 and #281 that were all very related)
514 - Implement semaphores (fix #303 and #291)
515 - When creating an actor from a function and its parameters,
516 move-only parameters are not allowed anymore, as it would prevent
517 the actor to be restartable if its parameters are consumed on run.
518 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
519 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
520 IO operations that can be started, waited for, or canceled as a regular
522 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
526 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
527 - New host variable: 'core_count'
530 - Remove xbt_os_thread_specific features
531 - Remove portability wrapper to condition variables
532 - Remove xbt_os_thread_yield()
535 - MPICH collective selector now mimics MPICH 3.3b
536 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
537 - MPI_Init(NULL,NULL) is now allowed
538 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
539 - Command-line arguments should now be handled properly in Fortran simulations
540 - MPI Attributes and related callbacks should work in Fortran
541 - Apps using C + Fortran should now work
542 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
543 - Support MPI_Error_String
546 - Due to an internal bug, Msg.run() must now be your last line.
547 We hope to fix it in a future release, and we are sorry for the inconvenience.
550 - #22: Process autorestart seem to only work with CAS01 cpus
551 - #56: Feature request: dynamic SMPI replay
552 - #93: simgrid should not eat --help
553 - #111: How to change the loopback link characteristics?
554 - #116: Platform cluster radical
555 - #117: Platform cluster/host order
556 - #129: Replay_multiple on steroid
557 - #131: Java : autostart does not work
558 - #143: Setting a breakpoint at a given time
559 - #235: Network energy models should be integrated
560 - #264: Add ptask L07 resource tracing
561 - #271: Dynamic and manual replay of SMPI traces
562 - #279: Breakpoints option uses time deltas instead of absolute time
563 - #281: Daemonized actor no longer daemonized after an auto-restart
564 - #282: TIT convertor erase traces when using absolute path in the trace list file
565 - #285: segfault when a s4u actor kills itself with msg process tracing activated
566 - #286: Pajé traces are not exposing the number of cores
567 - #287: Command-line parsing should stop when encountering '--'
568 - #288: MPI_Init(NULL, NULL)
569 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
570 - #290: Method to check if a mailbox has a message ready to be consumed
571 - #291: [s4u] semaphores do not seem available yet
572 - #294: Alltoall collective implementation of mpich
573 - #295: Starting a Host as OFF is not supported
574 - #296: DTD too permissive
575 - #299: Add a s4u barrier example
577 ----------------------------------------------------------------------------
579 SimGrid (3.20) Released June 24. 2018
581 The proxy snake_case() release.
584 - The whole API is now using snake_case() instead of mixing it with camelCase.
585 Compatibility wrappers in place for one year.
586 - We took the opportunity to increase the consistency of the API.
587 Please report any remaining glitches.
588 - New function simgrid::s4u::Host::get_actor_count:
589 Returns the number of actors running on a specific host.
592 - SMPI is now tested with ~45 proxy apps from various sources, with none or
593 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
594 - Replay: The replay file has been re-written in C++.
595 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
596 supported. They are stored in the trace and used when replayed.
597 - Basic support of MPI_Cancel. Robustness not guaranteed.
598 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
599 (only for MPI_COMM_TYPE_SHARED).
600 - New option: smpi/privatize-libs, to add external shared libs to be privatized
601 by SMPI. They will be copied locally and loaded separately by each process.
602 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
603 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
604 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
605 with internal index optind with multiple processes. Only works if MPI_Init has
607 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
608 syntax without preprocessing tricks.
611 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
612 behavior tracing is no longer limited to MSG
613 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
614 this means that traces now need to be updated or re-obtained; both Irecv
615 and Isend lines in your traces have now in total 5 mandatory fields:
616 <rankid> <command> <to/from rankid> <tag> <size>
617 To update your traces, it suffices to add a 0 for the tag here.
618 - TIT now also supports waiting for a distinct request via MPI_Wait.
619 Wait/Test now wait for a specific request, not just the last one that was
620 issued. This unfortunately means another update, because we need to
621 identify which request you want to wait for. We do this via the
622 triplet (sender, receiver, tag), which needs to be added:
623 <rankid> <command> <sender> <receiver> <tag>
624 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
625 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
628 - The deprecation of MSG is ongoing (but this should not impact you).
629 Many MSG functions are now simple wrappers on the C API of S4U. If
630 you wish to convert your code to S4U, find the S4U counterparts of
631 your MSG calls in src/msg/msg_legacy.cpp.
632 - MSG can still be used, but won't evolve anymore.
635 - Allow to run the Link energy plugin from the command line with
636 --cfg=plugin:link_energy
637 - Rename Energy plugin into host_energy
638 - Rename Load plugin into host_load
641 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
642 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
643 - Improve performance when handling timeouts of simix synchros.
646 - Config: the C API is now deprecated (will be removed in 3.23), and
647 the C++ API has been slightly improved.
650 - Fix several build issues on OSX.
651 - Move simgrid_config.h to simgrid/config.h (old header still working)
654 - #143: Setting a breakpoint at a given time
655 - #150: Inconsistent event names in SMPI replay
656 - #258: daemonized actors hang after all non-daemonized actors have completed
657 - #267: Linker error on unit_tmgr
658 - #269: SMPI: tracing of MPI_Wait/all/any broken
659 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
661 ----------------------------------------------------------------------------
663 SimGrid (3.19.1) Released March 22 2018
665 The Fixed ABI Release.
667 As you may know, we are currently refactoring SimGrid in deep.
668 Upcoming SimGrid4 will be really different from SimGrid3: modular,
669 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
671 Our goal is to smooth this transition, with backward compatibility
672 and automatic update paths, while still progressing toward SimGrid4.
674 SimGrid remains open during works: The last pure SimGrid3 release was
675 v3.12 while all subsequent versions are usable alpha versions of
676 SimGrid4: Existing interfaces remain unchanged, but the new S4U
677 interface is budding and the internals are deeply reorganized.
679 Since 2015, we work hard to reduce the changes to public APIs. When
680 we need to rename a public library symbol in S4U, we let your
681 compiler issue an explicative warning when you use the deprecated
682 function. These messages remain for four releases, i.e. for one full
683 year, before turning into an error. Starting with v3.15, your can
684 also adapt to API changes with the SIMGRID_VERSION macro, that is
685 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
687 Starting with this v3.19.1, our commitment to reduce the changes to
688 the public interfaces is extended from the API to the ABI: a program
689 using only MSG or SimDag and compiled against a given version of
690 simgrid can probably be used with a later version of SimGrid without
691 recompilation. We will do our best... but don't expect too much of
692 it, that's a really difficult goal during such profund refactoring.
694 The difference between v3.19 and v3.19.1 is that the former was
695 accidentally breaking the ABI of MSG, while the later is restoring
698 S4U and kernel APIs will still evolve until SimGrid4, with one-year
699 deprecation warnings as currently. In fact, cleaning up these
700 interfaces and converting them to snake_case() is one release goal of
701 v3.20. But don't worry, we are working to smooth this upgrade path.
703 In summary, new projects should start with S4U to benefit of the
704 future, but old MSG projects should still be usable with no change.
706 ----------------------------------------------------------------------------
708 SimGrid (3.19) Released March 20 2018
710 The Moscovitly-cold Spring release.
712 SMPI user-visible improvements:
713 - Support of MPI_Type_create_subarray
714 - Unimplemented MPI calls now abort (instead of issuing a warning)
715 Some harmless calls around errhandlers still only issue a warning.
716 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
717 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
718 is now possible (mmap privatization was refactored to this end).
719 Adding new SMPI instances should also be technically doable now.
721 SMPI internal cleanups:
722 - Removed index notion from SMPI -- no more "getPid() - 1"!
723 In tracing files, actors are called rank-<PROCESS-ID> but these are
724 NOT the real ranks (tracing several communicators is not supported yet).
725 - Deprecate smpi_process_index() (will be removed in 3.22)
726 - Better testing on classical proxy apps, through an external project:
727 https://github.com/simgrid/SMPI-proxy-apps/
730 - Use a graphical TOC to make it easier to find the documentation you need
731 - Revamp the MSG tutorial
734 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
735 - Remove parameter of MSG_process_killall().
736 Resetting the PID was bogus anyway (several actors could have the same PID).
739 - Execution->setHost() can be called after start(), to migrate it.
740 - Comm::test_any() is now implemented.
741 - s4u::Actor now has onCreation() and onDestruction() signals.
742 - Install some kernel header files for the users' plugins and more.
743 Warning, their API is really not stable yet.
744 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
745 include/simgrid/kernel/routing/*Zone.hpp
748 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
750 Build System and other cleanups:
751 - Remove unused run-time parameter "tracing/onelink-only".
752 - SimGrid now works with intel compilers (MC still buggy)
753 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
754 - tesh kills subprocesses on timeout (but not on windows)
757 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
760 - #194: Feature request: simgrid::s4u::Comm::test_any()
761 - #245: migrating an actor does not migrate its execution
762 - #253: Feature Request: expose clusters as objects
763 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
764 - #255: Tesh broken on Windows
765 - #256: Modernize FindSimGrid.cmake
766 - #257: Fix (ab)use of CMake install
768 ----------------------------------------------------------------------------
770 SimGrid (3.18) Released December 24 2017
772 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
774 MSG IS NOW DEPRECATED. Please use S4U for new projects.
775 - Support for MSG should not be removed from SimGrid before 2020,
776 but future evolutions will be limited to the integration of
777 user-provided patches.
780 - s4u::onDeadlock() signal, run before stopping the simulation.
781 - s4u::Exec: asynchronous executions (abstraction of background
783 - s4u::Host->getLoad() returns the achieved speed in flops/s
786 - Link::name() is deprecated and will be removed in v3.21, use
787 Link::getCname() instead.
788 - Mailbox::getName() changed to return a std::string, use
789 Mailbox::getCname() instead to get a char*.
790 - Storage::getName() changed to return a std::string, use
791 Storage::getCname() instead to get a char*.
792 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
793 to have both allocation and deallocation of the map in user space.
796 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
797 The amount of flops remaining to do is a vector, not a scalar.
798 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
799 name implies on both sequential and parallel tasks.
800 - Both changes fix GitHub's #223 using PR #237 as a basis.
801 Thanks Michael Mercier.
802 - Most examples were converted to S4U and hidden elsewhere at tests.
803 Rationale: we still want MSG to work; we want newcomers to use S4U.
806 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
807 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
810 - Switch to the faster dlopen privatization mechanism by default
811 - Documentation improvments
814 - Live migration is getting moved to a plugin. Dirty page tracking is
815 the first part of this plugin. This imply that VM migration is now
816 only possible if one this function is called:
817 - C/MSG: MSG_vm_live_migration_plugin_init()
818 - C/C++: sg_vm_live_migration_plugin_init()
819 - Java: Msg.liveMigrationInit()
820 For C and C++, "simgrid/plugins/live_migration.h" also has to be
824 - Define class simgrid::xbt::Path to manage file names.
825 - Removed unused functions:
826 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
827 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
828 xbt_graph_export_graphviz()
829 - xbt/str.h: xbt_str_join()
830 - Remove unused datatypes:
831 - xbt/heap.h: use std::priority_queue or boost::heap instead
832 - xbt/swag.h: use boost::intrusive::list instead
835 - New link_energy plugin for the consumption of the links.
836 - All of the operations on files and storage contents have been
837 packaged into a plugin (src/plugins/file_system). The current
838 public interface can be found in
839 include/simgrid/plugins/file_system.h
840 To use these functions you now have to initialize the plugin by
841 calling MSG_storage_file_system_init() just after calling
842 MSG_init() or sg_storage_file_system_init() just after creating
846 - Remove the undocumented/untested tag <include>
849 - Remove viva specific tracing as the tool is no longer maintained
852 - #248: Kill of finished processes leads to segfault
853 - #240: xbt_cond_wait_timeout should gracefully return for C apps
854 - #239: please implement signal s4u::onDeadlock()
855 - #230: segfaults when exit() without run()
856 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
857 - #223: MSG_task_get_flops_amount() not working with parallel tasks
858 - #222: Actor::kill() doesn't really kill and segfaults
859 - #221: odd LMM warning when killing an actor
860 - #120: Memory leak when the processes are forcefully killed
862 ----------------------------------------------------------------------------
864 SimGrid (3.17) Released October 8 2017
866 The Drained Leaks release: (almost) no known leaks despite the tests.
868 Even very long simulations will preserve your memory: our testsuite
869 stresses SimGrid for over 45mn on fast machines for a coverage of
870 over 80%, and there is only one single known leak, of about 4kb.
874 - Comm.detach(): start and forget about asynchronous emission. A cleanup
875 handler may be given to free resources if the comm cannot be completed.
876 - this_actor::send(mailbox) is now mailbox->put()
877 - New: simgrid::s4u::Comm::wait_all()
878 - New: Host.execute() for remote executions.
881 - Unused option network/sender-gap is removed.
884 - Deprecate MSG_task_isend_with_matching(): This unused feature
885 really complicates our internals. Will be removed in v3.20.
888 - Improved context termination. It is now possible to free resources, even
889 when a process is forcibly killed.
890 - ContextBoost: add support for Boost versions above 1.61.
893 - Bring back run-time option --cfg=exception/cutpath to remove exception
895 - Removed unused functions:
896 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
897 xbt_str_rtrim(), xbt_str_trim().
898 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
901 - Removed header files obsolete since SimGrid 3.12:
902 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
904 - #3: SD_exit should be made optional
905 - #120: Memory leak when the processes are forcefully killed
906 - #159: Threading test regression in Actor refcounting
907 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
908 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
909 - #186: Actor::killAll() segfaults if some process is blocked on wait()
910 - #191: VM migration and pstate
911 - #192: Updating the energy consumptions of all hosts crashes with VMs
912 - #195: All actors have PID=0 in the logs
913 - #204: Sometimes segfault with thread contexts and mmap privatization
914 - #222: Actor::kill() doesn't really kill and segfaults
915 - #225: Actor::kill() doesn't really kill when victims are doing a join()
917 ----------------------------------------------------------------------------
919 SimGrid (3.16) Released June 22. 2017.
921 The Blooming Spring Release: developments are budding.
924 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
925 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
926 - Java JAR file should be finally fully working on Mac OSX too.
928 XML platforms: Switch to platform v4.1 format.
929 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
930 - <zone> can be used as a synonym for the now deprecated <as>
931 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
932 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
933 - <actor> can be used as a synonym for the now deprecated <process>
934 - state_file and avail_file periodicity is now easier to express
935 (check the documentation) the old behavior should still work.
936 * Storage: not backward compatible, but it seems that nobody ever used it
937 - Remove attribute 'content_type' of <storage_type>: was never used
938 - Make attribute 'model' of <storage_type> optional: for future usage
939 - Remove Bconnection model property: was never (in)validated. Replaced by
940 the maximum of the read and write bandwidth as a resource constraint.
943 - New and Backwards Compatibility break:
944 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
945 When one wants to get the list of tasks whose states have changed during a
946 simulation round, s/he has to allocate and free a dynar and use it as
947 argument to this function. The former SD_simulate (double how_long)
951 - Allow multicore VMs, with the correct sharing (unless you overcommit)
952 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
955 - New function to update the consumption of all hosts at once.
956 - Fix the model for multi-core hosts, linear in the amount of busy
957 cores with an abnormality for pIdle (see paper).
958 BUG: Still not satisfactory for ptask on mono-cores.
961 - The netzone are now available from the MSG API.
962 The old names still work, but are now deprecated.
965 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
966 - New option: smpi/keep-temps to not cleanup temp files
967 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
968 mallocs mode is used, allows to change the size of the fake file used
969 (default 1MB), to potentially limit the number of mappings for large runs.
970 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
971 - Fortran ifort and flang compilers support
972 - New RMA calls supported (experimental):
973 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
974 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
975 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
976 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
977 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
978 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
981 - New: this_actor::isMaestro() returns whether we are in kernel mode.
982 - Behind the scene, ActivityImpl are now automatically refcounted.
983 Removing this nasty bottleneck will greatly simplify our internals.
986 - Replay: New function xbt_replay_action_get():
987 Retrieve the function previously associated to an event type.
988 - DROPPED FUNCTION: xbt_str_varsubst()
989 - DROPPED MODULE: strbuff. We don't need it anymore.
990 - DROPPED MODULE: matrix. We don't need it anymore.
991 - DROPPED MODULE: lib. We don't need it anymore.
993 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
995 SimGrid (3.15) Released March 22. 2017
997 The Spring Release: continuous integration servers become green
998 We fixed even the transient bugs on all target architectures:
999 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1000 Windows, FreeBSD, NetBSD.
1002 - Rename NetCards to NetPoints.
1003 This was intended to help NS3 users, but that's not a netcard.
1004 That's a point in the routing algorithm, let's avoid wrong simplifications.
1007 - New: MSG_process_yield(). Stop and yield to other processes.
1008 - New: MSG_process_daemon(). Daemon processes are automatically killed
1009 when the last non-daemon process terminates
1010 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1011 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1012 to make room for the upcoming network energy plugin.
1013 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1016 - Ensure that an actor can kill itself with Process::exit()
1017 - Kill the obscure NativeException. Nobody want to survive the issues
1018 it denotes, so use JniException that is a RuntimeException (not to
1019 be caught explicitly).
1020 - Partial bug fix in initialization. SimGrid flags on command line were
1021 consumed at C level but stayed in the original Java String[] args.
1022 This could mess users' args[i] if SG flags were not put at the end of
1024 The SimGrid flags are now removed from the Java arguments. However,
1025 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1026 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1027 It might be if you have --log or --cfg flags in the command line.
1028 - Fix numerous memleaks all around the place. In particular, around VMs.
1032 - s4u::onPlatformCreated: right before the simulation starts
1033 - s4u::onSimulationEnd: right after the main simulation loop
1034 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1035 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1036 event from the availability_file changes the avail speed.
1037 - Links are now usable from s4u
1038 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1039 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1040 - New examples: The conversion of MSG examples to S4U has begun
1041 - Actors: create, kill, migration, and suspend
1042 - Applications: master-worker and token-ring
1043 - Action replay: communications and storage
1044 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1047 - Backwards Compatibility breaks
1048 - The SD_route_get_list and SD_route_get_size functions have been removed.
1049 They are replaced by the unique
1050 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1051 The route from src to dst is built in the links dynar whose size can
1052 be obtained with xbt_dynar_length.
1053 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1054 sg_host_route_bandwidth and sg_host_route_latency.
1055 Macros ensure the backwards compatibility, but you should fix your code
1058 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1059 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1060 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1061 We don't support 100% of it, but it should be enough. Contact us if not.
1062 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1063 naive and probably inaccurate)
1064 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1065 mapped onto a unique small file using some system magic.
1066 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1067 - Fortran: cleanups, fixes, support of user-added operations
1068 - RMA: MPI_Accumulate are applied in correct order.
1069 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1070 - Better support for MPI_IN_PLACE
1071 - Support for MPI_Win attrs and keyvals.
1072 - Support MPI_Comm_group_create, MPI_Type_size_x
1073 - MPI ops should be properly applied to all allowed types
1074 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1075 algorithm by a non failing one from mpich.
1078 - Kill the fifo data container: we don't use it anymore.
1081 - A <cluster> can now be created with different speed values to represent pstates.
1082 see examples/platforms/energy_cluster.xml for an example.
1084 SimGrid (3.14.159) Released December 28. 2016
1086 The previous release was only a crude approximation of a Pi release;
1087 this one should be a bit better.
1089 - Revert a last minute change that broke on Mac OSX.
1090 - Fix the changelog and NEWS files.
1091 - Improve the documentation, in particular of the routing module.
1092 - Integrate some patches flying around in Debian.
1094 SimGrid (3.14) Released December 24. 2016
1097 * The whole documentation was reworked and reorganized. There is
1098 still much room for improvement here, but we're on it.
1101 * We now need python3 for our testing infrastructure.
1102 * Model checking needs libevent
1105 * Now works on FreeBSD
1108 * Privatization now works on FreeBSD
1110 * Privatization is now activated by default in smpirun
1111 You can turn it off with -no-privatize if something goes wrong.
1113 * Call-location tracing for SMPI.
1114 You can add the exact location (filename / linenumber) of an MPI call to
1115 your trace files and slow down or speed up the simulation between two
1116 consecutive calls by using an adjustment file (see the documentation).
1118 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1119 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1120 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1121 values were ignored in some cases. The timings of these functions can now
1122 be significantly different.
1124 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1125 smpi/running-power is renamed to smpi/host-speed
1127 * smpi/grow-injected-times option to enable or disable multiplication of the
1128 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1129 make simulation less precise (but also much faster).
1131 * smpirun script should be (much) faster for large deployments.
1133 * SMPI tracing: fixed issue with poor matching of send/receives.
1135 * Replay: Fix broken waitall
1137 New functions and features
1138 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1140 Dropped / renamed functions and features
1141 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1142 - MSG_mailbox_is_empty() -> Mailbox::empty()
1143 - MSG_mailbox_front() -> Mailbox::front()
1144 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1145 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1146 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1147 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1148 * Don't pass the free_f parameter to property related functions:
1149 - MSG_host_set_property_value()
1150 - MSG_as_router_set_property_value()
1151 - MSG_storage_set_property_value()
1152 * VM properties. Since msg_vm_t are msg_host_t, just use
1153 MSG_host_get_property_value() and friends
1154 * VM I/O related things:
1155 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1156 - Unimplemented save/restore methods
1157 * MSG_as_router_get_property_value() was redundant with
1158 MSG_environment_as_get_property_value().
1159 - Removed MSG_as_router_*propert*() functions
1160 - Added MSG_environment_as_set_property_value() for consistency
1161 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1162 Well, they are still there for now, but deprecated with a warning.
1163 Please switch to xbt_dict_new_homogeneous() before this is removed
1165 * Task affinity. Its intended behavior (that was very badly tested
1166 and probably not really working) was deceiving what most users
1167 would have hoped here.
1168 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1169 * Option network/coordinates is now useless and should be dropped.
1172 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1175 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1176 * Added Dragonfly topology support
1179 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1180 EiBps, EBps, Eibps, Ebps for bandwidth.
1181 They may become useful to some lucky ones.
1184 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1185 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1186 * ex/async_waitAll: new example, on asynchronous communications
1189 * Memory usage should be decreased for simulations with a large number
1190 of processes. This also helps for SMPI.
1192 SimGrid (3.13) stable; urgency=low
1194 The Half Release, a.k.a. the Zealous Easter Trim.
1196 - We removed half of the lines, that were mostly experimental cruft.
1197 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1198 (not counting blanks and comments -- according to openhub.net)
1199 - The internals are now compiled in C++ (and will soon be clean C++)
1200 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1201 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1203 * Backwards Compatibility breaks
1204 - Removed Lua simulation bindings (switch to C or Java for that).
1205 Lua can still be used to describe platforms
1206 - Removed Java kernel plug-ins.
1207 Will be reintroduced after the ongoing major internals reorg.
1209 - the following functions were removed.
1210 They were too specific and should be reimplemented in a generic
1211 way, with filter function.
1212 - MSG_task_listen_from_host
1213 - MSG_mailbox_get_count_host_waiting_tasks
1214 - MSG_mailbox_put_with_timeout was removed.
1215 Please use MSG_task_send_with_timeout instead.
1217 - the SD_application_reinit function was removed. It has been a noop for a while.
1218 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1219 replaced by a more flexible mechanism.
1220 - The following functions thus do not exist anymore
1221 - SD_workstation_get_access_mode
1222 - SD_workstation_set_access_mode
1223 - SD_workstation_get_current_task
1224 - Basic estimation functions have been removed but can easily be replaced
1225 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1226 - SD_workstation_get_computation_time => amount / sg_host_speed()
1228 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1229 Use VM.getSpeed()*load/100 for the legacy behavior.
1231 - option enable_tracing was removed. It was not doing anything for a while.
1232 - In the ModelChecker:
1233 - the model-checker now ptraces the model-checked process which means
1234 you cannot use a debugger on the latter anymore (we might make this
1235 optional in the feature);
1236 - removed soft-dirty page tracking;
1237 - remove model-checked side snapshot management,
1238 MC_snapshot() and MC_compare_snapshot();
1239 - keep the MC_cut() function as a stub (it was not really working
1240 in the previous release).
1243 * All options are consistently in kebab-case. Old names are kept as alias.
1246 * Switch to platform v4 format.
1247 - Rename from 'power' to 'speed' the attributes describing the amount of
1248 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1249 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1250 - In <host> and <link>, attributes availability and state are gone.
1251 It was redundant with state and availability traces, and with peak values.
1252 - In <cluster>, attributes availability_file and state_file are gone.
1253 It was too complex and unused.
1254 - Kill <gpu>. Was not doing anything.
1255 - The DOCTYPE points to the right URL:
1256 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1257 (the file at this address now documents the changelog since its v1)
1258 - A warning is emitted for unit-less values (they are still accepted).
1259 - speed. Default: 'f' or 'flops'. Also defined:
1260 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1261 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1262 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1263 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1264 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1265 - latency. Default: 's' second. Also defined:
1266 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1268 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1269 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1272 * s4u::Host is now the preferred public interface to the Host features.
1273 sg_host_* functions are C bindings to the exact same behavior
1274 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1277 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1278 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1279 They were deprecated since a few years, and probably did not even compile anymore.
1282 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1283 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1284 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1285 encouraged to update their codes. A list of the modified functions can be found at the end of
1286 include/simgrid/simdag.h
1289 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1290 the lambda or closure passed as a parameter will run in kernel mode.
1291 All the callback functions should be rewritten to that interface at some point.
1294 * Reorganizing and cleaning the internals all around the place.
1297 * Remove old default barrier/bcast buggy algorithms (see #18407)
1298 * Various bug fixes to handle more codes
1299 * Remove the need for the --foreground option of smpirun (it is still
1300 accepted for backward compatibility).
1303 * Kill the setset data container: MC don't use it anymore.
1304 * Kill the queue data container: it made more sense with GRAS.
1305 * Kill the xbt_peer_t data type: it's useless without GRAS.
1306 * Kill rm_cb feature of config sets: it was never useful.
1307 * Kill graphxml parsing feature. It was not used.
1308 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1310 - xbt_dynar_sort_strings(), when the content is char*
1311 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1312 They throw exceptions on invalid input;
1313 * C++ support for declaring CLI flags (simgrid::config::Flag);
1314 * class for abstracting different signal backends (simgrid::xbt::signal).
1315 with no external dependencies (we need very simple signals).
1318 * refactoring and cleanup of the code;
1319 * ongoing process to cleanly separate the model-checking algorithms
1320 from the code model-checking support.
1322 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1324 SimGrid (3.12) stable; urgency=low
1326 The Facelift Release.
1329 * Require g++ v4.7 at least to not speak prehistorical C++.
1330 * Require Boost 1.48 (for signal2 component).
1331 * Java must be version 7 at least when activated.
1332 * Builds on Windows again (including Java bindings).
1333 * Tracing is now always enabled (no way to turn it out)
1334 * Remove GTNetS. It was not working anyway.
1335 * Various cleanups in the cmake scripts.
1336 * Move headers around to sort them out on installed systems:
1337 - instr/instr.h -> simgrid/instr.h
1338 - instr/jedule/* -> simgrid/jedule
1339 - simdag/datatypes.h was removed
1340 - simdag/simdag.h -> simgrid/simdag.h
1341 - msg/datatypes.h was removed
1342 - msg/msg.h -> simgrid/msg.h
1345 * Interface improvement:
1346 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1347 - Sanitize the interface in MSG_task_ module:
1348 - Merge two functions that were close enough but misleading:
1349 set_compute_duration(t) -> set_flops_amount(t)
1350 get_remaining_computation(t) -> get_flops_amount(t)
1351 - set_data_size(t) -> set_bytes_amount(t)
1352 get_data_size(t) -> get_bytes_amount(t)
1353 - Massive cleanups in the functions related to the energy
1354 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1355 - New: MSG_host_get_pstate()
1357 - msg/energy/onoff: switching hosts on and off
1359 * Interface improvement:
1360 - Rename Host.isAvail() to Host.isOn()
1361 - Rename Process.currentProcess() to Process.getCurrentProcess()
1362 - Rename Task.setDataSize() to Task.setBytesAmount()
1363 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1365 - #18874: Actually allows the GC to reclaim tasks
1369 - SIMIX_process_throw: raises an exception in a remote process
1370 * Refactoring: Separate sub-modules
1371 - libsmx: the public interface, as libc in a real system
1372 - popping: the strange dance that converts a user request into a kernel handling
1373 - smx_context_*: the virtualization mechanisms that embed the user code
1374 - smx_*: the handling of each simcalls
1375 * Interface cleanup:
1376 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1377 * Rename smx_action_t into smx_synchro_t, making explicit that these
1378 things are used to synchronize processes with their environment.
1379 For example, a communication is a sort of synchronization involving
1380 the communicating processes (that may block until the exchange) and
1381 the platform. The same can be said from computations, etc.
1383 - #18888: segfault when a process ends before its kill_time
1387 - 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.
1388 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1389 - MPI_Type_set_name, MPI_Type_get_name
1390 - MPI_*_c2f and MPI_*_f2c functions
1391 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1392 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1393 - Activate a lot of new tests from the mpich 3 testsuite
1395 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1396 - InfiniBand network model added: Based on the works of Jerome Vienne
1397 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1398 - When smpi/display_timing is set, also display global simulation time and application times
1399 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1400 * Collective communications
1401 - 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
1402 - 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/).
1403 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1405 - Replay now uses algorithms from wanted collective selector
1406 - Replay can be used with SMP-aware algorithms
1407 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1408 - 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.
1410 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1411 - Correctly use loopback on fat-tree clusters
1412 - Asynchronous small messages shouldn't trigger deadlocks anymore
1413 * Energy/DVFS cleanup and improvement
1414 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1415 - new: smpi_get_host_pstate()
1418 - "Full" network optimization flag was broken since Surf++
1419 - Better handling of precision flags in maxmin
1420 - Fix bug causing sometimes "Impossible" errors
1421 - Properly pass cluster properties to included hosts
1422 * Improvement of the Energy plugin.
1423 - Always update the consumption before returning that value
1424 - New property: watt_off to denote the dissipation when the host is off
1425 - New functions getWattMinAt and getWattMaxAt to retrieve the
1426 dissipation of pstates that we are not currently at.
1427 * Java: class NetworkLink renamed to Link
1428 * New function: simcall_process_get_kill_time()
1429 * Massive rename s/workstation/host/
1430 - That's intrusive, but that's good for the project consistency. Sorry.
1431 - Change config option "workstation/model" into "host/model"
1434 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1435 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1436 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1437 - xbt_dirname and xbt_basename for non-POSIX systems
1439 * The model checker now runs as a separate process.
1440 * The model checker runs must now be launched with the new simgrid-mc program.
1441 * Record/Replay: the MC can display a textual representation of a path in the
1442 execution graph. It can then be replayed outside of the model checker.
1444 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1446 SimGrid (3.11) stable; urgency=low
1451 * Normalizing pointers addresses tool for better diff between logs
1454 * Add cloud examples using new VMs
1455 - examples/msg/cloud/two_tasks_vm.tesh
1456 - examples/msg/cloud/simple_vm.tesh
1457 - examples/java/cloud/cloud.tesh
1458 - examples/java/cloud/migration/migration.tesh
1459 * Add java surf examples:
1460 - examples/java/surfPlugin/surf_plugin.tesh
1461 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1462 - examples/java/surfCpuModel/surf_cpu_model.tesh
1463 * Add SMPI+MSG example:
1464 - examples/smpi/smpi_msg_masterslave/
1469 - msg task destroy cancel test
1470 - msg_host on/off test
1471 * Move all tests in testsuite to teshsuite (adding tesh files)
1472 * Restructure teshsuites
1473 - one folder for each kind of test
1474 * Restructure AddTests.cmake
1476 - structure the order of tests (with sections)
1479 * Add virtual machine
1480 - creation of a VM on a PM
1481 - migration of a VM from a PM to another PM
1483 - MSG_process_join(msg_process_t process, double timeout)
1484 - msg_bar_t MSG_barrier_init(unsigned int count)
1485 - int MSG_barrier_wait(msg_bar_t barrier)
1486 - void MSG_barrier_destroy(msg_bar_t barrier)
1487 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1488 * New option "msg/debug_multiple_use" to help debugging when a task is used
1491 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1493 - Increase code coverage in test suites
1497 * Protect context stacks against stack overflow. The number of protected memory
1498 pages allocated on the top of each stack (1 by default) can be configured
1499 with the new command line option --cfg=contexts/guard_size:<value>.
1500 * Simcalls are now generated by a python script that
1501 - generates files included by SimGrid
1502 - checks that all the functions exist, and proposes prototypes
1504 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1505 host_set_data, host_get_data
1507 - simcall_process_join(smx_process_t process, double timeout)
1508 * Fix bug where sleeping processing could not be suspended.
1511 * Translate surf models from C to C++
1512 - Generic classes for all models: Model, Resource, Action
1513 - A generic interface for each kind of model (CPU, Network, Storage
1514 Workstation, WorkstationVM)
1516 * Translate surf routings from C to C++
1517 * Add callbacks using sigc++ or boost::signals2
1518 - Add callback functions for resource creation/destruction
1519 - Add callback functions for action state change
1520 - Handle Energy as a plugin
1521 * Replace swag by boost::intrusive
1522 * Add new routing models for clusters. For documentation, see
1523 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1524 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1525 parameters for cluster tag
1526 - Fat trees, with topology="FAT_TREE" and
1527 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1528 - see examples/platforms/torus_cluster.xml and
1529 examples/platforms/fat_tree_cluster.xml
1530 * More documentation
1533 * Hostfiles support host:nb_processes construct to deploy several processes on
1535 * Collective communication algorithms should not crash if used with
1536 improper number of nodes and report the error.
1537 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1538 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1539 MPI_Dims_create, MPI_Cart_sub are supported.
1540 * New interface to use SMPI programmatically (still depends on MSG for
1541 some parts, see examples/smpi/smpi_msg_masterslave):
1542 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1546 * Global variables privatization in MPI executables is now performed at runtime
1547 with the option smpi/privatize_global_variables (default:no).
1548 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1549 dynamic libraries loaded by the application are not privatized (static
1550 linking with these libraries is advised in this case)
1553 * Options defined in XML work correctly now.
1556 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1557 copied into simgrid.jar or not (ON by default). Use this option if you want
1558 to reduce the size of the installed simgrid.jar, *and* the native libraries
1559 are kept installed elsewhere.
1560 * Surf binding with SWIG (code generated in maintainer mode only):
1561 - plugin to handle callbacks
1562 - CPU model only for the moment
1565 * Supernovae build mode is definitively removed. It was used to improve
1566 inlining and inter-module optimizations. It is nowadays superseded by
1567 link-time optimizations commonly available in compilers.
1568 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1569 * Add boost dependency for surf++
1570 * Add new macro for tests
1571 - ADD_TESH(name <tesh_args>)
1572 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1576 - xbt_bar_t XBT_barrier_init(unsigned int count)
1577 - int XBT_barrier_wait(xbt_bar_t barrier)
1578 - void XBT_barrier_destroy(xbt_bar_t barrier)
1579 * Make the xbt_os_time module public
1581 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1583 SimGrid (3.10) stable; urgency=low
1585 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1588 * Reintegrate Java to the main archive as desynchronizing these
1589 package is not acceptable anymore (Java is now considered stable)
1590 * Add explicit synchronization facilities through semaphores
1591 * Bug fix: Task.setDataSize() only changed the C world, not the value
1592 cached in the Java world
1595 * Dramatically change the way files are handled. API and internals changed, but
1596 this part of MSG was not considered as production grade either.
1597 * Add explicit synchronization facilities through semaphores
1598 * Add a new function MSG_host_get_process_list()
1599 * Preliminary DVFS support (see examples/msg/energy/ for details)
1602 * SMPI is now included directly in the libsimgrid as the windows
1603 linker doesn't force us on splitting it anymore.
1604 * Improvements of the SMPI replay tool:
1605 - Most of the collective communications are now rooted in the same process as
1606 in the original application.
1607 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1608 used until now). Multiple data types can now be used in a trace.
1609 - The replay tool now supports traces produce either by TAU or a modified
1611 - Bug Fix: the compute part of the reduce action is now taken into account.
1612 - Gatherv collective is now supported
1613 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1614 of smpirun outputs time independent traces for the current run. One file
1615 is created per process. If too many processes are simulated, this behavior
1616 can be changed to one file for all processes by using the
1617 tracing/smpi/format/ti_one_file flag
1618 * smpirun generates the host file if needed (with given host count and platform)
1619 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1620 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1621 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1622 --cfg=smpi/coll_selector:(mpich/ompi)
1623 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1625 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1626 happening inside a collective SMPI call.
1627 * Fix the behavior of complex data types handling.
1628 * Make MPI_Wtime another synchronization point to take computations into
1630 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1631 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1632 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1633 (no privatization of global variables yet)
1634 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1637 * Verification of liveness properties is now available for SMPI applications
1638 (in addition to MSG applications)
1639 * Bugged examples using SMPI in examples/smpi/mc/
1640 * Add --cfg=model-check/visited option. Allows the verification of infinite
1641 programs. Detection of loops in the execution thanks to the system state
1642 comparison and reduction of the state space to explore. Can be combined with
1643 DPOR for safety properties.
1646 * Allow to change SimGrid configuration (see --help) within the code
1647 thanks to SD_config() as it can be done in MSG.
1648 * Add a new function SD_task_set_amount() upon user request.
1651 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1652 * Remove rule based routing (no more PCRE dependency)
1653 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1654 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1655 * Add a loopback_bw and loopback_lat options to cluster tag.
1656 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1657 routing. Now peers are encapsulated in an AS and have their own private
1658 router but this is transparent.
1661 * Our own implementation of getline is renamed xbt_getline, and gets
1662 used even if the OS provide a getline(). This should reduce the
1663 configuration complexity by using the same code on all platforms.
1664 * New type: xbt_cfg_elm_boolean.
1665 * Allow to use yes/no for boolean configuration options in the command line.
1666 * Allow to disable SimGrid cleanups at exit from command line option.
1667 There are situations where one may want a simulation to end with an exit.
1668 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1669 annoying when scripting around the simulator. Adding a
1670 --cfg=clean_atexit:no allows to circumvent this issue.
1673 * Lots of memory leaks were corrected in this release.
1674 There is no known memory leaks anymore, in all of our 600+ tests.
1675 * New command line option --version, to get SimGrid version information.
1676 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1678 * Supernovae builds are deprecated, and expected to be removed in the next
1681 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1683 SimGrid (3.9) stable; urgency=low
1685 The Grasgory release: GRAS is really dead now.
1687 * Complete overhaul of the internal host structures scheme.
1690 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1691 was considered as experimental and badly maintained since then.
1692 * Keeping it was thus a trap to our potential users, that could take
1693 it instead of MSG or SMPI by mistake despite is pity state.
1694 * GRAS seems to have very few users (if any), and no one volunteered
1695 to maintain it further. It also induces a lot of XBT code (for
1696 portability sake), that must be maintained too.
1697 * For all these reasons, we killed GRAS. If someone wants to revive it
1698 in the future, don't cry, our git history still remembers of GRAS.
1701 * Major overhaul. Merge our documentation again as time proved that
1702 splitting it was really not helping our users.
1703 * Further improve the developer documentation to help newcomers
1704 hacking on SimGrid itself. The user documentation (and in
1705 particular, the beginner documentation) is still in a sorry state.
1708 * Now works on Windows too!
1709 * Much more extensive test suite, from MPICH
1712 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1713 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1714 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1715 * Bug fix: let task be scheduled when the last dependency to be solved is
1716 a control dependency.
1717 * Remove SD_load_environment_script function.
1718 Use the C sg_platf function if you want to declare a platform
1722 * New function: MSG_process_get_number()
1723 * Old function documented: MSG_config()
1724 * Remove MSG_load_platform_script function
1725 Use the C sg_platf function if you want to declare a platform
1729 * Change the default value of the TCP_gamma constant (maximal size of TCP
1730 congestion window) to a more realistic 4MiB value. If you notice changes in
1731 your simulation results, you can fall back to the previous 20k tiny window
1732 by adding --cfg=network/TCP_gamma:20000 on command line.
1733 * (Hopefully) fix a bug wrt periodic availability and state traces
1734 * Bug fix: use default values at start when first event in availability/state
1735 trace is not at time 0.
1738 * remove the "new_" part of function name sg_platf_new_trace_connect
1739 (resulting in sg_platf_trace_connect), since it does not create
1743 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1744 use case where it's really mandatory, and maintaining it was a pain in
1746 * New: xbt_fifo_search(), search an item with a user-provided
1747 comparison function instead of dumb pointer comparison.
1750 * Fix the lua deployment:
1751 Use `simgrid.init_application()` before deployment instead of
1752 `simgrid.msg_register_application()` after.
1755 * Transfer the tracing files into the corresponding modules.
1757 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1759 SimGrid (3.8.1) stable; urgency=low
1761 The "we are told that some people want to also *install* the simgrid
1764 * Add missing file "tesh.1" to the archive.
1766 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1768 SimGrid (3.8) stable; urgency=low
1770 The Psssshiiiit release: SimGrid jumps into the Cloud.
1773 * Add an experimental interface to manipulate VMs. They are mainly
1774 process groups with very few intrinsic semantic, but they should
1775 allow you to build the semantic you want easily.
1776 * New function: MSG_host_set_property_value()
1777 * New function: MSG_process_on_exit(). To clean memory in all cases.
1778 * Bug fixes that made the host (and link) failures unusable.
1779 * Add a way to auto-restart process when the host in which they are
1780 executing comes back (ON_FAILURE="RESTART" on deployment file,
1781 MSG_process_auto_restart_set).
1782 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1783 please stop using the old ones, they are DEPRECATED.
1785 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1786 Please use MSG_init() instead. (reducing the amount of entry
1787 points in the library helps us).
1788 * Make it impossible to link against the wrong version of the lib
1789 * Deprecate MSG_clean(). No need to call it anymore.
1790 * Function MSG_get_host_number() is not deprecated anymore.
1793 * Split the doc into a user guide and a reference guide.
1794 * Start a developper guide to help people hacking on SimGrid.
1797 * Enable tracing by default. This modules rocks you should use it.
1798 * Remove option custom_flags. Now use environment variables CFLAGS
1800 * Use default cmake things to detect lua instead of home grown ones.
1801 * New option "enable_mallocators" to disable mallocators, for debugging
1802 purpose ("on" by default).
1805 * Bug fixes around the resource failures: don't let the processes
1806 survive the host they are running onto.
1807 * Add an interface to auto-restart processes when the host in which they are
1808 executing comes back.
1809 * Ensures that SIMIX_clean is called automatically. It's not part of
1810 the public interface anymore (bindings should be updated).
1813 * Bug fix for when SD_Simulate is called with a positive value: be careful
1814 when comparing doubles. Sometimes they are different for non significant
1816 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1817 parallel task whose initial work is distributed among host according
1818 to the Amdahl's law. Such tasks are created with a parameter alpha
1819 that corresponds to the non-parallelizable part of the computation.
1820 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1821 between two sets of workstations assuming a 1D block distribution (each
1822 workstation owns a similar share of data) on both sides.
1824 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1825 Data redistribution will be automatically scheduled once parent and child
1826 are both scheduled. The filling of computation_amount and
1827 communication_amount structures is now done seamlessly thanks to the chosen
1829 * New function SD_workstation_dump to display various information
1830 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1831 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1832 bandwidth on the route onto which the task is scheduled and the amount of
1834 To divide the nominal bandwidth by 2, the rate then has to be:
1835 rate = bandwidth/(2*amount)
1836 * Compute tasks that have failed can now be rescheduled and executed again
1837 (from their beginning)
1838 * Increasing source code coverage (src/simdag is now covered at 95.8%
1842 * Re-implement time-independent trace replay using SMPI (at the
1843 smpi_smp_* level) instead of MSG. This should replace
1844 examples/msg/actions/actions.c
1845 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1846 hindexed and structs)
1847 * Implement the exchange of non-contiguous data.
1848 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1849 * Correct behavior of smpi/sender_gap and set its default value to 0
1850 * Add option to asynchronously send small messages to allow better
1851 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1852 specifies the size in bytes under which messages will be asynchronously sent.
1853 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1854 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1857 * New C interface to define a platform: XML is now optional.
1858 For more info, please check include/simgrid/platf.h.
1859 * New interface to define random platforms from the C:
1860 For more info, please check include/simgrid/platf_generator.h and
1861 examples/msg/masterslave/masterslave_platfgen.c
1862 * Export a sg_cmdline dynar containing all the arguments we got from
1866 * Two new tracing options for adding comments to trace file so you
1867 can track your experiments (see --help-tracing for details).
1868 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
1869 * Adding the SimGrid version that generated the trace file as a comment.
1870 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1871 * Fix to avoid key clashes on Paje links
1872 * Other minor fixes related to the Paje specification
1875 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1876 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1877 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1878 * Greatly improve the robustness of mmalloc to user errors (such as
1879 using an area after freeing it, or freeing it twice)
1881 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1883 SimGrid-java (3.8.1) stable; urgency=low
1885 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1886 Warning: all methods are visible, but only some of them are
1887 implemented so far. Check the source (src/jtrace.c)
1888 for further information.
1889 * New module: org.simgrid.msg.File (SimGrid File management functions)
1890 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1891 * Change the meaning of Process.restart: now restart the process from
1892 the beginning, like MSG_process_restart in C.
1893 * Add Process.setAutoRestart: handling of process restart when failed
1895 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1896 you to retrieve the properties of the processes/hosts
1897 * Deprecate Msg.clean(): you can just forget about it now.
1898 * New function Process.getCount(), that only works when compiling
1899 with the not yet released version 3.9 of the C library.
1901 * New context factory based on Coroutines. It mandates a modified JVM
1902 but then, the simulations run about five times faster, and there is
1903 no limit to the amount of processes (beside of the available memory).
1905 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1907 SimGrid (3.7.1) stable; urgency=low
1910 * Restore the prototype of MSG_process_create_with_environment() to
1911 the pre-3.7 situation by removing the kill_time argument.
1912 * Add a MSG_process_set_kill_time() function instead.
1915 * Fix weird behaviors when dealing with parallel tasks.
1918 * Simgrid is now built as a dll.
1919 * Simgrid-java now works on Windows.
1920 * Simgrid-Java is now included into Windows package.
1923 * First pre-build package for MacOSX.
1926 * Fix compilation when using MSG_USE_DEPRECATED.
1927 * Fix some compilation issues on Macs and Windows.
1928 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1929 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1931 We discovered that the Lua console is broken, but we are missing the
1932 manpower to fix it right now. The problem existed in 3.7 too, so we
1933 are not blocking the release for that. Sorry if you depended on this
1934 feature, any help would be really welcome.
1936 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1939 SimGrid-java (3.7.1) stable; urgency=low
1941 The "Java aint got to be bloated and slow" release
1944 * Various internal cleanups and performance improvement
1945 Simulations are expected to run up to twice faster or so
1946 * Make Process.kill(process) an instance method, not a static one
1947 * User processes are not java.lang.Thread subclasses.
1948 This breaks the compatibility (sorry), but previous API was
1949 brain-dead, making it impossible to have non-trivial
1950 initializations in the process constructor.
1951 * Require a full constructor per Process sub-class.
1952 Kinda breaks the compatibility (sorry), but this allows a much more
1953 efficient way to launch the processes at simulation startup.
1954 * Do not embeed our version of semaphores, java 1.5 can be considered
1955 as sufficiently prevalent for us to not dupplicate its features.
1959 * Add examples for almost every part of the API
1960 We spotted and fixed a lot of bugs in the process
1962 * New module: asynchronous communication API
1963 * New function: Process.sleep()
1964 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1965 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1966 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1968 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1970 SimGrid (3.7) stable; urgency=low
1972 The "spring cleanups (before the next Big Project kicks in)" release.
1975 * We can specify the SMPI latency/bandwidth factor with command line
1976 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1977 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1978 You can also use the "config tag" from platform file by adding this line
1979 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1980 (see "example/platforms/tag_config.xml" to use "config tag").
1981 Note that the command line supersedes the platform file configuration.
1982 * Change the correction factors used in LMM model, according to
1983 the latest experiments described in INRIA RR-7821.
1984 Accuracy should be improved this way.
1985 * Use the partial invalidation optimization by default for the
1986 network too. Should produce the exact same results, only faster.
1987 * Major cleanup in surf to merge models and split some optimization
1988 mechanisms from the core of the models. As a result you can now
1989 specify which model to use (e.g., --cfg=network/model:LV08
1990 --cfg=cpu/model:Cas01) and which optimization mode to use
1991 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1992 Incompatible combinations should err at initialization. See
1993 --help-models for the list of all models and optimization modes.
1994 * The CLM03 workstation model was dropped for simplicity because it
1995 used the deprecated CM02 network model. Use default instead.
1996 * Rename the TCP_gamma configuration option to network/TCP_gamma
1997 * Rename the coordinates configuration option to
1998 network/coordinates, and document it
1999 * Use now crosstraffic keyword instead of the terribly misleading
2000 fullduplex keyword. It is activated by default now in the current
2001 default model, use --cfg=network/crosstraffic:0 to turn it off.
2002 * Ongoing refactoring the model parsing to make XML files optional
2003 See include/simgrid/platf.h for details (still to be completed)
2006 * Major overhaul of the documentation. Almost instructive now :/
2007 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2008 functions and friends. This interface was considered as
2009 deprecated since over 2 years, it's time to inform our users that it is.
2010 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2011 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2012 use these (crappy) functions in your code.
2013 These functions will be removed soon. Stop using them now.
2014 * Deprecate MSG_get_host_{table,number}
2015 Implement MSG_hosts_as_dynar() instead.
2016 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2017 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2020 * Stabilize the parallel execution mode of user contexts
2021 * Introduce configuration variables to control parallel execution:
2022 - contexts/synchro: Synchronization mode to use when running
2023 contexts in parallel (either futex, posix or busy_wait)
2024 - contexts/parallel_threshold: Minimal number of user contexts
2025 that must be part of a scheduling round to switch to parallel
2026 execution mode (raw contexts only)
2027 * Fix bugs that prevented to use suspend/resume along with
2028 synchronization structures.
2029 * Fix bugs in process termination that lead to invalid memory access
2030 in very specific conditions.
2033 * Introduce a parallel mode for the models (controlled by surf/nthreads
2034 configuration item). In our tests, running the models in parallel
2035 never lead to any speedups because they are so fast that the gain
2036 of computing each model in parallel does not amortizes the
2037 synchronization costs, even when ultra fast futexes are used.
2038 This is released anyway because YMMV.
2041 * Performance boost by using a swag internally to compute the set of
2042 tasks that are finished and should constitute the return value of
2046 * Enable it by default now that it is considered rather stable.
2049 * Documentation of the tracing functions.
2050 * Performance gains when tracing categorized/uncategorized resource
2051 utilization by avoiding calls to get route when updating resource
2052 variables. LMM constraints are being used instead.
2053 * API changed to set task categories. Use MSG_task_set_category instead
2054 of TRACE_msg_set_task_category, and SD_task_set_category instead
2055 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2056 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2057 * Examples for link user variables added (see at examples/msg/tracing/)
2058 * Deprecated function TRACE_msg_set_process_category completely removed
2059 * Trace header updated according to the latest Paje file format
2060 * Tracing network lazy updates, no longer obligate users to use full updates
2061 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2062 * Experimental: let user code declare/set/push/pop application states for hosts
2063 * API changed to allow the manual creation of graph configuration files
2064 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2067 * Improve the API of Lua MSG bindings, using the Lua spirit.
2068 * Each simulated process now lives in its own Lua world (globals are
2069 automatically duplicated). It helps writing simulators. Will allow
2070 to run Splay programs within SimGrid in the future.
2071 * Add a Chord example in Lua, equivalent to the MSG one.
2074 * Start the implementation of a solution to express temporal
2075 properties, not only local assertions. This is still an
2076 experimental work in progress, stay clear from it to be safe.
2080 - Add new runtime parameters --help-logs and --help-log-categories
2081 to display information about supported logging parameters and
2083 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2084 don't exists anymore.
2085 * Mallocators: allow value NULL for the reset function.
2087 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2088 create homogeneous dictionaries, where all the elements share the
2089 same free function. Non homogeneous dictionaries will be
2090 deprecated in the next release.
2091 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2092 - Multi-level dictionaries are deprecated.
2094 - new function xbt_dynar_search_or_negative() that is useful when
2095 you have less than 2 million elements in your dynar and don't
2096 want of the extra complexity of catching exceptions when the
2097 element is not found.
2099 - Make xbt_os_thread module (for thread portability) public.
2100 Documentation is still to come, sorry.
2102 - Cleanups and simplifications to make it maintainable again.
2103 - Exotic features (such as memalign and valloc) were removed.
2104 - The metadata were extended and improved so that the
2105 model-checker becomes able to explore and inspect the heaps.
2106 - This may induce a performance drop when enable_model-checking is
2107 ON in cmake (even if it's not used in the simulation), but it is
2108 necessary at this point to get MC working.
2110 Turn model-checking OFF if simulation performance matters to you.
2111 Not enabling it at runtime is not enough, disable it in cmake.
2113 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2116 SimGrid (3.6.2) stable; urgency=low
2118 The "Not coding new stuff allows to polish old things" release.
2121 * New bindings to the NS3 packet level simulator (experimental)
2122 * Use the raw (efficient) execution contexts instead of the sysv
2123 (portable) ones when possible.
2124 * libpcre is now mandatory in any cases since not using it led to
2125 severe performance loss and possibly other issues
2126 * Update the XML platforms:
2127 - G5K: include the latest machine in Nancy
2128 - GridPP and LCG: new platforms
2129 * Documentation was partially updated, at least (more to come)
2131 Bug fixes, cosmetics and small improvements
2132 * Free terminated processes before the end of the simulation to avoid
2133 exhausting the memory of users having very dynamic amount of
2135 * Bug fix and cosmetics about canceling non-running tasks
2136 * Bug fix about the dot loader's issues when using libcgraph
2139 * Create an installer for windows with nsis (amd64 and win32)
2140 - Add an hello world project to illustrate simgrid project creation.
2141 - Embed libpcre into the Simgrid installer to avoid
2142 its compilation burden
2143 * The raw execution contexts should work on Apple now
2144 * Port to Windows 64 bits
2145 - Sysv contexts now have an implementation for this arch
2146 - GRAS communication features now support this arch
2147 * Drop support for borland compiler on windows
2148 - this code was not maintained, and we kinda depend on gcc nowadays
2149 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2150 * Fix portability issue on unstable ubuntu: linker became picky on
2153 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2155 SimGrid (3.6.1) stable; urgency=low
2157 The "Oops, we broke Macs too" release
2160 * Fixed contexts detection so that raw ones are used when possible
2161 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2162 leads to a strange error, with user code segfaulting sometimes when
2163 the generated code is not perfectly aligned (which is not
2164 controllable from the user side, depends on the amount of code)
2167 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2169 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2171 SimGrid-java (3.6) unstable; urgency=low
2174 * Split of every thing from simgrid v3.5 into a separate package.
2176 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2178 SimGrid (3.6) stable; urgency=medium
2180 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2183 * Bindings now constitute their own package, separated from the main one.
2184 Rationale: reduce our maintenance nightmare by reducing the module coupling
2185 They will soon be released on their own on gforge.
2187 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2188 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2190 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2191 * It's not quite deprecated for now because we have no replacement,
2192 but it may soon become the case.
2195 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2196 * Fortran: New user-level cache variable to store the rank of the running
2197 process. This improves performance by an order of magnitude.
2198 * C: New coccinelle script to automatically locate and modify global and
2199 local static variables.
2200 * Improved SMPI network model with a sender-side gap to account for multiple
2204 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2205 only say if a communication is finished, no matter whether it succeeded or
2206 failed. You can call MSG_comm_get_status() to know the status of a finished
2208 * New function MSG_task_dsend() to send a task and detach it. When a
2209 communication is detached, you are never notified of its success or failure
2210 and the memory is released automatically once it is finished. This function
2211 is useful when you don't care about the end nor the success of a
2213 * Change the prototypes of action replay. Sorry for inconvenience,
2214 but this is really more efficient this way (and to adapt your code,
2215 you just have to fix the initialization, that shouldn't be too long)
2216 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2217 ever managed to do anything useful with it.
2218 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2219 takes a dynar of communications. It returns immediately and gives the
2220 index of a finished communication (if any).
2221 * New example: a basic implementation of the Chord P2P algorithm.
2224 * New model for multi-core CPUs. You can now use the core attribute to
2225 precise the number of cores of a host. This is a basic model. Every
2226 process running on the host receives at most the power provided in
2227 the DTD (throughput<=power). Total throughput of process cannot exceed
2229 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2230 router linked by an up-link and a down-link (possibly asymmetrical).
2231 This kind of pattern allows to easily build last-mile model style platforms.
2232 Aggregating such patterns in a rule-based AS is thus the technique of
2233 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2234 * New model for Vivaldi routing. We transformed the Vivaldi network model
2235 into a Vivaldi routing model (based on the rule-based model). This allows to
2236 combine Vivaldi based latencies with last-mile platforms.
2239 * Added a check for NaN of IEEE754 infinite in the double entries of
2241 * Introduce a new context factory "raw", highly inspirated from the
2242 ucontext factory, but using manually crafted functions in assembly to
2243 do the work in an efficient manner.
2244 * Allow to change the used context factory at run time, not only at
2245 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2246 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2247 contexts at runtime (only with raw contexts or ucontexts).
2248 * Completely rewrote this module to allow parallel execution of user
2249 processes. Use --cfg=contexts/nthreads:N to execute user processes
2250 with N parallel threads (the default is 1, meaning no parallelism).
2251 * Allow to decide dynamically between sequential and parallel modes.
2252 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2253 processes in parallel only when their number is greater than or equal to P
2255 * Added a check for NaN of IEEE754 infinite in the double entries of
2259 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2260 won't output the state of processes when interrupted with Ctrl-C
2261 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2262 NULL-terminated array. This may solve backward compatibility issues
2263 due to the change to return type of SD_simulate. See also:
2264 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2265 * Add new macros with variable number of arguments.
2266 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2267 - in xbt/asserts.h: xbt_assert
2268 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2269 - in xbt/ex.h: THROWF and RETHROWF.
2270 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2272 * Change xbt_die() to accept a format string with arguments, just like printf.
2273 * New data structure: xbt_lib_t, like a dict but more general and with better
2277 * New configuration options
2278 Options triva/categorized and triva/uncategorized can be used to generate
2279 graph configuration files for Triva visualization tool.
2280 * Configuration option tracing/platform is renamed to tracing/categorized
2281 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2282 * New TRACE_user_link_variable interface:
2283 User provides the name of the link and the tracing variable to attach to it
2284 * the declaration of tracing categories must be done after the environment
2286 * simpler tracing interface, just one way to declare categories
2287 TRACE_category or TRACE_category_with_color, it is up to you
2288 * links in the trace file are again identified by their names
2289 * trace contains the full platform hierarchy exactly as declared using the ASes
2290 * Options tracing/msg/[task|process]:1 groups the process by hosts
2291 for both cases, tasks and processes must have names that are unique during the simulation
2292 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2293 * The experimental option tracing/msg/volume is deprecated
2294 its functionality may be reincorporated if needed
2296 The tracing generates a trace file with unordered timestamped events,
2297 because of the way the core simulator (surf) works. A script available
2298 at the tools directory (fix-paje-trace.sh) can be used to put the events
2299 in order. We have changed the tracing so it can generate ordered timestamped
2300 events in the final trace, but depending on the simulator (and how much time
2301 is simulated) that can lead to a huge memory utilization. It is deactivated
2302 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2304 Build Infrastructure
2305 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2306 - We are trying hard to keep the API stable, but it may happen that
2307 some things change (we're a research project after all, not a
2308 nuclear plant operating system). If such things should happen, you
2309 could rely on that macro to adapt.
2310 - current value: 30600 for 3.06.00, aka 3.6
2311 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2312 a number that can be compared with SIMGRID_VERSION.
2313 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2314 assertions and verbose/debug logging events are disabled at compile time.
2316 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2318 SimGrid (3.5) stable; urgency=medium
2321 * New feature to any SimGrid-based simulator: Model-Checking
2322 Check SIN#1 for more details.
2325 * New Model SMPI (three-interval linear regression for correction factors)
2326 See RR-7426, available at http://hal.inria.fr/inria-00527150
2327 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2328 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2329 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2330 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2331 * See also src/smpi/README
2336 - Tracing API changes: TRACE_start and TRACE_end should not be called
2337 by user-code. They are automatically called by simulators created
2338 with SimDAG, MSG and SMPI if the toolkit is compiled with
2339 tracing_enabled=ON. Categories declaration and utilization remain the
2340 same for MSG and SimDag.
2341 - A function was added to the tracing API to declare categories with
2343 - TRACE_category_with_color (char *category, char *color)
2344 where color must be in the following format
2345 "%f %f %f", red, green, blue
2346 and red, green, blue are float values in the interval [0, 1]
2347 - User can specify NULL as color parameter, or continue calling
2348 TRACE_category (cat)
2349 On that case, the tracing system will define random colors
2350 - The following command-line options are supported:
2351 --cfg=tracing/filename:msg.trace
2352 --cfg=tracing:1 (activate tracing, needed to use others)
2353 --cfg=tracing/platform:1 (categorized resource use)
2354 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2355 --cfg=tracing/msg/task:1 (task creation)
2356 --cfg=tracing/msg/process:1 (process creation, migration)
2357 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2358 --cfg=tracing/smpi:1 (SMPI interface tracing)
2359 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2360 - examples of examples/msg/tracing updated
2362 - DAXLoader and DOTLoader functions can generate tasks with categories
2363 - A new function to attribute a category to SD tasks:
2364 TRACE_sd_set_task_category (SD_task_t task, char *category)
2365 * Tracing the MPI interface implemented by SMPI
2366 - Collective operations are traced with states
2367 - Point-to-Point operations are traced with states/links
2368 - Tracing activated by a parameter "-trace filename" passed
2369 to smpirun during execution (considering that simgrid
2370 is compiled with tracing enabled)
2371 - To run the simulation with gdb, the simulator
2372 accepts --cfg=tracing/smpi:1 to trace SMPI
2373 - tesh files to check if smpi tracing is ok
2374 - See examples/smpi/NAS/DT-trace
2375 * GTNetS tracing re-worked
2376 - adaptation to the tracing system of GTNets to cope
2377 with modifications regarding the fullduplex mode
2378 - new tesh files to check if gtnets tracing is ok
2381 * Asynchronous communications through the functions:
2382 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2383 * New function: MSG_load_platform_script()
2384 to make possible using a lua script instead of XML files to set up platforms
2385 * New function: MSG_set_function
2386 to associate functions to processes, used when bypassing the parser
2387 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2389 Platforms: Add some more examples in examples/platforms
2390 * Grid'5000: see www.grid5000.fr
2391 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2394 * Change the XML format. This is a very important modification. SimGrid
2395 3.5 introduces a new hierarchical format based on the notion of
2396 Autonomous Systems. Compatibility with old format is ensured
2397 through the perl script provided in the install bin directory
2398 bin/simgrid_update_xml.
2399 It is now possible to build platforms with specific routing
2400 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2401 connect several platforms together. We will try to provide soon
2402 set of realistic platforms exploiting these properties (have a
2403 look at examples/platforms/ for the moment).
2404 * Take the opportunity of the XML format change to be a good XML citizen:
2405 rename link:ctn to link_ctn and similar changes (also dealt with by
2407 * Add a new routing scheme (rule-based) using regular expressions. It
2408 enables to have an extremely low memory footprint when the
2409 underlying routing is simple and can be compactly described. You
2410 need to have libpcre4-dev (perl regular expressions) installed if
2411 you want to use this routing scheme.
2412 * Revive the cluster TAG and allow to easily and efficiently (both in
2413 term of memory and speed) connect clusters together. Have a look
2414 at teshsuite/simdag/platforms/ to see how this can be done. With
2415 this tag, you can create clusters with thousands of tasks at no
2416 cost (have a look at examples/platforms/).
2417 Note: clusters are implemented as ASes, so there is no need for an
2418 enclosing AS tag if you have only one cluster in your platform.
2419 * Add new generic functions in the public interface that allows the user
2420 to call SURF 'create_resource' methods from your code (same
2421 functionality as the XML bypass mechanism but with a much lighter
2423 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2424 that uses a piecewise linear approximation to produce better
2425 results when exchanging small messages.
2426 * Add a new parameter to handle correctly full duplex link and account
2427 for interferences between uplink and downlink communications
2428 (activate with --cfg=fullduplex:1).
2431 * Rename the SD_READY (all dependencies are satisfied and task is
2432 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2433 dependencies are satisfied) state.
2434 This prevents a confusion between the notion of "ready to schedule"
2435 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2436 simulated" (SD_RUNNABLE) used by the simulation kernel.
2437 * Change the way a task is considered as ready. Instead of removing
2438 dependencies when a task is done, a counter is decreased. This way,
2439 it is always possible to reach ancestors thanks to the
2440 SD_taks_get_parents function (even after the end of the simulation.)
2441 * Change the return type of SD_Simulate from (SD_task_t*) into
2442 xbt_dynar_t. This function was in handling a dynar internally and
2443 converted it into a NULL terminated array for historical reasons.
2444 * New function SD_dotload(char*) to load a DAG described in dot
2445 format. This loader and the corresponding examples require the
2446 installation of the graphviz library.
2447 * Fix a bug in the management of tasks of size 0 in the surf network
2448 models. This problem was only visible with SIMDAG and you should
2449 thus disregard results produced with earlier versions if you
2450 relied on this feature (some tasks were blocked because of this).
2451 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2452 even though your applications were doing only point-to-point
2453 communications and sequential computations. Now you can really use any
2454 model you want (of course, if you create real parallel tasks, which are
2455 not implemented in most models beside ptaskL07, this will abort).
2456 * Add an example that schedules a DAX on an heterogeneous platform
2457 using a Min-Min strategy.
2458 * New function SD_workstation_get_current_task() that returns the kind
2459 of task currently running on a workstation in the sequential access
2461 * Raise some warnings when unexecuted tasks remains at the end of the
2462 simulation. This is usually caused by cycles in the DAG.
2465 * New function: SIMIX_process_set_function() called by MSG_set_function
2466 * Change the underlying waiting queue in semaphores so that a process
2467 can wait on several of them simultaneously (as in waitany).
2468 * Fix the way to handle tokens in semaphores so that all access patterns
2469 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2470 * kill the dirty pimple SIMIX_message_sizes_output()
2471 Please use (proper) visualization instead
2474 * New data container: setset (set of sets of elements)
2475 * New module: mmalloc (mapped malloc, allowing to have several
2476 independent segments of malloc)
2477 * New function: xbt_dict_cursor_set_data()
2478 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2479 * New function: xbt_dynar_is_empty()
2480 * New function: xbt_fifo_get_last_item()
2481 * Fix xbt_dynar_shrink(): use the right element size.
2482 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2483 uninitialized areas during expand.
2484 * Fix semaphores: previous implementation was severely broken.
2485 * Use library init/fini functions for our initialization.
2486 - you can use logs and other feature as soon as you want in your
2487 code (even before the xbt_init / MSG_init)
2488 - xbt_exit is now a no-op and produce a warning when used.
2491 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2492 benefit from the latest improvement to the simulation kernel.
2493 * Kill measurement sockets for now. If you rely on them, sorry. This
2494 release is not for you. This feature will be reintroduced in the
2495 future, but we cannot delay the release any further.
2496 * New function: gras_msgtype_get_name().
2497 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2498 * Fix (at last) the pmm example: it should not randomly fail anymore.
2500 Build chain: bug fixes and overall polishing
2501 * Cmake is now stable enough. Hence, we killed the autotools.
2502 * Port to windows ( TM :)
2503 * Fix the 'make install' target.
2504 No need to use 'make install-simgrid' anymore
2505 * Introduce a 'make dist' target compiling a *source* archive
2506 'make package' compiles a binary archive
2507 * Compile java files only on need
2508 * Add --cd and --setenv command line options to tesh
2509 * Out of source builds are not fully supported yet, but we are close
2510 * Enable supernovae and optimization flags by default for our users
2513 * Add layer to set up environment directly from lua, without XML.
2514 * The effect of gras_stub_generator can be achieved through
2515 lua too (check examples/gras/console/ping_generator.lua)
2517 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2519 SimGrid (3.4.1) stable; urgency=low
2521 The "Polishing easter eggs is probably a good idea" release.
2522 This is a bug fixes release only.
2526 * Fix a bug preventing the tasks from begin garbage collected.
2529 * Fix a bug occurring when a host involved in a communication fails.
2530 This was not detected properly by the other peer involved in the
2531 communication. Now, it's reported as a network error.
2534 * Warn the user about loop dependencies in data flow of DAX files
2535 * Obey the control-flow dependencies of DAX files
2538 * Add option "enable_smpi" allowing to not compile SMPI.
2539 Probably useful for the (Mac) users experiencing a build error here
2540 * Improve the detection of lua5.1 and ruby1.8
2542 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2544 SimGrid (3.4) stable; urgency=low
2546 The "Easter in Cargese" release. Also known as (major changes):
2548 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2549 ~> bindings were greatly improved
2550 ~> new tracing infrastructure for better visualization introduced
2552 * the "Welcome to configury modernity" release.
2553 ~> we switched from autotools to cmake, and improved our cdash
2556 A more detailled list of changes follow (full detail in svn log).
2558 Java Bindings: Various Cleanups
2559 * (install java-gcj-compat-dev on debian-like to use them)
2560 * Remove put/get: no need to export deprecated interface in Java
2561 Use send/receive instead.
2562 * Cleanup the examples and add a README per directory
2563 * Remove example autoDestination (that's the only way to go now)
2564 * Remove example explicitDestination (was a plain copy of basic)
2565 * Make JniException a runtime exception, so that there is no need to
2566 declare the fact that you may encounter such a beast. I guess that
2567 nobody will ever want to survive such error.
2568 * Create specific errors for each MSG case of failure:
2569 host failure, transfer failure, timeout, task cancelled
2570 * Cleanup the exceptions that may get thrown by each function
2571 * Other internal cleanups in Java bindings. Performance still bad :/
2572 Ruby and Lua Bindings: create them
2573 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2574 * That's new and great, you should try them out.
2575 Same functionalities than Java bindings, only even less polished
2577 * Kill the useless "rate" argument of SD_task_get_execution_time()
2578 Everyone used to provide -1 as a value, it was not used, and the
2579 semantic of a possible use wasn't even clear.
2580 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2581 either as comm costs or compute costs to mean that there is no
2582 such thing for that specific task.
2583 * Add a SD_task_set_name() function
2584 * Fix SD_task_unschedule() on typed tasks
2585 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2586 * In DAX loader, accept useless 'level' attributes to <job> since
2587 LIGO DAGs have them (seem to be to ease graphical representation).
2589 * Add an example masterslave_mailbox.c using send/receive and not
2590 the deprecated put/get interface.
2591 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2592 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2594 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2595 (declare MSG_USE_DEPRECATED to still have the old name)
2596 * Add a MSG_task_set_data() function
2597 * About trace replay (see examples/msg/actions):
2599 - Allow to work with splitted trace files for each process
2600 Give the specific trace file as argument of each process,
2601 and call MSG_action_trace_run(NULL)
2602 You can still have one merged file for all processes.
2603 - Fix implementation of collective operations
2604 * Allow task_execute() on 0-sized tasks (closes #10063)
2606 * This is the first release of SimGrid where SMPI is not considered
2607 beta anymore (even if some corners should still be improved)
2608 * Port over the new SIMIX_network submodule (internal refactoring)
2609 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2610 * Implement more missing elements of the standard:
2612 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2614 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2615 MPI_Scan() MPI_Get_processor_name()
2616 - Added implementation of missing case for Alltoall (warning: it's
2617 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2618 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2619 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2620 SMPI_MPI_Allgatherv()
2621 * Bug fixes include:
2622 - MPI_Waitsome() was broken
2623 - Allow relative includes in smpicc
2624 - Command line cfg argument 'reference_speed' was ignored...
2625 - Some functions did not properly lead to auto-benching of user code
2626 - smpicc passes -O2 by default (just like openmpi one)
2628 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2629 * Bug fixes about timeouts during communications
2630 * add SIMIX_message_sizes_output() as a pimple to write to file the
2631 amount of messages per size. Use gnuplot to get histogram.
2632 Pimple because that's the only user-visible function of simix,
2633 defined directly in xbt.h (irk, sorry)
2635 - Add a SIMIX_sem_get_capacity() function
2636 - Fix interactions with processe resume/suspende
2637 - release_forever() was stupidly broken
2638 - Fix SIMIX_display_process_status() for processes in a semaphore
2639 - Make SIMIX_sem_block_onto() user-visible
2640 * Refactoring context stuff:
2641 - Use pseudo-OOP for better modularity
2642 - reimplement SIMIX_process_kill() without process_schedule() so
2643 that the latter can take as invariant that it is called from
2645 - Merge context_start into context_new for sake of simplicity
2647 * Add a Vivaldi network model, coded live during SUD'10 ;)
2648 * Rename configuration variables to start a hierarchy:
2649 o cpu_model -> cpu/model
2650 o network_model -> network/model
2651 o workstation_model -> workstation/model
2652 * New configuration variables:
2653 o network/bandwidth_factor: correction to bandwidth
2654 o network/latency_factor: correction to latency
2655 o netwotk/weight_S: correction to the weight of competing streams
2656 * Add a long description to the models, that users can see with such
2657 argument on the command line: --cfg=cpu/model:help
2658 * --help-models display the long description of all known models
2660 * config: add the ability to set a default value after registration
2661 Does not override any previously set value (e.g. from cmd line)
2662 * dict: allow to have integer key and data.
2663 When so, you need to use the following functions
2664 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2665 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2666 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2667 In contrary to regular dicts, the key is not malloced before copy.
2668 Mixing scalar and regular elements in the same dict is not tested
2670 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2671 Tracing for Visualization:
2672 * SimGrid is now instrumented in order to generate a trace file for
2673 visualization analysis: to use it, need to compile SimGrid with the
2674 "tracing" option enabled, and instrument the program using SimGrid with
2675 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2676 (among other functions).
2677 * The instrumentation only traces the platform utilization for now
2678 * Documentation to use the tracing functions and how to analyze the
2679 traces with the Triva tool is written.
2680 * More information about: SimGrid FAQ (in the section Tracing Simulations
2683 * We moved to cmake as default build system. Autotools support will
2684 be dropped soon. Check the FAQ for more info about how to use it.
2685 * Greatly improved our cdash/ctest interactions
2686 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2687 * Added memory checking tests with valgrind; lot of memleak fixing.
2688 This may be the first release of simgrid with so few memory issues
2689 * Added code coverage tests.
2690 Our coverage is still improvable, but at least we see it on cdash.
2692 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2694 SimGrid (3.3.4) stable; urgency=low
2696 The "Desktop Grid needs love too" release (also called Xmas release).
2698 Models improvements:
2699 * Major speedup in the maxmin system solving by using lazy evaluation
2700 Instead of solving completely the maxmin system at each iteration,
2701 only invalidate (and recompute) the modified parts.
2702 This new feature is enabled in default models but you can try to
2703 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2704 * Cas01 IMproved as default CPU model
2705 This CPU model is the same Cas01 model, but it uses the
2706 maxmin-selective-update flag and a heap structure to manage
2707 actions on SURF kernel.
2708 It reduces the complexity to find the next action to finish and,
2709 consequently, it's faster than the old Cas01.
2710 This is the new default CPU model (Cas01).
2711 * Rename the old Cas01 model to Cas01_fullupdate
2712 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2713 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2714 * CpuTI (CPU Trace Integration)
2715 A new CPU model whose objective is simulate faster when using
2716 availability trace files.
2717 Instead of using a full featured, over engineered maxmin system for
2718 CPU modeling, this model does the pre-integration of traces files
2719 to calculate the amount of CPU power available, and so, executes
2720 faster than the old CPU models.
2721 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2722 * Use LV08 as default network model since it gives better accuracy
2723 for small messages and shouldn't change things for big ones.
2724 Use --cfg=network_model:CM02 to get the previous behavior.
2727 ******************************************
2728 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2729 ******************************************
2730 * The new CPU model may changes simulations!
2731 The point is that events occurring at the exact same timestamp
2732 are not scheduled in the same order with the old and new
2733 version. This may be enough to completely change the execution
2734 of simulations in some cases.
2735 * The new network model will change simulations!
2736 This new model is more realistic than the previous one, so you
2737 should consider redoing your old experiments with this model.
2738 Sorry for the inconvenience.
2741 * Introduce the supernovae compilation mode
2742 When compiled that way, the whole SimGrid (or almost) is put in a
2743 single compilation unit and compiled in one shoot.
2744 This is to help gcc which has difficulties to inline stuff from one
2746 The speedup seem to be above 15%, althrough more tests are needed on
2747 amd64 to confirm that gain.
2750 * Port of MSG's mailbox on top of SIMIX network
2751 The put/get mechanism was greatly simplified on the way.
2754 * New SIMIX network module. Provides:
2755 - Mailbox: rendez-vous mechanism to find with who you want to speak
2756 - Synchronous send/recv: easier and hopefully faster since the
2757 logic is handled in the maestro process directly now
2758 - Asynchronous send/recv: you dreamt of it? It's here now
2759 Too bad that nobody cared enough to propagate the change to MSG.
2760 * Add semaphores as SIMIX synchronization mechanism.
2763 * new function SD_daxload(char*) to load a DAX file
2764 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2765 * Introduce typed tasks. Specify its kind and cost at creation.
2766 At scheduling, just give where it should be placed, and the cost
2767 for each involved resource is automatically computed.
2768 Existing constructors so far (more to come of course):
2769 - SD_task_create_comm_e2e() for end-to-end communication
2770 - SD_task_create_comp_seq() for sequential computation
2771 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2772 * new function SD_task_dump() for debugging display
2773 * new function SD_task_dotty(task,FILE*) writing to file the info
2774 about the task in dotty format
2775 * SD_task_dependency_exists() can now cope with having one of its
2776 arguments NULL. If so, it tests whether the other argument has any
2778 * Add getters on list of preceding/following tasks:
2779 SD_task_get_parents(task) and SD_task_get_children(task)
2780 * Add getters on amount of workstations and list:
2781 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2782 * Add getter on task kind: SD_task_get_kind(task)
2783 * Update the start_time and finish_time of tasks on completion/failure
2784 * Bugfix: Remove task from state swags when destroyed
2787 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2790 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2791 dynar with one separate thread per value of the dynar.
2792 * Change the prototype of xbt_thread_create(), sorry.
2793 Added a boolean parameter indicating whether we want to join this
2794 thread (used in SG only for now)
2795 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2798 * GTNetS wrappers should now be usable again (and betterly tested too)
2799 * Fix a major regression from 3.2 where the timeout provided to
2800 MSG_task_put_with_timeout() was used as absolute time before which
2801 the comm should be done.
2802 * Start to fix the <cluster> tag.
2803 - Internal links should be good now (beside of the loopback, which
2804 use the private link instead)
2805 - paths to the external world is still rather broken
2806 - the <route:multi> tag is just broken. Actually that's brain-dead.
2807 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2808 to make it less stupid
2809 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2810 * Fix a source-level compatibility glitch from 3.2: after defining
2811 MSG_USE_DEPRECATED, you can use the old name
2812 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2813 * Allow to compile from the SVN with automake 1.11
2814 * Fix some problems when using the "start_time" tag in deployment XMLs.
2815 * Fix #8569: XBT/synchro.h has redundant declarations
2816 * Fix #8563: MSG return values and exceptions
2817 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2818 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2819 Thanks to Robson Peixoto.
2820 * Fix a vicious bug in dictionaries inducing that some elements were
2821 not freed on xbt_dict_free()
2823 Portability report of this version:
2824 * Main portability targets:
2825 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2826 on (amd64/i386/ia64)
2827 - mac leopard on i386
2828 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2829 but nothing critical.
2830 * Other platforms: windows, AIX and others were not tested for this release
2832 Timing report of this version:
2833 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2834 scenario parameters). From 8h to a few seconds in desktop grid settings.
2835 * Supernovae brings about 25% speedup on i386.
2837 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2839 SimGrid (3.3.3) stable; urgency=low
2841 The "Need for Speed" release.
2843 The timings done to validate the 3.3.2 were faulty.
2844 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2846 The problem was a conversion from a manually handled vector to
2847 xbt_dynar_t on the critical path.
2848 xbt_dynar_foreach calls functions, inducing stack management crap.
2850 We inlined these functions and xbt_dynar_foreach is now breath taking.
2851 We also inlined xbt_swag_belong on the way.
2853 Here are some approximate speedup measurements (on master/slaves
2854 simulations lasting between 10s and 20s each):
2855 3.3.1 -> 3.3.2: about same performance
2856 3.3.2 -> 3.3.3: 40% speedup
2857 3.3.1 -> 3.3.3: 40% speedup
2858 3.3.1 with inline patch -> 3.3.3: 30% speedup
2860 Our reading is that the refactoring which occurred in 3.3.2 made us
2861 suffer much more from the xbt_dynar_foreach low performance, but
2862 once we solved this, this refactoring proved to be very performance
2863 effective. From the 40% speedup, somehow, 10% are due to the
2864 inlining and 30% to the refactoring.
2866 That's a pitty that gcc cannot inline functions placed in other files
2867 alone. We have to choose between:
2868 - break the encapsulation (by putting private data structures and
2869 accessors in headers files to help gcc)
2870 - live with low performance
2871 - switch to a decent compiler such as icc (not quite possible).
2873 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2875 SimGrid (3.3.2) stable; urgency=low
2877 The "Simplicity does not preceed complexity, but follows it" release.
2879 The main contributors of this release were (lexical order):
2880 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2883 * Extract the routing logic into its own object.
2884 (was dupplicated in network.c and workstation_LV07.c;
2885 Allows to implement other ways of storing that info)
2886 => kill now useless network_card concept
2887 - Use dynar to represent routes (instead of void** + int*)
2888 - kill link_set (use surf_network_model->resource_set instead)
2889 - Add a command-line option to choose the routing schema to use
2890 - Add three new models:
2891 * Floyd (shortest path computed at initialization)
2892 * Dijikstra (shortest path recomputed all the time)
2893 * Cached Dijikstra (shortest path computed on need)
2894 All these models where contributed by Silas De Munck, and are
2895 described in his ICCS09 paper.
2897 * Simplify model declaration
2898 (less redirections, less function to write when defining a model)
2899 - Factorize stuff between models:
2902 surf_model_resource_set(model)
2903 surf_model_resource_by_name(model, name)
2904 - Unify the types of models in s_surf_model_t (using an union)
2905 - Embeed fields of common_public directly into s_surf_model_t
2906 - Rename model methods:
2907 action_free ~> action_unref
2908 action_change_state ~> action_state_set
2909 action_get_state ~> action_state_get
2910 - Change model methods into functions:
2911 (model)->common_public->action_use ~> surf_action_ref
2913 * Implement a generic resource; use it as ancestor to specific ones
2914 (allows to kill duplicated code in models)
2915 Drawback: timer command don't need no name nor properties;
2916 workstation_CLM03 don't need no properties
2917 (but I guess we can live with those few bytes wasted)
2919 * Improve the action object model
2920 - implement a constructor avoiding dupplicated code about field
2921 initialization in generic_action part.
2923 * Kill the SDP model: it has an external dependency, is deprecated
2924 in flavor of modern lmm models, and didn't compile since a while
2927 * Relocation of the context module from XBT to SIMIX.
2928 (the context were decoupled from the simix processes, duplicating a lot of code)
2929 => a lot of code was factorized
2930 - less overhead is introduced during scheduling
2931 - simpler API for the context factory
2932 - the logic for process creation,destruction and manipulation was simplified
2933 * Simplification of the s_smx_process_t data structure.
2934 => accesing the simix level data associated to a process is faster now,
2935 and the code is a lot more readable.
2938 * Implement some more MPI primitives:
2939 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2940 -implementation: Bcast: flat or 2-ary tree (default),
2941 Barrier: 4-ary tree,
2943 Allreduce: Reduce then Bcast
2944 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2945 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2946 Alltoallv: flat tree, like ompi
2948 * Add support for optimized collectives (Bcast is now binomial by default)
2949 * Port smpirun and smpicc to OS X
2952 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2953 Shout out if you used it.
2956 * Display the list of still queued messages in SG mode when existing
2960 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2961 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2962 * Add xbt_str_from_file(FILE*)
2963 * Add xbt_dict_get_key achieving a linear reverse search
2964 * Remove the context module
2966 Portability report of this version:
2967 * Main portability targets:
2968 - Linux(debian)/x86/context
2969 - Linux(debian)/x86/pthreads
2970 - Linux(debian)/amd64/context
2971 - Linux(debian)/amd64/pthreads
2972 On these, we still have the eratic breakages of gras/pmm and
2973 amok/saturate_sg reported in previous version. We still think
2974 that the tests are the cause of the fault, not the tested code.
2976 - Mac OSX Leopard/x86/context
2977 Still false negative in tesh autotesting.
2978 Smpi still fails, but this time because readlink does not accept -f
2979 Everything seems to work properly beside of that.
2982 - AIX version 5.3 (only tested contexts this time)
2983 Smpi still fails there because mktemp is not installed.
2984 Everything seems to work properly beside of that.
2986 I managed to compile it for the first time, but several breakages.
2987 Won't delay the release for this exotic platform.
2989 * Windows: it's still lagging behind. If you want to help, please
2992 Timing report of this version:
2993 This version seem to be more than 5% faster than 3.3.1 (on linux
2994 64bits with contextes). The gain is less than expected, we are
2995 investigating this for next release.
2997 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2999 SimGrid (3.3.1) stable; urgency=low
3002 * Implement a --cfg-help to show existing configuration variables
3003 * Build chain do not require doxygen in maintainer mode
3006 * fix a bug on struct sizeof computation, which prevented the
3007 exchange of arrays of structs in some conditions
3008 - added a regression test about this in datadesc_usage
3009 * Allow the exchange of 0-long dynamic vectors.
3010 - for that, use -1 as indicator of dynamic size instead of 0
3011 - This implied to change any size from unsigned long to long,
3012 reducing a bit communication abilities, but I guess that with
3013 64bits being quite common, this is more than enough.
3014 - This also induce a protocol change, thus bumping network protocol
3015 version from 0 to 1 (if we have external users, we have to get
3016 clean on that point too ;)
3017 - added two regression tests about this in datadesc_usage
3018 * Be more verbose when propagating local exceptions
3019 This helps debugging.
3020 * Display the status of simulated processes when receiving SIGINT in
3024 * Allow to control the simulation from a trace file.
3025 New functions MSG_action_register() and MSG_action_trace_run()
3026 The first one allows to associate a function execution to each
3027 kind of action while the second one parses a trace file and
3028 triggers the corresponding actions within the system.
3029 For now, only a toy example is provided in examples/msg/actions
3030 * Add an example of process migration in examples/msg/migration
3031 * Fix a bug in task exchange which broke MSG_task_get_sender()
3032 Add a teshsuite regression test for that.
3033 [Bug: if MSG_task_get_sender() is called after sender exit,
3035 * Fix a bug which prevented suspend/resume to work properly
3036 * Display the status of simulated processes when receiving SIGINT
3037 This fixes a regression of v3.3. due to the introduction of SIMIX
3038 * Bug fixing in failure management:
3039 - trace could not start by a failure at time 0
3040 - failure during communications were not working
3043 * Add SIMIX_process_set_name() to change the name of the current
3044 process in the log messages.
3045 * Store smx_hosts in a dict since we only retrieve them by name
3046 * Move the configuration infrastructure to surf
3049 * Move the configuration infrastructure to surf
3052 * Massive internal cleanups:
3053 - Store internal structures on processes instead of hosts (allows
3054 to have more than one process per host, in addition of being more
3056 - Cleanup the initialization/finalization process
3057 - Kill a whole bunch of unneeded synchronization:
3058 processes run in exclusive manner within the simulator
3059 - Move queues from global tables to process data fields
3061 - now accept -platform and -hostfile arguments
3062 - Pass the right rank value to processes according to the hostfile
3063 * Compile the examples by default, and use them as regression tests
3064 * Implement MPI_Wtime()
3065 * Change the reference speed to a command line option
3068 * TCP_gamma can now be specified as command line option using
3069 --cfg=TCP_gamma:10000000.0
3070 * Change the --surf-path cmd line option into --cfg=path:
3073 * Also include strbuff from xbt.h public header
3074 * xbt_ex_display(): do not free the exception after displaying
3075 This allows to do more with the given exception afterward.
3076 Users should call xbt_ex_free() themselves.
3080 Portability report of this version:
3081 * Main portability targets:
3082 - Linux(debian)/x86/context
3083 - Linux(debian)/x86/pthreads
3084 - Linux(debian)/amd64/context
3085 - Linux(debian)/amd64/pthreads
3086 These targets fail about 1/10 of times on gras/pmm, but we believe
3087 that this is because of the test, not because of simgrid.
3088 amok/saturate_sg fails even more rarely, and the test may not be
3091 - Mac OSX Leopard/x86/context
3092 The test suite still spits tons of errors because some obscure
3093 force prevents us from removing the temporary directories
3094 arguing that they still contain some metadata I've never heard of.
3095 Smpi fails because seq is not installed.
3096 Everything seems to work properly beside of that.
3099 - AIX version 5.3 (both contexts and pthread)
3100 Smpi still fails there because mktemp is not installed.
3101 XML inclusions seems rosty on AIX.
3103 * Windows: it's still lagging behind. If you want to help, please
3106 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3108 SimGrid (3.3) stable; urgency=high
3112 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3113 [Malek Cherier & Mt]
3115 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3118 Until now, GRAS were using MSG as an interface to SURF. It was
3119 quite difficult because both interface have several differences
3120 (MSG channels vs GRAS sockets were the most notable point).
3122 This also opens the gate to SMPI (which should occur soon) and speed
3123 up simulations by to 40% (even if it were not the main goal).
3125 **************************************
3126 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3127 **************************************
3128 The point is that events occurring at the exact same timestamp are
3129 not scheduled in the same order with the old and new version. This
3130 may be enough to completely change the execution of simulations in
3131 some cases. Sorry for the inconvenience.
3133 * Cleanup and upgrade the XML format to push further scalability
3134 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3136 * Improve the testing infrastructure with tesh. Now a very large part of
3137 the code is tested not only by being run but also by checking that the
3138 output match an expected output [Mt].
3140 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3141 is really less memory-demanding, which should allow you to use
3142 larger files in SimGrid [AL].
3144 * Inform valgrind about our contextes, so that it becomes usable
3145 with the default (and more effecient) version of SimGrid
3146 [contributed by Sékou Diakite, many thanks]
3149 * Introduce a listener thread in charge of receiving incoming
3150 messages from the network. It allows to overlap communication and
3151 computation but most notably, it removes some stupid deadlocks due
3152 to the fact that so far, a process could not send and receive at
3153 the same time. This made most non trivial communication schema
3155 * Convert the PIDs from long int to int to match the MSG ones (and
3156 linux ones too) [Mt]
3157 * New function: gras_agent_spawn() to launch a new process on
3158 current host. Only working in simulation for now. [Mt]
3159 * New function: gras_os_hostport() returning a constant form (ie,
3160 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3163 * Make the backtrace of exceptions more human readable [Mt]
3164 * New module: xbt/str [Mt]
3165 a ton of string utility functions (split, join, printf to a newly
3166 allocated buffer, trim, etc)
3167 * New module: xbt/hash [Mt]
3168 SHA1 hashing algorithm (more to come if needed)
3169 * New module: xbt/synchro [Mt]
3170 synchronization tools (mutex and conditions) working the same way
3171 in simulation and in real life (mainly useful for GRAS, but not
3173 * New module: xbt/queue [Mt]
3174 classical producer/consumer synchronization scheme
3175 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3176 (using the classical functions will get serialized) [Mt]
3177 * Make dictionary internal table dynamic. No need to specify its size
3178 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3180 * Make sure the log channels are organized as a tree under windows
3181 (because of ANSI C compatibility issue, any channel were child of
3182 root directly) [Mt].
3185 * Cleaned many thing in surf and fixed a few bugs [AL].
3186 * Add a nice command line configuration mechanism to compose models [AL].
3187 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3188 the previous one (KCCFLN05). It relies on something that looks like
3189 a max-min sharing mechanism but cannot be written as such. A new solver
3190 was thus designed [AL].
3191 * Add a new solver to lmm. Based on Lagrange optimization and
3192 gradient-based descent, it enables to efficiently maximise systems s.a
3194 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3196 This solver enables to propose two new network models for TCP Reno and
3197 TCP Vegas based on Low's work. These models still need to be fully
3198 tested though [Pedro Velho].
3201 * Bug fix in SD_simulate. Now the time bound given as argument is
3203 * Use the new parallel task model (ptask_L07) as default.
3204 * Use the SURF command line configuration mechanism.
3205 * 0-size tasks (for synchronization) should now work.
3207 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3209 SimGrid (3.2) stable; urgency=high
3213 We still experience issues on this platform, but we believe that at
3214 least MSG is usable.
3216 GRAS API BREAKAGE (for simplification purpose, sorry):
3217 * the gras_msgtype_by_name is not used anymore. Instead of
3218 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3219 you can write (and must)
3220 gras_msg_send(toserver, "request", &request);
3221 - If you still want to pass a gras_msgtype_t to the function (to cache
3222 the type and avoid the lookup time), use the gras_msg_send_() variant.
3223 - Impacted functions:
3224 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3225 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3226 * The callbacks are now expected to return 0 when everything went well
3227 (just like the main() function)
3229 GRAS new features and improvements:
3230 * New module mechanism where user code can use per process globals [Mt]
3231 This is similar to gras_userdata_*() functions, but for libraries. It
3232 factorize some code developped over and over in the examples and AMOK.
3233 It has still to be documented and used (only amok/peermanagement is
3235 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3236 loss when they were small enough to fit into the buffer and sent quickly
3237 enough so that they can all get received in one shoot.
3238 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3239 if not found. Use the *_or_null() variant for the old semantic.
3240 * In gras_msg_handle, do not discard messages without callback.
3241 They are probably messages to be explicitly awaited later (ie, proofs of
3242 mis-synchronization in userland since they are sent before being awaited)
3243 No big deal usually.
3244 * gras_socket_meas_send/recv: semantic changed!
3245 The numerical arguments used to be (1) the total amount of data to send
3246 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3247 messages. This was need for the fool willing to send more than MAXINT
3248 bytes on quite fat pipes.
3251 * Do really rename the hostmanagement module to peermanagement. [Mt]
3252 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3253 but this is rather new and this was documented in the module
3254 documentation (poor excuses, I admit)
3255 * Bandwidth measurement semantic changed! This follows the changes to
3256 gras_socket_meas_send/recv explained above.
3259 * A sequential mode has been added to the workstations. When a workstation
3260 is in sequential mode, it can execute only one task, and the other tasks
3261 are waiting in a FIFO. [Christophe Thiery]
3264 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3265 model for SIMDAG. [Christophe Thiery]
3266 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3267 equal to 0, which caused some bad behaviors in
3268 saturated_constraint_set_update. I now use a threshold mechanism like in
3272 * When running manually src/testall, you select specific units [Mt]
3273 testall is the result of our cunit mechanism, and should replace all
3274 the scripty thingy around since bash don't run easily on billware.
3276 * A mallocator system has been added. [Christophe Thiery]
3277 Mallocators allow you to recycle your unused objects instead of freeing them
3278 and allocating new ones.
3280 Documentation update:
3281 * FAQ reworking + New FAQs:
3282 - "Valgrind spits tons of errors!" [Mt]
3283 - "How to repport bugs" [Mt]
3284 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3285 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3286 - Communication time measurement within MSG [AL]
3287 - I experience weird communication times when I change the latency [AL]
3288 * GRAS tutorial [Mt]
3290 - an introduction to the framework and to the used communication model
3291 - an initiatic tour introducing the most proheminent features:
3293 . Lesson 0: Installing GRAS
3294 . Lesson 1: Setting up your own project
3295 o Part 2: Message passing
3296 . Lesson 2: Exchanging simple messages
3297 . Lesson 3: Passing arguments to the processes (in SG)
3298 . Lesson 4: Attaching callbacks to messages
3299 . Lesson 5: Using globals in processes
3300 . Lesson 6: Logging information properly
3301 . Lesson 7: Using internal timers
3302 . Lesson 8: Handling errors through exceptions
3303 . Lesson 9: Exchanging simple data
3304 . Lesson 10: Remote Procedure Calling (RPC)
3305 . Lesson 11: Explicitly waiting for messages
3306 . Recapping of message passing features in GRAS
3307 - A HOWTO section containing:
3308 o HOWTO design a GRAS application
3309 More are due, of course. They will come latter. In the meanwhile, you can
3310 check the examples which are still here.
3312 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3314 SimGrid (3.1) stable; urgency=high
3318 There was a stack corruption somewhere, visible only when optimizing
3319 with these versions. [Vince]
3322 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3323 module that enabled to play with Directed Acyclic Graphs. It is built
3324 directly on top of SURF and provides an API rather close to the old
3325 SG. Some old codes using SG are currently under rewrite to check that
3326 all needful functions are provided. [Christophe Thiery]
3329 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3330 extension of the classical CLM03 model that gracefully handles
3331 failures. This is now the default model for MSG and GRAS. It doesn't
3332 handle parallel tasks yet though. [AL]
3333 * Bug fix: Weights were not correctly set in the network part.
3334 WARNING: This may have resulted in incorrect results with simulations
3335 where there are more than one flow on a given link. [AL]
3338 * After a (long ?) discussion on simgrid-devel, we have decided that the
3339 convention we had on units was stupid. That is why it has been decided
3340 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3341 WARNING: This means that all previous platform files will not work as
3342 such with this version! A warning is issued to ask users to update
3344 A conversion script can be found in the contrib module of the CVS, under
3345 the name contrib/platform_generation/surfxml_update.pl [MQ]
3348 * Bug fix: Processes were started in reverse order, wrt deployment file.
3349 WARNING: if your code relies on this bug, please fix it. [AL]
3350 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3351 being executed on two different locations. [AL]
3352 * Bug fix: Failures are now better supported thanks to Derrick's tests
3353 (there was many failure situations I hadn't thought of and that weren't
3354 correctly handled). [AL]
3355 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3359 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3360 Exception occurring on server-side are propagated back to client (!).
3362 API CHANGE: the callback changed their prototype. Change:
3363 int my_handler(gras_socket_t expeditor, void *payload_data) {
3365 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3366 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3368 * New! function: gras_msg_handleall to deal with all messages arriving
3369 within a given period.
3370 * New! function: gras_socket_server_range to get a server socket in a
3371 range of port numbers (ease to avoid port number conflicts) [MQ]
3372 * New! gras processes display their backtrace when they get a SIGUSR1
3373 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3374 Sweet to debug RL processes [MQ]
3378 - Do not force experiment sizes to be expressed in kb, or it becomes
3379 impossible to measure the latency this way (needs one byte-long tests)
3380 WARNING: this changes the amok_bw_* function semantic. [MQ]
3381 - Implements the link saturation stuff. [MQ]
3382 * Peer management module:
3383 New! module factorizing code that we wrote over and over [MQ].
3386 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3387 - Test units are placed directly into the library code, they get extracted
3388 automatically and placed into the src/testall binary.
3389 - Convert most of the XBT tests to this system.
3390 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3391 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3392 live in the same C code. This required to change a little bit the API
3393 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3394 * New module: sparse graph structure with basic algorithms (this is work
3395 in progress and the API is not considered to be frozen yet). [AL]
3396 * Display more information on backtraces: source line & function names are
3397 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3398 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3399 * Reworked a little bit some #include statements to load only required
3400 headers. Some user code that relied on SimGrid to include stdlib or
3401 stdio may need to include it by themselves. [AL]
3402 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3404 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3405 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3406 but I guess that almost nobody used those parts. [MQ]
3408 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3410 SimGrid (3.0.1) stable; urgency=low
3413 * Unfortunately, I had missed 5 misnamed functions:
3414 xbt_fifo_item_t xbt_fifo_newitem(void);
3415 void xbt_fifo_freeitem(xbt_fifo_item_t);
3416 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3417 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3418 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3419 They're now deprecated. Please use their new versions:
3420 xbt_fifo_item_t xbt_fifo_new_item(void);
3421 void xbt_fifo_free_item(xbt_fifo_item_t);
3422 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3423 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3424 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3426 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3427 * Documentation: xbt_log module unmercifully reworked [MQ]
3428 * Bugfix: there was a problem with the ending of contexts with
3429 the pthread backend. It caused some weird deadlock or behavior
3430 depending on the pthread implementation. [AL]
3431 * Bugfix: get the exceptions raised in the simulator repport where
3432 and why they come from when they are not caught in time [AL, MQ]
3435 * Bugfix: Do repport the error when two non-connected hosts try to
3436 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3439 * Add additionnal checkings on communications. Assert that two
3440 communicating hosts are connected by a set of links... [AL]
3443 * Add additionnal checkings on channel values in communication [AL]
3444 * New: MSG_task_get_source to see on which host a task was generated [HC]
3445 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3446 the number of tasks waiting to be received on channel and sent
3448 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3449 waits for the first task coming from a given host.. [AL]
3451 GRAS new functionnalities: [MQ]
3452 * Enhance the parsing macro to allow the size of multidimentional objects
3453 to be given thru annotations.
3454 * New example (and documentation): Matrix Multiplication a la RPC
3455 (as when I was young!) and fix a bunch of bugs found on the way.
3457 GRAS performance improvements: [MQ]
3459 * Reduce the amount of cbps creation/destruction by making it static to
3460 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3463 * Change libdata to a set so that we can search for stuff by ID (and thus
3464 reduce the insane amount of dict lookups)
3467 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3468 RL communication on top of SG since it's so uneffective.
3469 It may also be used for inter-thread communication in RL, one day.
3470 * Use gras_datadesc_copy() to exchange messages on top of SG
3472 - improve message exchange performance on top of SG
3473 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3474 * Don't exchange on the network the size of the used part of buffer,
3475 instead, specify the possible buffer size to read().
3477 - reduces the amount of read/write calls (one pair per exchange)
3478 - reduces the amount of exchanged data (the size)
3479 - allows to retrieve all arrived data on receiver side, if we don't need
3480 it right now (subsequent read will peek the buffer)
3481 - allows the receiver to proceed with the beginning of the stream before
3482 everything is arrived
3483 - make it possible to build an iov transport (using readv/writev)
3485 - take care of the data with non-stable storage (like stacked data),
3487 * If possible, TCP send uses vector I/O (when writev() is here)
3488 - Don't use it for receive since we send data sizes and data on the
3489 same stream, so we wouldn't be able to chain large amount of chunks
3490 before having to flush the stuff to read the size.
3491 * Rework the transport plugin mechanism to simplify it and reduce the
3492 amount of pointer dereferencement when searching for the right function
3495 * I guess that now, we do almost as few system calls as possible while
3496 doing as few data copy as possible.
3498 To improve it further, we could try to send all the sizes first and then
3499 all the data (to use iov on receiving size), but it's only a partial
3500 solution: when you have 2 dimensional data, the sizes of the second
3501 dimension is data of the first dimension, so you need 3 streams.
3503 I'm not sure the potential performance gains justify the coding burden.
3505 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3507 SimGrid (3.00) stable; urgency=high
3510 * New! Give the possibility to hijack the surf parser and thus bypass
3511 MSG_create_environment and MSG_launch_application. Have a look at
3512 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3514 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3516 SimGrid (2.96) unstable; urgency=low
3521 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3522 This deprecates the xbt_error_t mechanisms.
3523 It modifies (simplifies) all XBT and GRAS API.
3524 MSG API keeps unchanged (exceptions raised by XBT are caught from
3525 within MSG and masked with existing error handling facilities)
3528 * New! Add a FATPIPE model. [AL]
3529 * New! Add a parallel task model. [AL]
3530 * New! Add automatically a loopback interface (in the default
3531 network model) if none was precised.
3534 * Bugfix: MSG_process_resume now works with the current running process.
3536 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3537 * Modification of MSG_task_get_compute_duration. Once a task has been
3538 processed, the value returned by this function is now equal to 0. [AL]
3539 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3540 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3541 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3542 of a m_task. For now, MSG_task_cancel only works with computation
3544 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3545 of the processor (in Mflop/s) regardless of the current load on the
3547 * API Change: use proper naming convention for MSG_getClock and
3548 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3550 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3551 This function changes the priority of a computation task. This priority
3552 doesn't affect the transfer rate. A priority of 2 will make a task
3553 receive two times more cpu power than the other ones. This function
3554 has been added to suit the needs of Nguyen The Loc and hasn't been that
3555 much tested yet. So if it fails, please report it and send me your code.
3557 * API Change: removed all functions and types that were marked "deprecated"
3558 since many months. Renamed MSG_global_init_args to MSG_global_init.
3560 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3562 SimGrid (2.95) unstable; urgency=low
3565 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3566 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3567 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3570 * Bugfix: Allow absolute paths for platform description files [MQ]
3571 * Bugfix: do free the variables after use. Leads to drastic performance
3573 * Implement max_duration (ie, timeouts) on resources [AL]
3576 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3578 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3579 the needed support (timeouts) [AL]
3582 * Implement measurement sockets. You can now get the bandwidth between two
3583 hosts thanks to AMOK (see below). [MQ]
3584 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3585 dynar over the network (yeah) [MQ]
3586 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3587 * simple/static token-ring example. [Alexandre Colucci and MQ]
3588 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3589 in place before [MQ]
3592 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3593 type parsing. "short" and "long" modifiers were also ignored; other
3594 modifier (such as reference level) are still ignored. [MQ]
3595 * Embeed the buffer size within the buffer itself on SG. [MQ]
3596 That way, send() are atomic and cannot get intermixed anymore (at least
3597 the ones which are less than 100k; bigger messages still have the issue)
3598 * Array size pushed by the field, not by the field type (or each
3599 and every long int will push stuff to the cbps) [MQ]
3600 * use select() to sleep since it allows to portably sleep less than one
3603 GRAS (minor cleanups)
3604 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3607 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3608 * type category 'ignored' killed as it was never used and were difficult
3610 * whether a type can cycle or not is now a flag, leaving room for more
3611 flags (as "ignored", if we feel the need one day ;)
3612 * Rename raw sockets to measurement sockets since "raw" has another
3613 meaning in networking community.
3616 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3617 and offers features not belonging to GRAS but that most applications
3618 need. One day, it may be a set of plugins loadable at runtime.
3619 * New module: bandwidth
3620 bandwidth measurement between arbitrary nodes running this module. [MQ]
3622 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3624 SimGrid (2.94) unstable; urgency=low
3626 The first beta release of SimGrid 3 !
3630 * Update the main page and the FAQ. Adding references to gforge.
3633 * Add a gras_os_getpid function.
3636 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3637 * Extend the logs so that they also print PID, hostname, date, ... if
3639 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3640 and kill the old version which were in testsuite/
3641 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3645 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3646 for users. Instead, each programming environment (GRAS, MSG,...) use it
3647 when needed to provide such a feature to users.
3648 Don't shortcut the mechanism or you will also shortcut the virtualization
3649 you need on the simulator.
3653 * Cleanups in configury with regard to compile optimization/warning flags.
3654 Also add -fno-loop-optimize to any powerpc since it's the optimization
3655 killing gcc (< 3.4.0).
3656 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3657 needed by MSG examples complications
3658 * Borrow configury beautifications from PHP
3661 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3662 warning (thanks loris for stumbling into this one).
3663 * Bugfix: stop loading private headers (gras_config.h) from the public
3667 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3668 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3669 RL would imply a DNS resolver, which is *hard* to do in a portable way
3670 (and therefore delayed).
3671 * Implement a real timer mechanism and use it in timing macros. This allows
3672 to avoid rounding errors and get a 0.000005 sec precision in timing
3673 macros. While I was at it, various cleanups:
3674 - allow to declare more than one timed section per file (fix a stupid bug)
3675 - move some private declaration to the right place
3676 - merge conditional execution and timing macros into emulation module
3677 - document the module
3678 - make sure the module cleanups its mess on gras_exit
3679 * Documentation improvements:
3680 - (new) how to compile applications using GRAS
3681 - (new) emulation support (timing macros)
3683 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3685 SimGrid (2.93) unstable; urgency=low
3687 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3690 - Use Paje properly where used. Still to be sanitized properly.
3691 - Portability fix: Add an implementation of the contexts using pthread
3695 - Add xbt_procname(): returns the name of the current process.
3696 Use it to show the current process's name in all logging.
3698 - fix detection of older flex version and the reaction, since we do
3699 depend on modern ones (we use lex_destroy)
3700 - Better separation of SG and RL in the libs: remove all simulation code
3701 from libgras. As a result, this lib is now only 200k when stripped.
3702 Some of the xbt modules may also be duplicated (two sets and such) and
3703 should be cleaned/killed before SG3.
3704 - Insist on using xlC on AIX because of weird problems involving gcc there.
3705 - Cleanup the make remote stuff. This is now done by scripts
3706 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3707 mainly for our private use, but we're working on changing them to user
3710 - Bugfix: flush the socket on close only if there is some *output*.
3711 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3713 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3716 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3718 SimGrid (2.92) unstable; urgency=low
3720 Alpha 3 on the path to SimGrid 3
3724 - New! First try of benchmarking macros.
3725 - New! First try so that gras_stub_generator generate deployment and
3726 remote compilation helpers.
3728 - Bugfix: Initialization fix in msg_test.
3732 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3734 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3736 - API Change: gras_os_sleep to take the amount of seconds as a double.
3737 Accepting an int was error prone since it was the only location where
3738 seconds were coded as such. It leaded to damn rounding errors.
3739 - Bugfix: Hard to belive that timers ever worked before this.
3741 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3743 SimGrid (2.91) unstable; urgency=low
3745 Alpha 2 on the path to SimGrid 3
3749 - Bug fix in the lmm_solver.
3751 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3752 through the function MSG_paje_output.
3753 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3754 - It is possible to bound the rate of a communication in MSG with
3755 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3756 to write it in the changelog).
3757 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3762 - Various cleanups to the autotools stuff
3763 - Begin to move Gras examples to examples/gras/
3764 - Let make distcheck work again (yeah!)
3766 - documentation overhauled using doxygen.
3767 gtk-doc-tools is dead in SimGrid now.
3768 - Automatically extract all existing logging categories, and add the list
3769 to the documentation (long standing one, to say the less)
3771 - Cleanup the known architecture table. Reorder the entries to group what
3772 should be, and use a more consistent naming scheme.
3773 (some of the test dataset are still to be regenerated)
3774 - New! Allow library to register globals on each process just as userdata
3776 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3777 lookup time (for now).
3778 Use it in msg and trp.
3779 This cleans a lot the internals and helps enforcing privacy of the
3780 headers between the gras components.
3781 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3782 - Bugfix: gras_os_time was delirious in RL.
3783 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3785 - Reenable GRAS now that it works.
3787 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3789 SimGrid (2.90) unstable; urgency=low
3791 Alpha 1 on the path to SimGrid 3
3793 * It is a long time since the last release of SimGrid. I'm sorry about
3794 that but as I had told you, I was rewriting a lot of things. I apologize
3795 to those who had been reporting bugs to me and that I had not answered.
3796 If your bug is still in the new version, please tell me. Here is a
3797 summary of the main changes.
3799 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3800 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3801 portability, speed, and a lot more but you'll figure it out later.
3802 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3803 I wanted to release the new MSG as soon as possible and I have
3804 broken GRAS, which is the reason why, for now, only MSG is fully
3805 functional. A laconic description of these projects is available
3806 in the documentation.
3808 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3809 kernel optimized for our needs (called SURF but only the developers
3810 should use it). Hence, MSG is now roughly 30 times faster and I think
3811 that by rewriting a little bit MSG, I could event speed it up a little
3812 bit more. Beside the gain in speed, it is also much easier to encode a
3813 new platform model with SURF than it was with SG. More to come...
3815 * REVOLUTION 3: I have tried to change a little as possible the API of
3816 MSG but a few things really had to disappear. The main differences
3817 with the previous version are:
3818 1) no more m_links_t and the corresponding functions. Platforms are
3819 directly read from an XML description and cannot be hard-coded
3820 anymore. The same format is used for application deployment
3821 description. The new format is described in the documentation.
3822 Have a look in tools/platform_generation. There is a tiny script
3823 that converts from the old platform format to the new one. Concerning
3824 the application deployment format, parsing the old one is tricky.
3825 I think most of you should however be able to convert your files. If
3826 it is really an issue, I can write a C code that does the conversion.
3828 2) the toolbox tbx does not exist anymore. We now have a library
3829 with much more data-structures but without the hash-tables (we have
3830 dictionaries that are much faster).
3832 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3834 *****************************************************************************
3835 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3836 * don't feel like converting the oldies. (Mt) *
3837 *****************************************************************************
3840 Version 2.90: "the long awaited one"
3841 - Finished rewriting and debugging MSG. Rewrote the documentation.
3842 - disable GRAS for now since it needs to be ported to the newest SG
3845 - Finish the port to windows (using mingw32 for cross-compile)
3848 - Main loop and datastructures of SURF. A cpu resource object is
3849 functional. Surf can thus be used to create cpu's with variable
3850 performance on which you can execute some actions.
3852 2004-11-15 Martin Quinson
3853 - Port to ARM. Simply added the alignment and size descriptions. Should
3854 work, but the ARM machines are so slow that I didn't had the opportunity
3855 to 'make check' over there yet.
3857 2004-11-15 Arnaud Legrand
3858 - Trace manager now written. It uses a heap structure and is therefore
3859 expected to be efficient. It may however be speeded up (particularly
3860 when many events occur at the same date) by using red and black
3861 trees. One day maybe...
3862 - Max-min linear system solver written. It uses a sparse matrix
3863 structure taking advantage of its expected use. Most operations are
3864 O(1) and free/calloc are called as few as possible. The computation of
3865 the minimum could however be improved by using a red and black tree
3868 2004-11-03 Arnaud Legrand
3869 - Rename every gras_* function that was in xbt/ to its xbt_
3871 - Add a heap and a doubly-linked list to xbt
3872 - Added a dichotomy to the dictionaries. make check works as well before
3873 so I assume that the patch is correct. I do not know however if things
3874 run effectively faster than before now. :)
3876 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3877 SimGrid, and the version number is bumped to 2.x
3879 2004-10-29 Martin Quinson
3880 - Introduction of the remote errors.
3881 They are the result of a RMI/RPC on the remote machine.
3882 ErrCodes being scalar values, you can't get the host on which those
3883 errors did happen. Extending the error mechanism as in Gnome is possible.
3884 No idea yet whether it is a good idea.
3886 2004-10-28 Martin Quinson
3887 - Interface revolution: the Starred Structure Eradication.
3888 I used to do typedef struct {} toto_t; and then handle *toto_t.
3889 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3890 - ???_t is a valid type (builded with typedef)
3891 - s_toto_t is a structure (access to fields with .)
3892 - s_toto is a structure needing 'struct' keyword to be used
3893 - e_toto_t is an enum
3894 - toto_t is an 'object' (struct*)
3896 typedef struct s_toto {} s_toto_t, *toto_t;
3897 typedef enum {} e_toto_t;
3898 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3899 s_toto_t) is private.
3901 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3902 so that it can set the variable to NULL. It was so for dicts and sets,
3903 it changed for dynars.
3905 - Fix a bunch of memleaks in dict_remove
3906 - Fix a bug in sg/server_socket opening: it failed all the time.
3908 2004-10-07 Martin Quinson
3909 - Speed up dynar lookup operation a bit.
3911 gras_dynar_get is dead.
3913 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3914 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3915 (which gives you the address of the stored data).
3917 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3918 copy of the data using an affectation to do the job and not a memcpy.
3920 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3921 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3923 It does not really speedup the dynar test because they are
3924 setting elements all the time (and look them seldom). But the dict does
3925 far more lookup than setting.
3927 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3929 2004-10-05 Martin Quinson
3930 - Allow to (en/dis)able the cycle detection at run time.
3932 Whether we should check for cycle or not is now a property of each
3933 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3934 datadesc_cycle_unset allow to remove this property when previously set.
3936 Note that the cycle detection is off by default since it impacts the
3937 performance. Watch the data you feed GRAS with ;)
3939 This property is hereditary. Any element embedded in a structure having it
3940 set have it set for the time of this data exchange.
3942 You should set it both on sender and receiver side. If you don't set it on
3943 sender side, it will enter an endless loop. If you forget on receiver
3944 side, the cycles won't be recreated after communication.
3946 - Header reorganization.
3947 Kill gras_private.h, each submodule must load the headers it needs.
3949 2004-10-04 Martin Quinson
3950 - Interface revolution: do not try to survive to malloc failure.
3952 Now, gras_malloc and friends call gras_abort() on failure.
3953 As a conclusion, malloc_error is not a valid error anymore, and all
3954 functions for which it was the only gras_error_t return value are
3955 changed. They now return void, or there result directly.
3956 This simplify the API a lot.
3958 2004-09-29 Martin Quinson
3959 - Re-enable raw sockets.
3960 Created by gras_socket_{client,server}_ext;
3961 Used with gras_raw_{send,recv}
3964 It should allow to kill the last bits of gras first version soon.
3966 This is not completely satisfactory yet (duplicate code with
3967 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3970 - Simplify transport plugin (internal) interface by not passing any
3971 argument to _server and _client, but embedding them in the socket
3974 2004-09-28 Martin Quinson
3975 - Finish the port to AIX.
3976 autoconf was my problem (segfault within the malloc replacement
3977 function. No idea why)
3979 2004-09-16 Martin Quinson
3980 - Fix some size_t madness on 64bit architectures.
3982 2004-09-08 Martin Quinson
3983 - Reduce the number of system headers loaded, overload some more system
3984 calls (such as malloc to cast the result of the system one, and work
3986 - Fix and reintroduce the config support
3988 2004-09-07 Martin Quinson
3989 - Source code reorganization to allow Arnaud to surf all over there.
3990 - Allow to document the logging categories.
3991 - Remove all uppercase from logging categories and useless cleanup in names.
3993 2004-08-18 Martin Quinson
3994 Version 0.6.2 (protocol not changed; API changed)
3995 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3996 gras_error_t), and NULL when not found. Functions expecting a msgtype
3997 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3998 hopefully usefull message.
3999 - Portability to prehistoric sparcs again
4001 2004-08-17 Martin Quinson
4002 Version 0.6.1 (protocol not changed; ABI not changed)
4003 - prealloc some buffers to speed things up
4005 2004-08-11 Martin Quinson
4006 Version 0.6 (protocol not changed; ABI expended)
4007 - The parsing macro can deal with the references, provided that you add
4008 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4010 2004-08-09 Martin Quinson
4011 Version 0.5 (protocol not changed; ABI changed)
4012 - Allow to off turn the cycle detection code in data exchange at
4013 compilation time. It should be at run time, but I'm short of time (and
4014 the config stuff is still broken). That way, we keep dict out of the
4015 critical path, which is good because the performance is poor:
4016 - search not dichotomial yet
4017 - dynar give no way to access their content and memcpy everytime
4018 - In composed data description (struct, ref and so on), stop foolness of
4019 keeping the subtype's ID, but store the type itself. This keeps sets out
4020 of the critical path, which is good since they rely on dynar and
4021 dictionaries. The only loose of that is that we cannot detect the
4022 redeclaration of a structure/union with another content (but I'm not sure
4023 the code detected well this error before anyway). We still can detect
4024 the redefinition discrepancy for the other types.
4025 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4026 it breaks the code because of type-punning used all over the place).
4027 This breaks on all non-gcc architectures (for now).
4029 All those changes (plus the buffer of last time) allow me to gain 2 order
4030 of magnitude on cruel tests consisting of 800000 array of integers on two
4031 level of a hierarchical structure (200 secondes -> 4 secondes)
4034 - the selector of reference must now return the type it points to, not
4035 the ID of this type.
4037 2004-08-06 Martin Quinson
4038 Version 0.4 (protocol changed; ABI not changed)
4039 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4040 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4041 - Implement a buffer transport, and use it by default (it relies on tcp in
4042 real life and on sg in simulation).
4043 That's a bit hackish since I had a new field to the structure to store
4044 its data without interfering with the subtype ones. Inheritance
4045 is tricky in C. And that's a kind of reverse inheritance with one class
4046 derivating two classes. Or maybe a game with java interfaces. Anyway,
4047 that's damn hard in C (at least).
4048 Moreover, I got tired while trying to ensure plugin separation and
4049 genericity in SG mode. MSG wants me to do weird things, so let's go for
4050 cruel hacks (temporarily of course ;).
4051 See comment in transport_private.h:71
4052 - do not include all the _interface headers in private but in the files
4053 which really need them (to cut the compilation time when they are
4056 2004-07-26 Martin Quinson
4057 Version 0.3 (protocol not changed; ABI changed)
4058 - Major overhault of the datadesc interface to simplify it:
4059 - shorted the function names:
4060 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4061 - add a trivial way to push/pop integers into the cbps without malloc.
4062 This allows to make really generic sub_type description, which simply
4063 pop their size of the stack.
4064 - add a function gras_datadesc_ref_pop_arr() which does what users want
4065 most of the time: Declare a dynamic array (which pops its size of the
4066 stack) and declare a reference to it. Poor name, but anyway.
4067 - kill the post-send callback, add a post-receive one
4069 2004-07-23 Martin Quinson
4070 Version 0.2 (protocol changed; ABI changed)
4071 - add some testing for cpbs in the test cases, and fix some more bugs.
4072 This invalidate again the little64 data file, since I cannot regenerate
4074 - remove an awfull optimization in the logging stuff, allowing me to:
4075 - understand it again
4076 - learn gcc how to check that the argument match the provided format
4077 - fix all errors revealed by gcc after that
4078 - internal keys of dict are not \0 terminated. Deal with it properly in
4079 loggings instead of segfaulting when the user want to see the logs :-/
4081 2004-07-22 Martin Quinson
4082 - Fix some stupid bug preventing cbps (callback postit) from working
4084 2004-07-21 Martin Quinson
4085 - Some documentation cleanups
4086 - remove the useless last argument of msgtype_declare
4087 - rename the Virtu functions to fit into the 'os' namespace
4088 - move headers src/include -> src/include/gras/ and stop fooling with
4089 gras -> . symbolic link
4090 - make distcheck is now successful
4092 2004-07-19 Martin Quinson
4094 - Build shared library also
4095 - Install html doc to the right location
4096 - stop removing maintainer files in make clean
4097 - build tests only on make check
4099 2004-07-13 Martin Quinson
4101 - No major issue in previous version => change versioning schema
4102 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4103 corresponding dataset.
4105 2004-07-11 Martin Quinson
4107 - Get it working with any kind of structure (we can compute the padding
4108 bytes remotely for all the architectures I have access to)
4109 - Implement the structure parsing macro (still not quite robust/complete)
4110 - Improvement to the remote testing toysuite
4112 2004-07-10 Martin Quinson
4113 [autoconf mechanism]
4114 - get ride of a bunch of deprecated macros
4115 - actually run the test for two-compliment, not only compile it :-/
4116 - test whether the structures get packed (and bail out if yes. Damn.
4117 Alignment is a serious matter)
4118 - test whether the structures get compacted (but respecting the alignment
4119 constraints of each types)
4120 - test whether the array fields of structures can straddle alignment boundaries
4122 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4123 (took me 2 hours to find that bug, looking at the wrong place)
4124 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4125 before sending/receiving and are used to compute the offsets of fields
4126 - Given that padding size depend even on compiler options, keep track of
4127 alignment and aligned_size only for the current architecture. Not a big
4128 deal since we send structure fields one after the other (seems
4130 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4131 program, let it work on linux/gcc/little32. portability todo.
4133 2004-07-08 Martin Quinson
4134 - import and improve remote compilation support from FAST
4135 - make sure make check works on half a dozen of machines out there
4137 2004-07-07 Martin Quinson
4138 Let's say it's version 0.0.3 ;)
4139 - Implement conversions (yuhu!)
4140 - Let it work on solaris (beside conversion, of course)
4141 - Stupid me, using rand() to generate the conversion datatests in not wise.
4143 2004-07-06 Martin Quinson
4144 - Let make dist work, since I'm gonna need it to compile on remote hosts
4145 - Let Tests/datadesc_usage write the architecture on which the file was
4146 generated as first byte.
4147 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4148 architecture descriptions.
4149 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4150 on those architectures.
4151 - Optimization: send/recv array of scalar in one shoot
4153 2004-07-05 Martin Quinson
4154 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4156 - Plug a whole bunch of memleaks
4157 - each process now have to call gras_{init,exit}. One day, their log
4158 settings will be separated
4159 - Continue the code factorisation between SG, RL and common in Transport.
4161 2004-07-04 Martin Quinson
4163 - Redistribution between SG and RL.
4164 We wanna have to accept in SG, so move accepted related parts of RL in
4165 the common part. (more precisely, the dynar of all known sockets is no
4166 more a static in transport.c, but part of the process_data)
4168 [gras_stub_generator]
4169 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4171 2004-07-03 Martin Quinson
4172 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4173 them all up in one shot)
4175 - Fix the ugly case of reference to dynamic array.
4176 - New (semi-public) function gras_datadesc_size to allow the messaging
4177 layer to malloc the needed space for the buffer.
4179 - gras_socket_close now expect the socket to close (and not its address to
4180 put NULL in it after it). This is because the socket passed to handlers
4181 is one of their argument (=> not writable).
4183 - propagate the interface cleanup from last week in datadesc, ie remove a
4184 superfluous level of indirection. User pass adress of variable
4185 containing data (both when sending and receiving), and not of a variable
4186 being a pointer to the data. Let's say that I like it better ;)
4187 The price for that is constructs like "int msg=*(int*)payload" in
4188 handlers, but it's a fine price, IMHO.
4190 - Let it work in RL (yuhu)
4192 2004-06-21 Martin Quinson
4194 - porting SG plugin and SG select to new standards (works almost).
4195 - plug memleaks and fix bugs around.
4198 - cleanup the prototype of data recv and force users to specify when they
4199 want to handle references to objects. Test case working even for cycles.
4200 - plug memleaks. Valgrind is perfectly ok with this.
4202 2004-06-12 Martin Quinson
4204 - cleanup the separation between plugin and main code in plugin creation
4206 2004-06-11 Martin Quinson
4208 - Reput hook for raw sockets, needed for BW experiments
4209 - kill a few lines of dead code
4210 [Data description] Interface cleanup
4211 - gras_datadesc_by_name returns the searched type or NULL.
4212 That way, no variable is needed to use a type desc once, which makes
4214 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4215 two parameters were strings (field name, type name), leading to
4217 [Dicos] Interface cleanup
4218 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4219 This is consistant with the dynar API.
4221 2004-04-21 Martin Quinson
4223 - Porting to new standards.
4225 - interface cleanup.
4226 There is no bag anymore, no need to take extra provision to mask the
4227 pointers behind "ID".
4228 Better splitup of functions between files create/exchange/convert.
4229 This is still a bit artificial since convert and receive are so
4230 interleaved, but anyway.
4232 - add a queued message list to procdata (the ones not matching criteria
4234 - factorize some more code between SG and RL wrt procdata
4236 - use gras_exit in example to track memleaks
4237 - get rid of gs_example now that GS is properly integrated into gras
4238 - update run_test to integrate the latest tests (datadesc)
4240 - rename WARNINGn macros to WARNn since it prooved error-prone
4242 2004-04-19 Martin Quinson
4244 - register init/exit functions within gras module mechanism
4245 - send/receive function.
4246 Convertion is not implemented, but short-cutted if not needed.
4247 struct/array elements are sent one by one (instead of block-wise), but
4248 nobody really cares (yet). Get a prototype before optimizing.
4249 - tests (using a file socket) for DD send/receive on:
4250 - base types: int, float
4251 - array: fixed size, string (ie ref to dynamic string)
4252 - structure: homogeneous, heterogeneous
4253 - chained list, graph with cycle
4254 Believe it or not, valgrind is not too unhappy with the results. The
4255 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4259 - Bugfix when using a filename explicitly (instead of '-')
4261 2004-04-09 Martin Quinson
4263 - factorize more code between RL and SG in socket creation
4264 - Complete the implementation and tests of:
4266 o file (only in RL, and mainly for debugging)
4268 I lost 3 days to design a portable address resolver, and then decided
4269 that the prototype mainly have to run on my box.
4270 Addressing portability too early may be like optimizing too early :-/
4272 - use gras_init in the Tests instead of the crappy parse_log_opt
4273 (the latter function is removed)
4274 [Conditional execution]
4275 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4276 [Code reorganisation]
4277 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4278 Build examples against the RL library, since there is no way to disable
4279 its creation for now.
4281 For information, the beginning of coding on GRAS was back in june
4282 2003. I guess that every line has been rewritten at least twice since