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'
10 - Removed all that was related to the "Storage" type in all APIs and bindings.
11 "Disks" have been introduced more than a year ago to replace them for the better.
12 It's time to finish this replacement.
15 - Define new template functions Mailbox::get_unique(), returning a std::unique_ptr.
16 - Functions Mailbox::get() and Mailbox::get_async() are now templated with the
17 type of the pointee. Untyped functions are deprecated. Use Mailbox::get<void>()
18 or Mailbox::get_async<void>() if you really want to play with void*.
19 - Unify the interface of Activity::on_{start/activity}
20 - New function: Comm::get_dst_data()
21 - New function: Comm::set_payload_size() to set the amount of bytes to simulate.
22 - New functions: Comm::sendto_{init,async} to initiate a communication
23 on between two (possibly remote) hosts.
26 - Drop simgrid::xbt::demangle(). Please use boost::core::demangle() instead.
27 - Drop xbt_dynar_cursor_rm(), xbt_dynar_search_or_negative(),
28 xbt_str_split_quoted(), xbt_str_split_quoted_in_place().
29 - Deprecate xbt_dynar_to_array().
31 C binding and interface:
32 - Rename (lowercase) functions sg_actor_get_PID, sg_actor_get_PPID, sg_actor_by_PID.
35 - New section: Configuring the logs from the command line.
36 - New section: Programming API of logs.
38 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
39 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
40 - FG#37: Parallel tasks are limited to 1 core per host
41 - FG#62: Running "smpirun -replay" on large networks
42 - GH#343: Wrong consumed energy when an exec timeouts
44 ----------------------------------------------------------------------------
46 SimGrid (3.26) December 15. 2020
48 The Release release (the French lockdown was eased today).
50 Important user-visible changes:
51 - SimGrid now requires a compiler with C++14 support.
52 Sibling projects should upgrade their FindSimgrid.cmake
53 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
54 several users had difficulties to understand issues when using high bandwidth or
55 small latency events. The new value was already the default for SMPI and
56 should not cause too much performance hit. It can change some simulation
57 timings if simulation was skipping a lot of small events (which was actually
58 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
59 - On some network models (L07, CM02, LV08), an implicit loopback link is created
60 for local communication. Default bandwidth was 500MBps and latency was 15µs,
61 which was rather bad. Default values have been changed to 10GBps and 0 latency.
62 Options network/loopback-lat and network/loopback-bw have been added to adapt
63 these values if needed. Network and routing models already providing separate
64 loopbacks (clusters and such) are not impacted by this change.
65 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
66 metadata that are added to each MPI message by MPI libraries (tag,
67 src/dest, and communicator information). This may change simulation times
68 for very small message sizes, but should be more realistic.
71 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
72 - New plugins: wifi_energy and link_load. See the associated examples.
73 - New examples: wifi networks, and communication suspend.
76 - LTO is now enabled for Intel/clang compilers.
77 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
78 Setting it to "auto" will use all available cores, while setting it to n will
79 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
80 - Remove obsolete runtime option 'contexts/parallel-threshold'.
81 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
82 has been kept for compatibility.
83 - Finally remove obsolete snake_case() aliases for runtime options.
84 - Further improve the documentation.
87 - the on_exit() of each actor is also executed when the simulation deadlocks.
88 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
89 An example is provided for s4u::Comm but it should work with Exec and Io.
92 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
93 https://framagit.org/simgrid/SMPI-proxy-apps
94 - MPI/IO: supports names starting with "./"
95 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
97 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
98 building shared libraries, this may cause issues, so environment variable
99 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
100 - most temporary files should now be created in /tmp dir (or equivalent).
101 If this one does not allow execution of code (noexec flag), this may cause issues.
102 Please use another tmp directory (using TMPDIR or equivalent system variable)
106 - A large refactoring is ongoing, but this should be transparent for users so far.
109 - Drop xbt_str_split().
111 C binding and interface:
112 - The return type of the sg_actor_on_exit() callbacks is now 'void'
113 (int value was previously ignored)
114 - Many C functions were renamed to follow a common naming style
115 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
116 names are kept until version 3.30.
117 - Many MSG tests were converted to the new S4U's interface in C, that
118 was extended for that.
120 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
121 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
122 - FG#41: Add sg_actor_create C interface
123 - FG#43: xbt::random needs some care
124 - FG#48: The Impossible Did Happen (yet again)
125 - FG#50: Suspending an actor executed at the current timestamp fails
126 - FG#52: Zero-seconds timeout : "That's in the past already"
127 - FG#53: Crash while using ns-3 network model
128 - FG#54: How to suspend a comm?
129 - FG!22: Stochastic Profiles
130 - FG!24: Documentation and fix for xbt/random
131 - FG!35: Add a modeling hint for parallel links in doc
132 - FG!36: [xbt/random] Read/Write the state of the RNG
133 - FG!42: Link cumulated load plugin
134 - FG!44: Test condition variable wait_until+notify_all
135 - GF#18137: Allow different stack sizes?
136 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
137 - GH#139: Allow pthread creation in SMPI
138 - GH#336: Packet-level simulation using SMPI?
139 - GH#345: Error in the LMM while migrating a VM
140 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
141 - GH#352: pip install / python setup.py install fail to find pybind11
142 - GH!337: Fix link_energy plugin for wifi platforms
143 - GH!339: Add Mailbox set_receiver method to python binding
144 - GH!344: Cast hugepages macros parameters to int64
146 ----------------------------------------------------------------------------
148 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
150 The Palindrom Day Release.
152 Important user-visible changes:
153 - Improve the Python usability (stability and documentation).
154 - A nasty synchronization bug was ironed out, see also below.
155 - Python's doc was organized as a separate tree, now integrated with C++.
156 - C bindings of S4U were not part of the doc.
157 - The C++ doc was also improved as methods are now split by theme.
158 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
159 - This is mandatory to use the Java bindings.
160 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
161 - SimDAG++: Automatic dependencies on S4U activities (experimental)
162 - Some features are already implemented but not all of them
163 - Cannot block an activity until it's scheduled on a resource
164 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
165 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
168 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
169 - Drop support for Viva/Triva (old visualization tools), hence removed
170 TRACE_get_node_types() and TRACE_get_edge_types() functions.
173 - Require pybind11 version 2.4.
174 - Greatly improve locking during context switches with Python.
175 - Declaring Python callbacks to this_actor::on_exit is now working.
176 - No more tests are failing currently, even on win32.
179 - Actor: Merge signals on_migration_start/end into on_host_change
180 - Actor: Rename migrate() into set_host()
181 - Disk: Allow users to get the read and write nominal bandwidth values
182 - Exec: Implement wait_for(timeout)
183 - Io: Implement wait_for(timeout)
186 - Parse errors now raise a simgrid::ParseError that you may want to catch.
189 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
190 been renamed to fix a spelling error. As usual, the previous definitions are
191 kept with a deprecation warning. However, to avoid runtime errors, it is now
192 forbidden to override these deprecated methods in a derived class. Please use
193 the new names immediately if you need to override them.
196 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
197 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
198 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
199 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
200 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
201 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
202 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
203 - New module: random, providing classical random numbers generators.
206 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
207 allocations' size above which they are considered "shared" by default (as if
208 they were performed through SMPI_SHARED_MALLOC macros).
209 Default = 0 = disabled feature.
210 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
211 This can cause some troubles if codes are already overriding these. If this
212 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
213 help, but will make this feature unusable.
214 - Input check is now performed much more consistently and outputs precise
215 warnings for each error found
216 - New supported calls : MPI_File_set_view, MPI_File_get_view
217 - MPI I/O now only supports disks, as storage is discontinued.
218 - MPI I/O allows opening files without specifying absolute path
219 - Note : SMPI configuration options are now initialized by smpi_init_options()
220 call, which should be called if SMPI is not being used through smpirun.
222 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
223 - FG#9: Python bindings crashing
224 - FG#39: Missing s4u::Comm::wait_any_for example
225 - FG#42: Add support for ThreadSanitizer (TSan)
226 - FG!19: Removing RngStream
227 - FG!20: A module for RNG calls
228 - FG!21: Choice between ad-hoc and standard distributions implementations
230 - FG!25: Fix link in Mutex doc
231 - FG!26: Fix links in SMPI interface doc
232 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
233 - GH#305: Unscheduled tasks are still excuted
234 - GH#313: smpirun: manual outdated w.r.t. --help
235 - GH#321: [S4U] Get task remaining work ratio
236 - GH#323: Crash when an actor turn off his physical host
237 - GH#335: Missing links on dragonfly example svg
238 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
239 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
241 ----------------------------------------------------------------------------
243 SimGrid (3.24) October 9. 2019
245 The Clean Disk Release.
248 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
249 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
250 thus becomes the default resource to execute I/O operations. Examples
251 and the file system plugin have been modified to use disks instead of
252 storage elements. Storage can still be used but is doomed to disappear.
253 Users are thus adviced to update their simulators accordingly.
254 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
255 for consistency with pthread_barrier_wait()
256 - Host::get_englobing_zone() returns the englobing netzone
257 - Actor::on_destruction is now called in the destructor
258 Actor::on_termination new signal called when the actor terminates
260 - Global signals are now part of the Engine:
261 - on_platform_creation: after config settings, before the XML parsing
262 - on_platform_created: right after the XML parsing
263 - on_time_advance: each time the clock advances
264 - on_simulation_end: after simulation, before cleanups
265 - on_deadlock: as the name implies.
267 - sg_{actor,host,link}_{data,data_set}() now all exist.
268 Use them to attach user data to the object and retrieve it.
271 - Introduce an experimental Wifi model. It sounds reasonable
272 according to the state of the art, but it still has to be properly
273 validated, at least against ns-3 if not against reality.
274 - Improved the usability of ns-3. Several bugs were ironed out.
275 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
276 It is now expressed as 'idle:epsilon:allCores' to properly model the
277 consumption of non-whole tasks on mono-core hosts. More info in the doc.
280 - convert a new set of functions to the S4U C interface and move the old MSG
281 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
284 - Fortran bindings for DVFS have been removed.
285 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
286 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
287 - SMPI can now be selected by cmake's find_module(MPI) with
288 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
289 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
290 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
291 - trace-call-location can be used with TI traces, and replayed, and is compatible
292 with smpi/comp-adjustment-file.
293 - sleep events are now correctly traced/replayed.
294 - Default for trace-call-location is now to use file names and not full paths.
295 To revert to previous behavior (in case of collision of filenames), option
296 "smpi/trace-call-use-absolute-path" can be set to yes.
299 - Use the included xxHash as an hash implem when C++14 is usable.
300 - Option model-checker/hash was removed. This is always activated now.
301 - New option smpi/buffering controls the MPI buffering in MC mode.
302 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
303 This is useful to check for MPI compliance.
306 - New section on plugins: how to define a new one, and existing ones.
309 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
310 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
313 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
314 and <mount> tags. The rationale is finish to decouple what should be handle
315 by SimGrid kernel (i.e., resources and activities) from plugins built in
316 user space (e.g., the file system plugin). This new tag currently accepts
317 only three attributes (id, read_bw, and write_bw). All the other information
318 that was declared with the storage related tags now has to be expressed as
319 properties. An example of platform using this new tag is available at
320 examples/platforms/hosts_with_disks.xml
323 - 'expect signal' can now accept more than one potential signal.
325 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
326 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
327 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
328 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
329 - FG#31: per-actor data
330 - FG#34: SG_BARRIER_SERIAL_THREAD?
331 - FG#35: model-checker does not like buster-produced binaries
332 - FG!13: MC: complete workaround in the error msg seen on modern systems
333 - FG!15: execute_flops now logs compute
334 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
335 - FG!17: ns-3: unblock the right number of communications + others issues
336 - FG!18: Improving the performance of the ns-3 bindings
337 - GH#189: Energy consumption of parallel tasks
338 - GH#219: Error in the throughput of TCP transfer
339 - GH!330: Minor polishing of English: Configuring_Simgrid
340 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
341 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
342 - GH!334: Rework energy plugin, again
344 ----------------------------------------------------------------------------
346 SimGrid (3.23.2) July 8. 2019
349 - Nicer introduction page.
350 - Migrate the "Deploy your application" page to the new doc.
351 - Move Java as a subtree of MSG.
354 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
357 - Drop xbt_dynar_sort_strings().
360 - Really fix FG#26: Turning off a link should raise NetworkFailureException
361 - FG#27: Wrong exception thrown to wait_any when link is turned off
362 - GH#328: Java: Canceling multiple tasks in a single vm/host
364 ----------------------------------------------------------------------------
366 SimGrid (3.23) June 25. 2019
368 The Exotic Solstice Release.
371 - SunOS and Haiku OS support. Because exotic platforms are fun.
372 - Stop setting random seed with srand48() at initialization.
373 - Use addr2line as a fallback for stacktraces when backtrace is not available.
374 - Build option -Denable_documentation is now OFF by default.
375 - Network model 'NS3' was renamed into 'ns-3'.
378 - Simgrid can now hopefully be installed with pip.
381 - wait_any can now be used for asynchronous executions too.
384 - New log appenders: stdout and stderr. Use stdout for xbt_help.
385 - Drop xbt_dict_dump.
388 - 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
389 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
390 - MPI interface is now const correct for input parameters
391 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
392 - Fortran bindings for async collectives.
393 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
396 - Remove option 'model-check/record': Paths are recorded in any cases now.
397 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
398 always sparse. That's an immense gain in memory, and can even be
399 faster because of cache effects. So there is no need to clutter the
400 code to allow the user to go for the unefficient mode.
403 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
404 models proved to be more accurate than these old experiments.
406 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
407 - FG#1: Broken link in error messages
408 - FG#2: missing installation documentation
409 - FG#3: missing documentation in smpirun
410 - FG#6: Python bindings not available on PyPI
411 - FG#7: simple cmake call requires doxygen
412 - FG#8: make python bindings an optional dependency
413 - FG#10: Can not use MSG_process_set_data from SMPI any more
414 - FG#11: Auto-restart actors forget their on_exit behavior
415 - FG#12: -Denable_lto=OFF doesn't disable LTO
416 - FG#13: Installs unstripped file 'bin/graphicator'
417 - FG#14: Installs the empty directory 'doc/simgrid/html'
418 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
419 - FG#17: Dead link in doc (pls_ns3)
420 - FG#20: 'tesh --help' should return 0
421 - FG#21: Documentation link on http://simgrid.org/ broken
422 - FG#22: Debian installation instruction are broken
423 - FG#26: Turning off a link should raise NetworkFailureException exceptions
424 - GH#133: Java: a process can run on a VM even if its host is off
425 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
426 - GH#326: Valgrind-detected error for join() when energy plugin is activated
428 ----------------------------------------------------------------------------
430 SimGrid (3.22) April 1. 2019
432 The Easter Christmas Release (this one is somewhat late).
435 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
436 - Not complete yet: asynchronous activities (amongst others) are still missing
437 - Still ongoing: the interface may change in the future. We need more
438 testers! Please report any glitches.
439 - No new project using Java should start now. Please switch to Python.
442 - Some of the internal cleanups may lead to speed improvements:
443 - The hard limitation on the amount of simulated actors with
444 Java+Mac was removed. Now, the available memory is the only limit.
445 - Our refcounting was tidyied, leading to 10% speedups in some cases.
446 - We are still working on making our code robust to the actor kills
447 and hosts' churn. Things are improving, but it's not perfect yet.
448 - Replaced our own code to display a backtrace (that was forking addr2line)
449 with the Boost.Stacktrace library.
450 You won't see your backtraces without this optional dependency.
451 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
452 - Stop setting random seed with srand() at initialization.
455 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
456 XML file version remains 4.2 since old files are still compatible.
459 - Process termination which was broken at version 3.21 has been repaired.
460 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
461 - Hide the examples into examples/deprecated. New users should use Python.
464 - Drop MSG_process_create_from_stdfunc() from the API.
465 This C++ function was a pimple in the C API, made necessary at some
466 point by the Java bindings. This is fixed now.
467 - Hide the examples into examples/deprecated. New users should use S4U.
468 - MSG_process_create and MSG_process_attach now crash if the host used
469 to run this process is off.
470 - Fix the protype of MSG_process_on_exit()
471 Now use (int,void*) callbacks instead of (void*,void*) ones.
472 The implementation was ways too messy to actually work, I guess.
475 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
476 too often and provide more accurate timings.
477 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
478 - MPI_Alltoallw support
479 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
480 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
481 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
482 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
484 - Basic implementation of generalized requests (SMPI doesn't
485 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
488 - Drop sg_cmdline. Please use xbt_cmdline instead.
489 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
490 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
491 OsSemaphore is implemented in a portable way with C++11 threads.
492 This should allow much more threads to be created at the same time,
493 allowing Mac and Java users to simulate many more actors.
494 - Implement the 'thread' factory with std::thread instead of xbt ones.
495 It is not possible to set the stack size with threads anymore, but
496 -fsplit-stack is the way to go nowadays when using threads.
497 - Drop the xbt_os_thread_t module (now unused)
498 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
499 - Drop xbt_str_join_array().
500 - Drop cunit, use Catch2 instead.
503 - Many cleanups in the kernel::activity namespace. This was long
504 overdue, and shall open the path to many future endeavors.
507 - #132: Java: a process can not shut down its own host
508 - #220: S4U: detached send causes exception if sender terminates soon after sending
509 - #261: Document the parameters of parallel execution's constructor
510 - #300: [s4u] BarrierPtr is missing
511 - #314: SMPI args internal cleanup
512 - #316: Fix a bug related to the CPU utilization of multi-core VM
513 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
514 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
515 - #325: Turning off a host has different behavior on sleeping actors and computing actors
517 ----------------------------------------------------------------------------
519 SimGrid (3.21) October 3. 2018
521 The Restarting Documentation (TRD) Release.
524 - Convert the user manual to Sphinx for improved usability.
525 Unstable version now hosted on https://simgrid.frama.io/simgrid
526 Many glitches remain; stay tuned.
527 - Restrict the Doxygen reference API to the installed header files.
528 (documenting all internals this way is overkill + doxygen scales badly)
529 - New tutorial on S4U (highly inspired from the old MSG one, but with
530 a git repository to fork as a starting point, and a docker image)
531 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
532 - The Developper manual is still to be converted (not compiled until then)
533 - Some parts are still missing in this conversion (such as the
534 platform and deployment sections) while others were blindly converted
535 and would need more love (such as the configuration flags).
536 Things will certainly further improve in the future releases.
539 - Fully reimplement auto_restart mechanism that was utterly broken
540 (fix #22, #131 and #281 that were all very related)
541 - Implement semaphores (fix #303 and #291)
542 - When creating an actor from a function and its parameters,
543 move-only parameters are not allowed anymore, as it would prevent
544 the actor to be restartable if its parameters are consumed on run.
545 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
546 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
547 IO operations that can be started, waited for, or canceled as a regular
549 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
553 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
554 - New host variable: 'core_count'
557 - Remove xbt_os_thread_specific features
558 - Remove portability wrapper to condition variables
559 - Remove xbt_os_thread_yield()
562 - MPICH collective selector now mimics MPICH 3.3b
563 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
564 - MPI_Init(NULL,NULL) is now allowed
565 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
566 - Command-line arguments should now be handled properly in Fortran simulations
567 - MPI Attributes and related callbacks should work in Fortran
568 - Apps using C + Fortran should now work
569 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
570 - Support MPI_Error_String
573 - Due to an internal bug, Msg.run() must now be your last line.
574 We hope to fix it in a future release, and we are sorry for the inconvenience.
577 - #22: Process autorestart seem to only work with CAS01 cpus
578 - #56: Feature request: dynamic SMPI replay
579 - #93: simgrid should not eat --help
580 - #111: How to change the loopback link characteristics?
581 - #116: Platform cluster radical
582 - #117: Platform cluster/host order
583 - #129: Replay_multiple on steroid
584 - #131: Java : autostart does not work
585 - #143: Setting a breakpoint at a given time
586 - #235: Network energy models should be integrated
587 - #264: Add ptask L07 resource tracing
588 - #271: Dynamic and manual replay of SMPI traces
589 - #279: Breakpoints option uses time deltas instead of absolute time
590 - #281: Daemonized actor no longer daemonized after an auto-restart
591 - #282: TIT convertor erase traces when using absolute path in the trace list file
592 - #285: segfault when a s4u actor kills itself with msg process tracing activated
593 - #286: Pajé traces are not exposing the number of cores
594 - #287: Command-line parsing should stop when encountering '--'
595 - #288: MPI_Init(NULL, NULL)
596 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
597 - #290: Method to check if a mailbox has a message ready to be consumed
598 - #291: [s4u] semaphores do not seem available yet
599 - #294: Alltoall collective implementation of mpich
600 - #295: Starting a Host as OFF is not supported
601 - #296: DTD too permissive
602 - #299: Add a s4u barrier example
604 ----------------------------------------------------------------------------
606 SimGrid (3.20) Released June 24. 2018
608 The proxy snake_case() release.
611 - The whole API is now using snake_case() instead of mixing it with camelCase.
612 Compatibility wrappers in place for one year.
613 - We took the opportunity to increase the consistency of the API.
614 Please report any remaining glitches.
615 - New function simgrid::s4u::Host::get_actor_count:
616 Returns the number of actors running on a specific host.
619 - SMPI is now tested with ~45 proxy apps from various sources, with none or
620 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
621 - Replay: The replay file has been re-written in C++.
622 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
623 supported. They are stored in the trace and used when replayed.
624 - Basic support of MPI_Cancel. Robustness not guaranteed.
625 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
626 (only for MPI_COMM_TYPE_SHARED).
627 - New option: smpi/privatize-libs, to add external shared libs to be privatized
628 by SMPI. They will be copied locally and loaded separately by each process.
629 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
630 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
631 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
632 with internal index optind with multiple processes. Only works if MPI_Init has
634 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
635 syntax without preprocessing tricks.
638 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
639 behavior tracing is no longer limited to MSG
640 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
641 this means that traces now need to be updated or re-obtained; both Irecv
642 and Isend lines in your traces have now in total 5 mandatory fields:
643 <rankid> <command> <to/from rankid> <tag> <size>
644 To update your traces, it suffices to add a 0 for the tag here.
645 - TIT now also supports waiting for a distinct request via MPI_Wait.
646 Wait/Test now wait for a specific request, not just the last one that was
647 issued. This unfortunately means another update, because we need to
648 identify which request you want to wait for. We do this via the
649 triplet (sender, receiver, tag), which needs to be added:
650 <rankid> <command> <sender> <receiver> <tag>
651 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
652 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
655 - The deprecation of MSG is ongoing (but this should not impact you).
656 Many MSG functions are now simple wrappers on the C API of S4U. If
657 you wish to convert your code to S4U, find the S4U counterparts of
658 your MSG calls in src/msg/msg_legacy.cpp.
659 - MSG can still be used, but won't evolve anymore.
662 - Allow to run the Link energy plugin from the command line with
663 --cfg=plugin:link_energy
664 - Rename Energy plugin into host_energy
665 - Rename Load plugin into host_load
668 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
669 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
670 - Improve performance when handling timeouts of simix synchros.
673 - Config: the C API is now deprecated (will be removed in 3.23), and
674 the C++ API has been slightly improved.
677 - Fix several build issues on OSX.
678 - Move simgrid_config.h to simgrid/config.h (old header still working)
681 - #143: Setting a breakpoint at a given time
682 - #150: Inconsistent event names in SMPI replay
683 - #258: daemonized actors hang after all non-daemonized actors have completed
684 - #267: Linker error on unit_tmgr
685 - #269: SMPI: tracing of MPI_Wait/all/any broken
686 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
688 ----------------------------------------------------------------------------
690 SimGrid (3.19.1) Released March 22 2018
692 The Fixed ABI Release.
694 As you may know, we are currently refactoring SimGrid in deep.
695 Upcoming SimGrid4 will be really different from SimGrid3: modular,
696 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
698 Our goal is to smooth this transition, with backward compatibility
699 and automatic update paths, while still progressing toward SimGrid4.
701 SimGrid remains open during works: The last pure SimGrid3 release was
702 v3.12 while all subsequent versions are usable alpha versions of
703 SimGrid4: Existing interfaces remain unchanged, but the new S4U
704 interface is budding and the internals are deeply reorganized.
706 Since 2015, we work hard to reduce the changes to public APIs. When
707 we need to rename a public library symbol in S4U, we let your
708 compiler issue an explicative warning when you use the deprecated
709 function. These messages remain for four releases, i.e. for one full
710 year, before turning into an error. Starting with v3.15, your can
711 also adapt to API changes with the SIMGRID_VERSION macro, that is
712 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
714 Starting with this v3.19.1, our commitment to reduce the changes to
715 the public interfaces is extended from the API to the ABI: a program
716 using only MSG or SimDag and compiled against a given version of
717 simgrid can probably be used with a later version of SimGrid without
718 recompilation. We will do our best... but don't expect too much of
719 it, that's a really difficult goal during such profund refactoring.
721 The difference between v3.19 and v3.19.1 is that the former was
722 accidentally breaking the ABI of MSG, while the later is restoring
725 S4U and kernel APIs will still evolve until SimGrid4, with one-year
726 deprecation warnings as currently. In fact, cleaning up these
727 interfaces and converting them to snake_case() is one release goal of
728 v3.20. But don't worry, we are working to smooth this upgrade path.
730 In summary, new projects should start with S4U to benefit of the
731 future, but old MSG projects should still be usable with no change.
733 ----------------------------------------------------------------------------
735 SimGrid (3.19) Released March 20 2018
737 The Moscovitly-cold Spring release.
739 SMPI user-visible improvements:
740 - Support of MPI_Type_create_subarray
741 - Unimplemented MPI calls now abort (instead of issuing a warning)
742 Some harmless calls around errhandlers still only issue a warning.
743 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
744 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
745 is now possible (mmap privatization was refactored to this end).
746 Adding new SMPI instances should also be technically doable now.
748 SMPI internal cleanups:
749 - Removed index notion from SMPI -- no more "getPid() - 1"!
750 In tracing files, actors are called rank-<PROCESS-ID> but these are
751 NOT the real ranks (tracing several communicators is not supported yet).
752 - Deprecate smpi_process_index() (will be removed in 3.22)
753 - Better testing on classical proxy apps, through an external project:
754 https://github.com/simgrid/SMPI-proxy-apps/
757 - Use a graphical TOC to make it easier to find the documentation you need
758 - Revamp the MSG tutorial
761 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
762 - Remove parameter of MSG_process_killall().
763 Resetting the PID was bogus anyway (several actors could have the same PID).
766 - Execution->setHost() can be called after start(), to migrate it.
767 - Comm::test_any() is now implemented.
768 - s4u::Actor now has onCreation() and onDestruction() signals.
769 - Install some kernel header files for the users' plugins and more.
770 Warning, their API is really not stable yet.
771 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
772 include/simgrid/kernel/routing/*Zone.hpp
775 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
777 Build System and other cleanups:
778 - Remove unused run-time parameter "tracing/onelink-only".
779 - SimGrid now works with intel compilers (MC still buggy)
780 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
781 - tesh kills subprocesses on timeout (but not on windows)
784 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
787 - #194: Feature request: simgrid::s4u::Comm::test_any()
788 - #245: migrating an actor does not migrate its execution
789 - #253: Feature Request: expose clusters as objects
790 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
791 - #255: Tesh broken on Windows
792 - #256: Modernize FindSimGrid.cmake
793 - #257: Fix (ab)use of CMake install
795 ----------------------------------------------------------------------------
797 SimGrid (3.18) Released December 24 2017
799 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
801 MSG IS NOW DEPRECATED. Please use S4U for new projects.
802 - Support for MSG should not be removed from SimGrid before 2020,
803 but future evolutions will be limited to the integration of
804 user-provided patches.
807 - s4u::onDeadlock() signal, run before stopping the simulation.
808 - s4u::Exec: asynchronous executions (abstraction of background
810 - s4u::Host->getLoad() returns the achieved speed in flops/s
813 - Link::name() is deprecated and will be removed in v3.21, use
814 Link::getCname() instead.
815 - Mailbox::getName() changed to return a std::string, use
816 Mailbox::getCname() instead to get a char*.
817 - Storage::getName() changed to return a std::string, use
818 Storage::getCname() instead to get a char*.
819 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
820 to have both allocation and deallocation of the map in user space.
823 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
824 The amount of flops remaining to do is a vector, not a scalar.
825 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
826 name implies on both sequential and parallel tasks.
827 - Both changes fix GitHub's #223 using PR #237 as a basis.
828 Thanks Michael Mercier.
829 - Most examples were converted to S4U and hidden elsewhere at tests.
830 Rationale: we still want MSG to work; we want newcomers to use S4U.
833 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
834 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
837 - Switch to the faster dlopen privatization mechanism by default
838 - Documentation improvments
841 - Live migration is getting moved to a plugin. Dirty page tracking is
842 the first part of this plugin. This imply that VM migration is now
843 only possible if one this function is called:
844 - C/MSG: MSG_vm_live_migration_plugin_init()
845 - C/C++: sg_vm_live_migration_plugin_init()
846 - Java: Msg.liveMigrationInit()
847 For C and C++, "simgrid/plugins/live_migration.h" also has to be
851 - Define class simgrid::xbt::Path to manage file names.
852 - Removed unused functions:
853 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
854 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
855 xbt_graph_export_graphviz()
856 - xbt/str.h: xbt_str_join()
857 - Remove unused datatypes:
858 - xbt/heap.h: use std::priority_queue or boost::heap instead
859 - xbt/swag.h: use boost::intrusive::list instead
862 - New link_energy plugin for the consumption of the links.
863 - All of the operations on files and storage contents have been
864 packaged into a plugin (src/plugins/file_system). The current
865 public interface can be found in
866 include/simgrid/plugins/file_system.h
867 To use these functions you now have to initialize the plugin by
868 calling MSG_storage_file_system_init() just after calling
869 MSG_init() or sg_storage_file_system_init() just after creating
873 - Remove the undocumented/untested tag <include>
876 - Remove viva specific tracing as the tool is no longer maintained
879 - #248: Kill of finished processes leads to segfault
880 - #240: xbt_cond_wait_timeout should gracefully return for C apps
881 - #239: please implement signal s4u::onDeadlock()
882 - #230: segfaults when exit() without run()
883 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
884 - #223: MSG_task_get_flops_amount() not working with parallel tasks
885 - #222: Actor::kill() doesn't really kill and segfaults
886 - #221: odd LMM warning when killing an actor
887 - #120: Memory leak when the processes are forcefully killed
889 ----------------------------------------------------------------------------
891 SimGrid (3.17) Released October 8 2017
893 The Drained Leaks release: (almost) no known leaks despite the tests.
895 Even very long simulations will preserve your memory: our testsuite
896 stresses SimGrid for over 45mn on fast machines for a coverage of
897 over 80%, and there is only one single known leak, of about 4kb.
901 - Comm.detach(): start and forget about asynchronous emission. A cleanup
902 handler may be given to free resources if the comm cannot be completed.
903 - this_actor::send(mailbox) is now mailbox->put()
904 - New: simgrid::s4u::Comm::wait_all()
905 - New: Host.execute() for remote executions.
908 - Unused option network/sender-gap is removed.
911 - Deprecate MSG_task_isend_with_matching(): This unused feature
912 really complicates our internals. Will be removed in v3.20.
915 - Improved context termination. It is now possible to free resources, even
916 when a process is forcibly killed.
917 - ContextBoost: add support for Boost versions above 1.61.
920 - Bring back run-time option --cfg=exception/cutpath to remove exception
922 - Removed unused functions:
923 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
924 xbt_str_rtrim(), xbt_str_trim().
925 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
928 - Removed header files obsolete since SimGrid 3.12:
929 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
931 - #3: SD_exit should be made optional
932 - #120: Memory leak when the processes are forcefully killed
933 - #159: Threading test regression in Actor refcounting
934 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
935 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
936 - #186: Actor::killAll() segfaults if some process is blocked on wait()
937 - #191: VM migration and pstate
938 - #192: Updating the energy consumptions of all hosts crashes with VMs
939 - #195: All actors have PID=0 in the logs
940 - #204: Sometimes segfault with thread contexts and mmap privatization
941 - #222: Actor::kill() doesn't really kill and segfaults
942 - #225: Actor::kill() doesn't really kill when victims are doing a join()
944 ----------------------------------------------------------------------------
946 SimGrid (3.16) Released June 22. 2017.
948 The Blooming Spring Release: developments are budding.
951 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
952 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
953 - Java JAR file should be finally fully working on Mac OSX too.
955 XML platforms: Switch to platform v4.1 format.
956 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
957 - <zone> can be used as a synonym for the now deprecated <as>
958 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
959 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
960 - <actor> can be used as a synonym for the now deprecated <process>
961 - state_file and avail_file periodicity is now easier to express
962 (check the documentation) the old behavior should still work.
963 * Storage: not backward compatible, but it seems that nobody ever used it
964 - Remove attribute 'content_type' of <storage_type>: was never used
965 - Make attribute 'model' of <storage_type> optional: for future usage
966 - Remove Bconnection model property: was never (in)validated. Replaced by
967 the maximum of the read and write bandwidth as a resource constraint.
970 - New and Backwards Compatibility break:
971 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
972 When one wants to get the list of tasks whose states have changed during a
973 simulation round, s/he has to allocate and free a dynar and use it as
974 argument to this function. The former SD_simulate (double how_long)
978 - Allow multicore VMs, with the correct sharing (unless you overcommit)
979 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
982 - New function to update the consumption of all hosts at once.
983 - Fix the model for multi-core hosts, linear in the amount of busy
984 cores with an abnormality for pIdle (see paper).
985 BUG: Still not satisfactory for ptask on mono-cores.
988 - The netzone are now available from the MSG API.
989 The old names still work, but are now deprecated.
992 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
993 - New option: smpi/keep-temps to not cleanup temp files
994 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
995 mallocs mode is used, allows to change the size of the fake file used
996 (default 1MB), to potentially limit the number of mappings for large runs.
997 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
998 - Fortran ifort and flang compilers support
999 - New RMA calls supported (experimental):
1000 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
1001 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
1002 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
1003 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
1004 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
1005 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1008 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1009 - Behind the scene, ActivityImpl are now automatically refcounted.
1010 Removing this nasty bottleneck will greatly simplify our internals.
1013 - Replay: New function xbt_replay_action_get():
1014 Retrieve the function previously associated to an event type.
1015 - DROPPED FUNCTION: xbt_str_varsubst()
1016 - DROPPED MODULE: strbuff. We don't need it anymore.
1017 - DROPPED MODULE: matrix. We don't need it anymore.
1018 - DROPPED MODULE: lib. We don't need it anymore.
1020 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1022 SimGrid (3.15) Released March 22. 2017
1024 The Spring Release: continuous integration servers become green
1025 We fixed even the transient bugs on all target architectures:
1026 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1027 Windows, FreeBSD, NetBSD.
1029 - Rename NetCards to NetPoints.
1030 This was intended to help NS3 users, but that's not a netcard.
1031 That's a point in the routing algorithm, let's avoid wrong simplifications.
1034 - New: MSG_process_yield(). Stop and yield to other processes.
1035 - New: MSG_process_daemon(). Daemon processes are automatically killed
1036 when the last non-daemon process terminates
1037 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1038 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1039 to make room for the upcoming network energy plugin.
1040 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1043 - Ensure that an actor can kill itself with Process::exit()
1044 - Kill the obscure NativeException. Nobody want to survive the issues
1045 it denotes, so use JniException that is a RuntimeException (not to
1046 be caught explicitly).
1047 - Partial bug fix in initialization. SimGrid flags on command line were
1048 consumed at C level but stayed in the original Java String[] args.
1049 This could mess users' args[i] if SG flags were not put at the end of
1051 The SimGrid flags are now removed from the Java arguments. However,
1052 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1053 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1054 It might be if you have --log or --cfg flags in the command line.
1055 - Fix numerous memleaks all around the place. In particular, around VMs.
1059 - s4u::onPlatformCreated: right before the simulation starts
1060 - s4u::onSimulationEnd: right after the main simulation loop
1061 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1062 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1063 event from the availability_file changes the avail speed.
1064 - Links are now usable from s4u
1065 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1066 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1067 - New examples: The conversion of MSG examples to S4U has begun
1068 - Actors: create, kill, migration, and suspend
1069 - Applications: master-worker and token-ring
1070 - Action replay: communications and storage
1071 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1074 - Backwards Compatibility breaks
1075 - The SD_route_get_list and SD_route_get_size functions have been removed.
1076 They are replaced by the unique
1077 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1078 The route from src to dst is built in the links dynar whose size can
1079 be obtained with xbt_dynar_length.
1080 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1081 sg_host_route_bandwidth and sg_host_route_latency.
1082 Macros ensure the backwards compatibility, but you should fix your code
1085 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1086 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1087 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1088 We don't support 100% of it, but it should be enough. Contact us if not.
1089 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1090 naive and probably inaccurate)
1091 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1092 mapped onto a unique small file using some system magic.
1093 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1094 - Fortran: cleanups, fixes, support of user-added operations
1095 - RMA: MPI_Accumulate are applied in correct order.
1096 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1097 - Better support for MPI_IN_PLACE
1098 - Support for MPI_Win attrs and keyvals.
1099 - Support MPI_Comm_group_create, MPI_Type_size_x
1100 - MPI ops should be properly applied to all allowed types
1101 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1102 algorithm by a non failing one from mpich.
1105 - Kill the fifo data container: we don't use it anymore.
1108 - A <cluster> can now be created with different speed values to represent pstates.
1109 see examples/platforms/energy_cluster.xml for an example.
1111 SimGrid (3.14.159) Released December 28. 2016
1113 The previous release was only a crude approximation of a Pi release;
1114 this one should be a bit better.
1116 - Revert a last minute change that broke on Mac OSX.
1117 - Fix the changelog and NEWS files.
1118 - Improve the documentation, in particular of the routing module.
1119 - Integrate some patches flying around in Debian.
1121 SimGrid (3.14) Released December 24. 2016
1124 * The whole documentation was reworked and reorganized. There is
1125 still much room for improvement here, but we're on it.
1128 * We now need python3 for our testing infrastructure.
1129 * Model checking needs libevent
1132 * Now works on FreeBSD
1135 * Privatization now works on FreeBSD
1137 * Privatization is now activated by default in smpirun
1138 You can turn it off with -no-privatize if something goes wrong.
1140 * Call-location tracing for SMPI.
1141 You can add the exact location (filename / linenumber) of an MPI call to
1142 your trace files and slow down or speed up the simulation between two
1143 consecutive calls by using an adjustment file (see the documentation).
1145 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1146 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1147 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1148 values were ignored in some cases. The timings of these functions can now
1149 be significantly different.
1151 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1152 smpi/running-power is renamed to smpi/host-speed
1154 * smpi/grow-injected-times option to enable or disable multiplication of the
1155 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1156 make simulation less precise (but also much faster).
1158 * smpirun script should be (much) faster for large deployments.
1160 * SMPI tracing: fixed issue with poor matching of send/receives.
1162 * Replay: Fix broken waitall
1164 New functions and features
1165 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1167 Dropped / renamed functions and features
1168 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1169 - MSG_mailbox_is_empty() -> Mailbox::empty()
1170 - MSG_mailbox_front() -> Mailbox::front()
1171 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1172 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1173 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1174 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1175 * Don't pass the free_f parameter to property related functions:
1176 - MSG_host_set_property_value()
1177 - MSG_as_router_set_property_value()
1178 - MSG_storage_set_property_value()
1179 * VM properties. Since msg_vm_t are msg_host_t, just use
1180 MSG_host_get_property_value() and friends
1181 * VM I/O related things:
1182 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1183 - Unimplemented save/restore methods
1184 * MSG_as_router_get_property_value() was redundant with
1185 MSG_environment_as_get_property_value().
1186 - Removed MSG_as_router_*propert*() functions
1187 - Added MSG_environment_as_set_property_value() for consistency
1188 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1189 Well, they are still there for now, but deprecated with a warning.
1190 Please switch to xbt_dict_new_homogeneous() before this is removed
1192 * Task affinity. Its intended behavior (that was very badly tested
1193 and probably not really working) was deceiving what most users
1194 would have hoped here.
1195 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1196 * Option network/coordinates is now useless and should be dropped.
1199 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1202 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1203 * Added Dragonfly topology support
1206 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1207 EiBps, EBps, Eibps, Ebps for bandwidth.
1208 They may become useful to some lucky ones.
1211 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1212 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1213 * ex/async_waitAll: new example, on asynchronous communications
1216 * Memory usage should be decreased for simulations with a large number
1217 of processes. This also helps for SMPI.
1219 SimGrid (3.13) stable; urgency=low
1221 The Half Release, a.k.a. the Zealous Easter Trim.
1223 - We removed half of the lines, that were mostly experimental cruft.
1224 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1225 (not counting blanks and comments -- according to openhub.net)
1226 - The internals are now compiled in C++ (and will soon be clean C++)
1227 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1228 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1230 * Backwards Compatibility breaks
1231 - Removed Lua simulation bindings (switch to C or Java for that).
1232 Lua can still be used to describe platforms
1233 - Removed Java kernel plug-ins.
1234 Will be reintroduced after the ongoing major internals reorg.
1236 - the following functions were removed.
1237 They were too specific and should be reimplemented in a generic
1238 way, with filter function.
1239 - MSG_task_listen_from_host
1240 - MSG_mailbox_get_count_host_waiting_tasks
1241 - MSG_mailbox_put_with_timeout was removed.
1242 Please use MSG_task_send_with_timeout instead.
1244 - the SD_application_reinit function was removed. It has been a noop for a while.
1245 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1246 replaced by a more flexible mechanism.
1247 - The following functions thus do not exist anymore
1248 - SD_workstation_get_access_mode
1249 - SD_workstation_set_access_mode
1250 - SD_workstation_get_current_task
1251 - Basic estimation functions have been removed but can easily be replaced
1252 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1253 - SD_workstation_get_computation_time => amount / sg_host_speed()
1255 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1256 Use VM.getSpeed()*load/100 for the legacy behavior.
1258 - option enable_tracing was removed. It was not doing anything for a while.
1259 - In the ModelChecker:
1260 - the model-checker now ptraces the model-checked process which means
1261 you cannot use a debugger on the latter anymore (we might make this
1262 optional in the feature);
1263 - removed soft-dirty page tracking;
1264 - remove model-checked side snapshot management,
1265 MC_snapshot() and MC_compare_snapshot();
1266 - keep the MC_cut() function as a stub (it was not really working
1267 in the previous release).
1270 * All options are consistently in kebab-case. Old names are kept as alias.
1273 * Switch to platform v4 format.
1274 - Rename from 'power' to 'speed' the attributes describing the amount of
1275 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1276 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1277 - In <host> and <link>, attributes availability and state are gone.
1278 It was redundant with state and availability traces, and with peak values.
1279 - In <cluster>, attributes availability_file and state_file are gone.
1280 It was too complex and unused.
1281 - Kill <gpu>. Was not doing anything.
1282 - The DOCTYPE points to the right URL:
1283 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1284 (the file at this address now documents the changelog since its v1)
1285 - A warning is emitted for unit-less values (they are still accepted).
1286 - speed. Default: 'f' or 'flops'. Also defined:
1287 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1288 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1289 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1290 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1291 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1292 - latency. Default: 's' second. Also defined:
1293 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1295 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1296 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1299 * s4u::Host is now the preferred public interface to the Host features.
1300 sg_host_* functions are C bindings to the exact same behavior
1301 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1304 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1305 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1306 They were deprecated since a few years, and probably did not even compile anymore.
1309 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1310 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1311 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1312 encouraged to update their codes. A list of the modified functions can be found at the end of
1313 include/simgrid/simdag.h
1316 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1317 the lambda or closure passed as a parameter will run in kernel mode.
1318 All the callback functions should be rewritten to that interface at some point.
1321 * Reorganizing and cleaning the internals all around the place.
1324 * Remove old default barrier/bcast buggy algorithms (see #18407)
1325 * Various bug fixes to handle more codes
1326 * Remove the need for the --foreground option of smpirun (it is still
1327 accepted for backward compatibility).
1330 * Kill the setset data container: MC don't use it anymore.
1331 * Kill the queue data container: it made more sense with GRAS.
1332 * Kill the xbt_peer_t data type: it's useless without GRAS.
1333 * Kill rm_cb feature of config sets: it was never useful.
1334 * Kill graphxml parsing feature. It was not used.
1335 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1337 - xbt_dynar_sort_strings(), when the content is char*
1338 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1339 They throw exceptions on invalid input;
1340 * C++ support for declaring CLI flags (simgrid::config::Flag);
1341 * class for abstracting different signal backends (simgrid::xbt::signal).
1342 with no external dependencies (we need very simple signals).
1345 * refactoring and cleanup of the code;
1346 * ongoing process to cleanly separate the model-checking algorithms
1347 from the code model-checking support.
1349 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1351 SimGrid (3.12) stable; urgency=low
1353 The Facelift Release.
1356 * Require g++ v4.7 at least to not speak prehistorical C++.
1357 * Require Boost 1.48 (for signal2 component).
1358 * Java must be version 7 at least when activated.
1359 * Builds on Windows again (including Java bindings).
1360 * Tracing is now always enabled (no way to turn it out)
1361 * Remove GTNetS. It was not working anyway.
1362 * Various cleanups in the cmake scripts.
1363 * Move headers around to sort them out on installed systems:
1364 - instr/instr.h -> simgrid/instr.h
1365 - instr/jedule/* -> simgrid/jedule
1366 - simdag/datatypes.h was removed
1367 - simdag/simdag.h -> simgrid/simdag.h
1368 - msg/datatypes.h was removed
1369 - msg/msg.h -> simgrid/msg.h
1372 * Interface improvement:
1373 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1374 - Sanitize the interface in MSG_task_ module:
1375 - Merge two functions that were close enough but misleading:
1376 set_compute_duration(t) -> set_flops_amount(t)
1377 get_remaining_computation(t) -> get_flops_amount(t)
1378 - set_data_size(t) -> set_bytes_amount(t)
1379 get_data_size(t) -> get_bytes_amount(t)
1380 - Massive cleanups in the functions related to the energy
1381 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1382 - New: MSG_host_get_pstate()
1384 - msg/energy/onoff: switching hosts on and off
1386 * Interface improvement:
1387 - Rename Host.isAvail() to Host.isOn()
1388 - Rename Process.currentProcess() to Process.getCurrentProcess()
1389 - Rename Task.setDataSize() to Task.setBytesAmount()
1390 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1392 - #18874: Actually allows the GC to reclaim tasks
1396 - SIMIX_process_throw: raises an exception in a remote process
1397 * Refactoring: Separate sub-modules
1398 - libsmx: the public interface, as libc in a real system
1399 - popping: the strange dance that converts a user request into a kernel handling
1400 - smx_context_*: the virtualization mechanisms that embed the user code
1401 - smx_*: the handling of each simcalls
1402 * Interface cleanup:
1403 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1404 * Rename smx_action_t into smx_synchro_t, making explicit that these
1405 things are used to synchronize processes with their environment.
1406 For example, a communication is a sort of synchronization involving
1407 the communicating processes (that may block until the exchange) and
1408 the platform. The same can be said from computations, etc.
1410 - #18888: segfault when a process ends before its kill_time
1414 - 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.
1415 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1416 - MPI_Type_set_name, MPI_Type_get_name
1417 - MPI_*_c2f and MPI_*_f2c functions
1418 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1419 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1420 - Activate a lot of new tests from the mpich 3 testsuite
1422 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1423 - InfiniBand network model added: Based on the works of Jerome Vienne
1424 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1425 - When smpi/display_timing is set, also display global simulation time and application times
1426 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1427 * Collective communications
1428 - 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
1429 - 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/).
1430 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1432 - Replay now uses algorithms from wanted collective selector
1433 - Replay can be used with SMP-aware algorithms
1434 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1435 - 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.
1437 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1438 - Correctly use loopback on fat-tree clusters
1439 - Asynchronous small messages shouldn't trigger deadlocks anymore
1440 * Energy/DVFS cleanup and improvement
1441 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1442 - new: smpi_get_host_pstate()
1445 - "Full" network optimization flag was broken since Surf++
1446 - Better handling of precision flags in maxmin
1447 - Fix bug causing sometimes "Impossible" errors
1448 - Properly pass cluster properties to included hosts
1449 * Improvement of the Energy plugin.
1450 - Always update the consumption before returning that value
1451 - New property: watt_off to denote the dissipation when the host is off
1452 - New functions getWattMinAt and getWattMaxAt to retrieve the
1453 dissipation of pstates that we are not currently at.
1454 * Java: class NetworkLink renamed to Link
1455 * New function: simcall_process_get_kill_time()
1456 * Massive rename s/workstation/host/
1457 - That's intrusive, but that's good for the project consistency. Sorry.
1458 - Change config option "workstation/model" into "host/model"
1461 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1462 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1463 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1464 - xbt_dirname and xbt_basename for non-POSIX systems
1466 * The model checker now runs as a separate process.
1467 * The model checker runs must now be launched with the new simgrid-mc program.
1468 * Record/Replay: the MC can display a textual representation of a path in the
1469 execution graph. It can then be replayed outside of the model checker.
1471 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1473 SimGrid (3.11) stable; urgency=low
1478 * Normalizing pointers addresses tool for better diff between logs
1481 * Add cloud examples using new VMs
1482 - examples/msg/cloud/two_tasks_vm.tesh
1483 - examples/msg/cloud/simple_vm.tesh
1484 - examples/java/cloud/cloud.tesh
1485 - examples/java/cloud/migration/migration.tesh
1486 * Add java surf examples:
1487 - examples/java/surfPlugin/surf_plugin.tesh
1488 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1489 - examples/java/surfCpuModel/surf_cpu_model.tesh
1490 * Add SMPI+MSG example:
1491 - examples/smpi/smpi_msg_masterslave/
1496 - msg task destroy cancel test
1497 - msg_host on/off test
1498 * Move all tests in testsuite to teshsuite (adding tesh files)
1499 * Restructure teshsuites
1500 - one folder for each kind of test
1501 * Restructure AddTests.cmake
1503 - structure the order of tests (with sections)
1506 * Add virtual machine
1507 - creation of a VM on a PM
1508 - migration of a VM from a PM to another PM
1510 - MSG_process_join(msg_process_t process, double timeout)
1511 - msg_bar_t MSG_barrier_init(unsigned int count)
1512 - int MSG_barrier_wait(msg_bar_t barrier)
1513 - void MSG_barrier_destroy(msg_bar_t barrier)
1514 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1515 * New option "msg/debug_multiple_use" to help debugging when a task is used
1518 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1520 - Increase code coverage in test suites
1524 * Protect context stacks against stack overflow. The number of protected memory
1525 pages allocated on the top of each stack (1 by default) can be configured
1526 with the new command line option --cfg=contexts/guard_size:<value>.
1527 * Simcalls are now generated by a python script that
1528 - generates files included by SimGrid
1529 - checks that all the functions exist, and proposes prototypes
1531 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1532 host_set_data, host_get_data
1534 - simcall_process_join(smx_process_t process, double timeout)
1535 * Fix bug where sleeping processing could not be suspended.
1538 * Translate surf models from C to C++
1539 - Generic classes for all models: Model, Resource, Action
1540 - A generic interface for each kind of model (CPU, Network, Storage
1541 Workstation, WorkstationVM)
1543 * Translate surf routings from C to C++
1544 * Add callbacks using sigc++ or boost::signals2
1545 - Add callback functions for resource creation/destruction
1546 - Add callback functions for action state change
1547 - Handle Energy as a plugin
1548 * Replace swag by boost::intrusive
1549 * Add new routing models for clusters. For documentation, see
1550 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1551 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1552 parameters for cluster tag
1553 - Fat trees, with topology="FAT_TREE" and
1554 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1555 - see examples/platforms/torus_cluster.xml and
1556 examples/platforms/fat_tree_cluster.xml
1557 * More documentation
1560 * Hostfiles support host:nb_processes construct to deploy several processes on
1562 * Collective communication algorithms should not crash if used with
1563 improper number of nodes and report the error.
1564 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1565 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1566 MPI_Dims_create, MPI_Cart_sub are supported.
1567 * New interface to use SMPI programmatically (still depends on MSG for
1568 some parts, see examples/smpi/smpi_msg_masterslave):
1569 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1573 * Global variables privatization in MPI executables is now performed at runtime
1574 with the option smpi/privatize_global_variables (default:no).
1575 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1576 dynamic libraries loaded by the application are not privatized (static
1577 linking with these libraries is advised in this case)
1580 * Options defined in XML work correctly now.
1583 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1584 copied into simgrid.jar or not (ON by default). Use this option if you want
1585 to reduce the size of the installed simgrid.jar, *and* the native libraries
1586 are kept installed elsewhere.
1587 * Surf binding with SWIG (code generated in maintainer mode only):
1588 - plugin to handle callbacks
1589 - CPU model only for the moment
1592 * Supernovae build mode is definitively removed. It was used to improve
1593 inlining and inter-module optimizations. It is nowadays superseded by
1594 link-time optimizations commonly available in compilers.
1595 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1596 * Add boost dependency for surf++
1597 * Add new macro for tests
1598 - ADD_TESH(name <tesh_args>)
1599 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1603 - xbt_bar_t XBT_barrier_init(unsigned int count)
1604 - int XBT_barrier_wait(xbt_bar_t barrier)
1605 - void XBT_barrier_destroy(xbt_bar_t barrier)
1606 * Make the xbt_os_time module public
1608 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1610 SimGrid (3.10) stable; urgency=low
1612 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1615 * Reintegrate Java to the main archive as desynchronizing these
1616 package is not acceptable anymore (Java is now considered stable)
1617 * Add explicit synchronization facilities through semaphores
1618 * Bug fix: Task.setDataSize() only changed the C world, not the value
1619 cached in the Java world
1622 * Dramatically change the way files are handled. API and internals changed, but
1623 this part of MSG was not considered as production grade either.
1624 * Add explicit synchronization facilities through semaphores
1625 * Add a new function MSG_host_get_process_list()
1626 * Preliminary DVFS support (see examples/msg/energy/ for details)
1629 * SMPI is now included directly in the libsimgrid as the windows
1630 linker doesn't force us on splitting it anymore.
1631 * Improvements of the SMPI replay tool:
1632 - Most of the collective communications are now rooted in the same process as
1633 in the original application.
1634 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1635 used until now). Multiple data types can now be used in a trace.
1636 - The replay tool now supports traces produce either by TAU or a modified
1638 - Bug Fix: the compute part of the reduce action is now taken into account.
1639 - Gatherv collective is now supported
1640 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1641 of smpirun outputs time independent traces for the current run. One file
1642 is created per process. If too many processes are simulated, this behavior
1643 can be changed to one file for all processes by using the
1644 tracing/smpi/format/ti_one_file flag
1645 * smpirun generates the host file if needed (with given host count and platform)
1646 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1647 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1648 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1649 --cfg=smpi/coll_selector:(mpich/ompi)
1650 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1652 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1653 happening inside a collective SMPI call.
1654 * Fix the behavior of complex data types handling.
1655 * Make MPI_Wtime another synchronization point to take computations into
1657 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1658 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1659 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1660 (no privatization of global variables yet)
1661 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1664 * Verification of liveness properties is now available for SMPI applications
1665 (in addition to MSG applications)
1666 * Bugged examples using SMPI in examples/smpi/mc/
1667 * Add --cfg=model-check/visited option. Allows the verification of infinite
1668 programs. Detection of loops in the execution thanks to the system state
1669 comparison and reduction of the state space to explore. Can be combined with
1670 DPOR for safety properties.
1673 * Allow to change SimGrid configuration (see --help) within the code
1674 thanks to SD_config() as it can be done in MSG.
1675 * Add a new function SD_task_set_amount() upon user request.
1678 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1679 * Remove rule based routing (no more PCRE dependency)
1680 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1681 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1682 * Add a loopback_bw and loopback_lat options to cluster tag.
1683 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1684 routing. Now peers are encapsulated in an AS and have their own private
1685 router but this is transparent.
1688 * Our own implementation of getline is renamed xbt_getline, and gets
1689 used even if the OS provide a getline(). This should reduce the
1690 configuration complexity by using the same code on all platforms.
1691 * New type: xbt_cfg_elm_boolean.
1692 * Allow to use yes/no for boolean configuration options in the command line.
1693 * Allow to disable SimGrid cleanups at exit from command line option.
1694 There are situations where one may want a simulation to end with an exit.
1695 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1696 annoying when scripting around the simulator. Adding a
1697 --cfg=clean_atexit:no allows to circumvent this issue.
1700 * Lots of memory leaks were corrected in this release.
1701 There is no known memory leaks anymore, in all of our 600+ tests.
1702 * New command line option --version, to get SimGrid version information.
1703 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1705 * Supernovae builds are deprecated, and expected to be removed in the next
1708 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1710 SimGrid (3.9) stable; urgency=low
1712 The Grasgory release: GRAS is really dead now.
1714 * Complete overhaul of the internal host structures scheme.
1717 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1718 was considered as experimental and badly maintained since then.
1719 * Keeping it was thus a trap to our potential users, that could take
1720 it instead of MSG or SMPI by mistake despite is pity state.
1721 * GRAS seems to have very few users (if any), and no one volunteered
1722 to maintain it further. It also induces a lot of XBT code (for
1723 portability sake), that must be maintained too.
1724 * For all these reasons, we killed GRAS. If someone wants to revive it
1725 in the future, don't cry, our git history still remembers of GRAS.
1728 * Major overhaul. Merge our documentation again as time proved that
1729 splitting it was really not helping our users.
1730 * Further improve the developer documentation to help newcomers
1731 hacking on SimGrid itself. The user documentation (and in
1732 particular, the beginner documentation) is still in a sorry state.
1735 * Now works on Windows too!
1736 * Much more extensive test suite, from MPICH
1739 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1740 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1741 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1742 * Bug fix: let task be scheduled when the last dependency to be solved is
1743 a control dependency.
1744 * Remove SD_load_environment_script function.
1745 Use the C sg_platf function if you want to declare a platform
1749 * New function: MSG_process_get_number()
1750 * Old function documented: MSG_config()
1751 * Remove MSG_load_platform_script function
1752 Use the C sg_platf function if you want to declare a platform
1756 * Change the default value of the TCP_gamma constant (maximal size of TCP
1757 congestion window) to a more realistic 4MiB value. If you notice changes in
1758 your simulation results, you can fall back to the previous 20k tiny window
1759 by adding --cfg=network/TCP_gamma:20000 on command line.
1760 * (Hopefully) fix a bug wrt periodic availability and state traces
1761 * Bug fix: use default values at start when first event in availability/state
1762 trace is not at time 0.
1765 * remove the "new_" part of function name sg_platf_new_trace_connect
1766 (resulting in sg_platf_trace_connect), since it does not create
1770 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1771 use case where it's really mandatory, and maintaining it was a pain in
1773 * New: xbt_fifo_search(), search an item with a user-provided
1774 comparison function instead of dumb pointer comparison.
1777 * Fix the lua deployment:
1778 Use `simgrid.init_application()` before deployment instead of
1779 `simgrid.msg_register_application()` after.
1782 * Transfer the tracing files into the corresponding modules.
1784 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1786 SimGrid (3.8.1) stable; urgency=low
1788 The "we are told that some people want to also *install* the simgrid
1791 * Add missing file "tesh.1" to the archive.
1793 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1795 SimGrid (3.8) stable; urgency=low
1797 The Psssshiiiit release: SimGrid jumps into the Cloud.
1800 * Add an experimental interface to manipulate VMs. They are mainly
1801 process groups with very few intrinsic semantic, but they should
1802 allow you to build the semantic you want easily.
1803 * New function: MSG_host_set_property_value()
1804 * New function: MSG_process_on_exit(). To clean memory in all cases.
1805 * Bug fixes that made the host (and link) failures unusable.
1806 * Add a way to auto-restart process when the host in which they are
1807 executing comes back (ON_FAILURE="RESTART" on deployment file,
1808 MSG_process_auto_restart_set).
1809 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1810 please stop using the old ones, they are DEPRECATED.
1812 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1813 Please use MSG_init() instead. (reducing the amount of entry
1814 points in the library helps us).
1815 * Make it impossible to link against the wrong version of the lib
1816 * Deprecate MSG_clean(). No need to call it anymore.
1817 * Function MSG_get_host_number() is not deprecated anymore.
1820 * Split the doc into a user guide and a reference guide.
1821 * Start a developper guide to help people hacking on SimGrid.
1824 * Enable tracing by default. This modules rocks you should use it.
1825 * Remove option custom_flags. Now use environment variables CFLAGS
1827 * Use default cmake things to detect lua instead of home grown ones.
1828 * New option "enable_mallocators" to disable mallocators, for debugging
1829 purpose ("on" by default).
1832 * Bug fixes around the resource failures: don't let the processes
1833 survive the host they are running onto.
1834 * Add an interface to auto-restart processes when the host in which they are
1835 executing comes back.
1836 * Ensures that SIMIX_clean is called automatically. It's not part of
1837 the public interface anymore (bindings should be updated).
1840 * Bug fix for when SD_Simulate is called with a positive value: be careful
1841 when comparing doubles. Sometimes they are different for non significant
1843 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1844 parallel task whose initial work is distributed among host according
1845 to the Amdahl's law. Such tasks are created with a parameter alpha
1846 that corresponds to the non-parallelizable part of the computation.
1847 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1848 between two sets of workstations assuming a 1D block distribution (each
1849 workstation owns a similar share of data) on both sides.
1851 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1852 Data redistribution will be automatically scheduled once parent and child
1853 are both scheduled. The filling of computation_amount and
1854 communication_amount structures is now done seamlessly thanks to the chosen
1856 * New function SD_workstation_dump to display various information
1857 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1858 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1859 bandwidth on the route onto which the task is scheduled and the amount of
1861 To divide the nominal bandwidth by 2, the rate then has to be:
1862 rate = bandwidth/(2*amount)
1863 * Compute tasks that have failed can now be rescheduled and executed again
1864 (from their beginning)
1865 * Increasing source code coverage (src/simdag is now covered at 95.8%
1869 * Re-implement time-independent trace replay using SMPI (at the
1870 smpi_smp_* level) instead of MSG. This should replace
1871 examples/msg/actions/actions.c
1872 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1873 hindexed and structs)
1874 * Implement the exchange of non-contiguous data.
1875 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1876 * Correct behavior of smpi/sender_gap and set its default value to 0
1877 * Add option to asynchronously send small messages to allow better
1878 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1879 specifies the size in bytes under which messages will be asynchronously sent.
1880 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1881 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1884 * New C interface to define a platform: XML is now optional.
1885 For more info, please check include/simgrid/platf.h.
1886 * New interface to define random platforms from the C:
1887 For more info, please check include/simgrid/platf_generator.h and
1888 examples/msg/masterslave/masterslave_platfgen.c
1889 * Export a sg_cmdline dynar containing all the arguments we got from
1893 * Two new tracing options for adding comments to trace file so you
1894 can track your experiments (see --help-tracing for details).
1895 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
1896 * Adding the SimGrid version that generated the trace file as a comment.
1897 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1898 * Fix to avoid key clashes on Paje links
1899 * Other minor fixes related to the Paje specification
1902 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1903 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1904 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1905 * Greatly improve the robustness of mmalloc to user errors (such as
1906 using an area after freeing it, or freeing it twice)
1908 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1910 SimGrid-java (3.8.1) stable; urgency=low
1912 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1913 Warning: all methods are visible, but only some of them are
1914 implemented so far. Check the source (src/jtrace.c)
1915 for further information.
1916 * New module: org.simgrid.msg.File (SimGrid File management functions)
1917 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1918 * Change the meaning of Process.restart: now restart the process from
1919 the beginning, like MSG_process_restart in C.
1920 * Add Process.setAutoRestart: handling of process restart when failed
1922 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1923 you to retrieve the properties of the processes/hosts
1924 * Deprecate Msg.clean(): you can just forget about it now.
1925 * New function Process.getCount(), that only works when compiling
1926 with the not yet released version 3.9 of the C library.
1928 * New context factory based on Coroutines. It mandates a modified JVM
1929 but then, the simulations run about five times faster, and there is
1930 no limit to the amount of processes (beside of the available memory).
1932 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1934 SimGrid (3.7.1) stable; urgency=low
1937 * Restore the prototype of MSG_process_create_with_environment() to
1938 the pre-3.7 situation by removing the kill_time argument.
1939 * Add a MSG_process_set_kill_time() function instead.
1942 * Fix weird behaviors when dealing with parallel tasks.
1945 * Simgrid is now built as a dll.
1946 * Simgrid-java now works on Windows.
1947 * Simgrid-Java is now included into Windows package.
1950 * First pre-build package for MacOSX.
1953 * Fix compilation when using MSG_USE_DEPRECATED.
1954 * Fix some compilation issues on Macs and Windows.
1955 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1956 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1958 We discovered that the Lua console is broken, but we are missing the
1959 manpower to fix it right now. The problem existed in 3.7 too, so we
1960 are not blocking the release for that. Sorry if you depended on this
1961 feature, any help would be really welcome.
1963 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1966 SimGrid-java (3.7.1) stable; urgency=low
1968 The "Java aint got to be bloated and slow" release
1971 * Various internal cleanups and performance improvement
1972 Simulations are expected to run up to twice faster or so
1973 * Make Process.kill(process) an instance method, not a static one
1974 * User processes are not java.lang.Thread subclasses.
1975 This breaks the compatibility (sorry), but previous API was
1976 brain-dead, making it impossible to have non-trivial
1977 initializations in the process constructor.
1978 * Require a full constructor per Process sub-class.
1979 Kinda breaks the compatibility (sorry), but this allows a much more
1980 efficient way to launch the processes at simulation startup.
1981 * Do not embeed our version of semaphores, java 1.5 can be considered
1982 as sufficiently prevalent for us to not dupplicate its features.
1986 * Add examples for almost every part of the API
1987 We spotted and fixed a lot of bugs in the process
1989 * New module: asynchronous communication API
1990 * New function: Process.sleep()
1991 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1992 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1993 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1995 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1997 SimGrid (3.7) stable; urgency=low
1999 The "spring cleanups (before the next Big Project kicks in)" release.
2002 * We can specify the SMPI latency/bandwidth factor with command line
2003 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
2004 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
2005 You can also use the "config tag" from platform file by adding this line
2006 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
2007 (see "example/platforms/tag_config.xml" to use "config tag").
2008 Note that the command line supersedes the platform file configuration.
2009 * Change the correction factors used in LMM model, according to
2010 the latest experiments described in INRIA RR-7821.
2011 Accuracy should be improved this way.
2012 * Use the partial invalidation optimization by default for the
2013 network too. Should produce the exact same results, only faster.
2014 * Major cleanup in surf to merge models and split some optimization
2015 mechanisms from the core of the models. As a result you can now
2016 specify which model to use (e.g., --cfg=network/model:LV08
2017 --cfg=cpu/model:Cas01) and which optimization mode to use
2018 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2019 Incompatible combinations should err at initialization. See
2020 --help-models for the list of all models and optimization modes.
2021 * The CLM03 workstation model was dropped for simplicity because it
2022 used the deprecated CM02 network model. Use default instead.
2023 * Rename the TCP_gamma configuration option to network/TCP_gamma
2024 * Rename the coordinates configuration option to
2025 network/coordinates, and document it
2026 * Use now crosstraffic keyword instead of the terribly misleading
2027 fullduplex keyword. It is activated by default now in the current
2028 default model, use --cfg=network/crosstraffic:0 to turn it off.
2029 * Ongoing refactoring the model parsing to make XML files optional
2030 See include/simgrid/platf.h for details (still to be completed)
2033 * Major overhaul of the documentation. Almost instructive now :/
2034 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2035 functions and friends. This interface was considered as
2036 deprecated since over 2 years, it's time to inform our users that it is.
2037 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2038 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2039 use these (crappy) functions in your code.
2040 These functions will be removed soon. Stop using them now.
2041 * Deprecate MSG_get_host_{table,number}
2042 Implement MSG_hosts_as_dynar() instead.
2043 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2044 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2047 * Stabilize the parallel execution mode of user contexts
2048 * Introduce configuration variables to control parallel execution:
2049 - contexts/synchro: Synchronization mode to use when running
2050 contexts in parallel (either futex, posix or busy_wait)
2051 - contexts/parallel_threshold: Minimal number of user contexts
2052 that must be part of a scheduling round to switch to parallel
2053 execution mode (raw contexts only)
2054 * Fix bugs that prevented to use suspend/resume along with
2055 synchronization structures.
2056 * Fix bugs in process termination that lead to invalid memory access
2057 in very specific conditions.
2060 * Introduce a parallel mode for the models (controlled by surf/nthreads
2061 configuration item). In our tests, running the models in parallel
2062 never lead to any speedups because they are so fast that the gain
2063 of computing each model in parallel does not amortizes the
2064 synchronization costs, even when ultra fast futexes are used.
2065 This is released anyway because YMMV.
2068 * Performance boost by using a swag internally to compute the set of
2069 tasks that are finished and should constitute the return value of
2073 * Enable it by default now that it is considered rather stable.
2076 * Documentation of the tracing functions.
2077 * Performance gains when tracing categorized/uncategorized resource
2078 utilization by avoiding calls to get route when updating resource
2079 variables. LMM constraints are being used instead.
2080 * API changed to set task categories. Use MSG_task_set_category instead
2081 of TRACE_msg_set_task_category, and SD_task_set_category instead
2082 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2083 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2084 * Examples for link user variables added (see at examples/msg/tracing/)
2085 * Deprecated function TRACE_msg_set_process_category completely removed
2086 * Trace header updated according to the latest Paje file format
2087 * Tracing network lazy updates, no longer obligate users to use full updates
2088 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2089 * Experimental: let user code declare/set/push/pop application states for hosts
2090 * API changed to allow the manual creation of graph configuration files
2091 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2094 * Improve the API of Lua MSG bindings, using the Lua spirit.
2095 * Each simulated process now lives in its own Lua world (globals are
2096 automatically duplicated). It helps writing simulators. Will allow
2097 to run Splay programs within SimGrid in the future.
2098 * Add a Chord example in Lua, equivalent to the MSG one.
2101 * Start the implementation of a solution to express temporal
2102 properties, not only local assertions. This is still an
2103 experimental work in progress, stay clear from it to be safe.
2107 - Add new runtime parameters --help-logs and --help-log-categories
2108 to display information about supported logging parameters and
2110 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2111 don't exists anymore.
2112 * Mallocators: allow value NULL for the reset function.
2114 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2115 create homogeneous dictionaries, where all the elements share the
2116 same free function. Non homogeneous dictionaries will be
2117 deprecated in the next release.
2118 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2119 - Multi-level dictionaries are deprecated.
2121 - new function xbt_dynar_search_or_negative() that is useful when
2122 you have less than 2 million elements in your dynar and don't
2123 want of the extra complexity of catching exceptions when the
2124 element is not found.
2126 - Make xbt_os_thread module (for thread portability) public.
2127 Documentation is still to come, sorry.
2129 - Cleanups and simplifications to make it maintainable again.
2130 - Exotic features (such as memalign and valloc) were removed.
2131 - The metadata were extended and improved so that the
2132 model-checker becomes able to explore and inspect the heaps.
2133 - This may induce a performance drop when enable_model-checking is
2134 ON in cmake (even if it's not used in the simulation), but it is
2135 necessary at this point to get MC working.
2137 Turn model-checking OFF if simulation performance matters to you.
2138 Not enabling it at runtime is not enough, disable it in cmake.
2140 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2143 SimGrid (3.6.2) stable; urgency=low
2145 The "Not coding new stuff allows to polish old things" release.
2148 * New bindings to the NS3 packet level simulator (experimental)
2149 * Use the raw (efficient) execution contexts instead of the sysv
2150 (portable) ones when possible.
2151 * libpcre is now mandatory in any cases since not using it led to
2152 severe performance loss and possibly other issues
2153 * Update the XML platforms:
2154 - G5K: include the latest machine in Nancy
2155 - GridPP and LCG: new platforms
2156 * Documentation was partially updated, at least (more to come)
2158 Bug fixes, cosmetics and small improvements
2159 * Free terminated processes before the end of the simulation to avoid
2160 exhausting the memory of users having very dynamic amount of
2162 * Bug fix and cosmetics about canceling non-running tasks
2163 * Bug fix about the dot loader's issues when using libcgraph
2166 * Create an installer for windows with nsis (amd64 and win32)
2167 - Add an hello world project to illustrate simgrid project creation.
2168 - Embed libpcre into the Simgrid installer to avoid
2169 its compilation burden
2170 * The raw execution contexts should work on Apple now
2171 * Port to Windows 64 bits
2172 - Sysv contexts now have an implementation for this arch
2173 - GRAS communication features now support this arch
2174 * Drop support for borland compiler on windows
2175 - this code was not maintained, and we kinda depend on gcc nowadays
2176 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2177 * Fix portability issue on unstable ubuntu: linker became picky on
2180 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2182 SimGrid (3.6.1) stable; urgency=low
2184 The "Oops, we broke Macs too" release
2187 * Fixed contexts detection so that raw ones are used when possible
2188 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2189 leads to a strange error, with user code segfaulting sometimes when
2190 the generated code is not perfectly aligned (which is not
2191 controllable from the user side, depends on the amount of code)
2194 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2196 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2198 SimGrid-java (3.6) unstable; urgency=low
2201 * Split of every thing from simgrid v3.5 into a separate package.
2203 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2205 SimGrid (3.6) stable; urgency=medium
2207 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2210 * Bindings now constitute their own package, separated from the main one.
2211 Rationale: reduce our maintenance nightmare by reducing the module coupling
2212 They will soon be released on their own on gforge.
2214 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2215 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2217 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2218 * It's not quite deprecated for now because we have no replacement,
2219 but it may soon become the case.
2222 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2223 * Fortran: New user-level cache variable to store the rank of the running
2224 process. This improves performance by an order of magnitude.
2225 * C: New coccinelle script to automatically locate and modify global and
2226 local static variables.
2227 * Improved SMPI network model with a sender-side gap to account for multiple
2231 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2232 only say if a communication is finished, no matter whether it succeeded or
2233 failed. You can call MSG_comm_get_status() to know the status of a finished
2235 * New function MSG_task_dsend() to send a task and detach it. When a
2236 communication is detached, you are never notified of its success or failure
2237 and the memory is released automatically once it is finished. This function
2238 is useful when you don't care about the end nor the success of a
2240 * Change the prototypes of action replay. Sorry for inconvenience,
2241 but this is really more efficient this way (and to adapt your code,
2242 you just have to fix the initialization, that shouldn't be too long)
2243 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2244 ever managed to do anything useful with it.
2245 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2246 takes a dynar of communications. It returns immediately and gives the
2247 index of a finished communication (if any).
2248 * New example: a basic implementation of the Chord P2P algorithm.
2251 * New model for multi-core CPUs. You can now use the core attribute to
2252 precise the number of cores of a host. This is a basic model. Every
2253 process running on the host receives at most the power provided in
2254 the DTD (throughput<=power). Total throughput of process cannot exceed
2256 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2257 router linked by an up-link and a down-link (possibly asymmetrical).
2258 This kind of pattern allows to easily build last-mile model style platforms.
2259 Aggregating such patterns in a rule-based AS is thus the technique of
2260 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2261 * New model for Vivaldi routing. We transformed the Vivaldi network model
2262 into a Vivaldi routing model (based on the rule-based model). This allows to
2263 combine Vivaldi based latencies with last-mile platforms.
2266 * Added a check for NaN of IEEE754 infinite in the double entries of
2268 * Introduce a new context factory "raw", highly inspirated from the
2269 ucontext factory, but using manually crafted functions in assembly to
2270 do the work in an efficient manner.
2271 * Allow to change the used context factory at run time, not only at
2272 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2273 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2274 contexts at runtime (only with raw contexts or ucontexts).
2275 * Completely rewrote this module to allow parallel execution of user
2276 processes. Use --cfg=contexts/nthreads:N to execute user processes
2277 with N parallel threads (the default is 1, meaning no parallelism).
2278 * Allow to decide dynamically between sequential and parallel modes.
2279 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2280 processes in parallel only when their number is greater than or equal to P
2282 * Added a check for NaN of IEEE754 infinite in the double entries of
2286 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2287 won't output the state of processes when interrupted with Ctrl-C
2288 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2289 NULL-terminated array. This may solve backward compatibility issues
2290 due to the change to return type of SD_simulate. See also:
2291 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2292 * Add new macros with variable number of arguments.
2293 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2294 - in xbt/asserts.h: xbt_assert
2295 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2296 - in xbt/ex.h: THROWF and RETHROWF.
2297 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2299 * Change xbt_die() to accept a format string with arguments, just like printf.
2300 * New data structure: xbt_lib_t, like a dict but more general and with better
2304 * New configuration options
2305 Options triva/categorized and triva/uncategorized can be used to generate
2306 graph configuration files for Triva visualization tool.
2307 * Configuration option tracing/platform is renamed to tracing/categorized
2308 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2309 * New TRACE_user_link_variable interface:
2310 User provides the name of the link and the tracing variable to attach to it
2311 * the declaration of tracing categories must be done after the environment
2313 * simpler tracing interface, just one way to declare categories
2314 TRACE_category or TRACE_category_with_color, it is up to you
2315 * links in the trace file are again identified by their names
2316 * trace contains the full platform hierarchy exactly as declared using the ASes
2317 * Options tracing/msg/[task|process]:1 groups the process by hosts
2318 for both cases, tasks and processes must have names that are unique during the simulation
2319 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2320 * The experimental option tracing/msg/volume is deprecated
2321 its functionality may be reincorporated if needed
2323 The tracing generates a trace file with unordered timestamped events,
2324 because of the way the core simulator (surf) works. A script available
2325 at the tools directory (fix-paje-trace.sh) can be used to put the events
2326 in order. We have changed the tracing so it can generate ordered timestamped
2327 events in the final trace, but depending on the simulator (and how much time
2328 is simulated) that can lead to a huge memory utilization. It is deactivated
2329 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2331 Build Infrastructure
2332 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2333 - We are trying hard to keep the API stable, but it may happen that
2334 some things change (we're a research project after all, not a
2335 nuclear plant operating system). If such things should happen, you
2336 could rely on that macro to adapt.
2337 - current value: 30600 for 3.06.00, aka 3.6
2338 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2339 a number that can be compared with SIMGRID_VERSION.
2340 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2341 assertions and verbose/debug logging events are disabled at compile time.
2343 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2345 SimGrid (3.5) stable; urgency=medium
2348 * New feature to any SimGrid-based simulator: Model-Checking
2349 Check SIN#1 for more details.
2352 * New Model SMPI (three-interval linear regression for correction factors)
2353 See RR-7426, available at http://hal.inria.fr/inria-00527150
2354 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2355 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2356 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2357 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2358 * See also src/smpi/README
2363 - Tracing API changes: TRACE_start and TRACE_end should not be called
2364 by user-code. They are automatically called by simulators created
2365 with SimDAG, MSG and SMPI if the toolkit is compiled with
2366 tracing_enabled=ON. Categories declaration and utilization remain the
2367 same for MSG and SimDag.
2368 - A function was added to the tracing API to declare categories with
2370 - TRACE_category_with_color (char *category, char *color)
2371 where color must be in the following format
2372 "%f %f %f", red, green, blue
2373 and red, green, blue are float values in the interval [0, 1]
2374 - User can specify NULL as color parameter, or continue calling
2375 TRACE_category (cat)
2376 On that case, the tracing system will define random colors
2377 - The following command-line options are supported:
2378 --cfg=tracing/filename:msg.trace
2379 --cfg=tracing:1 (activate tracing, needed to use others)
2380 --cfg=tracing/platform:1 (categorized resource use)
2381 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2382 --cfg=tracing/msg/task:1 (task creation)
2383 --cfg=tracing/msg/process:1 (process creation, migration)
2384 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2385 --cfg=tracing/smpi:1 (SMPI interface tracing)
2386 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2387 - examples of examples/msg/tracing updated
2389 - DAXLoader and DOTLoader functions can generate tasks with categories
2390 - A new function to attribute a category to SD tasks:
2391 TRACE_sd_set_task_category (SD_task_t task, char *category)
2392 * Tracing the MPI interface implemented by SMPI
2393 - Collective operations are traced with states
2394 - Point-to-Point operations are traced with states/links
2395 - Tracing activated by a parameter "-trace filename" passed
2396 to smpirun during execution (considering that simgrid
2397 is compiled with tracing enabled)
2398 - To run the simulation with gdb, the simulator
2399 accepts --cfg=tracing/smpi:1 to trace SMPI
2400 - tesh files to check if smpi tracing is ok
2401 - See examples/smpi/NAS/DT-trace
2402 * GTNetS tracing re-worked
2403 - adaptation to the tracing system of GTNets to cope
2404 with modifications regarding the fullduplex mode
2405 - new tesh files to check if gtnets tracing is ok
2408 * Asynchronous communications through the functions:
2409 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2410 * New function: MSG_load_platform_script()
2411 to make possible using a lua script instead of XML files to set up platforms
2412 * New function: MSG_set_function
2413 to associate functions to processes, used when bypassing the parser
2414 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2416 Platforms: Add some more examples in examples/platforms
2417 * Grid'5000: see www.grid5000.fr
2418 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2421 * Change the XML format. This is a very important modification. SimGrid
2422 3.5 introduces a new hierarchical format based on the notion of
2423 Autonomous Systems. Compatibility with old format is ensured
2424 through the perl script provided in the install bin directory
2425 bin/simgrid_update_xml.
2426 It is now possible to build platforms with specific routing
2427 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2428 connect several platforms together. We will try to provide soon
2429 set of realistic platforms exploiting these properties (have a
2430 look at examples/platforms/ for the moment).
2431 * Take the opportunity of the XML format change to be a good XML citizen:
2432 rename link:ctn to link_ctn and similar changes (also dealt with by
2434 * Add a new routing scheme (rule-based) using regular expressions. It
2435 enables to have an extremely low memory footprint when the
2436 underlying routing is simple and can be compactly described. You
2437 need to have libpcre4-dev (perl regular expressions) installed if
2438 you want to use this routing scheme.
2439 * Revive the cluster TAG and allow to easily and efficiently (both in
2440 term of memory and speed) connect clusters together. Have a look
2441 at teshsuite/simdag/platforms/ to see how this can be done. With
2442 this tag, you can create clusters with thousands of tasks at no
2443 cost (have a look at examples/platforms/).
2444 Note: clusters are implemented as ASes, so there is no need for an
2445 enclosing AS tag if you have only one cluster in your platform.
2446 * Add new generic functions in the public interface that allows the user
2447 to call SURF 'create_resource' methods from your code (same
2448 functionality as the XML bypass mechanism but with a much lighter
2450 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2451 that uses a piecewise linear approximation to produce better
2452 results when exchanging small messages.
2453 * Add a new parameter to handle correctly full duplex link and account
2454 for interferences between uplink and downlink communications
2455 (activate with --cfg=fullduplex:1).
2458 * Rename the SD_READY (all dependencies are satisfied and task is
2459 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2460 dependencies are satisfied) state.
2461 This prevents a confusion between the notion of "ready to schedule"
2462 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2463 simulated" (SD_RUNNABLE) used by the simulation kernel.
2464 * Change the way a task is considered as ready. Instead of removing
2465 dependencies when a task is done, a counter is decreased. This way,
2466 it is always possible to reach ancestors thanks to the
2467 SD_taks_get_parents function (even after the end of the simulation.)
2468 * Change the return type of SD_Simulate from (SD_task_t*) into
2469 xbt_dynar_t. This function was in handling a dynar internally and
2470 converted it into a NULL terminated array for historical reasons.
2471 * New function SD_dotload(char*) to load a DAG described in dot
2472 format. This loader and the corresponding examples require the
2473 installation of the graphviz library.
2474 * Fix a bug in the management of tasks of size 0 in the surf network
2475 models. This problem was only visible with SIMDAG and you should
2476 thus disregard results produced with earlier versions if you
2477 relied on this feature (some tasks were blocked because of this).
2478 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2479 even though your applications were doing only point-to-point
2480 communications and sequential computations. Now you can really use any
2481 model you want (of course, if you create real parallel tasks, which are
2482 not implemented in most models beside ptaskL07, this will abort).
2483 * Add an example that schedules a DAX on an heterogeneous platform
2484 using a Min-Min strategy.
2485 * New function SD_workstation_get_current_task() that returns the kind
2486 of task currently running on a workstation in the sequential access
2488 * Raise some warnings when unexecuted tasks remains at the end of the
2489 simulation. This is usually caused by cycles in the DAG.
2492 * New function: SIMIX_process_set_function() called by MSG_set_function
2493 * Change the underlying waiting queue in semaphores so that a process
2494 can wait on several of them simultaneously (as in waitany).
2495 * Fix the way to handle tokens in semaphores so that all access patterns
2496 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2497 * kill the dirty pimple SIMIX_message_sizes_output()
2498 Please use (proper) visualization instead
2501 * New data container: setset (set of sets of elements)
2502 * New module: mmalloc (mapped malloc, allowing to have several
2503 independent segments of malloc)
2504 * New function: xbt_dict_cursor_set_data()
2505 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2506 * New function: xbt_dynar_is_empty()
2507 * New function: xbt_fifo_get_last_item()
2508 * Fix xbt_dynar_shrink(): use the right element size.
2509 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2510 uninitialized areas during expand.
2511 * Fix semaphores: previous implementation was severely broken.
2512 * Use library init/fini functions for our initialization.
2513 - you can use logs and other feature as soon as you want in your
2514 code (even before the xbt_init / MSG_init)
2515 - xbt_exit is now a no-op and produce a warning when used.
2518 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2519 benefit from the latest improvement to the simulation kernel.
2520 * Kill measurement sockets for now. If you rely on them, sorry. This
2521 release is not for you. This feature will be reintroduced in the
2522 future, but we cannot delay the release any further.
2523 * New function: gras_msgtype_get_name().
2524 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2525 * Fix (at last) the pmm example: it should not randomly fail anymore.
2527 Build chain: bug fixes and overall polishing
2528 * Cmake is now stable enough. Hence, we killed the autotools.
2529 * Port to windows ( TM :)
2530 * Fix the 'make install' target.
2531 No need to use 'make install-simgrid' anymore
2532 * Introduce a 'make dist' target compiling a *source* archive
2533 'make package' compiles a binary archive
2534 * Compile java files only on need
2535 * Add --cd and --setenv command line options to tesh
2536 * Out of source builds are not fully supported yet, but we are close
2537 * Enable supernovae and optimization flags by default for our users
2540 * Add layer to set up environment directly from lua, without XML.
2541 * The effect of gras_stub_generator can be achieved through
2542 lua too (check examples/gras/console/ping_generator.lua)
2544 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2546 SimGrid (3.4.1) stable; urgency=low
2548 The "Polishing easter eggs is probably a good idea" release.
2549 This is a bug fixes release only.
2553 * Fix a bug preventing the tasks from begin garbage collected.
2556 * Fix a bug occurring when a host involved in a communication fails.
2557 This was not detected properly by the other peer involved in the
2558 communication. Now, it's reported as a network error.
2561 * Warn the user about loop dependencies in data flow of DAX files
2562 * Obey the control-flow dependencies of DAX files
2565 * Add option "enable_smpi" allowing to not compile SMPI.
2566 Probably useful for the (Mac) users experiencing a build error here
2567 * Improve the detection of lua5.1 and ruby1.8
2569 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2571 SimGrid (3.4) stable; urgency=low
2573 The "Easter in Cargese" release. Also known as (major changes):
2575 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2576 ~> bindings were greatly improved
2577 ~> new tracing infrastructure for better visualization introduced
2579 * the "Welcome to configury modernity" release.
2580 ~> we switched from autotools to cmake, and improved our cdash
2583 A more detailled list of changes follow (full detail in svn log).
2585 Java Bindings: Various Cleanups
2586 * (install java-gcj-compat-dev on debian-like to use them)
2587 * Remove put/get: no need to export deprecated interface in Java
2588 Use send/receive instead.
2589 * Cleanup the examples and add a README per directory
2590 * Remove example autoDestination (that's the only way to go now)
2591 * Remove example explicitDestination (was a plain copy of basic)
2592 * Make JniException a runtime exception, so that there is no need to
2593 declare the fact that you may encounter such a beast. I guess that
2594 nobody will ever want to survive such error.
2595 * Create specific errors for each MSG case of failure:
2596 host failure, transfer failure, timeout, task cancelled
2597 * Cleanup the exceptions that may get thrown by each function
2598 * Other internal cleanups in Java bindings. Performance still bad :/
2599 Ruby and Lua Bindings: create them
2600 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2601 * That's new and great, you should try them out.
2602 Same functionalities than Java bindings, only even less polished
2604 * Kill the useless "rate" argument of SD_task_get_execution_time()
2605 Everyone used to provide -1 as a value, it was not used, and the
2606 semantic of a possible use wasn't even clear.
2607 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2608 either as comm costs or compute costs to mean that there is no
2609 such thing for that specific task.
2610 * Add a SD_task_set_name() function
2611 * Fix SD_task_unschedule() on typed tasks
2612 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2613 * In DAX loader, accept useless 'level' attributes to <job> since
2614 LIGO DAGs have them (seem to be to ease graphical representation).
2616 * Add an example masterslave_mailbox.c using send/receive and not
2617 the deprecated put/get interface.
2618 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2619 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2621 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2622 (declare MSG_USE_DEPRECATED to still have the old name)
2623 * Add a MSG_task_set_data() function
2624 * About trace replay (see examples/msg/actions):
2626 - Allow to work with splitted trace files for each process
2627 Give the specific trace file as argument of each process,
2628 and call MSG_action_trace_run(NULL)
2629 You can still have one merged file for all processes.
2630 - Fix implementation of collective operations
2631 * Allow task_execute() on 0-sized tasks (closes #10063)
2633 * This is the first release of SimGrid where SMPI is not considered
2634 beta anymore (even if some corners should still be improved)
2635 * Port over the new SIMIX_network submodule (internal refactoring)
2636 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2637 * Implement more missing elements of the standard:
2639 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2641 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2642 MPI_Scan() MPI_Get_processor_name()
2643 - Added implementation of missing case for Alltoall (warning: it's
2644 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2645 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2646 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2647 SMPI_MPI_Allgatherv()
2648 * Bug fixes include:
2649 - MPI_Waitsome() was broken
2650 - Allow relative includes in smpicc
2651 - Command line cfg argument 'reference_speed' was ignored...
2652 - Some functions did not properly lead to auto-benching of user code
2653 - smpicc passes -O2 by default (just like openmpi one)
2655 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2656 * Bug fixes about timeouts during communications
2657 * add SIMIX_message_sizes_output() as a pimple to write to file the
2658 amount of messages per size. Use gnuplot to get histogram.
2659 Pimple because that's the only user-visible function of simix,
2660 defined directly in xbt.h (irk, sorry)
2662 - Add a SIMIX_sem_get_capacity() function
2663 - Fix interactions with processe resume/suspende
2664 - release_forever() was stupidly broken
2665 - Fix SIMIX_display_process_status() for processes in a semaphore
2666 - Make SIMIX_sem_block_onto() user-visible
2667 * Refactoring context stuff:
2668 - Use pseudo-OOP for better modularity
2669 - reimplement SIMIX_process_kill() without process_schedule() so
2670 that the latter can take as invariant that it is called from
2672 - Merge context_start into context_new for sake of simplicity
2674 * Add a Vivaldi network model, coded live during SUD'10 ;)
2675 * Rename configuration variables to start a hierarchy:
2676 o cpu_model -> cpu/model
2677 o network_model -> network/model
2678 o workstation_model -> workstation/model
2679 * New configuration variables:
2680 o network/bandwidth_factor: correction to bandwidth
2681 o network/latency_factor: correction to latency
2682 o netwotk/weight_S: correction to the weight of competing streams
2683 * Add a long description to the models, that users can see with such
2684 argument on the command line: --cfg=cpu/model:help
2685 * --help-models display the long description of all known models
2687 * config: add the ability to set a default value after registration
2688 Does not override any previously set value (e.g. from cmd line)
2689 * dict: allow to have integer key and data.
2690 When so, you need to use the following functions
2691 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2692 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2693 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2694 In contrary to regular dicts, the key is not malloced before copy.
2695 Mixing scalar and regular elements in the same dict is not tested
2697 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2698 Tracing for Visualization:
2699 * SimGrid is now instrumented in order to generate a trace file for
2700 visualization analysis: to use it, need to compile SimGrid with the
2701 "tracing" option enabled, and instrument the program using SimGrid with
2702 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2703 (among other functions).
2704 * The instrumentation only traces the platform utilization for now
2705 * Documentation to use the tracing functions and how to analyze the
2706 traces with the Triva tool is written.
2707 * More information about: SimGrid FAQ (in the section Tracing Simulations
2710 * We moved to cmake as default build system. Autotools support will
2711 be dropped soon. Check the FAQ for more info about how to use it.
2712 * Greatly improved our cdash/ctest interactions
2713 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2714 * Added memory checking tests with valgrind; lot of memleak fixing.
2715 This may be the first release of simgrid with so few memory issues
2716 * Added code coverage tests.
2717 Our coverage is still improvable, but at least we see it on cdash.
2719 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2721 SimGrid (3.3.4) stable; urgency=low
2723 The "Desktop Grid needs love too" release (also called Xmas release).
2725 Models improvements:
2726 * Major speedup in the maxmin system solving by using lazy evaluation
2727 Instead of solving completely the maxmin system at each iteration,
2728 only invalidate (and recompute) the modified parts.
2729 This new feature is enabled in default models but you can try to
2730 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2731 * Cas01 IMproved as default CPU model
2732 This CPU model is the same Cas01 model, but it uses the
2733 maxmin-selective-update flag and a heap structure to manage
2734 actions on SURF kernel.
2735 It reduces the complexity to find the next action to finish and,
2736 consequently, it's faster than the old Cas01.
2737 This is the new default CPU model (Cas01).
2738 * Rename the old Cas01 model to Cas01_fullupdate
2739 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2740 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2741 * CpuTI (CPU Trace Integration)
2742 A new CPU model whose objective is simulate faster when using
2743 availability trace files.
2744 Instead of using a full featured, over engineered maxmin system for
2745 CPU modeling, this model does the pre-integration of traces files
2746 to calculate the amount of CPU power available, and so, executes
2747 faster than the old CPU models.
2748 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2749 * Use LV08 as default network model since it gives better accuracy
2750 for small messages and shouldn't change things for big ones.
2751 Use --cfg=network_model:CM02 to get the previous behavior.
2754 ******************************************
2755 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2756 ******************************************
2757 * The new CPU model may changes simulations!
2758 The point is that events occurring at the exact same timestamp
2759 are not scheduled in the same order with the old and new
2760 version. This may be enough to completely change the execution
2761 of simulations in some cases.
2762 * The new network model will change simulations!
2763 This new model is more realistic than the previous one, so you
2764 should consider redoing your old experiments with this model.
2765 Sorry for the inconvenience.
2768 * Introduce the supernovae compilation mode
2769 When compiled that way, the whole SimGrid (or almost) is put in a
2770 single compilation unit and compiled in one shoot.
2771 This is to help gcc which has difficulties to inline stuff from one
2773 The speedup seem to be above 15%, althrough more tests are needed on
2774 amd64 to confirm that gain.
2777 * Port of MSG's mailbox on top of SIMIX network
2778 The put/get mechanism was greatly simplified on the way.
2781 * New SIMIX network module. Provides:
2782 - Mailbox: rendez-vous mechanism to find with who you want to speak
2783 - Synchronous send/recv: easier and hopefully faster since the
2784 logic is handled in the maestro process directly now
2785 - Asynchronous send/recv: you dreamt of it? It's here now
2786 Too bad that nobody cared enough to propagate the change to MSG.
2787 * Add semaphores as SIMIX synchronization mechanism.
2790 * new function SD_daxload(char*) to load a DAX file
2791 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2792 * Introduce typed tasks. Specify its kind and cost at creation.
2793 At scheduling, just give where it should be placed, and the cost
2794 for each involved resource is automatically computed.
2795 Existing constructors so far (more to come of course):
2796 - SD_task_create_comm_e2e() for end-to-end communication
2797 - SD_task_create_comp_seq() for sequential computation
2798 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2799 * new function SD_task_dump() for debugging display
2800 * new function SD_task_dotty(task,FILE*) writing to file the info
2801 about the task in dotty format
2802 * SD_task_dependency_exists() can now cope with having one of its
2803 arguments NULL. If so, it tests whether the other argument has any
2805 * Add getters on list of preceding/following tasks:
2806 SD_task_get_parents(task) and SD_task_get_children(task)
2807 * Add getters on amount of workstations and list:
2808 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2809 * Add getter on task kind: SD_task_get_kind(task)
2810 * Update the start_time and finish_time of tasks on completion/failure
2811 * Bugfix: Remove task from state swags when destroyed
2814 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2817 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2818 dynar with one separate thread per value of the dynar.
2819 * Change the prototype of xbt_thread_create(), sorry.
2820 Added a boolean parameter indicating whether we want to join this
2821 thread (used in SG only for now)
2822 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2825 * GTNetS wrappers should now be usable again (and betterly tested too)
2826 * Fix a major regression from 3.2 where the timeout provided to
2827 MSG_task_put_with_timeout() was used as absolute time before which
2828 the comm should be done.
2829 * Start to fix the <cluster> tag.
2830 - Internal links should be good now (beside of the loopback, which
2831 use the private link instead)
2832 - paths to the external world is still rather broken
2833 - the <route:multi> tag is just broken. Actually that's brain-dead.
2834 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2835 to make it less stupid
2836 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2837 * Fix a source-level compatibility glitch from 3.2: after defining
2838 MSG_USE_DEPRECATED, you can use the old name
2839 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2840 * Allow to compile from the SVN with automake 1.11
2841 * Fix some problems when using the "start_time" tag in deployment XMLs.
2842 * Fix #8569: XBT/synchro.h has redundant declarations
2843 * Fix #8563: MSG return values and exceptions
2844 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2845 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2846 Thanks to Robson Peixoto.
2847 * Fix a vicious bug in dictionaries inducing that some elements were
2848 not freed on xbt_dict_free()
2850 Portability report of this version:
2851 * Main portability targets:
2852 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2853 on (amd64/i386/ia64)
2854 - mac leopard on i386
2855 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2856 but nothing critical.
2857 * Other platforms: windows, AIX and others were not tested for this release
2859 Timing report of this version:
2860 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2861 scenario parameters). From 8h to a few seconds in desktop grid settings.
2862 * Supernovae brings about 25% speedup on i386.
2864 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2866 SimGrid (3.3.3) stable; urgency=low
2868 The "Need for Speed" release.
2870 The timings done to validate the 3.3.2 were faulty.
2871 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2873 The problem was a conversion from a manually handled vector to
2874 xbt_dynar_t on the critical path.
2875 xbt_dynar_foreach calls functions, inducing stack management crap.
2877 We inlined these functions and xbt_dynar_foreach is now breath taking.
2878 We also inlined xbt_swag_belong on the way.
2880 Here are some approximate speedup measurements (on master/slaves
2881 simulations lasting between 10s and 20s each):
2882 3.3.1 -> 3.3.2: about same performance
2883 3.3.2 -> 3.3.3: 40% speedup
2884 3.3.1 -> 3.3.3: 40% speedup
2885 3.3.1 with inline patch -> 3.3.3: 30% speedup
2887 Our reading is that the refactoring which occurred in 3.3.2 made us
2888 suffer much more from the xbt_dynar_foreach low performance, but
2889 once we solved this, this refactoring proved to be very performance
2890 effective. From the 40% speedup, somehow, 10% are due to the
2891 inlining and 30% to the refactoring.
2893 That's a pitty that gcc cannot inline functions placed in other files
2894 alone. We have to choose between:
2895 - break the encapsulation (by putting private data structures and
2896 accessors in headers files to help gcc)
2897 - live with low performance
2898 - switch to a decent compiler such as icc (not quite possible).
2900 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2902 SimGrid (3.3.2) stable; urgency=low
2904 The "Simplicity does not preceed complexity, but follows it" release.
2906 The main contributors of this release were (lexical order):
2907 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2910 * Extract the routing logic into its own object.
2911 (was dupplicated in network.c and workstation_LV07.c;
2912 Allows to implement other ways of storing that info)
2913 => kill now useless network_card concept
2914 - Use dynar to represent routes (instead of void** + int*)
2915 - kill link_set (use surf_network_model->resource_set instead)
2916 - Add a command-line option to choose the routing schema to use
2917 - Add three new models:
2918 * Floyd (shortest path computed at initialization)
2919 * Dijikstra (shortest path recomputed all the time)
2920 * Cached Dijikstra (shortest path computed on need)
2921 All these models where contributed by Silas De Munck, and are
2922 described in his ICCS09 paper.
2924 * Simplify model declaration
2925 (less redirections, less function to write when defining a model)
2926 - Factorize stuff between models:
2929 surf_model_resource_set(model)
2930 surf_model_resource_by_name(model, name)
2931 - Unify the types of models in s_surf_model_t (using an union)
2932 - Embeed fields of common_public directly into s_surf_model_t
2933 - Rename model methods:
2934 action_free ~> action_unref
2935 action_change_state ~> action_state_set
2936 action_get_state ~> action_state_get
2937 - Change model methods into functions:
2938 (model)->common_public->action_use ~> surf_action_ref
2940 * Implement a generic resource; use it as ancestor to specific ones
2941 (allows to kill duplicated code in models)
2942 Drawback: timer command don't need no name nor properties;
2943 workstation_CLM03 don't need no properties
2944 (but I guess we can live with those few bytes wasted)
2946 * Improve the action object model
2947 - implement a constructor avoiding dupplicated code about field
2948 initialization in generic_action part.
2950 * Kill the SDP model: it has an external dependency, is deprecated
2951 in flavor of modern lmm models, and didn't compile since a while
2954 * Relocation of the context module from XBT to SIMIX.
2955 (the context were decoupled from the simix processes, duplicating a lot of code)
2956 => a lot of code was factorized
2957 - less overhead is introduced during scheduling
2958 - simpler API for the context factory
2959 - the logic for process creation,destruction and manipulation was simplified
2960 * Simplification of the s_smx_process_t data structure.
2961 => accesing the simix level data associated to a process is faster now,
2962 and the code is a lot more readable.
2965 * Implement some more MPI primitives:
2966 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2967 -implementation: Bcast: flat or 2-ary tree (default),
2968 Barrier: 4-ary tree,
2970 Allreduce: Reduce then Bcast
2971 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2972 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2973 Alltoallv: flat tree, like ompi
2975 * Add support for optimized collectives (Bcast is now binomial by default)
2976 * Port smpirun and smpicc to OS X
2979 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2980 Shout out if you used it.
2983 * Display the list of still queued messages in SG mode when existing
2987 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2988 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2989 * Add xbt_str_from_file(FILE*)
2990 * Add xbt_dict_get_key achieving a linear reverse search
2991 * Remove the context module
2993 Portability report of this version:
2994 * Main portability targets:
2995 - Linux(debian)/x86/context
2996 - Linux(debian)/x86/pthreads
2997 - Linux(debian)/amd64/context
2998 - Linux(debian)/amd64/pthreads
2999 On these, we still have the eratic breakages of gras/pmm and
3000 amok/saturate_sg reported in previous version. We still think
3001 that the tests are the cause of the fault, not the tested code.
3003 - Mac OSX Leopard/x86/context
3004 Still false negative in tesh autotesting.
3005 Smpi still fails, but this time because readlink does not accept -f
3006 Everything seems to work properly beside of that.
3009 - AIX version 5.3 (only tested contexts this time)
3010 Smpi still fails there because mktemp is not installed.
3011 Everything seems to work properly beside of that.
3013 I managed to compile it for the first time, but several breakages.
3014 Won't delay the release for this exotic platform.
3016 * Windows: it's still lagging behind. If you want to help, please
3019 Timing report of this version:
3020 This version seem to be more than 5% faster than 3.3.1 (on linux
3021 64bits with contextes). The gain is less than expected, we are
3022 investigating this for next release.
3024 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3026 SimGrid (3.3.1) stable; urgency=low
3029 * Implement a --cfg-help to show existing configuration variables
3030 * Build chain do not require doxygen in maintainer mode
3033 * fix a bug on struct sizeof computation, which prevented the
3034 exchange of arrays of structs in some conditions
3035 - added a regression test about this in datadesc_usage
3036 * Allow the exchange of 0-long dynamic vectors.
3037 - for that, use -1 as indicator of dynamic size instead of 0
3038 - This implied to change any size from unsigned long to long,
3039 reducing a bit communication abilities, but I guess that with
3040 64bits being quite common, this is more than enough.
3041 - This also induce a protocol change, thus bumping network protocol
3042 version from 0 to 1 (if we have external users, we have to get
3043 clean on that point too ;)
3044 - added two regression tests about this in datadesc_usage
3045 * Be more verbose when propagating local exceptions
3046 This helps debugging.
3047 * Display the status of simulated processes when receiving SIGINT in
3051 * Allow to control the simulation from a trace file.
3052 New functions MSG_action_register() and MSG_action_trace_run()
3053 The first one allows to associate a function execution to each
3054 kind of action while the second one parses a trace file and
3055 triggers the corresponding actions within the system.
3056 For now, only a toy example is provided in examples/msg/actions
3057 * Add an example of process migration in examples/msg/migration
3058 * Fix a bug in task exchange which broke MSG_task_get_sender()
3059 Add a teshsuite regression test for that.
3060 [Bug: if MSG_task_get_sender() is called after sender exit,
3062 * Fix a bug which prevented suspend/resume to work properly
3063 * Display the status of simulated processes when receiving SIGINT
3064 This fixes a regression of v3.3. due to the introduction of SIMIX
3065 * Bug fixing in failure management:
3066 - trace could not start by a failure at time 0
3067 - failure during communications were not working
3070 * Add SIMIX_process_set_name() to change the name of the current
3071 process in the log messages.
3072 * Store smx_hosts in a dict since we only retrieve them by name
3073 * Move the configuration infrastructure to surf
3076 * Move the configuration infrastructure to surf
3079 * Massive internal cleanups:
3080 - Store internal structures on processes instead of hosts (allows
3081 to have more than one process per host, in addition of being more
3083 - Cleanup the initialization/finalization process
3084 - Kill a whole bunch of unneeded synchronization:
3085 processes run in exclusive manner within the simulator
3086 - Move queues from global tables to process data fields
3088 - now accept -platform and -hostfile arguments
3089 - Pass the right rank value to processes according to the hostfile
3090 * Compile the examples by default, and use them as regression tests
3091 * Implement MPI_Wtime()
3092 * Change the reference speed to a command line option
3095 * TCP_gamma can now be specified as command line option using
3096 --cfg=TCP_gamma:10000000.0
3097 * Change the --surf-path cmd line option into --cfg=path:
3100 * Also include strbuff from xbt.h public header
3101 * xbt_ex_display(): do not free the exception after displaying
3102 This allows to do more with the given exception afterward.
3103 Users should call xbt_ex_free() themselves.
3107 Portability report of this version:
3108 * Main portability targets:
3109 - Linux(debian)/x86/context
3110 - Linux(debian)/x86/pthreads
3111 - Linux(debian)/amd64/context
3112 - Linux(debian)/amd64/pthreads
3113 These targets fail about 1/10 of times on gras/pmm, but we believe
3114 that this is because of the test, not because of simgrid.
3115 amok/saturate_sg fails even more rarely, and the test may not be
3118 - Mac OSX Leopard/x86/context
3119 The test suite still spits tons of errors because some obscure
3120 force prevents us from removing the temporary directories
3121 arguing that they still contain some metadata I've never heard of.
3122 Smpi fails because seq is not installed.
3123 Everything seems to work properly beside of that.
3126 - AIX version 5.3 (both contexts and pthread)
3127 Smpi still fails there because mktemp is not installed.
3128 XML inclusions seems rosty on AIX.
3130 * Windows: it's still lagging behind. If you want to help, please
3133 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3135 SimGrid (3.3) stable; urgency=high
3139 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3140 [Malek Cherier & Mt]
3142 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3145 Until now, GRAS were using MSG as an interface to SURF. It was
3146 quite difficult because both interface have several differences
3147 (MSG channels vs GRAS sockets were the most notable point).
3149 This also opens the gate to SMPI (which should occur soon) and speed
3150 up simulations by to 40% (even if it were not the main goal).
3152 **************************************
3153 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3154 **************************************
3155 The point is that events occurring at the exact same timestamp are
3156 not scheduled in the same order with the old and new version. This
3157 may be enough to completely change the execution of simulations in
3158 some cases. Sorry for the inconvenience.
3160 * Cleanup and upgrade the XML format to push further scalability
3161 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3163 * Improve the testing infrastructure with tesh. Now a very large part of
3164 the code is tested not only by being run but also by checking that the
3165 output match an expected output [Mt].
3167 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3168 is really less memory-demanding, which should allow you to use
3169 larger files in SimGrid [AL].
3171 * Inform valgrind about our contextes, so that it becomes usable
3172 with the default (and more effecient) version of SimGrid
3173 [contributed by Sékou Diakite, many thanks]
3176 * Introduce a listener thread in charge of receiving incoming
3177 messages from the network. It allows to overlap communication and
3178 computation but most notably, it removes some stupid deadlocks due
3179 to the fact that so far, a process could not send and receive at
3180 the same time. This made most non trivial communication schema
3182 * Convert the PIDs from long int to int to match the MSG ones (and
3183 linux ones too) [Mt]
3184 * New function: gras_agent_spawn() to launch a new process on
3185 current host. Only working in simulation for now. [Mt]
3186 * New function: gras_os_hostport() returning a constant form (ie,
3187 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3190 * Make the backtrace of exceptions more human readable [Mt]
3191 * New module: xbt/str [Mt]
3192 a ton of string utility functions (split, join, printf to a newly
3193 allocated buffer, trim, etc)
3194 * New module: xbt/hash [Mt]
3195 SHA1 hashing algorithm (more to come if needed)
3196 * New module: xbt/synchro [Mt]
3197 synchronization tools (mutex and conditions) working the same way
3198 in simulation and in real life (mainly useful for GRAS, but not
3200 * New module: xbt/queue [Mt]
3201 classical producer/consumer synchronization scheme
3202 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3203 (using the classical functions will get serialized) [Mt]
3204 * Make dictionary internal table dynamic. No need to specify its size
3205 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3207 * Make sure the log channels are organized as a tree under windows
3208 (because of ANSI C compatibility issue, any channel were child of
3209 root directly) [Mt].
3212 * Cleaned many thing in surf and fixed a few bugs [AL].
3213 * Add a nice command line configuration mechanism to compose models [AL].
3214 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3215 the previous one (KCCFLN05). It relies on something that looks like
3216 a max-min sharing mechanism but cannot be written as such. A new solver
3217 was thus designed [AL].
3218 * Add a new solver to lmm. Based on Lagrange optimization and
3219 gradient-based descent, it enables to efficiently maximise systems s.a
3221 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3223 This solver enables to propose two new network models for TCP Reno and
3224 TCP Vegas based on Low's work. These models still need to be fully
3225 tested though [Pedro Velho].
3228 * Bug fix in SD_simulate. Now the time bound given as argument is
3230 * Use the new parallel task model (ptask_L07) as default.
3231 * Use the SURF command line configuration mechanism.
3232 * 0-size tasks (for synchronization) should now work.
3234 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3236 SimGrid (3.2) stable; urgency=high
3240 We still experience issues on this platform, but we believe that at
3241 least MSG is usable.
3243 GRAS API BREAKAGE (for simplification purpose, sorry):
3244 * the gras_msgtype_by_name is not used anymore. Instead of
3245 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3246 you can write (and must)
3247 gras_msg_send(toserver, "request", &request);
3248 - If you still want to pass a gras_msgtype_t to the function (to cache
3249 the type and avoid the lookup time), use the gras_msg_send_() variant.
3250 - Impacted functions:
3251 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3252 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3253 * The callbacks are now expected to return 0 when everything went well
3254 (just like the main() function)
3256 GRAS new features and improvements:
3257 * New module mechanism where user code can use per process globals [Mt]
3258 This is similar to gras_userdata_*() functions, but for libraries. It
3259 factorize some code developped over and over in the examples and AMOK.
3260 It has still to be documented and used (only amok/peermanagement is
3262 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3263 loss when they were small enough to fit into the buffer and sent quickly
3264 enough so that they can all get received in one shoot.
3265 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3266 if not found. Use the *_or_null() variant for the old semantic.
3267 * In gras_msg_handle, do not discard messages without callback.
3268 They are probably messages to be explicitly awaited later (ie, proofs of
3269 mis-synchronization in userland since they are sent before being awaited)
3270 No big deal usually.
3271 * gras_socket_meas_send/recv: semantic changed!
3272 The numerical arguments used to be (1) the total amount of data to send
3273 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3274 messages. This was need for the fool willing to send more than MAXINT
3275 bytes on quite fat pipes.
3278 * Do really rename the hostmanagement module to peermanagement. [Mt]
3279 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3280 but this is rather new and this was documented in the module
3281 documentation (poor excuses, I admit)
3282 * Bandwidth measurement semantic changed! This follows the changes to
3283 gras_socket_meas_send/recv explained above.
3286 * A sequential mode has been added to the workstations. When a workstation
3287 is in sequential mode, it can execute only one task, and the other tasks
3288 are waiting in a FIFO. [Christophe Thiery]
3291 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3292 model for SIMDAG. [Christophe Thiery]
3293 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3294 equal to 0, which caused some bad behaviors in
3295 saturated_constraint_set_update. I now use a threshold mechanism like in
3299 * When running manually src/testall, you select specific units [Mt]
3300 testall is the result of our cunit mechanism, and should replace all
3301 the scripty thingy around since bash don't run easily on billware.
3303 * A mallocator system has been added. [Christophe Thiery]
3304 Mallocators allow you to recycle your unused objects instead of freeing them
3305 and allocating new ones.
3307 Documentation update:
3308 * FAQ reworking + New FAQs:
3309 - "Valgrind spits tons of errors!" [Mt]
3310 - "How to repport bugs" [Mt]
3311 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3312 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3313 - Communication time measurement within MSG [AL]
3314 - I experience weird communication times when I change the latency [AL]
3315 * GRAS tutorial [Mt]
3317 - an introduction to the framework and to the used communication model
3318 - an initiatic tour introducing the most proheminent features:
3320 . Lesson 0: Installing GRAS
3321 . Lesson 1: Setting up your own project
3322 o Part 2: Message passing
3323 . Lesson 2: Exchanging simple messages
3324 . Lesson 3: Passing arguments to the processes (in SG)
3325 . Lesson 4: Attaching callbacks to messages
3326 . Lesson 5: Using globals in processes
3327 . Lesson 6: Logging information properly
3328 . Lesson 7: Using internal timers
3329 . Lesson 8: Handling errors through exceptions
3330 . Lesson 9: Exchanging simple data
3331 . Lesson 10: Remote Procedure Calling (RPC)
3332 . Lesson 11: Explicitly waiting for messages
3333 . Recapping of message passing features in GRAS
3334 - A HOWTO section containing:
3335 o HOWTO design a GRAS application
3336 More are due, of course. They will come latter. In the meanwhile, you can
3337 check the examples which are still here.
3339 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3341 SimGrid (3.1) stable; urgency=high
3345 There was a stack corruption somewhere, visible only when optimizing
3346 with these versions. [Vince]
3349 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3350 module that enabled to play with Directed Acyclic Graphs. It is built
3351 directly on top of SURF and provides an API rather close to the old
3352 SG. Some old codes using SG are currently under rewrite to check that
3353 all needful functions are provided. [Christophe Thiery]
3356 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3357 extension of the classical CLM03 model that gracefully handles
3358 failures. This is now the default model for MSG and GRAS. It doesn't
3359 handle parallel tasks yet though. [AL]
3360 * Bug fix: Weights were not correctly set in the network part.
3361 WARNING: This may have resulted in incorrect results with simulations
3362 where there are more than one flow on a given link. [AL]
3365 * After a (long ?) discussion on simgrid-devel, we have decided that the
3366 convention we had on units was stupid. That is why it has been decided
3367 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3368 WARNING: This means that all previous platform files will not work as
3369 such with this version! A warning is issued to ask users to update
3371 A conversion script can be found in the contrib module of the CVS, under
3372 the name contrib/platform_generation/surfxml_update.pl [MQ]
3375 * Bug fix: Processes were started in reverse order, wrt deployment file.
3376 WARNING: if your code relies on this bug, please fix it. [AL]
3377 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3378 being executed on two different locations. [AL]
3379 * Bug fix: Failures are now better supported thanks to Derrick's tests
3380 (there was many failure situations I hadn't thought of and that weren't
3381 correctly handled). [AL]
3382 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3386 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3387 Exception occurring on server-side are propagated back to client (!).
3389 API CHANGE: the callback changed their prototype. Change:
3390 int my_handler(gras_socket_t expeditor, void *payload_data) {
3392 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3393 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3395 * New! function: gras_msg_handleall to deal with all messages arriving
3396 within a given period.
3397 * New! function: gras_socket_server_range to get a server socket in a
3398 range of port numbers (ease to avoid port number conflicts) [MQ]
3399 * New! gras processes display their backtrace when they get a SIGUSR1
3400 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3401 Sweet to debug RL processes [MQ]
3405 - Do not force experiment sizes to be expressed in kb, or it becomes
3406 impossible to measure the latency this way (needs one byte-long tests)
3407 WARNING: this changes the amok_bw_* function semantic. [MQ]
3408 - Implements the link saturation stuff. [MQ]
3409 * Peer management module:
3410 New! module factorizing code that we wrote over and over [MQ].
3413 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3414 - Test units are placed directly into the library code, they get extracted
3415 automatically and placed into the src/testall binary.
3416 - Convert most of the XBT tests to this system.
3417 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3418 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3419 live in the same C code. This required to change a little bit the API
3420 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3421 * New module: sparse graph structure with basic algorithms (this is work
3422 in progress and the API is not considered to be frozen yet). [AL]
3423 * Display more information on backtraces: source line & function names are
3424 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3425 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3426 * Reworked a little bit some #include statements to load only required
3427 headers. Some user code that relied on SimGrid to include stdlib or
3428 stdio may need to include it by themselves. [AL]
3429 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3431 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3432 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3433 but I guess that almost nobody used those parts. [MQ]
3435 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3437 SimGrid (3.0.1) stable; urgency=low
3440 * Unfortunately, I had missed 5 misnamed functions:
3441 xbt_fifo_item_t xbt_fifo_newitem(void);
3442 void xbt_fifo_freeitem(xbt_fifo_item_t);
3443 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3444 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3445 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3446 They're now deprecated. Please use their new versions:
3447 xbt_fifo_item_t xbt_fifo_new_item(void);
3448 void xbt_fifo_free_item(xbt_fifo_item_t);
3449 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3450 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3451 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3453 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3454 * Documentation: xbt_log module unmercifully reworked [MQ]
3455 * Bugfix: there was a problem with the ending of contexts with
3456 the pthread backend. It caused some weird deadlock or behavior
3457 depending on the pthread implementation. [AL]
3458 * Bugfix: get the exceptions raised in the simulator repport where
3459 and why they come from when they are not caught in time [AL, MQ]
3462 * Bugfix: Do repport the error when two non-connected hosts try to
3463 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3466 * Add additionnal checkings on communications. Assert that two
3467 communicating hosts are connected by a set of links... [AL]
3470 * Add additionnal checkings on channel values in communication [AL]
3471 * New: MSG_task_get_source to see on which host a task was generated [HC]
3472 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3473 the number of tasks waiting to be received on channel and sent
3475 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3476 waits for the first task coming from a given host.. [AL]
3478 GRAS new functionnalities: [MQ]
3479 * Enhance the parsing macro to allow the size of multidimentional objects
3480 to be given thru annotations.
3481 * New example (and documentation): Matrix Multiplication a la RPC
3482 (as when I was young!) and fix a bunch of bugs found on the way.
3484 GRAS performance improvements: [MQ]
3486 * Reduce the amount of cbps creation/destruction by making it static to
3487 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3490 * Change libdata to a set so that we can search for stuff by ID (and thus
3491 reduce the insane amount of dict lookups)
3494 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3495 RL communication on top of SG since it's so uneffective.
3496 It may also be used for inter-thread communication in RL, one day.
3497 * Use gras_datadesc_copy() to exchange messages on top of SG
3499 - improve message exchange performance on top of SG
3500 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3501 * Don't exchange on the network the size of the used part of buffer,
3502 instead, specify the possible buffer size to read().
3504 - reduces the amount of read/write calls (one pair per exchange)
3505 - reduces the amount of exchanged data (the size)
3506 - allows to retrieve all arrived data on receiver side, if we don't need
3507 it right now (subsequent read will peek the buffer)
3508 - allows the receiver to proceed with the beginning of the stream before
3509 everything is arrived
3510 - make it possible to build an iov transport (using readv/writev)
3512 - take care of the data with non-stable storage (like stacked data),
3514 * If possible, TCP send uses vector I/O (when writev() is here)
3515 - Don't use it for receive since we send data sizes and data on the
3516 same stream, so we wouldn't be able to chain large amount of chunks
3517 before having to flush the stuff to read the size.
3518 * Rework the transport plugin mechanism to simplify it and reduce the
3519 amount of pointer dereferencement when searching for the right function
3522 * I guess that now, we do almost as few system calls as possible while
3523 doing as few data copy as possible.
3525 To improve it further, we could try to send all the sizes first and then
3526 all the data (to use iov on receiving size), but it's only a partial
3527 solution: when you have 2 dimensional data, the sizes of the second
3528 dimension is data of the first dimension, so you need 3 streams.
3530 I'm not sure the potential performance gains justify the coding burden.
3532 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3534 SimGrid (3.00) stable; urgency=high
3537 * New! Give the possibility to hijack the surf parser and thus bypass
3538 MSG_create_environment and MSG_launch_application. Have a look at
3539 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3541 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3543 SimGrid (2.96) unstable; urgency=low
3548 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3549 This deprecates the xbt_error_t mechanisms.
3550 It modifies (simplifies) all XBT and GRAS API.
3551 MSG API keeps unchanged (exceptions raised by XBT are caught from
3552 within MSG and masked with existing error handling facilities)
3555 * New! Add a FATPIPE model. [AL]
3556 * New! Add a parallel task model. [AL]
3557 * New! Add automatically a loopback interface (in the default
3558 network model) if none was precised.
3561 * Bugfix: MSG_process_resume now works with the current running process.
3563 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3564 * Modification of MSG_task_get_compute_duration. Once a task has been
3565 processed, the value returned by this function is now equal to 0. [AL]
3566 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3567 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3568 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3569 of a m_task. For now, MSG_task_cancel only works with computation
3571 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3572 of the processor (in Mflop/s) regardless of the current load on the
3574 * API Change: use proper naming convention for MSG_getClock and
3575 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3577 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3578 This function changes the priority of a computation task. This priority
3579 doesn't affect the transfer rate. A priority of 2 will make a task
3580 receive two times more cpu power than the other ones. This function
3581 has been added to suit the needs of Nguyen The Loc and hasn't been that
3582 much tested yet. So if it fails, please report it and send me your code.
3584 * API Change: removed all functions and types that were marked "deprecated"
3585 since many months. Renamed MSG_global_init_args to MSG_global_init.
3587 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3589 SimGrid (2.95) unstable; urgency=low
3592 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3593 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3594 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3597 * Bugfix: Allow absolute paths for platform description files [MQ]
3598 * Bugfix: do free the variables after use. Leads to drastic performance
3600 * Implement max_duration (ie, timeouts) on resources [AL]
3603 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3605 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3606 the needed support (timeouts) [AL]
3609 * Implement measurement sockets. You can now get the bandwidth between two
3610 hosts thanks to AMOK (see below). [MQ]
3611 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3612 dynar over the network (yeah) [MQ]
3613 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3614 * simple/static token-ring example. [Alexandre Colucci and MQ]
3615 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3616 in place before [MQ]
3619 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3620 type parsing. "short" and "long" modifiers were also ignored; other
3621 modifier (such as reference level) are still ignored. [MQ]
3622 * Embeed the buffer size within the buffer itself on SG. [MQ]
3623 That way, send() are atomic and cannot get intermixed anymore (at least
3624 the ones which are less than 100k; bigger messages still have the issue)
3625 * Array size pushed by the field, not by the field type (or each
3626 and every long int will push stuff to the cbps) [MQ]
3627 * use select() to sleep since it allows to portably sleep less than one
3630 GRAS (minor cleanups)
3631 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3634 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3635 * type category 'ignored' killed as it was never used and were difficult
3637 * whether a type can cycle or not is now a flag, leaving room for more
3638 flags (as "ignored", if we feel the need one day ;)
3639 * Rename raw sockets to measurement sockets since "raw" has another
3640 meaning in networking community.
3643 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3644 and offers features not belonging to GRAS but that most applications
3645 need. One day, it may be a set of plugins loadable at runtime.
3646 * New module: bandwidth
3647 bandwidth measurement between arbitrary nodes running this module. [MQ]
3649 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3651 SimGrid (2.94) unstable; urgency=low
3653 The first beta release of SimGrid 3 !
3657 * Update the main page and the FAQ. Adding references to gforge.
3660 * Add a gras_os_getpid function.
3663 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3664 * Extend the logs so that they also print PID, hostname, date, ... if
3666 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3667 and kill the old version which were in testsuite/
3668 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3672 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3673 for users. Instead, each programming environment (GRAS, MSG,...) use it
3674 when needed to provide such a feature to users.
3675 Don't shortcut the mechanism or you will also shortcut the virtualization
3676 you need on the simulator.
3680 * Cleanups in configury with regard to compile optimization/warning flags.
3681 Also add -fno-loop-optimize to any powerpc since it's the optimization
3682 killing gcc (< 3.4.0).
3683 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3684 needed by MSG examples complications
3685 * Borrow configury beautifications from PHP
3688 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3689 warning (thanks loris for stumbling into this one).
3690 * Bugfix: stop loading private headers (gras_config.h) from the public
3694 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3695 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3696 RL would imply a DNS resolver, which is *hard* to do in a portable way
3697 (and therefore delayed).
3698 * Implement a real timer mechanism and use it in timing macros. This allows
3699 to avoid rounding errors and get a 0.000005 sec precision in timing
3700 macros. While I was at it, various cleanups:
3701 - allow to declare more than one timed section per file (fix a stupid bug)
3702 - move some private declaration to the right place
3703 - merge conditional execution and timing macros into emulation module
3704 - document the module
3705 - make sure the module cleanups its mess on gras_exit
3706 * Documentation improvements:
3707 - (new) how to compile applications using GRAS
3708 - (new) emulation support (timing macros)
3710 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3712 SimGrid (2.93) unstable; urgency=low
3714 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3717 - Use Paje properly where used. Still to be sanitized properly.
3718 - Portability fix: Add an implementation of the contexts using pthread
3722 - Add xbt_procname(): returns the name of the current process.
3723 Use it to show the current process's name in all logging.
3725 - fix detection of older flex version and the reaction, since we do
3726 depend on modern ones (we use lex_destroy)
3727 - Better separation of SG and RL in the libs: remove all simulation code
3728 from libgras. As a result, this lib is now only 200k when stripped.
3729 Some of the xbt modules may also be duplicated (two sets and such) and
3730 should be cleaned/killed before SG3.
3731 - Insist on using xlC on AIX because of weird problems involving gcc there.
3732 - Cleanup the make remote stuff. This is now done by scripts
3733 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3734 mainly for our private use, but we're working on changing them to user
3737 - Bugfix: flush the socket on close only if there is some *output*.
3738 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3740 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3743 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3745 SimGrid (2.92) unstable; urgency=low
3747 Alpha 3 on the path to SimGrid 3
3751 - New! First try of benchmarking macros.
3752 - New! First try so that gras_stub_generator generate deployment and
3753 remote compilation helpers.
3755 - Bugfix: Initialization fix in msg_test.
3759 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3761 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3763 - API Change: gras_os_sleep to take the amount of seconds as a double.
3764 Accepting an int was error prone since it was the only location where
3765 seconds were coded as such. It leaded to damn rounding errors.
3766 - Bugfix: Hard to belive that timers ever worked before this.
3768 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3770 SimGrid (2.91) unstable; urgency=low
3772 Alpha 2 on the path to SimGrid 3
3776 - Bug fix in the lmm_solver.
3778 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3779 through the function MSG_paje_output.
3780 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3781 - It is possible to bound the rate of a communication in MSG with
3782 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3783 to write it in the changelog).
3784 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3789 - Various cleanups to the autotools stuff
3790 - Begin to move Gras examples to examples/gras/
3791 - Let make distcheck work again (yeah!)
3793 - documentation overhauled using doxygen.
3794 gtk-doc-tools is dead in SimGrid now.
3795 - Automatically extract all existing logging categories, and add the list
3796 to the documentation (long standing one, to say the less)
3798 - Cleanup the known architecture table. Reorder the entries to group what
3799 should be, and use a more consistent naming scheme.
3800 (some of the test dataset are still to be regenerated)
3801 - New! Allow library to register globals on each process just as userdata
3803 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3804 lookup time (for now).
3805 Use it in msg and trp.
3806 This cleans a lot the internals and helps enforcing privacy of the
3807 headers between the gras components.
3808 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3809 - Bugfix: gras_os_time was delirious in RL.
3810 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3812 - Reenable GRAS now that it works.
3814 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3816 SimGrid (2.90) unstable; urgency=low
3818 Alpha 1 on the path to SimGrid 3
3820 * It is a long time since the last release of SimGrid. I'm sorry about
3821 that but as I had told you, I was rewriting a lot of things. I apologize
3822 to those who had been reporting bugs to me and that I had not answered.
3823 If your bug is still in the new version, please tell me. Here is a
3824 summary of the main changes.
3826 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3827 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3828 portability, speed, and a lot more but you'll figure it out later.
3829 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3830 I wanted to release the new MSG as soon as possible and I have
3831 broken GRAS, which is the reason why, for now, only MSG is fully
3832 functional. A laconic description of these projects is available
3833 in the documentation.
3835 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3836 kernel optimized for our needs (called SURF but only the developers
3837 should use it). Hence, MSG is now roughly 30 times faster and I think
3838 that by rewriting a little bit MSG, I could event speed it up a little
3839 bit more. Beside the gain in speed, it is also much easier to encode a
3840 new platform model with SURF than it was with SG. More to come...
3842 * REVOLUTION 3: I have tried to change a little as possible the API of
3843 MSG but a few things really had to disappear. The main differences
3844 with the previous version are:
3845 1) no more m_links_t and the corresponding functions. Platforms are
3846 directly read from an XML description and cannot be hard-coded
3847 anymore. The same format is used for application deployment
3848 description. The new format is described in the documentation.
3849 Have a look in tools/platform_generation. There is a tiny script
3850 that converts from the old platform format to the new one. Concerning
3851 the application deployment format, parsing the old one is tricky.
3852 I think most of you should however be able to convert your files. If
3853 it is really an issue, I can write a C code that does the conversion.
3855 2) the toolbox tbx does not exist anymore. We now have a library
3856 with much more data-structures but without the hash-tables (we have
3857 dictionaries that are much faster).
3859 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3861 *****************************************************************************
3862 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3863 * don't feel like converting the oldies. (Mt) *
3864 *****************************************************************************
3867 Version 2.90: "the long awaited one"
3868 - Finished rewriting and debugging MSG. Rewrote the documentation.
3869 - disable GRAS for now since it needs to be ported to the newest SG
3872 - Finish the port to windows (using mingw32 for cross-compile)
3875 - Main loop and datastructures of SURF. A cpu resource object is
3876 functional. Surf can thus be used to create cpu's with variable
3877 performance on which you can execute some actions.
3879 2004-11-15 Martin Quinson
3880 - Port to ARM. Simply added the alignment and size descriptions. Should
3881 work, but the ARM machines are so slow that I didn't had the opportunity
3882 to 'make check' over there yet.
3884 2004-11-15 Arnaud Legrand
3885 - Trace manager now written. It uses a heap structure and is therefore
3886 expected to be efficient. It may however be speeded up (particularly
3887 when many events occur at the same date) by using red and black
3888 trees. One day maybe...
3889 - Max-min linear system solver written. It uses a sparse matrix
3890 structure taking advantage of its expected use. Most operations are
3891 O(1) and free/calloc are called as few as possible. The computation of
3892 the minimum could however be improved by using a red and black tree
3895 2004-11-03 Arnaud Legrand
3896 - Rename every gras_* function that was in xbt/ to its xbt_
3898 - Add a heap and a doubly-linked list to xbt
3899 - Added a dichotomy to the dictionaries. make check works as well before
3900 so I assume that the patch is correct. I do not know however if things
3901 run effectively faster than before now. :)
3903 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3904 SimGrid, and the version number is bumped to 2.x
3906 2004-10-29 Martin Quinson
3907 - Introduction of the remote errors.
3908 They are the result of a RMI/RPC on the remote machine.
3909 ErrCodes being scalar values, you can't get the host on which those
3910 errors did happen. Extending the error mechanism as in Gnome is possible.
3911 No idea yet whether it is a good idea.
3913 2004-10-28 Martin Quinson
3914 - Interface revolution: the Starred Structure Eradication.
3915 I used to do typedef struct {} toto_t; and then handle *toto_t.
3916 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3917 - ???_t is a valid type (builded with typedef)
3918 - s_toto_t is a structure (access to fields with .)
3919 - s_toto is a structure needing 'struct' keyword to be used
3920 - e_toto_t is an enum
3921 - toto_t is an 'object' (struct*)
3923 typedef struct s_toto {} s_toto_t, *toto_t;
3924 typedef enum {} e_toto_t;
3925 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3926 s_toto_t) is private.
3928 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3929 so that it can set the variable to NULL. It was so for dicts and sets,
3930 it changed for dynars.
3932 - Fix a bunch of memleaks in dict_remove
3933 - Fix a bug in sg/server_socket opening: it failed all the time.
3935 2004-10-07 Martin Quinson
3936 - Speed up dynar lookup operation a bit.
3938 gras_dynar_get is dead.
3940 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3941 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3942 (which gives you the address of the stored data).
3944 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3945 copy of the data using an affectation to do the job and not a memcpy.
3947 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3948 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3950 It does not really speedup the dynar test because they are
3951 setting elements all the time (and look them seldom). But the dict does
3952 far more lookup than setting.
3954 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3956 2004-10-05 Martin Quinson
3957 - Allow to (en/dis)able the cycle detection at run time.
3959 Whether we should check for cycle or not is now a property of each
3960 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3961 datadesc_cycle_unset allow to remove this property when previously set.
3963 Note that the cycle detection is off by default since it impacts the
3964 performance. Watch the data you feed GRAS with ;)
3966 This property is hereditary. Any element embedded in a structure having it
3967 set have it set for the time of this data exchange.
3969 You should set it both on sender and receiver side. If you don't set it on
3970 sender side, it will enter an endless loop. If you forget on receiver
3971 side, the cycles won't be recreated after communication.
3973 - Header reorganization.
3974 Kill gras_private.h, each submodule must load the headers it needs.
3976 2004-10-04 Martin Quinson
3977 - Interface revolution: do not try to survive to malloc failure.
3979 Now, gras_malloc and friends call gras_abort() on failure.
3980 As a conclusion, malloc_error is not a valid error anymore, and all
3981 functions for which it was the only gras_error_t return value are
3982 changed. They now return void, or there result directly.
3983 This simplify the API a lot.
3985 2004-09-29 Martin Quinson
3986 - Re-enable raw sockets.
3987 Created by gras_socket_{client,server}_ext;
3988 Used with gras_raw_{send,recv}
3991 It should allow to kill the last bits of gras first version soon.
3993 This is not completely satisfactory yet (duplicate code with
3994 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3997 - Simplify transport plugin (internal) interface by not passing any
3998 argument to _server and _client, but embedding them in the socket
4001 2004-09-28 Martin Quinson
4002 - Finish the port to AIX.
4003 autoconf was my problem (segfault within the malloc replacement
4004 function. No idea why)
4006 2004-09-16 Martin Quinson
4007 - Fix some size_t madness on 64bit architectures.
4009 2004-09-08 Martin Quinson
4010 - Reduce the number of system headers loaded, overload some more system
4011 calls (such as malloc to cast the result of the system one, and work
4013 - Fix and reintroduce the config support
4015 2004-09-07 Martin Quinson
4016 - Source code reorganization to allow Arnaud to surf all over there.
4017 - Allow to document the logging categories.
4018 - Remove all uppercase from logging categories and useless cleanup in names.
4020 2004-08-18 Martin Quinson
4021 Version 0.6.2 (protocol not changed; API changed)
4022 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4023 gras_error_t), and NULL when not found. Functions expecting a msgtype
4024 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4025 hopefully usefull message.
4026 - Portability to prehistoric sparcs again
4028 2004-08-17 Martin Quinson
4029 Version 0.6.1 (protocol not changed; ABI not changed)
4030 - prealloc some buffers to speed things up
4032 2004-08-11 Martin Quinson
4033 Version 0.6 (protocol not changed; ABI expended)
4034 - The parsing macro can deal with the references, provided that you add
4035 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4037 2004-08-09 Martin Quinson
4038 Version 0.5 (protocol not changed; ABI changed)
4039 - Allow to off turn the cycle detection code in data exchange at
4040 compilation time. It should be at run time, but I'm short of time (and
4041 the config stuff is still broken). That way, we keep dict out of the
4042 critical path, which is good because the performance is poor:
4043 - search not dichotomial yet
4044 - dynar give no way to access their content and memcpy everytime
4045 - In composed data description (struct, ref and so on), stop foolness of
4046 keeping the subtype's ID, but store the type itself. This keeps sets out
4047 of the critical path, which is good since they rely on dynar and
4048 dictionaries. The only loose of that is that we cannot detect the
4049 redeclaration of a structure/union with another content (but I'm not sure
4050 the code detected well this error before anyway). We still can detect
4051 the redefinition discrepancy for the other types.
4052 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4053 it breaks the code because of type-punning used all over the place).
4054 This breaks on all non-gcc architectures (for now).
4056 All those changes (plus the buffer of last time) allow me to gain 2 order
4057 of magnitude on cruel tests consisting of 800000 array of integers on two
4058 level of a hierarchical structure (200 secondes -> 4 secondes)
4061 - the selector of reference must now return the type it points to, not
4062 the ID of this type.
4064 2004-08-06 Martin Quinson
4065 Version 0.4 (protocol changed; ABI not changed)
4066 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4067 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4068 - Implement a buffer transport, and use it by default (it relies on tcp in
4069 real life and on sg in simulation).
4070 That's a bit hackish since I had a new field to the structure to store
4071 its data without interfering with the subtype ones. Inheritance
4072 is tricky in C. And that's a kind of reverse inheritance with one class
4073 derivating two classes. Or maybe a game with java interfaces. Anyway,
4074 that's damn hard in C (at least).
4075 Moreover, I got tired while trying to ensure plugin separation and
4076 genericity in SG mode. MSG wants me to do weird things, so let's go for
4077 cruel hacks (temporarily of course ;).
4078 See comment in transport_private.h:71
4079 - do not include all the _interface headers in private but in the files
4080 which really need them (to cut the compilation time when they are
4083 2004-07-26 Martin Quinson
4084 Version 0.3 (protocol not changed; ABI changed)
4085 - Major overhault of the datadesc interface to simplify it:
4086 - shorted the function names:
4087 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4088 - add a trivial way to push/pop integers into the cbps without malloc.
4089 This allows to make really generic sub_type description, which simply
4090 pop their size of the stack.
4091 - add a function gras_datadesc_ref_pop_arr() which does what users want
4092 most of the time: Declare a dynamic array (which pops its size of the
4093 stack) and declare a reference to it. Poor name, but anyway.
4094 - kill the post-send callback, add a post-receive one
4096 2004-07-23 Martin Quinson
4097 Version 0.2 (protocol changed; ABI changed)
4098 - add some testing for cpbs in the test cases, and fix some more bugs.
4099 This invalidate again the little64 data file, since I cannot regenerate
4101 - remove an awfull optimization in the logging stuff, allowing me to:
4102 - understand it again
4103 - learn gcc how to check that the argument match the provided format
4104 - fix all errors revealed by gcc after that
4105 - internal keys of dict are not \0 terminated. Deal with it properly in
4106 loggings instead of segfaulting when the user want to see the logs :-/
4108 2004-07-22 Martin Quinson
4109 - Fix some stupid bug preventing cbps (callback postit) from working
4111 2004-07-21 Martin Quinson
4112 - Some documentation cleanups
4113 - remove the useless last argument of msgtype_declare
4114 - rename the Virtu functions to fit into the 'os' namespace
4115 - move headers src/include -> src/include/gras/ and stop fooling with
4116 gras -> . symbolic link
4117 - make distcheck is now successful
4119 2004-07-19 Martin Quinson
4121 - Build shared library also
4122 - Install html doc to the right location
4123 - stop removing maintainer files in make clean
4124 - build tests only on make check
4126 2004-07-13 Martin Quinson
4128 - No major issue in previous version => change versioning schema
4129 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4130 corresponding dataset.
4132 2004-07-11 Martin Quinson
4134 - Get it working with any kind of structure (we can compute the padding
4135 bytes remotely for all the architectures I have access to)
4136 - Implement the structure parsing macro (still not quite robust/complete)
4137 - Improvement to the remote testing toysuite
4139 2004-07-10 Martin Quinson
4140 [autoconf mechanism]
4141 - get ride of a bunch of deprecated macros
4142 - actually run the test for two-compliment, not only compile it :-/
4143 - test whether the structures get packed (and bail out if yes. Damn.
4144 Alignment is a serious matter)
4145 - test whether the structures get compacted (but respecting the alignment
4146 constraints of each types)
4147 - test whether the array fields of structures can straddle alignment boundaries
4149 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4150 (took me 2 hours to find that bug, looking at the wrong place)
4151 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4152 before sending/receiving and are used to compute the offsets of fields
4153 - Given that padding size depend even on compiler options, keep track of
4154 alignment and aligned_size only for the current architecture. Not a big
4155 deal since we send structure fields one after the other (seems
4157 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4158 program, let it work on linux/gcc/little32. portability todo.
4160 2004-07-08 Martin Quinson
4161 - import and improve remote compilation support from FAST
4162 - make sure make check works on half a dozen of machines out there
4164 2004-07-07 Martin Quinson
4165 Let's say it's version 0.0.3 ;)
4166 - Implement conversions (yuhu!)
4167 - Let it work on solaris (beside conversion, of course)
4168 - Stupid me, using rand() to generate the conversion datatests in not wise.
4170 2004-07-06 Martin Quinson
4171 - Let make dist work, since I'm gonna need it to compile on remote hosts
4172 - Let Tests/datadesc_usage write the architecture on which the file was
4173 generated as first byte.
4174 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4175 architecture descriptions.
4176 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4177 on those architectures.
4178 - Optimization: send/recv array of scalar in one shoot
4180 2004-07-05 Martin Quinson
4181 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4183 - Plug a whole bunch of memleaks
4184 - each process now have to call gras_{init,exit}. One day, their log
4185 settings will be separated
4186 - Continue the code factorisation between SG, RL and common in Transport.
4188 2004-07-04 Martin Quinson
4190 - Redistribution between SG and RL.
4191 We wanna have to accept in SG, so move accepted related parts of RL in
4192 the common part. (more precisely, the dynar of all known sockets is no
4193 more a static in transport.c, but part of the process_data)
4195 [gras_stub_generator]
4196 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4198 2004-07-03 Martin Quinson
4199 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4200 them all up in one shot)
4202 - Fix the ugly case of reference to dynamic array.
4203 - New (semi-public) function gras_datadesc_size to allow the messaging
4204 layer to malloc the needed space for the buffer.
4206 - gras_socket_close now expect the socket to close (and not its address to
4207 put NULL in it after it). This is because the socket passed to handlers
4208 is one of their argument (=> not writable).
4210 - propagate the interface cleanup from last week in datadesc, ie remove a
4211 superfluous level of indirection. User pass adress of variable
4212 containing data (both when sending and receiving), and not of a variable
4213 being a pointer to the data. Let's say that I like it better ;)
4214 The price for that is constructs like "int msg=*(int*)payload" in
4215 handlers, but it's a fine price, IMHO.
4217 - Let it work in RL (yuhu)
4219 2004-06-21 Martin Quinson
4221 - porting SG plugin and SG select to new standards (works almost).
4222 - plug memleaks and fix bugs around.
4225 - cleanup the prototype of data recv and force users to specify when they
4226 want to handle references to objects. Test case working even for cycles.
4227 - plug memleaks. Valgrind is perfectly ok with this.
4229 2004-06-12 Martin Quinson
4231 - cleanup the separation between plugin and main code in plugin creation
4233 2004-06-11 Martin Quinson
4235 - Reput hook for raw sockets, needed for BW experiments
4236 - kill a few lines of dead code
4237 [Data description] Interface cleanup
4238 - gras_datadesc_by_name returns the searched type or NULL.
4239 That way, no variable is needed to use a type desc once, which makes
4241 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4242 two parameters were strings (field name, type name), leading to
4244 [Dicos] Interface cleanup
4245 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4246 This is consistant with the dynar API.
4248 2004-04-21 Martin Quinson
4250 - Porting to new standards.
4252 - interface cleanup.
4253 There is no bag anymore, no need to take extra provision to mask the
4254 pointers behind "ID".
4255 Better splitup of functions between files create/exchange/convert.
4256 This is still a bit artificial since convert and receive are so
4257 interleaved, but anyway.
4259 - add a queued message list to procdata (the ones not matching criteria
4261 - factorize some more code between SG and RL wrt procdata
4263 - use gras_exit in example to track memleaks
4264 - get rid of gs_example now that GS is properly integrated into gras
4265 - update run_test to integrate the latest tests (datadesc)
4267 - rename WARNINGn macros to WARNn since it prooved error-prone
4269 2004-04-19 Martin Quinson
4271 - register init/exit functions within gras module mechanism
4272 - send/receive function.
4273 Convertion is not implemented, but short-cutted if not needed.
4274 struct/array elements are sent one by one (instead of block-wise), but
4275 nobody really cares (yet). Get a prototype before optimizing.
4276 - tests (using a file socket) for DD send/receive on:
4277 - base types: int, float
4278 - array: fixed size, string (ie ref to dynamic string)
4279 - structure: homogeneous, heterogeneous
4280 - chained list, graph with cycle
4281 Believe it or not, valgrind is not too unhappy with the results. The
4282 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4286 - Bugfix when using a filename explicitly (instead of '-')
4288 2004-04-09 Martin Quinson
4290 - factorize more code between RL and SG in socket creation
4291 - Complete the implementation and tests of:
4293 o file (only in RL, and mainly for debugging)
4295 I lost 3 days to design a portable address resolver, and then decided
4296 that the prototype mainly have to run on my box.
4297 Addressing portability too early may be like optimizing too early :-/
4299 - use gras_init in the Tests instead of the crappy parse_log_opt
4300 (the latter function is removed)
4301 [Conditional execution]
4302 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4303 [Code reorganisation]
4304 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4305 Build examples against the RL library, since there is no way to disable
4306 its creation for now.
4308 For information, the beginning of coding on GRAS was back in june
4309 2003. I guess that every line has been rewritten at least twice since