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.
29 - New section: Configuring the logs from the command line.
30 - New section: Programming API of logs.
32 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
33 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
34 - FG#37: Parallel tasks are limited to 1 core per host
36 ----------------------------------------------------------------------------
38 SimGrid (3.26) December 15. 2020
40 The Release release (the French lockdown was eased today).
42 Important user-visible changes:
43 - SimGrid now requires a compiler with C++14 support.
44 Sibling projects should upgrade their FindSimgrid.cmake
45 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
46 several users had difficulties to understand issues when using high bandwidth or
47 small latency events. The new value was already the default for SMPI and
48 should not cause too much performance hit. It can change some simulation
49 timings if simulation was skipping a lot of small events (which was actually
50 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
51 - On some network models (L07, CM02, LV08), an implicit loopback link is created
52 for local communication. Default bandwidth was 500MBps and latency was 15µs,
53 which was rather bad. Default values have been changed to 10GBps and 0 latency.
54 Options network/loopback-lat and network/loopback-bw have been added to adapt
55 these values if needed. Network and routing models already providing separate
56 loopbacks (clusters and such) are not impacted by this change.
57 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
58 metadata that are added to each MPI message by MPI libraries (tag,
59 src/dest, and communicator information). This may change simulation times
60 for very small message sizes, but should be more realistic.
63 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
64 - New plugins: wifi_energy and link_load. See the associated examples.
65 - New examples: wifi networks, and communication suspend.
68 - LTO is now enabled for Intel/clang compilers.
69 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
70 Setting it to "auto" will use all available cores, while setting it to n will
71 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
72 - Remove obsolete runtime option 'contexts/parallel-threshold'.
73 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
74 has been kept for compatibility.
75 - Finally remove obsolete snake_case() aliases for runtime options.
76 - Further improve the documentation.
79 - the on_exit() of each actor is also executed when the simulation deadlocks.
80 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
81 An example is provided for s4u::Comm but it should work with Exec and Io.
84 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
85 https://framagit.org/simgrid/SMPI-proxy-apps
86 - MPI/IO: supports names starting with "./"
87 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
89 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
90 building shared libraries, this may cause issues, so environment variable
91 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
92 - most temporary files should now be created in /tmp dir (or equivalent).
93 If this one does not allow execution of code (noexec flag), this may cause issues.
94 Please use another tmp directory (using TMPDIR or equivalent system variable)
98 - A large refactoring is ongoing, but this should be transparent for users so far.
101 - Drop xbt_str_split().
103 C binding and interface:
104 - The return type of the sg_actor_on_exit() callbacks is now 'void'
105 (int value was previously ignored)
106 - Many C functions were renamed to follow a common naming style
107 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
108 names are kept until version 3.30.
109 - Many MSG tests were converted to the new S4U's interface in C, that
110 was extended for that.
112 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
113 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
114 - FG#41: Add sg_actor_create C interface
115 - FG#43: xbt::random needs some care
116 - FG#48: The Impossible Did Happen (yet again)
117 - FG#50: Suspending an actor executed at the current timestamp fails
118 - FG#52: Zero-seconds timeout : "That's in the past already"
119 - FG#53: Crash while using ns-3 network model
120 - FG#54: How to suspend a comm?
121 - FG!22: Stochastic Profiles
122 - FG!24: Documentation and fix for xbt/random
123 - FG!35: Add a modeling hint for parallel links in doc
124 - FG!36: [xbt/random] Read/Write the state of the RNG
125 - FG!42: Link cumulated load plugin
126 - FG!44: Test condition variable wait_until+notify_all
127 - GF#18137: Allow different stack sizes?
128 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
129 - GH#139: Allow pthread creation in SMPI
130 - GH#336: Packet-level simulation using SMPI?
131 - GH#345: Error in the LMM while migrating a VM
132 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
133 - GH#352: pip install / python setup.py install fail to find pybind11
134 - GH!337: Fix link_energy plugin for wifi platforms
135 - GH!339: Add Mailbox set_receiver method to python binding
136 - GH!344: Cast hugepages macros parameters to int64
138 ----------------------------------------------------------------------------
140 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
142 The Palindrom Day Release.
144 Important user-visible changes:
145 - Improve the Python usability (stability and documentation).
146 - A nasty synchronization bug was ironed out, see also below.
147 - Python's doc was organized as a separate tree, now integrated with C++.
148 - C bindings of S4U were not part of the doc.
149 - The C++ doc was also improved as methods are now split by theme.
150 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
151 - This is mandatory to use the Java bindings.
152 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
153 - SimDAG++: Automatic dependencies on S4U activities (experimental)
154 - Some features are already implemented but not all of them
155 - Cannot block an activity until it's scheduled on a resource
156 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
157 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
160 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
161 - Drop support for Viva/Triva (old visualization tools), hence removed
162 TRACE_get_node_types() and TRACE_get_edge_types() functions.
165 - Require pybind11 version 2.4.
166 - Greatly improve locking during context switches with Python.
167 - Declaring Python callbacks to this_actor::on_exit is now working.
168 - No more tests are failing currently, even on win32.
171 - Actor: Merge signals on_migration_start/end into on_host_change
172 - Actor: Rename migrate() into set_host()
173 - Disk: Allow users to get the read and write nominal bandwidth values
174 - Exec: Implement wait_for(timeout)
175 - Io: Implement wait_for(timeout)
178 - Parse errors now raise a simgrid::ParseError that you may want to catch.
181 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
182 been renamed to fix a spelling error. As usual, the previous definitions are
183 kept with a deprecation warning. However, to avoid runtime errors, it is now
184 forbidden to override these deprecated methods in a derived class. Please use
185 the new names immediately if you need to override them.
188 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
189 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
190 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
191 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
192 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
193 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
194 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
195 - New module: random, providing classical random numbers generators.
198 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
199 allocations' size above which they are considered "shared" by default (as if
200 they were performed through SMPI_SHARED_MALLOC macros).
201 Default = 0 = disabled feature.
202 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
203 This can cause some troubles if codes are already overriding these. If this
204 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
205 help, but will make this feature unusable.
206 - Input check is now performed much more consistently and outputs precise
207 warnings for each error found
208 - New supported calls : MPI_File_set_view, MPI_File_get_view
209 - MPI I/O now only supports disks, as storage is discontinued.
210 - MPI I/O allows opening files without specifying absolute path
211 - Note : SMPI configuration options are now initialized by smpi_init_options()
212 call, which should be called if SMPI is not being used through smpirun.
214 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
215 - FG#9: Python bindings crashing
216 - FG#39: Missing s4u::Comm::wait_any_for example
217 - FG#42: Add support for ThreadSanitizer (TSan)
218 - FG!19: Removing RngStream
219 - FG!20: A module for RNG calls
220 - FG!21: Choice between ad-hoc and standard distributions implementations
222 - FG!25: Fix link in Mutex doc
223 - FG!26: Fix links in SMPI interface doc
224 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
225 - GH#305: Unscheduled tasks are still excuted
226 - GH#313: smpirun: manual outdated w.r.t. --help
227 - GH#321: [S4U] Get task remaining work ratio
228 - GH#323: Crash when an actor turn off his physical host
229 - GH#335: Missing links on dragonfly example svg
230 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
231 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
233 ----------------------------------------------------------------------------
235 SimGrid (3.24) October 9. 2019
237 The Clean Disk Release.
240 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
241 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
242 thus becomes the default resource to execute I/O operations. Examples
243 and the file system plugin have been modified to use disks instead of
244 storage elements. Storage can still be used but is doomed to disappear.
245 Users are thus adviced to update their simulators accordingly.
246 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
247 for consistency with pthread_barrier_wait()
248 - Host::get_englobing_zone() returns the englobing netzone
249 - Actor::on_destruction is now called in the destructor
250 Actor::on_termination new signal called when the actor terminates
252 - Global signals are now part of the Engine:
253 - on_platform_creation: after config settings, before the XML parsing
254 - on_platform_created: right after the XML parsing
255 - on_time_advance: each time the clock advances
256 - on_simulation_end: after simulation, before cleanups
257 - on_deadlock: as the name implies.
259 - sg_{actor,host,link}_{data,data_set}() now all exist.
260 Use them to attach user data to the object and retrieve it.
263 - Introduce an experimental Wifi model. It sounds reasonable
264 according to the state of the art, but it still has to be properly
265 validated, at least against ns-3 if not against reality.
266 - Improved the usability of ns-3. Several bugs were ironed out.
267 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
268 It is now expressed as 'idle:epsilon:allCores' to properly model the
269 consumption of non-whole tasks on mono-core hosts. More info in the doc.
272 - convert a new set of functions to the S4U C interface and move the old MSG
273 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
276 - Fortran bindings for DVFS have been removed.
277 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
278 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
279 - SMPI can now be selected by cmake's find_module(MPI) with
280 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
281 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
282 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
283 - trace-call-location can be used with TI traces, and replayed, and is compatible
284 with smpi/comp-adjustment-file.
285 - sleep events are now correctly traced/replayed.
286 - Default for trace-call-location is now to use file names and not full paths.
287 To revert to previous behavior (in case of collision of filenames), option
288 "smpi/trace-call-use-absolute-path" can be set to yes.
291 - Use the included xxHash as an hash implem when C++14 is usable.
292 - Option model-checker/hash was removed. This is always activated now.
293 - New option smpi/buffering controls the MPI buffering in MC mode.
294 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
295 This is useful to check for MPI compliance.
298 - New section on plugins: how to define a new one, and existing ones.
301 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
302 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
305 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
306 and <mount> tags. The rationale is finish to decouple what should be handle
307 by SimGrid kernel (i.e., resources and activities) from plugins built in
308 user space (e.g., the file system plugin). This new tag currently accepts
309 only three attributes (id, read_bw, and write_bw). All the other information
310 that was declared with the storage related tags now has to be expressed as
311 properties. An example of platform using this new tag is available at
312 examples/platforms/hosts_with_disks.xml
315 - 'expect signal' can now accept more than one potential signal.
317 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
318 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
319 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
320 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
321 - FG#31: per-actor data
322 - FG#34: SG_BARRIER_SERIAL_THREAD?
323 - FG#35: model-checker does not like buster-produced binaries
324 - FG!13: MC: complete workaround in the error msg seen on modern systems
325 - FG!15: execute_flops now logs compute
326 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
327 - FG!17: ns-3: unblock the right number of communications + others issues
328 - FG!18: Improving the performance of the ns-3 bindings
329 - GH#189: Energy consumption of parallel tasks
330 - GH#219: Error in the throughput of TCP transfer
331 - GH!330: Minor polishing of English: Configuring_Simgrid
332 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
333 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
334 - GH!334: Rework energy plugin, again
336 ----------------------------------------------------------------------------
338 SimGrid (3.23.2) July 8. 2019
341 - Nicer introduction page.
342 - Migrate the "Deploy your application" page to the new doc.
343 - Move Java as a subtree of MSG.
346 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
349 - Drop xbt_dynar_sort_strings().
352 - Really fix FG#26: Turning off a link should raise NetworkFailureException
353 - FG#27: Wrong exception thrown to wait_any when link is turned off
354 - GH#328: Java: Canceling multiple tasks in a single vm/host
356 ----------------------------------------------------------------------------
358 SimGrid (3.23) June 25. 2019
360 The Exotic Solstice Release.
363 - SunOS and Haiku OS support. Because exotic platforms are fun.
364 - Stop setting random seed with srand48() at initialization.
365 - Use addr2line as a fallback for stacktraces when backtrace is not available.
366 - Build option -Denable_documentation is now OFF by default.
367 - Network model 'NS3' was renamed into 'ns-3'.
370 - Simgrid can now hopefully be installed with pip.
373 - wait_any can now be used for asynchronous executions too.
376 - New log appenders: stdout and stderr. Use stdout for xbt_help.
377 - Drop xbt_dict_dump.
380 - 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
381 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
382 - MPI interface is now const correct for input parameters
383 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
384 - Fortran bindings for async collectives.
385 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
388 - Remove option 'model-check/record': Paths are recorded in any cases now.
389 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
390 always sparse. That's an immense gain in memory, and can even be
391 faster because of cache effects. So there is no need to clutter the
392 code to allow the user to go for the unefficient mode.
395 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
396 models proved to be more accurate than these old experiments.
398 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
399 - FG#1: Broken link in error messages
400 - FG#2: missing installation documentation
401 - FG#3: missing documentation in smpirun
402 - FG#6: Python bindings not available on PyPI
403 - FG#7: simple cmake call requires doxygen
404 - FG#8: make python bindings an optional dependency
405 - FG#10: Can not use MSG_process_set_data from SMPI any more
406 - FG#11: Auto-restart actors forget their on_exit behavior
407 - FG#12: -Denable_lto=OFF doesn't disable LTO
408 - FG#13: Installs unstripped file 'bin/graphicator'
409 - FG#14: Installs the empty directory 'doc/simgrid/html'
410 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
411 - FG#17: Dead link in doc (pls_ns3)
412 - FG#20: 'tesh --help' should return 0
413 - FG#21: Documentation link on http://simgrid.org/ broken
414 - FG#22: Debian installation instruction are broken
415 - FG#26: Turning off a link should raise NetworkFailureException exceptions
416 - GH#133: Java: a process can run on a VM even if its host is off
417 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
418 - GH#326: Valgrind-detected error for join() when energy plugin is activated
420 ----------------------------------------------------------------------------
422 SimGrid (3.22) April 1. 2019
424 The Easter Christmas Release (this one is somewhat late).
427 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
428 - Not complete yet: asynchronous activities (amongst others) are still missing
429 - Still ongoing: the interface may change in the future. We need more
430 testers! Please report any glitches.
431 - No new project using Java should start now. Please switch to Python.
434 - Some of the internal cleanups may lead to speed improvements:
435 - The hard limitation on the amount of simulated actors with
436 Java+Mac was removed. Now, the available memory is the only limit.
437 - Our refcounting was tidyied, leading to 10% speedups in some cases.
438 - We are still working on making our code robust to the actor kills
439 and hosts' churn. Things are improving, but it's not perfect yet.
440 - Replaced our own code to display a backtrace (that was forking addr2line)
441 with the Boost.Stacktrace library.
442 You won't see your backtraces without this optional dependency.
443 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
444 - Stop setting random seed with srand() at initialization.
447 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
448 XML file version remains 4.2 since old files are still compatible.
451 - Process termination which was broken at version 3.21 has been repaired.
452 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
453 - Hide the examples into examples/deprecated. New users should use Python.
456 - Drop MSG_process_create_from_stdfunc() from the API.
457 This C++ function was a pimple in the C API, made necessary at some
458 point by the Java bindings. This is fixed now.
459 - Hide the examples into examples/deprecated. New users should use S4U.
460 - MSG_process_create and MSG_process_attach now crash if the host used
461 to run this process is off.
462 - Fix the protype of MSG_process_on_exit()
463 Now use (int,void*) callbacks instead of (void*,void*) ones.
464 The implementation was ways too messy to actually work, I guess.
467 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
468 too often and provide more accurate timings.
469 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
470 - MPI_Alltoallw support
471 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
472 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
473 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
474 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
476 - Basic implementation of generalized requests (SMPI doesn't
477 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
480 - Drop sg_cmdline. Please use xbt_cmdline instead.
481 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
482 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
483 OsSemaphore is implemented in a portable way with C++11 threads.
484 This should allow much more threads to be created at the same time,
485 allowing Mac and Java users to simulate many more actors.
486 - Implement the 'thread' factory with std::thread instead of xbt ones.
487 It is not possible to set the stack size with threads anymore, but
488 -fsplit-stack is the way to go nowadays when using threads.
489 - Drop the xbt_os_thread_t module (now unused)
490 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
491 - Drop xbt_str_join_array().
492 - Drop cunit, use Catch2 instead.
495 - Many cleanups in the kernel::activity namespace. This was long
496 overdue, and shall open the path to many future endeavors.
499 - #132: Java: a process can not shut down its own host
500 - #220: S4U: detached send causes exception if sender terminates soon after sending
501 - #261: Document the parameters of parallel execution's constructor
502 - #300: [s4u] BarrierPtr is missing
503 - #314: SMPI args internal cleanup
504 - #316: Fix a bug related to the CPU utilization of multi-core VM
505 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
506 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
507 - #325: Turning off a host has different behavior on sleeping actors and computing actors
509 ----------------------------------------------------------------------------
511 SimGrid (3.21) October 3. 2018
513 The Restarting Documentation (TRD) Release.
516 - Convert the user manual to Sphinx for improved usability.
517 Unstable version now hosted on https://simgrid.frama.io/simgrid
518 Many glitches remain; stay tuned.
519 - Restrict the Doxygen reference API to the installed header files.
520 (documenting all internals this way is overkill + doxygen scales badly)
521 - New tutorial on S4U (highly inspired from the old MSG one, but with
522 a git repository to fork as a starting point, and a docker image)
523 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
524 - The Developper manual is still to be converted (not compiled until then)
525 - Some parts are still missing in this conversion (such as the
526 platform and deployment sections) while others were blindly converted
527 and would need more love (such as the configuration flags).
528 Things will certainly further improve in the future releases.
531 - Fully reimplement auto_restart mechanism that was utterly broken
532 (fix #22, #131 and #281 that were all very related)
533 - Implement semaphores (fix #303 and #291)
534 - When creating an actor from a function and its parameters,
535 move-only parameters are not allowed anymore, as it would prevent
536 the actor to be restartable if its parameters are consumed on run.
537 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
538 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
539 IO operations that can be started, waited for, or canceled as a regular
541 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
545 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
546 - New host variable: 'core_count'
549 - Remove xbt_os_thread_specific features
550 - Remove portability wrapper to condition variables
551 - Remove xbt_os_thread_yield()
554 - MPICH collective selector now mimics MPICH 3.3b
555 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
556 - MPI_Init(NULL,NULL) is now allowed
557 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
558 - Command-line arguments should now be handled properly in Fortran simulations
559 - MPI Attributes and related callbacks should work in Fortran
560 - Apps using C + Fortran should now work
561 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
562 - Support MPI_Error_String
565 - Due to an internal bug, Msg.run() must now be your last line.
566 We hope to fix it in a future release, and we are sorry for the inconvenience.
569 - #22: Process autorestart seem to only work with CAS01 cpus
570 - #56: Feature request: dynamic SMPI replay
571 - #93: simgrid should not eat --help
572 - #111: How to change the loopback link characteristics?
573 - #116: Platform cluster radical
574 - #117: Platform cluster/host order
575 - #129: Replay_multiple on steroid
576 - #131: Java : autostart does not work
577 - #143: Setting a breakpoint at a given time
578 - #235: Network energy models should be integrated
579 - #264: Add ptask L07 resource tracing
580 - #271: Dynamic and manual replay of SMPI traces
581 - #279: Breakpoints option uses time deltas instead of absolute time
582 - #281: Daemonized actor no longer daemonized after an auto-restart
583 - #282: TIT convertor erase traces when using absolute path in the trace list file
584 - #285: segfault when a s4u actor kills itself with msg process tracing activated
585 - #286: Pajé traces are not exposing the number of cores
586 - #287: Command-line parsing should stop when encountering '--'
587 - #288: MPI_Init(NULL, NULL)
588 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
589 - #290: Method to check if a mailbox has a message ready to be consumed
590 - #291: [s4u] semaphores do not seem available yet
591 - #294: Alltoall collective implementation of mpich
592 - #295: Starting a Host as OFF is not supported
593 - #296: DTD too permissive
594 - #299: Add a s4u barrier example
596 ----------------------------------------------------------------------------
598 SimGrid (3.20) Released June 24. 2018
600 The proxy snake_case() release.
603 - The whole API is now using snake_case() instead of mixing it with camelCase.
604 Compatibility wrappers in place for one year.
605 - We took the opportunity to increase the consistency of the API.
606 Please report any remaining glitches.
607 - New function simgrid::s4u::Host::get_actor_count:
608 Returns the number of actors running on a specific host.
611 - SMPI is now tested with ~45 proxy apps from various sources, with none or
612 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
613 - Replay: The replay file has been re-written in C++.
614 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
615 supported. They are stored in the trace and used when replayed.
616 - Basic support of MPI_Cancel. Robustness not guaranteed.
617 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
618 (only for MPI_COMM_TYPE_SHARED).
619 - New option: smpi/privatize-libs, to add external shared libs to be privatized
620 by SMPI. They will be copied locally and loaded separately by each process.
621 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
622 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
623 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
624 with internal index optind with multiple processes. Only works if MPI_Init has
626 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
627 syntax without preprocessing tricks.
630 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
631 behavior tracing is no longer limited to MSG
632 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
633 this means that traces now need to be updated or re-obtained; both Irecv
634 and Isend lines in your traces have now in total 5 mandatory fields:
635 <rankid> <command> <to/from rankid> <tag> <size>
636 To update your traces, it suffices to add a 0 for the tag here.
637 - TIT now also supports waiting for a distinct request via MPI_Wait.
638 Wait/Test now wait for a specific request, not just the last one that was
639 issued. This unfortunately means another update, because we need to
640 identify which request you want to wait for. We do this via the
641 triplet (sender, receiver, tag), which needs to be added:
642 <rankid> <command> <sender> <receiver> <tag>
643 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
644 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
647 - The deprecation of MSG is ongoing (but this should not impact you).
648 Many MSG functions are now simple wrappers on the C API of S4U. If
649 you wish to convert your code to S4U, find the S4U counterparts of
650 your MSG calls in src/msg/msg_legacy.cpp.
651 - MSG can still be used, but won't evolve anymore.
654 - Allow to run the Link energy plugin from the command line with
655 --cfg=plugin:link_energy
656 - Rename Energy plugin into host_energy
657 - Rename Load plugin into host_load
660 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
661 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
662 - Improve performance when handling timeouts of simix synchros.
665 - Config: the C API is now deprecated (will be removed in 3.23), and
666 the C++ API has been slightly improved.
669 - Fix several build issues on OSX.
670 - Move simgrid_config.h to simgrid/config.h (old header still working)
673 - #143: Setting a breakpoint at a given time
674 - #150: Inconsistent event names in SMPI replay
675 - #258: daemonized actors hang after all non-daemonized actors have completed
676 - #267: Linker error on unit_tmgr
677 - #269: SMPI: tracing of MPI_Wait/all/any broken
678 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
680 ----------------------------------------------------------------------------
682 SimGrid (3.19.1) Released March 22 2018
684 The Fixed ABI Release.
686 As you may know, we are currently refactoring SimGrid in deep.
687 Upcoming SimGrid4 will be really different from SimGrid3: modular,
688 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
690 Our goal is to smooth this transition, with backward compatibility
691 and automatic update paths, while still progressing toward SimGrid4.
693 SimGrid remains open during works: The last pure SimGrid3 release was
694 v3.12 while all subsequent versions are usable alpha versions of
695 SimGrid4: Existing interfaces remain unchanged, but the new S4U
696 interface is budding and the internals are deeply reorganized.
698 Since 2015, we work hard to reduce the changes to public APIs. When
699 we need to rename a public library symbol in S4U, we let your
700 compiler issue an explicative warning when you use the deprecated
701 function. These messages remain for four releases, i.e. for one full
702 year, before turning into an error. Starting with v3.15, your can
703 also adapt to API changes with the SIMGRID_VERSION macro, that is
704 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
706 Starting with this v3.19.1, our commitment to reduce the changes to
707 the public interfaces is extended from the API to the ABI: a program
708 using only MSG or SimDag and compiled against a given version of
709 simgrid can probably be used with a later version of SimGrid without
710 recompilation. We will do our best... but don't expect too much of
711 it, that's a really difficult goal during such profund refactoring.
713 The difference between v3.19 and v3.19.1 is that the former was
714 accidentally breaking the ABI of MSG, while the later is restoring
717 S4U and kernel APIs will still evolve until SimGrid4, with one-year
718 deprecation warnings as currently. In fact, cleaning up these
719 interfaces and converting them to snake_case() is one release goal of
720 v3.20. But don't worry, we are working to smooth this upgrade path.
722 In summary, new projects should start with S4U to benefit of the
723 future, but old MSG projects should still be usable with no change.
725 ----------------------------------------------------------------------------
727 SimGrid (3.19) Released March 20 2018
729 The Moscovitly-cold Spring release.
731 SMPI user-visible improvements:
732 - Support of MPI_Type_create_subarray
733 - Unimplemented MPI calls now abort (instead of issuing a warning)
734 Some harmless calls around errhandlers still only issue a warning.
735 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
736 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
737 is now possible (mmap privatization was refactored to this end).
738 Adding new SMPI instances should also be technically doable now.
740 SMPI internal cleanups:
741 - Removed index notion from SMPI -- no more "getPid() - 1"!
742 In tracing files, actors are called rank-<PROCESS-ID> but these are
743 NOT the real ranks (tracing several communicators is not supported yet).
744 - Deprecate smpi_process_index() (will be removed in 3.22)
745 - Better testing on classical proxy apps, through an external project:
746 https://github.com/simgrid/SMPI-proxy-apps/
749 - Use a graphical TOC to make it easier to find the documentation you need
750 - Revamp the MSG tutorial
753 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
754 - Remove parameter of MSG_process_killall().
755 Resetting the PID was bogus anyway (several actors could have the same PID).
758 - Execution->setHost() can be called after start(), to migrate it.
759 - Comm::test_any() is now implemented.
760 - s4u::Actor now has onCreation() and onDestruction() signals.
761 - Install some kernel header files for the users' plugins and more.
762 Warning, their API is really not stable yet.
763 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
764 include/simgrid/kernel/routing/*Zone.hpp
767 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
769 Build System and other cleanups:
770 - Remove unused run-time parameter "tracing/onelink-only".
771 - SimGrid now works with intel compilers (MC still buggy)
772 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
773 - tesh kills subprocesses on timeout (but not on windows)
776 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
779 - #194: Feature request: simgrid::s4u::Comm::test_any()
780 - #245: migrating an actor does not migrate its execution
781 - #253: Feature Request: expose clusters as objects
782 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
783 - #255: Tesh broken on Windows
784 - #256: Modernize FindSimGrid.cmake
785 - #257: Fix (ab)use of CMake install
787 ----------------------------------------------------------------------------
789 SimGrid (3.18) Released December 24 2017
791 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
793 MSG IS NOW DEPRECATED. Please use S4U for new projects.
794 - Support for MSG should not be removed from SimGrid before 2020,
795 but future evolutions will be limited to the integration of
796 user-provided patches.
799 - s4u::onDeadlock() signal, run before stopping the simulation.
800 - s4u::Exec: asynchronous executions (abstraction of background
802 - s4u::Host->getLoad() returns the achieved speed in flops/s
805 - Link::name() is deprecated and will be removed in v3.21, use
806 Link::getCname() instead.
807 - Mailbox::getName() changed to return a std::string, use
808 Mailbox::getCname() instead to get a char*.
809 - Storage::getName() changed to return a std::string, use
810 Storage::getCname() instead to get a char*.
811 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
812 to have both allocation and deallocation of the map in user space.
815 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
816 The amount of flops remaining to do is a vector, not a scalar.
817 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
818 name implies on both sequential and parallel tasks.
819 - Both changes fix GitHub's #223 using PR #237 as a basis.
820 Thanks Michael Mercier.
821 - Most examples were converted to S4U and hidden elsewhere at tests.
822 Rationale: we still want MSG to work; we want newcomers to use S4U.
825 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
826 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
829 - Switch to the faster dlopen privatization mechanism by default
830 - Documentation improvments
833 - Live migration is getting moved to a plugin. Dirty page tracking is
834 the first part of this plugin. This imply that VM migration is now
835 only possible if one this function is called:
836 - C/MSG: MSG_vm_live_migration_plugin_init()
837 - C/C++: sg_vm_live_migration_plugin_init()
838 - Java: Msg.liveMigrationInit()
839 For C and C++, "simgrid/plugins/live_migration.h" also has to be
843 - Define class simgrid::xbt::Path to manage file names.
844 - Removed unused functions:
845 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
846 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
847 xbt_graph_export_graphviz()
848 - xbt/str.h: xbt_str_join()
849 - Remove unused datatypes:
850 - xbt/heap.h: use std::priority_queue or boost::heap instead
851 - xbt/swag.h: use boost::intrusive::list instead
854 - New link_energy plugin for the consumption of the links.
855 - All of the operations on files and storage contents have been
856 packaged into a plugin (src/plugins/file_system). The current
857 public interface can be found in
858 include/simgrid/plugins/file_system.h
859 To use these functions you now have to initialize the plugin by
860 calling MSG_storage_file_system_init() just after calling
861 MSG_init() or sg_storage_file_system_init() just after creating
865 - Remove the undocumented/untested tag <include>
868 - Remove viva specific tracing as the tool is no longer maintained
871 - #248: Kill of finished processes leads to segfault
872 - #240: xbt_cond_wait_timeout should gracefully return for C apps
873 - #239: please implement signal s4u::onDeadlock()
874 - #230: segfaults when exit() without run()
875 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
876 - #223: MSG_task_get_flops_amount() not working with parallel tasks
877 - #222: Actor::kill() doesn't really kill and segfaults
878 - #221: odd LMM warning when killing an actor
879 - #120: Memory leak when the processes are forcefully killed
881 ----------------------------------------------------------------------------
883 SimGrid (3.17) Released October 8 2017
885 The Drained Leaks release: (almost) no known leaks despite the tests.
887 Even very long simulations will preserve your memory: our testsuite
888 stresses SimGrid for over 45mn on fast machines for a coverage of
889 over 80%, and there is only one single known leak, of about 4kb.
893 - Comm.detach(): start and forget about asynchronous emission. A cleanup
894 handler may be given to free resources if the comm cannot be completed.
895 - this_actor::send(mailbox) is now mailbox->put()
896 - New: simgrid::s4u::Comm::wait_all()
897 - New: Host.execute() for remote executions.
900 - Unused option network/sender-gap is removed.
903 - Deprecate MSG_task_isend_with_matching(): This unused feature
904 really complicates our internals. Will be removed in v3.20.
907 - Improved context termination. It is now possible to free resources, even
908 when a process is forcibly killed.
909 - ContextBoost: add support for Boost versions above 1.61.
912 - Bring back run-time option --cfg=exception/cutpath to remove exception
914 - Removed unused functions:
915 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
916 xbt_str_rtrim(), xbt_str_trim().
917 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
920 - Removed header files obsolete since SimGrid 3.12:
921 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
923 - #3: SD_exit should be made optional
924 - #120: Memory leak when the processes are forcefully killed
925 - #159: Threading test regression in Actor refcounting
926 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
927 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
928 - #186: Actor::killAll() segfaults if some process is blocked on wait()
929 - #191: VM migration and pstate
930 - #192: Updating the energy consumptions of all hosts crashes with VMs
931 - #195: All actors have PID=0 in the logs
932 - #204: Sometimes segfault with thread contexts and mmap privatization
933 - #222: Actor::kill() doesn't really kill and segfaults
934 - #225: Actor::kill() doesn't really kill when victims are doing a join()
936 ----------------------------------------------------------------------------
938 SimGrid (3.16) Released June 22. 2017.
940 The Blooming Spring Release: developments are budding.
943 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
944 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
945 - Java JAR file should be finally fully working on Mac OSX too.
947 XML platforms: Switch to platform v4.1 format.
948 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
949 - <zone> can be used as a synonym for the now deprecated <as>
950 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
951 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
952 - <actor> can be used as a synonym for the now deprecated <process>
953 - state_file and avail_file periodicity is now easier to express
954 (check the documentation) the old behavior should still work.
955 * Storage: not backward compatible, but it seems that nobody ever used it
956 - Remove attribute 'content_type' of <storage_type>: was never used
957 - Make attribute 'model' of <storage_type> optional: for future usage
958 - Remove Bconnection model property: was never (in)validated. Replaced by
959 the maximum of the read and write bandwidth as a resource constraint.
962 - New and Backwards Compatibility break:
963 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
964 When one wants to get the list of tasks whose states have changed during a
965 simulation round, s/he has to allocate and free a dynar and use it as
966 argument to this function. The former SD_simulate (double how_long)
970 - Allow multicore VMs, with the correct sharing (unless you overcommit)
971 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
974 - New function to update the consumption of all hosts at once.
975 - Fix the model for multi-core hosts, linear in the amount of busy
976 cores with an abnormality for pIdle (see paper).
977 BUG: Still not satisfactory for ptask on mono-cores.
980 - The netzone are now available from the MSG API.
981 The old names still work, but are now deprecated.
984 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
985 - New option: smpi/keep-temps to not cleanup temp files
986 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
987 mallocs mode is used, allows to change the size of the fake file used
988 (default 1MB), to potentially limit the number of mappings for large runs.
989 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
990 - Fortran ifort and flang compilers support
991 - New RMA calls supported (experimental):
992 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
993 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
994 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
995 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
996 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
997 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
1000 - New: this_actor::isMaestro() returns whether we are in kernel mode.
1001 - Behind the scene, ActivityImpl are now automatically refcounted.
1002 Removing this nasty bottleneck will greatly simplify our internals.
1005 - Replay: New function xbt_replay_action_get():
1006 Retrieve the function previously associated to an event type.
1007 - DROPPED FUNCTION: xbt_str_varsubst()
1008 - DROPPED MODULE: strbuff. We don't need it anymore.
1009 - DROPPED MODULE: matrix. We don't need it anymore.
1010 - DROPPED MODULE: lib. We don't need it anymore.
1012 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1014 SimGrid (3.15) Released March 22. 2017
1016 The Spring Release: continuous integration servers become green
1017 We fixed even the transient bugs on all target architectures:
1018 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1019 Windows, FreeBSD, NetBSD.
1021 - Rename NetCards to NetPoints.
1022 This was intended to help NS3 users, but that's not a netcard.
1023 That's a point in the routing algorithm, let's avoid wrong simplifications.
1026 - New: MSG_process_yield(). Stop and yield to other processes.
1027 - New: MSG_process_daemon(). Daemon processes are automatically killed
1028 when the last non-daemon process terminates
1029 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1030 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1031 to make room for the upcoming network energy plugin.
1032 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1035 - Ensure that an actor can kill itself with Process::exit()
1036 - Kill the obscure NativeException. Nobody want to survive the issues
1037 it denotes, so use JniException that is a RuntimeException (not to
1038 be caught explicitly).
1039 - Partial bug fix in initialization. SimGrid flags on command line were
1040 consumed at C level but stayed in the original Java String[] args.
1041 This could mess users' args[i] if SG flags were not put at the end of
1043 The SimGrid flags are now removed from the Java arguments. However,
1044 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1045 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1046 It might be if you have --log or --cfg flags in the command line.
1047 - Fix numerous memleaks all around the place. In particular, around VMs.
1051 - s4u::onPlatformCreated: right before the simulation starts
1052 - s4u::onSimulationEnd: right after the main simulation loop
1053 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1054 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1055 event from the availability_file changes the avail speed.
1056 - Links are now usable from s4u
1057 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1058 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1059 - New examples: The conversion of MSG examples to S4U has begun
1060 - Actors: create, kill, migration, and suspend
1061 - Applications: master-worker and token-ring
1062 - Action replay: communications and storage
1063 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1066 - Backwards Compatibility breaks
1067 - The SD_route_get_list and SD_route_get_size functions have been removed.
1068 They are replaced by the unique
1069 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1070 The route from src to dst is built in the links dynar whose size can
1071 be obtained with xbt_dynar_length.
1072 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1073 sg_host_route_bandwidth and sg_host_route_latency.
1074 Macros ensure the backwards compatibility, but you should fix your code
1077 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1078 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1079 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1080 We don't support 100% of it, but it should be enough. Contact us if not.
1081 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1082 naive and probably inaccurate)
1083 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1084 mapped onto a unique small file using some system magic.
1085 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1086 - Fortran: cleanups, fixes, support of user-added operations
1087 - RMA: MPI_Accumulate are applied in correct order.
1088 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1089 - Better support for MPI_IN_PLACE
1090 - Support for MPI_Win attrs and keyvals.
1091 - Support MPI_Comm_group_create, MPI_Type_size_x
1092 - MPI ops should be properly applied to all allowed types
1093 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1094 algorithm by a non failing one from mpich.
1097 - Kill the fifo data container: we don't use it anymore.
1100 - A <cluster> can now be created with different speed values to represent pstates.
1101 see examples/platforms/energy_cluster.xml for an example.
1103 SimGrid (3.14.159) Released December 28. 2016
1105 The previous release was only a crude approximation of a Pi release;
1106 this one should be a bit better.
1108 - Revert a last minute change that broke on Mac OSX.
1109 - Fix the changelog and NEWS files.
1110 - Improve the documentation, in particular of the routing module.
1111 - Integrate some patches flying around in Debian.
1113 SimGrid (3.14) Released December 24. 2016
1116 * The whole documentation was reworked and reorganized. There is
1117 still much room for improvement here, but we're on it.
1120 * We now need python3 for our testing infrastructure.
1121 * Model checking needs libevent
1124 * Now works on FreeBSD
1127 * Privatization now works on FreeBSD
1129 * Privatization is now activated by default in smpirun
1130 You can turn it off with -no-privatize if something goes wrong.
1132 * Call-location tracing for SMPI.
1133 You can add the exact location (filename / linenumber) of an MPI call to
1134 your trace files and slow down or speed up the simulation between two
1135 consecutive calls by using an adjustment file (see the documentation).
1137 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1138 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1139 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1140 values were ignored in some cases. The timings of these functions can now
1141 be significantly different.
1143 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1144 smpi/running-power is renamed to smpi/host-speed
1146 * smpi/grow-injected-times option to enable or disable multiplication of the
1147 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1148 make simulation less precise (but also much faster).
1150 * smpirun script should be (much) faster for large deployments.
1152 * SMPI tracing: fixed issue with poor matching of send/receives.
1154 * Replay: Fix broken waitall
1156 New functions and features
1157 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1159 Dropped / renamed functions and features
1160 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1161 - MSG_mailbox_is_empty() -> Mailbox::empty()
1162 - MSG_mailbox_front() -> Mailbox::front()
1163 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1164 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1165 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1166 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1167 * Don't pass the free_f parameter to property related functions:
1168 - MSG_host_set_property_value()
1169 - MSG_as_router_set_property_value()
1170 - MSG_storage_set_property_value()
1171 * VM properties. Since msg_vm_t are msg_host_t, just use
1172 MSG_host_get_property_value() and friends
1173 * VM I/O related things:
1174 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1175 - Unimplemented save/restore methods
1176 * MSG_as_router_get_property_value() was redundant with
1177 MSG_environment_as_get_property_value().
1178 - Removed MSG_as_router_*propert*() functions
1179 - Added MSG_environment_as_set_property_value() for consistency
1180 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1181 Well, they are still there for now, but deprecated with a warning.
1182 Please switch to xbt_dict_new_homogeneous() before this is removed
1184 * Task affinity. Its intended behavior (that was very badly tested
1185 and probably not really working) was deceiving what most users
1186 would have hoped here.
1187 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1188 * Option network/coordinates is now useless and should be dropped.
1191 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1194 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1195 * Added Dragonfly topology support
1198 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1199 EiBps, EBps, Eibps, Ebps for bandwidth.
1200 They may become useful to some lucky ones.
1203 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1204 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1205 * ex/async_waitAll: new example, on asynchronous communications
1208 * Memory usage should be decreased for simulations with a large number
1209 of processes. This also helps for SMPI.
1211 SimGrid (3.13) stable; urgency=low
1213 The Half Release, a.k.a. the Zealous Easter Trim.
1215 - We removed half of the lines, that were mostly experimental cruft.
1216 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1217 (not counting blanks and comments -- according to openhub.net)
1218 - The internals are now compiled in C++ (and will soon be clean C++)
1219 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1220 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1222 * Backwards Compatibility breaks
1223 - Removed Lua simulation bindings (switch to C or Java for that).
1224 Lua can still be used to describe platforms
1225 - Removed Java kernel plug-ins.
1226 Will be reintroduced after the ongoing major internals reorg.
1228 - the following functions were removed.
1229 They were too specific and should be reimplemented in a generic
1230 way, with filter function.
1231 - MSG_task_listen_from_host
1232 - MSG_mailbox_get_count_host_waiting_tasks
1233 - MSG_mailbox_put_with_timeout was removed.
1234 Please use MSG_task_send_with_timeout instead.
1236 - the SD_application_reinit function was removed. It has been a noop for a while.
1237 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1238 replaced by a more flexible mechanism.
1239 - The following functions thus do not exist anymore
1240 - SD_workstation_get_access_mode
1241 - SD_workstation_set_access_mode
1242 - SD_workstation_get_current_task
1243 - Basic estimation functions have been removed but can easily be replaced
1244 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1245 - SD_workstation_get_computation_time => amount / sg_host_speed()
1247 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1248 Use VM.getSpeed()*load/100 for the legacy behavior.
1250 - option enable_tracing was removed. It was not doing anything for a while.
1251 - In the ModelChecker:
1252 - the model-checker now ptraces the model-checked process which means
1253 you cannot use a debugger on the latter anymore (we might make this
1254 optional in the feature);
1255 - removed soft-dirty page tracking;
1256 - remove model-checked side snapshot management,
1257 MC_snapshot() and MC_compare_snapshot();
1258 - keep the MC_cut() function as a stub (it was not really working
1259 in the previous release).
1262 * All options are consistently in kebab-case. Old names are kept as alias.
1265 * Switch to platform v4 format.
1266 - Rename from 'power' to 'speed' the attributes describing the amount of
1267 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1268 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1269 - In <host> and <link>, attributes availability and state are gone.
1270 It was redundant with state and availability traces, and with peak values.
1271 - In <cluster>, attributes availability_file and state_file are gone.
1272 It was too complex and unused.
1273 - Kill <gpu>. Was not doing anything.
1274 - The DOCTYPE points to the right URL:
1275 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1276 (the file at this address now documents the changelog since its v1)
1277 - A warning is emitted for unit-less values (they are still accepted).
1278 - speed. Default: 'f' or 'flops'. Also defined:
1279 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1280 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1281 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1282 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1283 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1284 - latency. Default: 's' second. Also defined:
1285 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1287 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1288 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1291 * s4u::Host is now the preferred public interface to the Host features.
1292 sg_host_* functions are C bindings to the exact same behavior
1293 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1296 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1297 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1298 They were deprecated since a few years, and probably did not even compile anymore.
1301 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1302 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1303 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1304 encouraged to update their codes. A list of the modified functions can be found at the end of
1305 include/simgrid/simdag.h
1308 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1309 the lambda or closure passed as a parameter will run in kernel mode.
1310 All the callback functions should be rewritten to that interface at some point.
1313 * Reorganizing and cleaning the internals all around the place.
1316 * Remove old default barrier/bcast buggy algorithms (see #18407)
1317 * Various bug fixes to handle more codes
1318 * Remove the need for the --foreground option of smpirun (it is still
1319 accepted for backward compatibility).
1322 * Kill the setset data container: MC don't use it anymore.
1323 * Kill the queue data container: it made more sense with GRAS.
1324 * Kill the xbt_peer_t data type: it's useless without GRAS.
1325 * Kill rm_cb feature of config sets: it was never useful.
1326 * Kill graphxml parsing feature. It was not used.
1327 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1329 - xbt_dynar_sort_strings(), when the content is char*
1330 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1331 They throw exceptions on invalid input;
1332 * C++ support for declaring CLI flags (simgrid::config::Flag);
1333 * class for abstracting different signal backends (simgrid::xbt::signal).
1334 with no external dependencies (we need very simple signals).
1337 * refactoring and cleanup of the code;
1338 * ongoing process to cleanly separate the model-checking algorithms
1339 from the code model-checking support.
1341 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1343 SimGrid (3.12) stable; urgency=low
1345 The Facelift Release.
1348 * Require g++ v4.7 at least to not speak prehistorical C++.
1349 * Require Boost 1.48 (for signal2 component).
1350 * Java must be version 7 at least when activated.
1351 * Builds on Windows again (including Java bindings).
1352 * Tracing is now always enabled (no way to turn it out)
1353 * Remove GTNetS. It was not working anyway.
1354 * Various cleanups in the cmake scripts.
1355 * Move headers around to sort them out on installed systems:
1356 - instr/instr.h -> simgrid/instr.h
1357 - instr/jedule/* -> simgrid/jedule
1358 - simdag/datatypes.h was removed
1359 - simdag/simdag.h -> simgrid/simdag.h
1360 - msg/datatypes.h was removed
1361 - msg/msg.h -> simgrid/msg.h
1364 * Interface improvement:
1365 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1366 - Sanitize the interface in MSG_task_ module:
1367 - Merge two functions that were close enough but misleading:
1368 set_compute_duration(t) -> set_flops_amount(t)
1369 get_remaining_computation(t) -> get_flops_amount(t)
1370 - set_data_size(t) -> set_bytes_amount(t)
1371 get_data_size(t) -> get_bytes_amount(t)
1372 - Massive cleanups in the functions related to the energy
1373 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1374 - New: MSG_host_get_pstate()
1376 - msg/energy/onoff: switching hosts on and off
1378 * Interface improvement:
1379 - Rename Host.isAvail() to Host.isOn()
1380 - Rename Process.currentProcess() to Process.getCurrentProcess()
1381 - Rename Task.setDataSize() to Task.setBytesAmount()
1382 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1384 - #18874: Actually allows the GC to reclaim tasks
1388 - SIMIX_process_throw: raises an exception in a remote process
1389 * Refactoring: Separate sub-modules
1390 - libsmx: the public interface, as libc in a real system
1391 - popping: the strange dance that converts a user request into a kernel handling
1392 - smx_context_*: the virtualization mechanisms that embed the user code
1393 - smx_*: the handling of each simcalls
1394 * Interface cleanup:
1395 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1396 * Rename smx_action_t into smx_synchro_t, making explicit that these
1397 things are used to synchronize processes with their environment.
1398 For example, a communication is a sort of synchronization involving
1399 the communicating processes (that may block until the exchange) and
1400 the platform. The same can be said from computations, etc.
1402 - #18888: segfault when a process ends before its kill_time
1406 - 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.
1407 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1408 - MPI_Type_set_name, MPI_Type_get_name
1409 - MPI_*_c2f and MPI_*_f2c functions
1410 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1411 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1412 - Activate a lot of new tests from the mpich 3 testsuite
1414 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1415 - InfiniBand network model added: Based on the works of Jerome Vienne
1416 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1417 - When smpi/display_timing is set, also display global simulation time and application times
1418 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1419 * Collective communications
1420 - 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
1421 - 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/).
1422 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1424 - Replay now uses algorithms from wanted collective selector
1425 - Replay can be used with SMP-aware algorithms
1426 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1427 - 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.
1429 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1430 - Correctly use loopback on fat-tree clusters
1431 - Asynchronous small messages shouldn't trigger deadlocks anymore
1432 * Energy/DVFS cleanup and improvement
1433 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1434 - new: smpi_get_host_pstate()
1437 - "Full" network optimization flag was broken since Surf++
1438 - Better handling of precision flags in maxmin
1439 - Fix bug causing sometimes "Impossible" errors
1440 - Properly pass cluster properties to included hosts
1441 * Improvement of the Energy plugin.
1442 - Always update the consumption before returning that value
1443 - New property: watt_off to denote the dissipation when the host is off
1444 - New functions getWattMinAt and getWattMaxAt to retrieve the
1445 dissipation of pstates that we are not currently at.
1446 * Java: class NetworkLink renamed to Link
1447 * New function: simcall_process_get_kill_time()
1448 * Massive rename s/workstation/host/
1449 - That's intrusive, but that's good for the project consistency. Sorry.
1450 - Change config option "workstation/model" into "host/model"
1453 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1454 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1455 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1456 - xbt_dirname and xbt_basename for non-POSIX systems
1458 * The model checker now runs as a separate process.
1459 * The model checker runs must now be launched with the new simgrid-mc program.
1460 * Record/Replay: the MC can display a textual representation of a path in the
1461 execution graph. It can then be replayed outside of the model checker.
1463 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1465 SimGrid (3.11) stable; urgency=low
1470 * Normalizing pointers addresses tool for better diff between logs
1473 * Add cloud examples using new VMs
1474 - examples/msg/cloud/two_tasks_vm.tesh
1475 - examples/msg/cloud/simple_vm.tesh
1476 - examples/java/cloud/cloud.tesh
1477 - examples/java/cloud/migration/migration.tesh
1478 * Add java surf examples:
1479 - examples/java/surfPlugin/surf_plugin.tesh
1480 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1481 - examples/java/surfCpuModel/surf_cpu_model.tesh
1482 * Add SMPI+MSG example:
1483 - examples/smpi/smpi_msg_masterslave/
1488 - msg task destroy cancel test
1489 - msg_host on/off test
1490 * Move all tests in testsuite to teshsuite (adding tesh files)
1491 * Restructure teshsuites
1492 - one folder for each kind of test
1493 * Restructure AddTests.cmake
1495 - structure the order of tests (with sections)
1498 * Add virtual machine
1499 - creation of a VM on a PM
1500 - migration of a VM from a PM to another PM
1502 - MSG_process_join(msg_process_t process, double timeout)
1503 - msg_bar_t MSG_barrier_init(unsigned int count)
1504 - int MSG_barrier_wait(msg_bar_t barrier)
1505 - void MSG_barrier_destroy(msg_bar_t barrier)
1506 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1507 * New option "msg/debug_multiple_use" to help debugging when a task is used
1510 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1512 - Increase code coverage in test suites
1516 * Protect context stacks against stack overflow. The number of protected memory
1517 pages allocated on the top of each stack (1 by default) can be configured
1518 with the new command line option --cfg=contexts/guard_size:<value>.
1519 * Simcalls are now generated by a python script that
1520 - generates files included by SimGrid
1521 - checks that all the functions exist, and proposes prototypes
1523 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1524 host_set_data, host_get_data
1526 - simcall_process_join(smx_process_t process, double timeout)
1527 * Fix bug where sleeping processing could not be suspended.
1530 * Translate surf models from C to C++
1531 - Generic classes for all models: Model, Resource, Action
1532 - A generic interface for each kind of model (CPU, Network, Storage
1533 Workstation, WorkstationVM)
1535 * Translate surf routings from C to C++
1536 * Add callbacks using sigc++ or boost::signals2
1537 - Add callback functions for resource creation/destruction
1538 - Add callback functions for action state change
1539 - Handle Energy as a plugin
1540 * Replace swag by boost::intrusive
1541 * Add new routing models for clusters. For documentation, see
1542 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1543 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1544 parameters for cluster tag
1545 - Fat trees, with topology="FAT_TREE" and
1546 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1547 - see examples/platforms/torus_cluster.xml and
1548 examples/platforms/fat_tree_cluster.xml
1549 * More documentation
1552 * Hostfiles support host:nb_processes construct to deploy several processes on
1554 * Collective communication algorithms should not crash if used with
1555 improper number of nodes and report the error.
1556 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1557 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1558 MPI_Dims_create, MPI_Cart_sub are supported.
1559 * New interface to use SMPI programmatically (still depends on MSG for
1560 some parts, see examples/smpi/smpi_msg_masterslave):
1561 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1565 * Global variables privatization in MPI executables is now performed at runtime
1566 with the option smpi/privatize_global_variables (default:no).
1567 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1568 dynamic libraries loaded by the application are not privatized (static
1569 linking with these libraries is advised in this case)
1572 * Options defined in XML work correctly now.
1575 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1576 copied into simgrid.jar or not (ON by default). Use this option if you want
1577 to reduce the size of the installed simgrid.jar, *and* the native libraries
1578 are kept installed elsewhere.
1579 * Surf binding with SWIG (code generated in maintainer mode only):
1580 - plugin to handle callbacks
1581 - CPU model only for the moment
1584 * Supernovae build mode is definitively removed. It was used to improve
1585 inlining and inter-module optimizations. It is nowadays superseded by
1586 link-time optimizations commonly available in compilers.
1587 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1588 * Add boost dependency for surf++
1589 * Add new macro for tests
1590 - ADD_TESH(name <tesh_args>)
1591 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1595 - xbt_bar_t XBT_barrier_init(unsigned int count)
1596 - int XBT_barrier_wait(xbt_bar_t barrier)
1597 - void XBT_barrier_destroy(xbt_bar_t barrier)
1598 * Make the xbt_os_time module public
1600 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1602 SimGrid (3.10) stable; urgency=low
1604 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1607 * Reintegrate Java to the main archive as desynchronizing these
1608 package is not acceptable anymore (Java is now considered stable)
1609 * Add explicit synchronization facilities through semaphores
1610 * Bug fix: Task.setDataSize() only changed the C world, not the value
1611 cached in the Java world
1614 * Dramatically change the way files are handled. API and internals changed, but
1615 this part of MSG was not considered as production grade either.
1616 * Add explicit synchronization facilities through semaphores
1617 * Add a new function MSG_host_get_process_list()
1618 * Preliminary DVFS support (see examples/msg/energy/ for details)
1621 * SMPI is now included directly in the libsimgrid as the windows
1622 linker doesn't force us on splitting it anymore.
1623 * Improvements of the SMPI replay tool:
1624 - Most of the collective communications are now rooted in the same process as
1625 in the original application.
1626 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1627 used until now). Multiple data types can now be used in a trace.
1628 - The replay tool now supports traces produce either by TAU or a modified
1630 - Bug Fix: the compute part of the reduce action is now taken into account.
1631 - Gatherv collective is now supported
1632 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1633 of smpirun outputs time independent traces for the current run. One file
1634 is created per process. If too many processes are simulated, this behavior
1635 can be changed to one file for all processes by using the
1636 tracing/smpi/format/ti_one_file flag
1637 * smpirun generates the host file if needed (with given host count and platform)
1638 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1639 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1640 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1641 --cfg=smpi/coll_selector:(mpich/ompi)
1642 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1644 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1645 happening inside a collective SMPI call.
1646 * Fix the behavior of complex data types handling.
1647 * Make MPI_Wtime another synchronization point to take computations into
1649 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1650 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1651 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1652 (no privatization of global variables yet)
1653 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1656 * Verification of liveness properties is now available for SMPI applications
1657 (in addition to MSG applications)
1658 * Bugged examples using SMPI in examples/smpi/mc/
1659 * Add --cfg=model-check/visited option. Allows the verification of infinite
1660 programs. Detection of loops in the execution thanks to the system state
1661 comparison and reduction of the state space to explore. Can be combined with
1662 DPOR for safety properties.
1665 * Allow to change SimGrid configuration (see --help) within the code
1666 thanks to SD_config() as it can be done in MSG.
1667 * Add a new function SD_task_set_amount() upon user request.
1670 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1671 * Remove rule based routing (no more PCRE dependency)
1672 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1673 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1674 * Add a loopback_bw and loopback_lat options to cluster tag.
1675 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1676 routing. Now peers are encapsulated in an AS and have their own private
1677 router but this is transparent.
1680 * Our own implementation of getline is renamed xbt_getline, and gets
1681 used even if the OS provide a getline(). This should reduce the
1682 configuration complexity by using the same code on all platforms.
1683 * New type: xbt_cfg_elm_boolean.
1684 * Allow to use yes/no for boolean configuration options in the command line.
1685 * Allow to disable SimGrid cleanups at exit from command line option.
1686 There are situations where one may want a simulation to end with an exit.
1687 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1688 annoying when scripting around the simulator. Adding a
1689 --cfg=clean_atexit:no allows to circumvent this issue.
1692 * Lots of memory leaks were corrected in this release.
1693 There is no known memory leaks anymore, in all of our 600+ tests.
1694 * New command line option --version, to get SimGrid version information.
1695 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1697 * Supernovae builds are deprecated, and expected to be removed in the next
1700 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1702 SimGrid (3.9) stable; urgency=low
1704 The Grasgory release: GRAS is really dead now.
1706 * Complete overhaul of the internal host structures scheme.
1709 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1710 was considered as experimental and badly maintained since then.
1711 * Keeping it was thus a trap to our potential users, that could take
1712 it instead of MSG or SMPI by mistake despite is pity state.
1713 * GRAS seems to have very few users (if any), and no one volunteered
1714 to maintain it further. It also induces a lot of XBT code (for
1715 portability sake), that must be maintained too.
1716 * For all these reasons, we killed GRAS. If someone wants to revive it
1717 in the future, don't cry, our git history still remembers of GRAS.
1720 * Major overhaul. Merge our documentation again as time proved that
1721 splitting it was really not helping our users.
1722 * Further improve the developer documentation to help newcomers
1723 hacking on SimGrid itself. The user documentation (and in
1724 particular, the beginner documentation) is still in a sorry state.
1727 * Now works on Windows too!
1728 * Much more extensive test suite, from MPICH
1731 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1732 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1733 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1734 * Bug fix: let task be scheduled when the last dependency to be solved is
1735 a control dependency.
1736 * Remove SD_load_environment_script function.
1737 Use the C sg_platf function if you want to declare a platform
1741 * New function: MSG_process_get_number()
1742 * Old function documented: MSG_config()
1743 * Remove MSG_load_platform_script function
1744 Use the C sg_platf function if you want to declare a platform
1748 * Change the default value of the TCP_gamma constant (maximal size of TCP
1749 congestion window) to a more realistic 4MiB value. If you notice changes in
1750 your simulation results, you can fall back to the previous 20k tiny window
1751 by adding --cfg=network/TCP_gamma:20000 on command line.
1752 * (Hopefully) fix a bug wrt periodic availability and state traces
1753 * Bug fix: use default values at start when first event in availability/state
1754 trace is not at time 0.
1757 * remove the "new_" part of function name sg_platf_new_trace_connect
1758 (resulting in sg_platf_trace_connect), since it does not create
1762 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1763 use case where it's really mandatory, and maintaining it was a pain in
1765 * New: xbt_fifo_search(), search an item with a user-provided
1766 comparison function instead of dumb pointer comparison.
1769 * Fix the lua deployment:
1770 Use `simgrid.init_application()` before deployment instead of
1771 `simgrid.msg_register_application()` after.
1774 * Transfer the tracing files into the corresponding modules.
1776 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1778 SimGrid (3.8.1) stable; urgency=low
1780 The "we are told that some people want to also *install* the simgrid
1783 * Add missing file "tesh.1" to the archive.
1785 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1787 SimGrid (3.8) stable; urgency=low
1789 The Psssshiiiit release: SimGrid jumps into the Cloud.
1792 * Add an experimental interface to manipulate VMs. They are mainly
1793 process groups with very few intrinsic semantic, but they should
1794 allow you to build the semantic you want easily.
1795 * New function: MSG_host_set_property_value()
1796 * New function: MSG_process_on_exit(). To clean memory in all cases.
1797 * Bug fixes that made the host (and link) failures unusable.
1798 * Add a way to auto-restart process when the host in which they are
1799 executing comes back (ON_FAILURE="RESTART" on deployment file,
1800 MSG_process_auto_restart_set).
1801 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1802 please stop using the old ones, they are DEPRECATED.
1804 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1805 Please use MSG_init() instead. (reducing the amount of entry
1806 points in the library helps us).
1807 * Make it impossible to link against the wrong version of the lib
1808 * Deprecate MSG_clean(). No need to call it anymore.
1809 * Function MSG_get_host_number() is not deprecated anymore.
1812 * Split the doc into a user guide and a reference guide.
1813 * Start a developper guide to help people hacking on SimGrid.
1816 * Enable tracing by default. This modules rocks you should use it.
1817 * Remove option custom_flags. Now use environment variables CFLAGS
1819 * Use default cmake things to detect lua instead of home grown ones.
1820 * New option "enable_mallocators" to disable mallocators, for debugging
1821 purpose ("on" by default).
1824 * Bug fixes around the resource failures: don't let the processes
1825 survive the host they are running onto.
1826 * Add an interface to auto-restart processes when the host in which they are
1827 executing comes back.
1828 * Ensures that SIMIX_clean is called automatically. It's not part of
1829 the public interface anymore (bindings should be updated).
1832 * Bug fix for when SD_Simulate is called with a positive value: be careful
1833 when comparing doubles. Sometimes they are different for non significant
1835 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1836 parallel task whose initial work is distributed among host according
1837 to the Amdahl's law. Such tasks are created with a parameter alpha
1838 that corresponds to the non-parallelizable part of the computation.
1839 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1840 between two sets of workstations assuming a 1D block distribution (each
1841 workstation owns a similar share of data) on both sides.
1843 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1844 Data redistribution will be automatically scheduled once parent and child
1845 are both scheduled. The filling of computation_amount and
1846 communication_amount structures is now done seamlessly thanks to the chosen
1848 * New function SD_workstation_dump to display various information
1849 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1850 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1851 bandwidth on the route onto which the task is scheduled and the amount of
1853 To divide the nominal bandwidth by 2, the rate then has to be:
1854 rate = bandwidth/(2*amount)
1855 * Compute tasks that have failed can now be rescheduled and executed again
1856 (from their beginning)
1857 * Increasing source code coverage (src/simdag is now covered at 95.8%
1861 * Re-implement time-independent trace replay using SMPI (at the
1862 smpi_smp_* level) instead of MSG. This should replace
1863 examples/msg/actions/actions.c
1864 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1865 hindexed and structs)
1866 * Implement the exchange of non-contiguous data.
1867 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1868 * Correct behavior of smpi/sender_gap and set its default value to 0
1869 * Add option to asynchronously send small messages to allow better
1870 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1871 specifies the size in bytes under which messages will be asynchronously sent.
1872 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1873 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1876 * New C interface to define a platform: XML is now optional.
1877 For more info, please check include/simgrid/platf.h.
1878 * New interface to define random platforms from the C:
1879 For more info, please check include/simgrid/platf_generator.h and
1880 examples/msg/masterslave/masterslave_platfgen.c
1881 * Export a sg_cmdline dynar containing all the arguments we got from
1885 * Two new tracing options for adding comments to trace file so you
1886 can track your experiments (see --help-tracing for details).
1887 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
1888 * Adding the SimGrid version that generated the trace file as a comment.
1889 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1890 * Fix to avoid key clashes on Paje links
1891 * Other minor fixes related to the Paje specification
1894 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1895 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1896 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1897 * Greatly improve the robustness of mmalloc to user errors (such as
1898 using an area after freeing it, or freeing it twice)
1900 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1902 SimGrid-java (3.8.1) stable; urgency=low
1904 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1905 Warning: all methods are visible, but only some of them are
1906 implemented so far. Check the source (src/jtrace.c)
1907 for further information.
1908 * New module: org.simgrid.msg.File (SimGrid File management functions)
1909 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1910 * Change the meaning of Process.restart: now restart the process from
1911 the beginning, like MSG_process_restart in C.
1912 * Add Process.setAutoRestart: handling of process restart when failed
1914 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1915 you to retrieve the properties of the processes/hosts
1916 * Deprecate Msg.clean(): you can just forget about it now.
1917 * New function Process.getCount(), that only works when compiling
1918 with the not yet released version 3.9 of the C library.
1920 * New context factory based on Coroutines. It mandates a modified JVM
1921 but then, the simulations run about five times faster, and there is
1922 no limit to the amount of processes (beside of the available memory).
1924 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1926 SimGrid (3.7.1) stable; urgency=low
1929 * Restore the prototype of MSG_process_create_with_environment() to
1930 the pre-3.7 situation by removing the kill_time argument.
1931 * Add a MSG_process_set_kill_time() function instead.
1934 * Fix weird behaviors when dealing with parallel tasks.
1937 * Simgrid is now built as a dll.
1938 * Simgrid-java now works on Windows.
1939 * Simgrid-Java is now included into Windows package.
1942 * First pre-build package for MacOSX.
1945 * Fix compilation when using MSG_USE_DEPRECATED.
1946 * Fix some compilation issues on Macs and Windows.
1947 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1948 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1950 We discovered that the Lua console is broken, but we are missing the
1951 manpower to fix it right now. The problem existed in 3.7 too, so we
1952 are not blocking the release for that. Sorry if you depended on this
1953 feature, any help would be really welcome.
1955 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1958 SimGrid-java (3.7.1) stable; urgency=low
1960 The "Java aint got to be bloated and slow" release
1963 * Various internal cleanups and performance improvement
1964 Simulations are expected to run up to twice faster or so
1965 * Make Process.kill(process) an instance method, not a static one
1966 * User processes are not java.lang.Thread subclasses.
1967 This breaks the compatibility (sorry), but previous API was
1968 brain-dead, making it impossible to have non-trivial
1969 initializations in the process constructor.
1970 * Require a full constructor per Process sub-class.
1971 Kinda breaks the compatibility (sorry), but this allows a much more
1972 efficient way to launch the processes at simulation startup.
1973 * Do not embeed our version of semaphores, java 1.5 can be considered
1974 as sufficiently prevalent for us to not dupplicate its features.
1978 * Add examples for almost every part of the API
1979 We spotted and fixed a lot of bugs in the process
1981 * New module: asynchronous communication API
1982 * New function: Process.sleep()
1983 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1984 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1985 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1987 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1989 SimGrid (3.7) stable; urgency=low
1991 The "spring cleanups (before the next Big Project kicks in)" release.
1994 * We can specify the SMPI latency/bandwidth factor with command line
1995 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1996 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1997 You can also use the "config tag" from platform file by adding this line
1998 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1999 (see "example/platforms/tag_config.xml" to use "config tag").
2000 Note that the command line supersedes the platform file configuration.
2001 * Change the correction factors used in LMM model, according to
2002 the latest experiments described in INRIA RR-7821.
2003 Accuracy should be improved this way.
2004 * Use the partial invalidation optimization by default for the
2005 network too. Should produce the exact same results, only faster.
2006 * Major cleanup in surf to merge models and split some optimization
2007 mechanisms from the core of the models. As a result you can now
2008 specify which model to use (e.g., --cfg=network/model:LV08
2009 --cfg=cpu/model:Cas01) and which optimization mode to use
2010 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2011 Incompatible combinations should err at initialization. See
2012 --help-models for the list of all models and optimization modes.
2013 * The CLM03 workstation model was dropped for simplicity because it
2014 used the deprecated CM02 network model. Use default instead.
2015 * Rename the TCP_gamma configuration option to network/TCP_gamma
2016 * Rename the coordinates configuration option to
2017 network/coordinates, and document it
2018 * Use now crosstraffic keyword instead of the terribly misleading
2019 fullduplex keyword. It is activated by default now in the current
2020 default model, use --cfg=network/crosstraffic:0 to turn it off.
2021 * Ongoing refactoring the model parsing to make XML files optional
2022 See include/simgrid/platf.h for details (still to be completed)
2025 * Major overhaul of the documentation. Almost instructive now :/
2026 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2027 functions and friends. This interface was considered as
2028 deprecated since over 2 years, it's time to inform our users that it is.
2029 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2030 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2031 use these (crappy) functions in your code.
2032 These functions will be removed soon. Stop using them now.
2033 * Deprecate MSG_get_host_{table,number}
2034 Implement MSG_hosts_as_dynar() instead.
2035 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2036 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2039 * Stabilize the parallel execution mode of user contexts
2040 * Introduce configuration variables to control parallel execution:
2041 - contexts/synchro: Synchronization mode to use when running
2042 contexts in parallel (either futex, posix or busy_wait)
2043 - contexts/parallel_threshold: Minimal number of user contexts
2044 that must be part of a scheduling round to switch to parallel
2045 execution mode (raw contexts only)
2046 * Fix bugs that prevented to use suspend/resume along with
2047 synchronization structures.
2048 * Fix bugs in process termination that lead to invalid memory access
2049 in very specific conditions.
2052 * Introduce a parallel mode for the models (controlled by surf/nthreads
2053 configuration item). In our tests, running the models in parallel
2054 never lead to any speedups because they are so fast that the gain
2055 of computing each model in parallel does not amortizes the
2056 synchronization costs, even when ultra fast futexes are used.
2057 This is released anyway because YMMV.
2060 * Performance boost by using a swag internally to compute the set of
2061 tasks that are finished and should constitute the return value of
2065 * Enable it by default now that it is considered rather stable.
2068 * Documentation of the tracing functions.
2069 * Performance gains when tracing categorized/uncategorized resource
2070 utilization by avoiding calls to get route when updating resource
2071 variables. LMM constraints are being used instead.
2072 * API changed to set task categories. Use MSG_task_set_category instead
2073 of TRACE_msg_set_task_category, and SD_task_set_category instead
2074 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2075 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2076 * Examples for link user variables added (see at examples/msg/tracing/)
2077 * Deprecated function TRACE_msg_set_process_category completely removed
2078 * Trace header updated according to the latest Paje file format
2079 * Tracing network lazy updates, no longer obligate users to use full updates
2080 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2081 * Experimental: let user code declare/set/push/pop application states for hosts
2082 * API changed to allow the manual creation of graph configuration files
2083 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2086 * Improve the API of Lua MSG bindings, using the Lua spirit.
2087 * Each simulated process now lives in its own Lua world (globals are
2088 automatically duplicated). It helps writing simulators. Will allow
2089 to run Splay programs within SimGrid in the future.
2090 * Add a Chord example in Lua, equivalent to the MSG one.
2093 * Start the implementation of a solution to express temporal
2094 properties, not only local assertions. This is still an
2095 experimental work in progress, stay clear from it to be safe.
2099 - Add new runtime parameters --help-logs and --help-log-categories
2100 to display information about supported logging parameters and
2102 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2103 don't exists anymore.
2104 * Mallocators: allow value NULL for the reset function.
2106 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2107 create homogeneous dictionaries, where all the elements share the
2108 same free function. Non homogeneous dictionaries will be
2109 deprecated in the next release.
2110 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2111 - Multi-level dictionaries are deprecated.
2113 - new function xbt_dynar_search_or_negative() that is useful when
2114 you have less than 2 million elements in your dynar and don't
2115 want of the extra complexity of catching exceptions when the
2116 element is not found.
2118 - Make xbt_os_thread module (for thread portability) public.
2119 Documentation is still to come, sorry.
2121 - Cleanups and simplifications to make it maintainable again.
2122 - Exotic features (such as memalign and valloc) were removed.
2123 - The metadata were extended and improved so that the
2124 model-checker becomes able to explore and inspect the heaps.
2125 - This may induce a performance drop when enable_model-checking is
2126 ON in cmake (even if it's not used in the simulation), but it is
2127 necessary at this point to get MC working.
2129 Turn model-checking OFF if simulation performance matters to you.
2130 Not enabling it at runtime is not enough, disable it in cmake.
2132 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2135 SimGrid (3.6.2) stable; urgency=low
2137 The "Not coding new stuff allows to polish old things" release.
2140 * New bindings to the NS3 packet level simulator (experimental)
2141 * Use the raw (efficient) execution contexts instead of the sysv
2142 (portable) ones when possible.
2143 * libpcre is now mandatory in any cases since not using it led to
2144 severe performance loss and possibly other issues
2145 * Update the XML platforms:
2146 - G5K: include the latest machine in Nancy
2147 - GridPP and LCG: new platforms
2148 * Documentation was partially updated, at least (more to come)
2150 Bug fixes, cosmetics and small improvements
2151 * Free terminated processes before the end of the simulation to avoid
2152 exhausting the memory of users having very dynamic amount of
2154 * Bug fix and cosmetics about canceling non-running tasks
2155 * Bug fix about the dot loader's issues when using libcgraph
2158 * Create an installer for windows with nsis (amd64 and win32)
2159 - Add an hello world project to illustrate simgrid project creation.
2160 - Embed libpcre into the Simgrid installer to avoid
2161 its compilation burden
2162 * The raw execution contexts should work on Apple now
2163 * Port to Windows 64 bits
2164 - Sysv contexts now have an implementation for this arch
2165 - GRAS communication features now support this arch
2166 * Drop support for borland compiler on windows
2167 - this code was not maintained, and we kinda depend on gcc nowadays
2168 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2169 * Fix portability issue on unstable ubuntu: linker became picky on
2172 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2174 SimGrid (3.6.1) stable; urgency=low
2176 The "Oops, we broke Macs too" release
2179 * Fixed contexts detection so that raw ones are used when possible
2180 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2181 leads to a strange error, with user code segfaulting sometimes when
2182 the generated code is not perfectly aligned (which is not
2183 controllable from the user side, depends on the amount of code)
2186 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2188 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2190 SimGrid-java (3.6) unstable; urgency=low
2193 * Split of every thing from simgrid v3.5 into a separate package.
2195 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2197 SimGrid (3.6) stable; urgency=medium
2199 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2202 * Bindings now constitute their own package, separated from the main one.
2203 Rationale: reduce our maintenance nightmare by reducing the module coupling
2204 They will soon be released on their own on gforge.
2206 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2207 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2209 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2210 * It's not quite deprecated for now because we have no replacement,
2211 but it may soon become the case.
2214 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2215 * Fortran: New user-level cache variable to store the rank of the running
2216 process. This improves performance by an order of magnitude.
2217 * C: New coccinelle script to automatically locate and modify global and
2218 local static variables.
2219 * Improved SMPI network model with a sender-side gap to account for multiple
2223 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2224 only say if a communication is finished, no matter whether it succeeded or
2225 failed. You can call MSG_comm_get_status() to know the status of a finished
2227 * New function MSG_task_dsend() to send a task and detach it. When a
2228 communication is detached, you are never notified of its success or failure
2229 and the memory is released automatically once it is finished. This function
2230 is useful when you don't care about the end nor the success of a
2232 * Change the prototypes of action replay. Sorry for inconvenience,
2233 but this is really more efficient this way (and to adapt your code,
2234 you just have to fix the initialization, that shouldn't be too long)
2235 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2236 ever managed to do anything useful with it.
2237 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2238 takes a dynar of communications. It returns immediately and gives the
2239 index of a finished communication (if any).
2240 * New example: a basic implementation of the Chord P2P algorithm.
2243 * New model for multi-core CPUs. You can now use the core attribute to
2244 precise the number of cores of a host. This is a basic model. Every
2245 process running on the host receives at most the power provided in
2246 the DTD (throughput<=power). Total throughput of process cannot exceed
2248 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2249 router linked by an up-link and a down-link (possibly asymmetrical).
2250 This kind of pattern allows to easily build last-mile model style platforms.
2251 Aggregating such patterns in a rule-based AS is thus the technique of
2252 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2253 * New model for Vivaldi routing. We transformed the Vivaldi network model
2254 into a Vivaldi routing model (based on the rule-based model). This allows to
2255 combine Vivaldi based latencies with last-mile platforms.
2258 * Added a check for NaN of IEEE754 infinite in the double entries of
2260 * Introduce a new context factory "raw", highly inspirated from the
2261 ucontext factory, but using manually crafted functions in assembly to
2262 do the work in an efficient manner.
2263 * Allow to change the used context factory at run time, not only at
2264 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2265 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2266 contexts at runtime (only with raw contexts or ucontexts).
2267 * Completely rewrote this module to allow parallel execution of user
2268 processes. Use --cfg=contexts/nthreads:N to execute user processes
2269 with N parallel threads (the default is 1, meaning no parallelism).
2270 * Allow to decide dynamically between sequential and parallel modes.
2271 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2272 processes in parallel only when their number is greater than or equal to P
2274 * Added a check for NaN of IEEE754 infinite in the double entries of
2278 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2279 won't output the state of processes when interrupted with Ctrl-C
2280 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2281 NULL-terminated array. This may solve backward compatibility issues
2282 due to the change to return type of SD_simulate. See also:
2283 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2284 * Add new macros with variable number of arguments.
2285 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2286 - in xbt/asserts.h: xbt_assert
2287 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2288 - in xbt/ex.h: THROWF and RETHROWF.
2289 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2291 * Change xbt_die() to accept a format string with arguments, just like printf.
2292 * New data structure: xbt_lib_t, like a dict but more general and with better
2296 * New configuration options
2297 Options triva/categorized and triva/uncategorized can be used to generate
2298 graph configuration files for Triva visualization tool.
2299 * Configuration option tracing/platform is renamed to tracing/categorized
2300 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2301 * New TRACE_user_link_variable interface:
2302 User provides the name of the link and the tracing variable to attach to it
2303 * the declaration of tracing categories must be done after the environment
2305 * simpler tracing interface, just one way to declare categories
2306 TRACE_category or TRACE_category_with_color, it is up to you
2307 * links in the trace file are again identified by their names
2308 * trace contains the full platform hierarchy exactly as declared using the ASes
2309 * Options tracing/msg/[task|process]:1 groups the process by hosts
2310 for both cases, tasks and processes must have names that are unique during the simulation
2311 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2312 * The experimental option tracing/msg/volume is deprecated
2313 its functionality may be reincorporated if needed
2315 The tracing generates a trace file with unordered timestamped events,
2316 because of the way the core simulator (surf) works. A script available
2317 at the tools directory (fix-paje-trace.sh) can be used to put the events
2318 in order. We have changed the tracing so it can generate ordered timestamped
2319 events in the final trace, but depending on the simulator (and how much time
2320 is simulated) that can lead to a huge memory utilization. It is deactivated
2321 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2323 Build Infrastructure
2324 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2325 - We are trying hard to keep the API stable, but it may happen that
2326 some things change (we're a research project after all, not a
2327 nuclear plant operating system). If such things should happen, you
2328 could rely on that macro to adapt.
2329 - current value: 30600 for 3.06.00, aka 3.6
2330 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2331 a number that can be compared with SIMGRID_VERSION.
2332 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2333 assertions and verbose/debug logging events are disabled at compile time.
2335 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2337 SimGrid (3.5) stable; urgency=medium
2340 * New feature to any SimGrid-based simulator: Model-Checking
2341 Check SIN#1 for more details.
2344 * New Model SMPI (three-interval linear regression for correction factors)
2345 See RR-7426, available at http://hal.inria.fr/inria-00527150
2346 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2347 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2348 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2349 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2350 * See also src/smpi/README
2355 - Tracing API changes: TRACE_start and TRACE_end should not be called
2356 by user-code. They are automatically called by simulators created
2357 with SimDAG, MSG and SMPI if the toolkit is compiled with
2358 tracing_enabled=ON. Categories declaration and utilization remain the
2359 same for MSG and SimDag.
2360 - A function was added to the tracing API to declare categories with
2362 - TRACE_category_with_color (char *category, char *color)
2363 where color must be in the following format
2364 "%f %f %f", red, green, blue
2365 and red, green, blue are float values in the interval [0, 1]
2366 - User can specify NULL as color parameter, or continue calling
2367 TRACE_category (cat)
2368 On that case, the tracing system will define random colors
2369 - The following command-line options are supported:
2370 --cfg=tracing/filename:msg.trace
2371 --cfg=tracing:1 (activate tracing, needed to use others)
2372 --cfg=tracing/platform:1 (categorized resource use)
2373 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2374 --cfg=tracing/msg/task:1 (task creation)
2375 --cfg=tracing/msg/process:1 (process creation, migration)
2376 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2377 --cfg=tracing/smpi:1 (SMPI interface tracing)
2378 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2379 - examples of examples/msg/tracing updated
2381 - DAXLoader and DOTLoader functions can generate tasks with categories
2382 - A new function to attribute a category to SD tasks:
2383 TRACE_sd_set_task_category (SD_task_t task, char *category)
2384 * Tracing the MPI interface implemented by SMPI
2385 - Collective operations are traced with states
2386 - Point-to-Point operations are traced with states/links
2387 - Tracing activated by a parameter "-trace filename" passed
2388 to smpirun during execution (considering that simgrid
2389 is compiled with tracing enabled)
2390 - To run the simulation with gdb, the simulator
2391 accepts --cfg=tracing/smpi:1 to trace SMPI
2392 - tesh files to check if smpi tracing is ok
2393 - See examples/smpi/NAS/DT-trace
2394 * GTNetS tracing re-worked
2395 - adaptation to the tracing system of GTNets to cope
2396 with modifications regarding the fullduplex mode
2397 - new tesh files to check if gtnets tracing is ok
2400 * Asynchronous communications through the functions:
2401 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2402 * New function: MSG_load_platform_script()
2403 to make possible using a lua script instead of XML files to set up platforms
2404 * New function: MSG_set_function
2405 to associate functions to processes, used when bypassing the parser
2406 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2408 Platforms: Add some more examples in examples/platforms
2409 * Grid'5000: see www.grid5000.fr
2410 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2413 * Change the XML format. This is a very important modification. SimGrid
2414 3.5 introduces a new hierarchical format based on the notion of
2415 Autonomous Systems. Compatibility with old format is ensured
2416 through the perl script provided in the install bin directory
2417 bin/simgrid_update_xml.
2418 It is now possible to build platforms with specific routing
2419 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2420 connect several platforms together. We will try to provide soon
2421 set of realistic platforms exploiting these properties (have a
2422 look at examples/platforms/ for the moment).
2423 * Take the opportunity of the XML format change to be a good XML citizen:
2424 rename link:ctn to link_ctn and similar changes (also dealt with by
2426 * Add a new routing scheme (rule-based) using regular expressions. It
2427 enables to have an extremely low memory footprint when the
2428 underlying routing is simple and can be compactly described. You
2429 need to have libpcre4-dev (perl regular expressions) installed if
2430 you want to use this routing scheme.
2431 * Revive the cluster TAG and allow to easily and efficiently (both in
2432 term of memory and speed) connect clusters together. Have a look
2433 at teshsuite/simdag/platforms/ to see how this can be done. With
2434 this tag, you can create clusters with thousands of tasks at no
2435 cost (have a look at examples/platforms/).
2436 Note: clusters are implemented as ASes, so there is no need for an
2437 enclosing AS tag if you have only one cluster in your platform.
2438 * Add new generic functions in the public interface that allows the user
2439 to call SURF 'create_resource' methods from your code (same
2440 functionality as the XML bypass mechanism but with a much lighter
2442 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2443 that uses a piecewise linear approximation to produce better
2444 results when exchanging small messages.
2445 * Add a new parameter to handle correctly full duplex link and account
2446 for interferences between uplink and downlink communications
2447 (activate with --cfg=fullduplex:1).
2450 * Rename the SD_READY (all dependencies are satisfied and task is
2451 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2452 dependencies are satisfied) state.
2453 This prevents a confusion between the notion of "ready to schedule"
2454 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2455 simulated" (SD_RUNNABLE) used by the simulation kernel.
2456 * Change the way a task is considered as ready. Instead of removing
2457 dependencies when a task is done, a counter is decreased. This way,
2458 it is always possible to reach ancestors thanks to the
2459 SD_taks_get_parents function (even after the end of the simulation.)
2460 * Change the return type of SD_Simulate from (SD_task_t*) into
2461 xbt_dynar_t. This function was in handling a dynar internally and
2462 converted it into a NULL terminated array for historical reasons.
2463 * New function SD_dotload(char*) to load a DAG described in dot
2464 format. This loader and the corresponding examples require the
2465 installation of the graphviz library.
2466 * Fix a bug in the management of tasks of size 0 in the surf network
2467 models. This problem was only visible with SIMDAG and you should
2468 thus disregard results produced with earlier versions if you
2469 relied on this feature (some tasks were blocked because of this).
2470 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2471 even though your applications were doing only point-to-point
2472 communications and sequential computations. Now you can really use any
2473 model you want (of course, if you create real parallel tasks, which are
2474 not implemented in most models beside ptaskL07, this will abort).
2475 * Add an example that schedules a DAX on an heterogeneous platform
2476 using a Min-Min strategy.
2477 * New function SD_workstation_get_current_task() that returns the kind
2478 of task currently running on a workstation in the sequential access
2480 * Raise some warnings when unexecuted tasks remains at the end of the
2481 simulation. This is usually caused by cycles in the DAG.
2484 * New function: SIMIX_process_set_function() called by MSG_set_function
2485 * Change the underlying waiting queue in semaphores so that a process
2486 can wait on several of them simultaneously (as in waitany).
2487 * Fix the way to handle tokens in semaphores so that all access patterns
2488 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2489 * kill the dirty pimple SIMIX_message_sizes_output()
2490 Please use (proper) visualization instead
2493 * New data container: setset (set of sets of elements)
2494 * New module: mmalloc (mapped malloc, allowing to have several
2495 independent segments of malloc)
2496 * New function: xbt_dict_cursor_set_data()
2497 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2498 * New function: xbt_dynar_is_empty()
2499 * New function: xbt_fifo_get_last_item()
2500 * Fix xbt_dynar_shrink(): use the right element size.
2501 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2502 uninitialized areas during expand.
2503 * Fix semaphores: previous implementation was severely broken.
2504 * Use library init/fini functions for our initialization.
2505 - you can use logs and other feature as soon as you want in your
2506 code (even before the xbt_init / MSG_init)
2507 - xbt_exit is now a no-op and produce a warning when used.
2510 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2511 benefit from the latest improvement to the simulation kernel.
2512 * Kill measurement sockets for now. If you rely on them, sorry. This
2513 release is not for you. This feature will be reintroduced in the
2514 future, but we cannot delay the release any further.
2515 * New function: gras_msgtype_get_name().
2516 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2517 * Fix (at last) the pmm example: it should not randomly fail anymore.
2519 Build chain: bug fixes and overall polishing
2520 * Cmake is now stable enough. Hence, we killed the autotools.
2521 * Port to windows ( TM :)
2522 * Fix the 'make install' target.
2523 No need to use 'make install-simgrid' anymore
2524 * Introduce a 'make dist' target compiling a *source* archive
2525 'make package' compiles a binary archive
2526 * Compile java files only on need
2527 * Add --cd and --setenv command line options to tesh
2528 * Out of source builds are not fully supported yet, but we are close
2529 * Enable supernovae and optimization flags by default for our users
2532 * Add layer to set up environment directly from lua, without XML.
2533 * The effect of gras_stub_generator can be achieved through
2534 lua too (check examples/gras/console/ping_generator.lua)
2536 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2538 SimGrid (3.4.1) stable; urgency=low
2540 The "Polishing easter eggs is probably a good idea" release.
2541 This is a bug fixes release only.
2545 * Fix a bug preventing the tasks from begin garbage collected.
2548 * Fix a bug occurring when a host involved in a communication fails.
2549 This was not detected properly by the other peer involved in the
2550 communication. Now, it's reported as a network error.
2553 * Warn the user about loop dependencies in data flow of DAX files
2554 * Obey the control-flow dependencies of DAX files
2557 * Add option "enable_smpi" allowing to not compile SMPI.
2558 Probably useful for the (Mac) users experiencing a build error here
2559 * Improve the detection of lua5.1 and ruby1.8
2561 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2563 SimGrid (3.4) stable; urgency=low
2565 The "Easter in Cargese" release. Also known as (major changes):
2567 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2568 ~> bindings were greatly improved
2569 ~> new tracing infrastructure for better visualization introduced
2571 * the "Welcome to configury modernity" release.
2572 ~> we switched from autotools to cmake, and improved our cdash
2575 A more detailled list of changes follow (full detail in svn log).
2577 Java Bindings: Various Cleanups
2578 * (install java-gcj-compat-dev on debian-like to use them)
2579 * Remove put/get: no need to export deprecated interface in Java
2580 Use send/receive instead.
2581 * Cleanup the examples and add a README per directory
2582 * Remove example autoDestination (that's the only way to go now)
2583 * Remove example explicitDestination (was a plain copy of basic)
2584 * Make JniException a runtime exception, so that there is no need to
2585 declare the fact that you may encounter such a beast. I guess that
2586 nobody will ever want to survive such error.
2587 * Create specific errors for each MSG case of failure:
2588 host failure, transfer failure, timeout, task cancelled
2589 * Cleanup the exceptions that may get thrown by each function
2590 * Other internal cleanups in Java bindings. Performance still bad :/
2591 Ruby and Lua Bindings: create them
2592 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2593 * That's new and great, you should try them out.
2594 Same functionalities than Java bindings, only even less polished
2596 * Kill the useless "rate" argument of SD_task_get_execution_time()
2597 Everyone used to provide -1 as a value, it was not used, and the
2598 semantic of a possible use wasn't even clear.
2599 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2600 either as comm costs or compute costs to mean that there is no
2601 such thing for that specific task.
2602 * Add a SD_task_set_name() function
2603 * Fix SD_task_unschedule() on typed tasks
2604 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2605 * In DAX loader, accept useless 'level' attributes to <job> since
2606 LIGO DAGs have them (seem to be to ease graphical representation).
2608 * Add an example masterslave_mailbox.c using send/receive and not
2609 the deprecated put/get interface.
2610 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2611 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2613 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2614 (declare MSG_USE_DEPRECATED to still have the old name)
2615 * Add a MSG_task_set_data() function
2616 * About trace replay (see examples/msg/actions):
2618 - Allow to work with splitted trace files for each process
2619 Give the specific trace file as argument of each process,
2620 and call MSG_action_trace_run(NULL)
2621 You can still have one merged file for all processes.
2622 - Fix implementation of collective operations
2623 * Allow task_execute() on 0-sized tasks (closes #10063)
2625 * This is the first release of SimGrid where SMPI is not considered
2626 beta anymore (even if some corners should still be improved)
2627 * Port over the new SIMIX_network submodule (internal refactoring)
2628 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2629 * Implement more missing elements of the standard:
2631 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2633 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2634 MPI_Scan() MPI_Get_processor_name()
2635 - Added implementation of missing case for Alltoall (warning: it's
2636 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2637 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2638 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2639 SMPI_MPI_Allgatherv()
2640 * Bug fixes include:
2641 - MPI_Waitsome() was broken
2642 - Allow relative includes in smpicc
2643 - Command line cfg argument 'reference_speed' was ignored...
2644 - Some functions did not properly lead to auto-benching of user code
2645 - smpicc passes -O2 by default (just like openmpi one)
2647 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2648 * Bug fixes about timeouts during communications
2649 * add SIMIX_message_sizes_output() as a pimple to write to file the
2650 amount of messages per size. Use gnuplot to get histogram.
2651 Pimple because that's the only user-visible function of simix,
2652 defined directly in xbt.h (irk, sorry)
2654 - Add a SIMIX_sem_get_capacity() function
2655 - Fix interactions with processe resume/suspende
2656 - release_forever() was stupidly broken
2657 - Fix SIMIX_display_process_status() for processes in a semaphore
2658 - Make SIMIX_sem_block_onto() user-visible
2659 * Refactoring context stuff:
2660 - Use pseudo-OOP for better modularity
2661 - reimplement SIMIX_process_kill() without process_schedule() so
2662 that the latter can take as invariant that it is called from
2664 - Merge context_start into context_new for sake of simplicity
2666 * Add a Vivaldi network model, coded live during SUD'10 ;)
2667 * Rename configuration variables to start a hierarchy:
2668 o cpu_model -> cpu/model
2669 o network_model -> network/model
2670 o workstation_model -> workstation/model
2671 * New configuration variables:
2672 o network/bandwidth_factor: correction to bandwidth
2673 o network/latency_factor: correction to latency
2674 o netwotk/weight_S: correction to the weight of competing streams
2675 * Add a long description to the models, that users can see with such
2676 argument on the command line: --cfg=cpu/model:help
2677 * --help-models display the long description of all known models
2679 * config: add the ability to set a default value after registration
2680 Does not override any previously set value (e.g. from cmd line)
2681 * dict: allow to have integer key and data.
2682 When so, you need to use the following functions
2683 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2684 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2685 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2686 In contrary to regular dicts, the key is not malloced before copy.
2687 Mixing scalar and regular elements in the same dict is not tested
2689 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2690 Tracing for Visualization:
2691 * SimGrid is now instrumented in order to generate a trace file for
2692 visualization analysis: to use it, need to compile SimGrid with the
2693 "tracing" option enabled, and instrument the program using SimGrid with
2694 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2695 (among other functions).
2696 * The instrumentation only traces the platform utilization for now
2697 * Documentation to use the tracing functions and how to analyze the
2698 traces with the Triva tool is written.
2699 * More information about: SimGrid FAQ (in the section Tracing Simulations
2702 * We moved to cmake as default build system. Autotools support will
2703 be dropped soon. Check the FAQ for more info about how to use it.
2704 * Greatly improved our cdash/ctest interactions
2705 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2706 * Added memory checking tests with valgrind; lot of memleak fixing.
2707 This may be the first release of simgrid with so few memory issues
2708 * Added code coverage tests.
2709 Our coverage is still improvable, but at least we see it on cdash.
2711 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2713 SimGrid (3.3.4) stable; urgency=low
2715 The "Desktop Grid needs love too" release (also called Xmas release).
2717 Models improvements:
2718 * Major speedup in the maxmin system solving by using lazy evaluation
2719 Instead of solving completely the maxmin system at each iteration,
2720 only invalidate (and recompute) the modified parts.
2721 This new feature is enabled in default models but you can try to
2722 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2723 * Cas01 IMproved as default CPU model
2724 This CPU model is the same Cas01 model, but it uses the
2725 maxmin-selective-update flag and a heap structure to manage
2726 actions on SURF kernel.
2727 It reduces the complexity to find the next action to finish and,
2728 consequently, it's faster than the old Cas01.
2729 This is the new default CPU model (Cas01).
2730 * Rename the old Cas01 model to Cas01_fullupdate
2731 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2732 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2733 * CpuTI (CPU Trace Integration)
2734 A new CPU model whose objective is simulate faster when using
2735 availability trace files.
2736 Instead of using a full featured, over engineered maxmin system for
2737 CPU modeling, this model does the pre-integration of traces files
2738 to calculate the amount of CPU power available, and so, executes
2739 faster than the old CPU models.
2740 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2741 * Use LV08 as default network model since it gives better accuracy
2742 for small messages and shouldn't change things for big ones.
2743 Use --cfg=network_model:CM02 to get the previous behavior.
2746 ******************************************
2747 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2748 ******************************************
2749 * The new CPU model may changes simulations!
2750 The point is that events occurring at the exact same timestamp
2751 are not scheduled in the same order with the old and new
2752 version. This may be enough to completely change the execution
2753 of simulations in some cases.
2754 * The new network model will change simulations!
2755 This new model is more realistic than the previous one, so you
2756 should consider redoing your old experiments with this model.
2757 Sorry for the inconvenience.
2760 * Introduce the supernovae compilation mode
2761 When compiled that way, the whole SimGrid (or almost) is put in a
2762 single compilation unit and compiled in one shoot.
2763 This is to help gcc which has difficulties to inline stuff from one
2765 The speedup seem to be above 15%, althrough more tests are needed on
2766 amd64 to confirm that gain.
2769 * Port of MSG's mailbox on top of SIMIX network
2770 The put/get mechanism was greatly simplified on the way.
2773 * New SIMIX network module. Provides:
2774 - Mailbox: rendez-vous mechanism to find with who you want to speak
2775 - Synchronous send/recv: easier and hopefully faster since the
2776 logic is handled in the maestro process directly now
2777 - Asynchronous send/recv: you dreamt of it? It's here now
2778 Too bad that nobody cared enough to propagate the change to MSG.
2779 * Add semaphores as SIMIX synchronization mechanism.
2782 * new function SD_daxload(char*) to load a DAX file
2783 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2784 * Introduce typed tasks. Specify its kind and cost at creation.
2785 At scheduling, just give where it should be placed, and the cost
2786 for each involved resource is automatically computed.
2787 Existing constructors so far (more to come of course):
2788 - SD_task_create_comm_e2e() for end-to-end communication
2789 - SD_task_create_comp_seq() for sequential computation
2790 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2791 * new function SD_task_dump() for debugging display
2792 * new function SD_task_dotty(task,FILE*) writing to file the info
2793 about the task in dotty format
2794 * SD_task_dependency_exists() can now cope with having one of its
2795 arguments NULL. If so, it tests whether the other argument has any
2797 * Add getters on list of preceding/following tasks:
2798 SD_task_get_parents(task) and SD_task_get_children(task)
2799 * Add getters on amount of workstations and list:
2800 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2801 * Add getter on task kind: SD_task_get_kind(task)
2802 * Update the start_time and finish_time of tasks on completion/failure
2803 * Bugfix: Remove task from state swags when destroyed
2806 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2809 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2810 dynar with one separate thread per value of the dynar.
2811 * Change the prototype of xbt_thread_create(), sorry.
2812 Added a boolean parameter indicating whether we want to join this
2813 thread (used in SG only for now)
2814 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2817 * GTNetS wrappers should now be usable again (and betterly tested too)
2818 * Fix a major regression from 3.2 where the timeout provided to
2819 MSG_task_put_with_timeout() was used as absolute time before which
2820 the comm should be done.
2821 * Start to fix the <cluster> tag.
2822 - Internal links should be good now (beside of the loopback, which
2823 use the private link instead)
2824 - paths to the external world is still rather broken
2825 - the <route:multi> tag is just broken. Actually that's brain-dead.
2826 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2827 to make it less stupid
2828 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2829 * Fix a source-level compatibility glitch from 3.2: after defining
2830 MSG_USE_DEPRECATED, you can use the old name
2831 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2832 * Allow to compile from the SVN with automake 1.11
2833 * Fix some problems when using the "start_time" tag in deployment XMLs.
2834 * Fix #8569: XBT/synchro.h has redundant declarations
2835 * Fix #8563: MSG return values and exceptions
2836 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2837 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2838 Thanks to Robson Peixoto.
2839 * Fix a vicious bug in dictionaries inducing that some elements were
2840 not freed on xbt_dict_free()
2842 Portability report of this version:
2843 * Main portability targets:
2844 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2845 on (amd64/i386/ia64)
2846 - mac leopard on i386
2847 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2848 but nothing critical.
2849 * Other platforms: windows, AIX and others were not tested for this release
2851 Timing report of this version:
2852 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2853 scenario parameters). From 8h to a few seconds in desktop grid settings.
2854 * Supernovae brings about 25% speedup on i386.
2856 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2858 SimGrid (3.3.3) stable; urgency=low
2860 The "Need for Speed" release.
2862 The timings done to validate the 3.3.2 were faulty.
2863 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2865 The problem was a conversion from a manually handled vector to
2866 xbt_dynar_t on the critical path.
2867 xbt_dynar_foreach calls functions, inducing stack management crap.
2869 We inlined these functions and xbt_dynar_foreach is now breath taking.
2870 We also inlined xbt_swag_belong on the way.
2872 Here are some approximate speedup measurements (on master/slaves
2873 simulations lasting between 10s and 20s each):
2874 3.3.1 -> 3.3.2: about same performance
2875 3.3.2 -> 3.3.3: 40% speedup
2876 3.3.1 -> 3.3.3: 40% speedup
2877 3.3.1 with inline patch -> 3.3.3: 30% speedup
2879 Our reading is that the refactoring which occurred in 3.3.2 made us
2880 suffer much more from the xbt_dynar_foreach low performance, but
2881 once we solved this, this refactoring proved to be very performance
2882 effective. From the 40% speedup, somehow, 10% are due to the
2883 inlining and 30% to the refactoring.
2885 That's a pitty that gcc cannot inline functions placed in other files
2886 alone. We have to choose between:
2887 - break the encapsulation (by putting private data structures and
2888 accessors in headers files to help gcc)
2889 - live with low performance
2890 - switch to a decent compiler such as icc (not quite possible).
2892 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2894 SimGrid (3.3.2) stable; urgency=low
2896 The "Simplicity does not preceed complexity, but follows it" release.
2898 The main contributors of this release were (lexical order):
2899 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2902 * Extract the routing logic into its own object.
2903 (was dupplicated in network.c and workstation_LV07.c;
2904 Allows to implement other ways of storing that info)
2905 => kill now useless network_card concept
2906 - Use dynar to represent routes (instead of void** + int*)
2907 - kill link_set (use surf_network_model->resource_set instead)
2908 - Add a command-line option to choose the routing schema to use
2909 - Add three new models:
2910 * Floyd (shortest path computed at initialization)
2911 * Dijikstra (shortest path recomputed all the time)
2912 * Cached Dijikstra (shortest path computed on need)
2913 All these models where contributed by Silas De Munck, and are
2914 described in his ICCS09 paper.
2916 * Simplify model declaration
2917 (less redirections, less function to write when defining a model)
2918 - Factorize stuff between models:
2921 surf_model_resource_set(model)
2922 surf_model_resource_by_name(model, name)
2923 - Unify the types of models in s_surf_model_t (using an union)
2924 - Embeed fields of common_public directly into s_surf_model_t
2925 - Rename model methods:
2926 action_free ~> action_unref
2927 action_change_state ~> action_state_set
2928 action_get_state ~> action_state_get
2929 - Change model methods into functions:
2930 (model)->common_public->action_use ~> surf_action_ref
2932 * Implement a generic resource; use it as ancestor to specific ones
2933 (allows to kill duplicated code in models)
2934 Drawback: timer command don't need no name nor properties;
2935 workstation_CLM03 don't need no properties
2936 (but I guess we can live with those few bytes wasted)
2938 * Improve the action object model
2939 - implement a constructor avoiding dupplicated code about field
2940 initialization in generic_action part.
2942 * Kill the SDP model: it has an external dependency, is deprecated
2943 in flavor of modern lmm models, and didn't compile since a while
2946 * Relocation of the context module from XBT to SIMIX.
2947 (the context were decoupled from the simix processes, duplicating a lot of code)
2948 => a lot of code was factorized
2949 - less overhead is introduced during scheduling
2950 - simpler API for the context factory
2951 - the logic for process creation,destruction and manipulation was simplified
2952 * Simplification of the s_smx_process_t data structure.
2953 => accesing the simix level data associated to a process is faster now,
2954 and the code is a lot more readable.
2957 * Implement some more MPI primitives:
2958 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2959 -implementation: Bcast: flat or 2-ary tree (default),
2960 Barrier: 4-ary tree,
2962 Allreduce: Reduce then Bcast
2963 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2964 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2965 Alltoallv: flat tree, like ompi
2967 * Add support for optimized collectives (Bcast is now binomial by default)
2968 * Port smpirun and smpicc to OS X
2971 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2972 Shout out if you used it.
2975 * Display the list of still queued messages in SG mode when existing
2979 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2980 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2981 * Add xbt_str_from_file(FILE*)
2982 * Add xbt_dict_get_key achieving a linear reverse search
2983 * Remove the context module
2985 Portability report of this version:
2986 * Main portability targets:
2987 - Linux(debian)/x86/context
2988 - Linux(debian)/x86/pthreads
2989 - Linux(debian)/amd64/context
2990 - Linux(debian)/amd64/pthreads
2991 On these, we still have the eratic breakages of gras/pmm and
2992 amok/saturate_sg reported in previous version. We still think
2993 that the tests are the cause of the fault, not the tested code.
2995 - Mac OSX Leopard/x86/context
2996 Still false negative in tesh autotesting.
2997 Smpi still fails, but this time because readlink does not accept -f
2998 Everything seems to work properly beside of that.
3001 - AIX version 5.3 (only tested contexts this time)
3002 Smpi still fails there because mktemp is not installed.
3003 Everything seems to work properly beside of that.
3005 I managed to compile it for the first time, but several breakages.
3006 Won't delay the release for this exotic platform.
3008 * Windows: it's still lagging behind. If you want to help, please
3011 Timing report of this version:
3012 This version seem to be more than 5% faster than 3.3.1 (on linux
3013 64bits with contextes). The gain is less than expected, we are
3014 investigating this for next release.
3016 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3018 SimGrid (3.3.1) stable; urgency=low
3021 * Implement a --cfg-help to show existing configuration variables
3022 * Build chain do not require doxygen in maintainer mode
3025 * fix a bug on struct sizeof computation, which prevented the
3026 exchange of arrays of structs in some conditions
3027 - added a regression test about this in datadesc_usage
3028 * Allow the exchange of 0-long dynamic vectors.
3029 - for that, use -1 as indicator of dynamic size instead of 0
3030 - This implied to change any size from unsigned long to long,
3031 reducing a bit communication abilities, but I guess that with
3032 64bits being quite common, this is more than enough.
3033 - This also induce a protocol change, thus bumping network protocol
3034 version from 0 to 1 (if we have external users, we have to get
3035 clean on that point too ;)
3036 - added two regression tests about this in datadesc_usage
3037 * Be more verbose when propagating local exceptions
3038 This helps debugging.
3039 * Display the status of simulated processes when receiving SIGINT in
3043 * Allow to control the simulation from a trace file.
3044 New functions MSG_action_register() and MSG_action_trace_run()
3045 The first one allows to associate a function execution to each
3046 kind of action while the second one parses a trace file and
3047 triggers the corresponding actions within the system.
3048 For now, only a toy example is provided in examples/msg/actions
3049 * Add an example of process migration in examples/msg/migration
3050 * Fix a bug in task exchange which broke MSG_task_get_sender()
3051 Add a teshsuite regression test for that.
3052 [Bug: if MSG_task_get_sender() is called after sender exit,
3054 * Fix a bug which prevented suspend/resume to work properly
3055 * Display the status of simulated processes when receiving SIGINT
3056 This fixes a regression of v3.3. due to the introduction of SIMIX
3057 * Bug fixing in failure management:
3058 - trace could not start by a failure at time 0
3059 - failure during communications were not working
3062 * Add SIMIX_process_set_name() to change the name of the current
3063 process in the log messages.
3064 * Store smx_hosts in a dict since we only retrieve them by name
3065 * Move the configuration infrastructure to surf
3068 * Move the configuration infrastructure to surf
3071 * Massive internal cleanups:
3072 - Store internal structures on processes instead of hosts (allows
3073 to have more than one process per host, in addition of being more
3075 - Cleanup the initialization/finalization process
3076 - Kill a whole bunch of unneeded synchronization:
3077 processes run in exclusive manner within the simulator
3078 - Move queues from global tables to process data fields
3080 - now accept -platform and -hostfile arguments
3081 - Pass the right rank value to processes according to the hostfile
3082 * Compile the examples by default, and use them as regression tests
3083 * Implement MPI_Wtime()
3084 * Change the reference speed to a command line option
3087 * TCP_gamma can now be specified as command line option using
3088 --cfg=TCP_gamma:10000000.0
3089 * Change the --surf-path cmd line option into --cfg=path:
3092 * Also include strbuff from xbt.h public header
3093 * xbt_ex_display(): do not free the exception after displaying
3094 This allows to do more with the given exception afterward.
3095 Users should call xbt_ex_free() themselves.
3099 Portability report of this version:
3100 * Main portability targets:
3101 - Linux(debian)/x86/context
3102 - Linux(debian)/x86/pthreads
3103 - Linux(debian)/amd64/context
3104 - Linux(debian)/amd64/pthreads
3105 These targets fail about 1/10 of times on gras/pmm, but we believe
3106 that this is because of the test, not because of simgrid.
3107 amok/saturate_sg fails even more rarely, and the test may not be
3110 - Mac OSX Leopard/x86/context
3111 The test suite still spits tons of errors because some obscure
3112 force prevents us from removing the temporary directories
3113 arguing that they still contain some metadata I've never heard of.
3114 Smpi fails because seq is not installed.
3115 Everything seems to work properly beside of that.
3118 - AIX version 5.3 (both contexts and pthread)
3119 Smpi still fails there because mktemp is not installed.
3120 XML inclusions seems rosty on AIX.
3122 * Windows: it's still lagging behind. If you want to help, please
3125 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3127 SimGrid (3.3) stable; urgency=high
3131 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3132 [Malek Cherier & Mt]
3134 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3137 Until now, GRAS were using MSG as an interface to SURF. It was
3138 quite difficult because both interface have several differences
3139 (MSG channels vs GRAS sockets were the most notable point).
3141 This also opens the gate to SMPI (which should occur soon) and speed
3142 up simulations by to 40% (even if it were not the main goal).
3144 **************************************
3145 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3146 **************************************
3147 The point is that events occurring at the exact same timestamp are
3148 not scheduled in the same order with the old and new version. This
3149 may be enough to completely change the execution of simulations in
3150 some cases. Sorry for the inconvenience.
3152 * Cleanup and upgrade the XML format to push further scalability
3153 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3155 * Improve the testing infrastructure with tesh. Now a very large part of
3156 the code is tested not only by being run but also by checking that the
3157 output match an expected output [Mt].
3159 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3160 is really less memory-demanding, which should allow you to use
3161 larger files in SimGrid [AL].
3163 * Inform valgrind about our contextes, so that it becomes usable
3164 with the default (and more effecient) version of SimGrid
3165 [contributed by Sékou Diakite, many thanks]
3168 * Introduce a listener thread in charge of receiving incoming
3169 messages from the network. It allows to overlap communication and
3170 computation but most notably, it removes some stupid deadlocks due
3171 to the fact that so far, a process could not send and receive at
3172 the same time. This made most non trivial communication schema
3174 * Convert the PIDs from long int to int to match the MSG ones (and
3175 linux ones too) [Mt]
3176 * New function: gras_agent_spawn() to launch a new process on
3177 current host. Only working in simulation for now. [Mt]
3178 * New function: gras_os_hostport() returning a constant form (ie,
3179 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3182 * Make the backtrace of exceptions more human readable [Mt]
3183 * New module: xbt/str [Mt]
3184 a ton of string utility functions (split, join, printf to a newly
3185 allocated buffer, trim, etc)
3186 * New module: xbt/hash [Mt]
3187 SHA1 hashing algorithm (more to come if needed)
3188 * New module: xbt/synchro [Mt]
3189 synchronization tools (mutex and conditions) working the same way
3190 in simulation and in real life (mainly useful for GRAS, but not
3192 * New module: xbt/queue [Mt]
3193 classical producer/consumer synchronization scheme
3194 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3195 (using the classical functions will get serialized) [Mt]
3196 * Make dictionary internal table dynamic. No need to specify its size
3197 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3199 * Make sure the log channels are organized as a tree under windows
3200 (because of ANSI C compatibility issue, any channel were child of
3201 root directly) [Mt].
3204 * Cleaned many thing in surf and fixed a few bugs [AL].
3205 * Add a nice command line configuration mechanism to compose models [AL].
3206 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3207 the previous one (KCCFLN05). It relies on something that looks like
3208 a max-min sharing mechanism but cannot be written as such. A new solver
3209 was thus designed [AL].
3210 * Add a new solver to lmm. Based on Lagrange optimization and
3211 gradient-based descent, it enables to efficiently maximise systems s.a
3213 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3215 This solver enables to propose two new network models for TCP Reno and
3216 TCP Vegas based on Low's work. These models still need to be fully
3217 tested though [Pedro Velho].
3220 * Bug fix in SD_simulate. Now the time bound given as argument is
3222 * Use the new parallel task model (ptask_L07) as default.
3223 * Use the SURF command line configuration mechanism.
3224 * 0-size tasks (for synchronization) should now work.
3226 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3228 SimGrid (3.2) stable; urgency=high
3232 We still experience issues on this platform, but we believe that at
3233 least MSG is usable.
3235 GRAS API BREAKAGE (for simplification purpose, sorry):
3236 * the gras_msgtype_by_name is not used anymore. Instead of
3237 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3238 you can write (and must)
3239 gras_msg_send(toserver, "request", &request);
3240 - If you still want to pass a gras_msgtype_t to the function (to cache
3241 the type and avoid the lookup time), use the gras_msg_send_() variant.
3242 - Impacted functions:
3243 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3244 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3245 * The callbacks are now expected to return 0 when everything went well
3246 (just like the main() function)
3248 GRAS new features and improvements:
3249 * New module mechanism where user code can use per process globals [Mt]
3250 This is similar to gras_userdata_*() functions, but for libraries. It
3251 factorize some code developped over and over in the examples and AMOK.
3252 It has still to be documented and used (only amok/peermanagement is
3254 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3255 loss when they were small enough to fit into the buffer and sent quickly
3256 enough so that they can all get received in one shoot.
3257 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3258 if not found. Use the *_or_null() variant for the old semantic.
3259 * In gras_msg_handle, do not discard messages without callback.
3260 They are probably messages to be explicitly awaited later (ie, proofs of
3261 mis-synchronization in userland since they are sent before being awaited)
3262 No big deal usually.
3263 * gras_socket_meas_send/recv: semantic changed!
3264 The numerical arguments used to be (1) the total amount of data to send
3265 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3266 messages. This was need for the fool willing to send more than MAXINT
3267 bytes on quite fat pipes.
3270 * Do really rename the hostmanagement module to peermanagement. [Mt]
3271 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3272 but this is rather new and this was documented in the module
3273 documentation (poor excuses, I admit)
3274 * Bandwidth measurement semantic changed! This follows the changes to
3275 gras_socket_meas_send/recv explained above.
3278 * A sequential mode has been added to the workstations. When a workstation
3279 is in sequential mode, it can execute only one task, and the other tasks
3280 are waiting in a FIFO. [Christophe Thiery]
3283 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3284 model for SIMDAG. [Christophe Thiery]
3285 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3286 equal to 0, which caused some bad behaviors in
3287 saturated_constraint_set_update. I now use a threshold mechanism like in
3291 * When running manually src/testall, you select specific units [Mt]
3292 testall is the result of our cunit mechanism, and should replace all
3293 the scripty thingy around since bash don't run easily on billware.
3295 * A mallocator system has been added. [Christophe Thiery]
3296 Mallocators allow you to recycle your unused objects instead of freeing them
3297 and allocating new ones.
3299 Documentation update:
3300 * FAQ reworking + New FAQs:
3301 - "Valgrind spits tons of errors!" [Mt]
3302 - "How to repport bugs" [Mt]
3303 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3304 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3305 - Communication time measurement within MSG [AL]
3306 - I experience weird communication times when I change the latency [AL]
3307 * GRAS tutorial [Mt]
3309 - an introduction to the framework and to the used communication model
3310 - an initiatic tour introducing the most proheminent features:
3312 . Lesson 0: Installing GRAS
3313 . Lesson 1: Setting up your own project
3314 o Part 2: Message passing
3315 . Lesson 2: Exchanging simple messages
3316 . Lesson 3: Passing arguments to the processes (in SG)
3317 . Lesson 4: Attaching callbacks to messages
3318 . Lesson 5: Using globals in processes
3319 . Lesson 6: Logging information properly
3320 . Lesson 7: Using internal timers
3321 . Lesson 8: Handling errors through exceptions
3322 . Lesson 9: Exchanging simple data
3323 . Lesson 10: Remote Procedure Calling (RPC)
3324 . Lesson 11: Explicitly waiting for messages
3325 . Recapping of message passing features in GRAS
3326 - A HOWTO section containing:
3327 o HOWTO design a GRAS application
3328 More are due, of course. They will come latter. In the meanwhile, you can
3329 check the examples which are still here.
3331 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3333 SimGrid (3.1) stable; urgency=high
3337 There was a stack corruption somewhere, visible only when optimizing
3338 with these versions. [Vince]
3341 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3342 module that enabled to play with Directed Acyclic Graphs. It is built
3343 directly on top of SURF and provides an API rather close to the old
3344 SG. Some old codes using SG are currently under rewrite to check that
3345 all needful functions are provided. [Christophe Thiery]
3348 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3349 extension of the classical CLM03 model that gracefully handles
3350 failures. This is now the default model for MSG and GRAS. It doesn't
3351 handle parallel tasks yet though. [AL]
3352 * Bug fix: Weights were not correctly set in the network part.
3353 WARNING: This may have resulted in incorrect results with simulations
3354 where there are more than one flow on a given link. [AL]
3357 * After a (long ?) discussion on simgrid-devel, we have decided that the
3358 convention we had on units was stupid. That is why it has been decided
3359 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3360 WARNING: This means that all previous platform files will not work as
3361 such with this version! A warning is issued to ask users to update
3363 A conversion script can be found in the contrib module of the CVS, under
3364 the name contrib/platform_generation/surfxml_update.pl [MQ]
3367 * Bug fix: Processes were started in reverse order, wrt deployment file.
3368 WARNING: if your code relies on this bug, please fix it. [AL]
3369 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3370 being executed on two different locations. [AL]
3371 * Bug fix: Failures are now better supported thanks to Derrick's tests
3372 (there was many failure situations I hadn't thought of and that weren't
3373 correctly handled). [AL]
3374 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3378 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3379 Exception occurring on server-side are propagated back to client (!).
3381 API CHANGE: the callback changed their prototype. Change:
3382 int my_handler(gras_socket_t expeditor, void *payload_data) {
3384 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3385 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3387 * New! function: gras_msg_handleall to deal with all messages arriving
3388 within a given period.
3389 * New! function: gras_socket_server_range to get a server socket in a
3390 range of port numbers (ease to avoid port number conflicts) [MQ]
3391 * New! gras processes display their backtrace when they get a SIGUSR1
3392 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3393 Sweet to debug RL processes [MQ]
3397 - Do not force experiment sizes to be expressed in kb, or it becomes
3398 impossible to measure the latency this way (needs one byte-long tests)
3399 WARNING: this changes the amok_bw_* function semantic. [MQ]
3400 - Implements the link saturation stuff. [MQ]
3401 * Peer management module:
3402 New! module factorizing code that we wrote over and over [MQ].
3405 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3406 - Test units are placed directly into the library code, they get extracted
3407 automatically and placed into the src/testall binary.
3408 - Convert most of the XBT tests to this system.
3409 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3410 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3411 live in the same C code. This required to change a little bit the API
3412 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3413 * New module: sparse graph structure with basic algorithms (this is work
3414 in progress and the API is not considered to be frozen yet). [AL]
3415 * Display more information on backtraces: source line & function names are
3416 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3417 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3418 * Reworked a little bit some #include statements to load only required
3419 headers. Some user code that relied on SimGrid to include stdlib or
3420 stdio may need to include it by themselves. [AL]
3421 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3423 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3424 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3425 but I guess that almost nobody used those parts. [MQ]
3427 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3429 SimGrid (3.0.1) stable; urgency=low
3432 * Unfortunately, I had missed 5 misnamed functions:
3433 xbt_fifo_item_t xbt_fifo_newitem(void);
3434 void xbt_fifo_freeitem(xbt_fifo_item_t);
3435 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3436 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3437 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3438 They're now deprecated. Please use their new versions:
3439 xbt_fifo_item_t xbt_fifo_new_item(void);
3440 void xbt_fifo_free_item(xbt_fifo_item_t);
3441 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3442 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3443 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3445 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3446 * Documentation: xbt_log module unmercifully reworked [MQ]
3447 * Bugfix: there was a problem with the ending of contexts with
3448 the pthread backend. It caused some weird deadlock or behavior
3449 depending on the pthread implementation. [AL]
3450 * Bugfix: get the exceptions raised in the simulator repport where
3451 and why they come from when they are not caught in time [AL, MQ]
3454 * Bugfix: Do repport the error when two non-connected hosts try to
3455 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3458 * Add additionnal checkings on communications. Assert that two
3459 communicating hosts are connected by a set of links... [AL]
3462 * Add additionnal checkings on channel values in communication [AL]
3463 * New: MSG_task_get_source to see on which host a task was generated [HC]
3464 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3465 the number of tasks waiting to be received on channel and sent
3467 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3468 waits for the first task coming from a given host.. [AL]
3470 GRAS new functionnalities: [MQ]
3471 * Enhance the parsing macro to allow the size of multidimentional objects
3472 to be given thru annotations.
3473 * New example (and documentation): Matrix Multiplication a la RPC
3474 (as when I was young!) and fix a bunch of bugs found on the way.
3476 GRAS performance improvements: [MQ]
3478 * Reduce the amount of cbps creation/destruction by making it static to
3479 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3482 * Change libdata to a set so that we can search for stuff by ID (and thus
3483 reduce the insane amount of dict lookups)
3486 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3487 RL communication on top of SG since it's so uneffective.
3488 It may also be used for inter-thread communication in RL, one day.
3489 * Use gras_datadesc_copy() to exchange messages on top of SG
3491 - improve message exchange performance on top of SG
3492 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3493 * Don't exchange on the network the size of the used part of buffer,
3494 instead, specify the possible buffer size to read().
3496 - reduces the amount of read/write calls (one pair per exchange)
3497 - reduces the amount of exchanged data (the size)
3498 - allows to retrieve all arrived data on receiver side, if we don't need
3499 it right now (subsequent read will peek the buffer)
3500 - allows the receiver to proceed with the beginning of the stream before
3501 everything is arrived
3502 - make it possible to build an iov transport (using readv/writev)
3504 - take care of the data with non-stable storage (like stacked data),
3506 * If possible, TCP send uses vector I/O (when writev() is here)
3507 - Don't use it for receive since we send data sizes and data on the
3508 same stream, so we wouldn't be able to chain large amount of chunks
3509 before having to flush the stuff to read the size.
3510 * Rework the transport plugin mechanism to simplify it and reduce the
3511 amount of pointer dereferencement when searching for the right function
3514 * I guess that now, we do almost as few system calls as possible while
3515 doing as few data copy as possible.
3517 To improve it further, we could try to send all the sizes first and then
3518 all the data (to use iov on receiving size), but it's only a partial
3519 solution: when you have 2 dimensional data, the sizes of the second
3520 dimension is data of the first dimension, so you need 3 streams.
3522 I'm not sure the potential performance gains justify the coding burden.
3524 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3526 SimGrid (3.00) stable; urgency=high
3529 * New! Give the possibility to hijack the surf parser and thus bypass
3530 MSG_create_environment and MSG_launch_application. Have a look at
3531 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3533 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3535 SimGrid (2.96) unstable; urgency=low
3540 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3541 This deprecates the xbt_error_t mechanisms.
3542 It modifies (simplifies) all XBT and GRAS API.
3543 MSG API keeps unchanged (exceptions raised by XBT are caught from
3544 within MSG and masked with existing error handling facilities)
3547 * New! Add a FATPIPE model. [AL]
3548 * New! Add a parallel task model. [AL]
3549 * New! Add automatically a loopback interface (in the default
3550 network model) if none was precised.
3553 * Bugfix: MSG_process_resume now works with the current running process.
3555 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3556 * Modification of MSG_task_get_compute_duration. Once a task has been
3557 processed, the value returned by this function is now equal to 0. [AL]
3558 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3559 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3560 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3561 of a m_task. For now, MSG_task_cancel only works with computation
3563 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3564 of the processor (in Mflop/s) regardless of the current load on the
3566 * API Change: use proper naming convention for MSG_getClock and
3567 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3569 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3570 This function changes the priority of a computation task. This priority
3571 doesn't affect the transfer rate. A priority of 2 will make a task
3572 receive two times more cpu power than the other ones. This function
3573 has been added to suit the needs of Nguyen The Loc and hasn't been that
3574 much tested yet. So if it fails, please report it and send me your code.
3576 * API Change: removed all functions and types that were marked "deprecated"
3577 since many months. Renamed MSG_global_init_args to MSG_global_init.
3579 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3581 SimGrid (2.95) unstable; urgency=low
3584 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3585 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3586 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3589 * Bugfix: Allow absolute paths for platform description files [MQ]
3590 * Bugfix: do free the variables after use. Leads to drastic performance
3592 * Implement max_duration (ie, timeouts) on resources [AL]
3595 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3597 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3598 the needed support (timeouts) [AL]
3601 * Implement measurement sockets. You can now get the bandwidth between two
3602 hosts thanks to AMOK (see below). [MQ]
3603 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3604 dynar over the network (yeah) [MQ]
3605 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3606 * simple/static token-ring example. [Alexandre Colucci and MQ]
3607 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3608 in place before [MQ]
3611 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3612 type parsing. "short" and "long" modifiers were also ignored; other
3613 modifier (such as reference level) are still ignored. [MQ]
3614 * Embeed the buffer size within the buffer itself on SG. [MQ]
3615 That way, send() are atomic and cannot get intermixed anymore (at least
3616 the ones which are less than 100k; bigger messages still have the issue)
3617 * Array size pushed by the field, not by the field type (or each
3618 and every long int will push stuff to the cbps) [MQ]
3619 * use select() to sleep since it allows to portably sleep less than one
3622 GRAS (minor cleanups)
3623 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3626 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3627 * type category 'ignored' killed as it was never used and were difficult
3629 * whether a type can cycle or not is now a flag, leaving room for more
3630 flags (as "ignored", if we feel the need one day ;)
3631 * Rename raw sockets to measurement sockets since "raw" has another
3632 meaning in networking community.
3635 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3636 and offers features not belonging to GRAS but that most applications
3637 need. One day, it may be a set of plugins loadable at runtime.
3638 * New module: bandwidth
3639 bandwidth measurement between arbitrary nodes running this module. [MQ]
3641 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3643 SimGrid (2.94) unstable; urgency=low
3645 The first beta release of SimGrid 3 !
3649 * Update the main page and the FAQ. Adding references to gforge.
3652 * Add a gras_os_getpid function.
3655 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3656 * Extend the logs so that they also print PID, hostname, date, ... if
3658 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3659 and kill the old version which were in testsuite/
3660 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3664 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3665 for users. Instead, each programming environment (GRAS, MSG,...) use it
3666 when needed to provide such a feature to users.
3667 Don't shortcut the mechanism or you will also shortcut the virtualization
3668 you need on the simulator.
3672 * Cleanups in configury with regard to compile optimization/warning flags.
3673 Also add -fno-loop-optimize to any powerpc since it's the optimization
3674 killing gcc (< 3.4.0).
3675 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3676 needed by MSG examples complications
3677 * Borrow configury beautifications from PHP
3680 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3681 warning (thanks loris for stumbling into this one).
3682 * Bugfix: stop loading private headers (gras_config.h) from the public
3686 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3687 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3688 RL would imply a DNS resolver, which is *hard* to do in a portable way
3689 (and therefore delayed).
3690 * Implement a real timer mechanism and use it in timing macros. This allows
3691 to avoid rounding errors and get a 0.000005 sec precision in timing
3692 macros. While I was at it, various cleanups:
3693 - allow to declare more than one timed section per file (fix a stupid bug)
3694 - move some private declaration to the right place
3695 - merge conditional execution and timing macros into emulation module
3696 - document the module
3697 - make sure the module cleanups its mess on gras_exit
3698 * Documentation improvements:
3699 - (new) how to compile applications using GRAS
3700 - (new) emulation support (timing macros)
3702 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3704 SimGrid (2.93) unstable; urgency=low
3706 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3709 - Use Paje properly where used. Still to be sanitized properly.
3710 - Portability fix: Add an implementation of the contexts using pthread
3714 - Add xbt_procname(): returns the name of the current process.
3715 Use it to show the current process's name in all logging.
3717 - fix detection of older flex version and the reaction, since we do
3718 depend on modern ones (we use lex_destroy)
3719 - Better separation of SG and RL in the libs: remove all simulation code
3720 from libgras. As a result, this lib is now only 200k when stripped.
3721 Some of the xbt modules may also be duplicated (two sets and such) and
3722 should be cleaned/killed before SG3.
3723 - Insist on using xlC on AIX because of weird problems involving gcc there.
3724 - Cleanup the make remote stuff. This is now done by scripts
3725 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3726 mainly for our private use, but we're working on changing them to user
3729 - Bugfix: flush the socket on close only if there is some *output*.
3730 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3732 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3735 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3737 SimGrid (2.92) unstable; urgency=low
3739 Alpha 3 on the path to SimGrid 3
3743 - New! First try of benchmarking macros.
3744 - New! First try so that gras_stub_generator generate deployment and
3745 remote compilation helpers.
3747 - Bugfix: Initialization fix in msg_test.
3751 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3753 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3755 - API Change: gras_os_sleep to take the amount of seconds as a double.
3756 Accepting an int was error prone since it was the only location where
3757 seconds were coded as such. It leaded to damn rounding errors.
3758 - Bugfix: Hard to belive that timers ever worked before this.
3760 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3762 SimGrid (2.91) unstable; urgency=low
3764 Alpha 2 on the path to SimGrid 3
3768 - Bug fix in the lmm_solver.
3770 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3771 through the function MSG_paje_output.
3772 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3773 - It is possible to bound the rate of a communication in MSG with
3774 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3775 to write it in the changelog).
3776 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3781 - Various cleanups to the autotools stuff
3782 - Begin to move Gras examples to examples/gras/
3783 - Let make distcheck work again (yeah!)
3785 - documentation overhauled using doxygen.
3786 gtk-doc-tools is dead in SimGrid now.
3787 - Automatically extract all existing logging categories, and add the list
3788 to the documentation (long standing one, to say the less)
3790 - Cleanup the known architecture table. Reorder the entries to group what
3791 should be, and use a more consistent naming scheme.
3792 (some of the test dataset are still to be regenerated)
3793 - New! Allow library to register globals on each process just as userdata
3795 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3796 lookup time (for now).
3797 Use it in msg and trp.
3798 This cleans a lot the internals and helps enforcing privacy of the
3799 headers between the gras components.
3800 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3801 - Bugfix: gras_os_time was delirious in RL.
3802 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3804 - Reenable GRAS now that it works.
3806 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3808 SimGrid (2.90) unstable; urgency=low
3810 Alpha 1 on the path to SimGrid 3
3812 * It is a long time since the last release of SimGrid. I'm sorry about
3813 that but as I had told you, I was rewriting a lot of things. I apologize
3814 to those who had been reporting bugs to me and that I had not answered.
3815 If your bug is still in the new version, please tell me. Here is a
3816 summary of the main changes.
3818 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3819 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3820 portability, speed, and a lot more but you'll figure it out later.
3821 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3822 I wanted to release the new MSG as soon as possible and I have
3823 broken GRAS, which is the reason why, for now, only MSG is fully
3824 functional. A laconic description of these projects is available
3825 in the documentation.
3827 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3828 kernel optimized for our needs (called SURF but only the developers
3829 should use it). Hence, MSG is now roughly 30 times faster and I think
3830 that by rewriting a little bit MSG, I could event speed it up a little
3831 bit more. Beside the gain in speed, it is also much easier to encode a
3832 new platform model with SURF than it was with SG. More to come...
3834 * REVOLUTION 3: I have tried to change a little as possible the API of
3835 MSG but a few things really had to disappear. The main differences
3836 with the previous version are:
3837 1) no more m_links_t and the corresponding functions. Platforms are
3838 directly read from an XML description and cannot be hard-coded
3839 anymore. The same format is used for application deployment
3840 description. The new format is described in the documentation.
3841 Have a look in tools/platform_generation. There is a tiny script
3842 that converts from the old platform format to the new one. Concerning
3843 the application deployment format, parsing the old one is tricky.
3844 I think most of you should however be able to convert your files. If
3845 it is really an issue, I can write a C code that does the conversion.
3847 2) the toolbox tbx does not exist anymore. We now have a library
3848 with much more data-structures but without the hash-tables (we have
3849 dictionaries that are much faster).
3851 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3853 *****************************************************************************
3854 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3855 * don't feel like converting the oldies. (Mt) *
3856 *****************************************************************************
3859 Version 2.90: "the long awaited one"
3860 - Finished rewriting and debugging MSG. Rewrote the documentation.
3861 - disable GRAS for now since it needs to be ported to the newest SG
3864 - Finish the port to windows (using mingw32 for cross-compile)
3867 - Main loop and datastructures of SURF. A cpu resource object is
3868 functional. Surf can thus be used to create cpu's with variable
3869 performance on which you can execute some actions.
3871 2004-11-15 Martin Quinson
3872 - Port to ARM. Simply added the alignment and size descriptions. Should
3873 work, but the ARM machines are so slow that I didn't had the opportunity
3874 to 'make check' over there yet.
3876 2004-11-15 Arnaud Legrand
3877 - Trace manager now written. It uses a heap structure and is therefore
3878 expected to be efficient. It may however be speeded up (particularly
3879 when many events occur at the same date) by using red and black
3880 trees. One day maybe...
3881 - Max-min linear system solver written. It uses a sparse matrix
3882 structure taking advantage of its expected use. Most operations are
3883 O(1) and free/calloc are called as few as possible. The computation of
3884 the minimum could however be improved by using a red and black tree
3887 2004-11-03 Arnaud Legrand
3888 - Rename every gras_* function that was in xbt/ to its xbt_
3890 - Add a heap and a doubly-linked list to xbt
3891 - Added a dichotomy to the dictionaries. make check works as well before
3892 so I assume that the patch is correct. I do not know however if things
3893 run effectively faster than before now. :)
3895 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3896 SimGrid, and the version number is bumped to 2.x
3898 2004-10-29 Martin Quinson
3899 - Introduction of the remote errors.
3900 They are the result of a RMI/RPC on the remote machine.
3901 ErrCodes being scalar values, you can't get the host on which those
3902 errors did happen. Extending the error mechanism as in Gnome is possible.
3903 No idea yet whether it is a good idea.
3905 2004-10-28 Martin Quinson
3906 - Interface revolution: the Starred Structure Eradication.
3907 I used to do typedef struct {} toto_t; and then handle *toto_t.
3908 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3909 - ???_t is a valid type (builded with typedef)
3910 - s_toto_t is a structure (access to fields with .)
3911 - s_toto is a structure needing 'struct' keyword to be used
3912 - e_toto_t is an enum
3913 - toto_t is an 'object' (struct*)
3915 typedef struct s_toto {} s_toto_t, *toto_t;
3916 typedef enum {} e_toto_t;
3917 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3918 s_toto_t) is private.
3920 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3921 so that it can set the variable to NULL. It was so for dicts and sets,
3922 it changed for dynars.
3924 - Fix a bunch of memleaks in dict_remove
3925 - Fix a bug in sg/server_socket opening: it failed all the time.
3927 2004-10-07 Martin Quinson
3928 - Speed up dynar lookup operation a bit.
3930 gras_dynar_get is dead.
3932 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3933 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3934 (which gives you the address of the stored data).
3936 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3937 copy of the data using an affectation to do the job and not a memcpy.
3939 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3940 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3942 It does not really speedup the dynar test because they are
3943 setting elements all the time (and look them seldom). But the dict does
3944 far more lookup than setting.
3946 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3948 2004-10-05 Martin Quinson
3949 - Allow to (en/dis)able the cycle detection at run time.
3951 Whether we should check for cycle or not is now a property of each
3952 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3953 datadesc_cycle_unset allow to remove this property when previously set.
3955 Note that the cycle detection is off by default since it impacts the
3956 performance. Watch the data you feed GRAS with ;)
3958 This property is hereditary. Any element embedded in a structure having it
3959 set have it set for the time of this data exchange.
3961 You should set it both on sender and receiver side. If you don't set it on
3962 sender side, it will enter an endless loop. If you forget on receiver
3963 side, the cycles won't be recreated after communication.
3965 - Header reorganization.
3966 Kill gras_private.h, each submodule must load the headers it needs.
3968 2004-10-04 Martin Quinson
3969 - Interface revolution: do not try to survive to malloc failure.
3971 Now, gras_malloc and friends call gras_abort() on failure.
3972 As a conclusion, malloc_error is not a valid error anymore, and all
3973 functions for which it was the only gras_error_t return value are
3974 changed. They now return void, or there result directly.
3975 This simplify the API a lot.
3977 2004-09-29 Martin Quinson
3978 - Re-enable raw sockets.
3979 Created by gras_socket_{client,server}_ext;
3980 Used with gras_raw_{send,recv}
3983 It should allow to kill the last bits of gras first version soon.
3985 This is not completely satisfactory yet (duplicate code with
3986 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3989 - Simplify transport plugin (internal) interface by not passing any
3990 argument to _server and _client, but embedding them in the socket
3993 2004-09-28 Martin Quinson
3994 - Finish the port to AIX.
3995 autoconf was my problem (segfault within the malloc replacement
3996 function. No idea why)
3998 2004-09-16 Martin Quinson
3999 - Fix some size_t madness on 64bit architectures.
4001 2004-09-08 Martin Quinson
4002 - Reduce the number of system headers loaded, overload some more system
4003 calls (such as malloc to cast the result of the system one, and work
4005 - Fix and reintroduce the config support
4007 2004-09-07 Martin Quinson
4008 - Source code reorganization to allow Arnaud to surf all over there.
4009 - Allow to document the logging categories.
4010 - Remove all uppercase from logging categories and useless cleanup in names.
4012 2004-08-18 Martin Quinson
4013 Version 0.6.2 (protocol not changed; API changed)
4014 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4015 gras_error_t), and NULL when not found. Functions expecting a msgtype
4016 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4017 hopefully usefull message.
4018 - Portability to prehistoric sparcs again
4020 2004-08-17 Martin Quinson
4021 Version 0.6.1 (protocol not changed; ABI not changed)
4022 - prealloc some buffers to speed things up
4024 2004-08-11 Martin Quinson
4025 Version 0.6 (protocol not changed; ABI expended)
4026 - The parsing macro can deal with the references, provided that you add
4027 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4029 2004-08-09 Martin Quinson
4030 Version 0.5 (protocol not changed; ABI changed)
4031 - Allow to off turn the cycle detection code in data exchange at
4032 compilation time. It should be at run time, but I'm short of time (and
4033 the config stuff is still broken). That way, we keep dict out of the
4034 critical path, which is good because the performance is poor:
4035 - search not dichotomial yet
4036 - dynar give no way to access their content and memcpy everytime
4037 - In composed data description (struct, ref and so on), stop foolness of
4038 keeping the subtype's ID, but store the type itself. This keeps sets out
4039 of the critical path, which is good since they rely on dynar and
4040 dictionaries. The only loose of that is that we cannot detect the
4041 redeclaration of a structure/union with another content (but I'm not sure
4042 the code detected well this error before anyway). We still can detect
4043 the redefinition discrepancy for the other types.
4044 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4045 it breaks the code because of type-punning used all over the place).
4046 This breaks on all non-gcc architectures (for now).
4048 All those changes (plus the buffer of last time) allow me to gain 2 order
4049 of magnitude on cruel tests consisting of 800000 array of integers on two
4050 level of a hierarchical structure (200 secondes -> 4 secondes)
4053 - the selector of reference must now return the type it points to, not
4054 the ID of this type.
4056 2004-08-06 Martin Quinson
4057 Version 0.4 (protocol changed; ABI not changed)
4058 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4059 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4060 - Implement a buffer transport, and use it by default (it relies on tcp in
4061 real life and on sg in simulation).
4062 That's a bit hackish since I had a new field to the structure to store
4063 its data without interfering with the subtype ones. Inheritance
4064 is tricky in C. And that's a kind of reverse inheritance with one class
4065 derivating two classes. Or maybe a game with java interfaces. Anyway,
4066 that's damn hard in C (at least).
4067 Moreover, I got tired while trying to ensure plugin separation and
4068 genericity in SG mode. MSG wants me to do weird things, so let's go for
4069 cruel hacks (temporarily of course ;).
4070 See comment in transport_private.h:71
4071 - do not include all the _interface headers in private but in the files
4072 which really need them (to cut the compilation time when they are
4075 2004-07-26 Martin Quinson
4076 Version 0.3 (protocol not changed; ABI changed)
4077 - Major overhault of the datadesc interface to simplify it:
4078 - shorted the function names:
4079 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4080 - add a trivial way to push/pop integers into the cbps without malloc.
4081 This allows to make really generic sub_type description, which simply
4082 pop their size of the stack.
4083 - add a function gras_datadesc_ref_pop_arr() which does what users want
4084 most of the time: Declare a dynamic array (which pops its size of the
4085 stack) and declare a reference to it. Poor name, but anyway.
4086 - kill the post-send callback, add a post-receive one
4088 2004-07-23 Martin Quinson
4089 Version 0.2 (protocol changed; ABI changed)
4090 - add some testing for cpbs in the test cases, and fix some more bugs.
4091 This invalidate again the little64 data file, since I cannot regenerate
4093 - remove an awfull optimization in the logging stuff, allowing me to:
4094 - understand it again
4095 - learn gcc how to check that the argument match the provided format
4096 - fix all errors revealed by gcc after that
4097 - internal keys of dict are not \0 terminated. Deal with it properly in
4098 loggings instead of segfaulting when the user want to see the logs :-/
4100 2004-07-22 Martin Quinson
4101 - Fix some stupid bug preventing cbps (callback postit) from working
4103 2004-07-21 Martin Quinson
4104 - Some documentation cleanups
4105 - remove the useless last argument of msgtype_declare
4106 - rename the Virtu functions to fit into the 'os' namespace
4107 - move headers src/include -> src/include/gras/ and stop fooling with
4108 gras -> . symbolic link
4109 - make distcheck is now successful
4111 2004-07-19 Martin Quinson
4113 - Build shared library also
4114 - Install html doc to the right location
4115 - stop removing maintainer files in make clean
4116 - build tests only on make check
4118 2004-07-13 Martin Quinson
4120 - No major issue in previous version => change versioning schema
4121 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4122 corresponding dataset.
4124 2004-07-11 Martin Quinson
4126 - Get it working with any kind of structure (we can compute the padding
4127 bytes remotely for all the architectures I have access to)
4128 - Implement the structure parsing macro (still not quite robust/complete)
4129 - Improvement to the remote testing toysuite
4131 2004-07-10 Martin Quinson
4132 [autoconf mechanism]
4133 - get ride of a bunch of deprecated macros
4134 - actually run the test for two-compliment, not only compile it :-/
4135 - test whether the structures get packed (and bail out if yes. Damn.
4136 Alignment is a serious matter)
4137 - test whether the structures get compacted (but respecting the alignment
4138 constraints of each types)
4139 - test whether the array fields of structures can straddle alignment boundaries
4141 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4142 (took me 2 hours to find that bug, looking at the wrong place)
4143 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4144 before sending/receiving and are used to compute the offsets of fields
4145 - Given that padding size depend even on compiler options, keep track of
4146 alignment and aligned_size only for the current architecture. Not a big
4147 deal since we send structure fields one after the other (seems
4149 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4150 program, let it work on linux/gcc/little32. portability todo.
4152 2004-07-08 Martin Quinson
4153 - import and improve remote compilation support from FAST
4154 - make sure make check works on half a dozen of machines out there
4156 2004-07-07 Martin Quinson
4157 Let's say it's version 0.0.3 ;)
4158 - Implement conversions (yuhu!)
4159 - Let it work on solaris (beside conversion, of course)
4160 - Stupid me, using rand() to generate the conversion datatests in not wise.
4162 2004-07-06 Martin Quinson
4163 - Let make dist work, since I'm gonna need it to compile on remote hosts
4164 - Let Tests/datadesc_usage write the architecture on which the file was
4165 generated as first byte.
4166 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4167 architecture descriptions.
4168 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4169 on those architectures.
4170 - Optimization: send/recv array of scalar in one shoot
4172 2004-07-05 Martin Quinson
4173 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4175 - Plug a whole bunch of memleaks
4176 - each process now have to call gras_{init,exit}. One day, their log
4177 settings will be separated
4178 - Continue the code factorisation between SG, RL and common in Transport.
4180 2004-07-04 Martin Quinson
4182 - Redistribution between SG and RL.
4183 We wanna have to accept in SG, so move accepted related parts of RL in
4184 the common part. (more precisely, the dynar of all known sockets is no
4185 more a static in transport.c, but part of the process_data)
4187 [gras_stub_generator]
4188 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4190 2004-07-03 Martin Quinson
4191 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4192 them all up in one shot)
4194 - Fix the ugly case of reference to dynamic array.
4195 - New (semi-public) function gras_datadesc_size to allow the messaging
4196 layer to malloc the needed space for the buffer.
4198 - gras_socket_close now expect the socket to close (and not its address to
4199 put NULL in it after it). This is because the socket passed to handlers
4200 is one of their argument (=> not writable).
4202 - propagate the interface cleanup from last week in datadesc, ie remove a
4203 superfluous level of indirection. User pass adress of variable
4204 containing data (both when sending and receiving), and not of a variable
4205 being a pointer to the data. Let's say that I like it better ;)
4206 The price for that is constructs like "int msg=*(int*)payload" in
4207 handlers, but it's a fine price, IMHO.
4209 - Let it work in RL (yuhu)
4211 2004-06-21 Martin Quinson
4213 - porting SG plugin and SG select to new standards (works almost).
4214 - plug memleaks and fix bugs around.
4217 - cleanup the prototype of data recv and force users to specify when they
4218 want to handle references to objects. Test case working even for cycles.
4219 - plug memleaks. Valgrind is perfectly ok with this.
4221 2004-06-12 Martin Quinson
4223 - cleanup the separation between plugin and main code in plugin creation
4225 2004-06-11 Martin Quinson
4227 - Reput hook for raw sockets, needed for BW experiments
4228 - kill a few lines of dead code
4229 [Data description] Interface cleanup
4230 - gras_datadesc_by_name returns the searched type or NULL.
4231 That way, no variable is needed to use a type desc once, which makes
4233 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4234 two parameters were strings (field name, type name), leading to
4236 [Dicos] Interface cleanup
4237 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4238 This is consistant with the dynar API.
4240 2004-04-21 Martin Quinson
4242 - Porting to new standards.
4244 - interface cleanup.
4245 There is no bag anymore, no need to take extra provision to mask the
4246 pointers behind "ID".
4247 Better splitup of functions between files create/exchange/convert.
4248 This is still a bit artificial since convert and receive are so
4249 interleaved, but anyway.
4251 - add a queued message list to procdata (the ones not matching criteria
4253 - factorize some more code between SG and RL wrt procdata
4255 - use gras_exit in example to track memleaks
4256 - get rid of gs_example now that GS is properly integrated into gras
4257 - update run_test to integrate the latest tests (datadesc)
4259 - rename WARNINGn macros to WARNn since it prooved error-prone
4261 2004-04-19 Martin Quinson
4263 - register init/exit functions within gras module mechanism
4264 - send/receive function.
4265 Convertion is not implemented, but short-cutted if not needed.
4266 struct/array elements are sent one by one (instead of block-wise), but
4267 nobody really cares (yet). Get a prototype before optimizing.
4268 - tests (using a file socket) for DD send/receive on:
4269 - base types: int, float
4270 - array: fixed size, string (ie ref to dynamic string)
4271 - structure: homogeneous, heterogeneous
4272 - chained list, graph with cycle
4273 Believe it or not, valgrind is not too unhappy with the results. The
4274 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4278 - Bugfix when using a filename explicitly (instead of '-')
4280 2004-04-09 Martin Quinson
4282 - factorize more code between RL and SG in socket creation
4283 - Complete the implementation and tests of:
4285 o file (only in RL, and mainly for debugging)
4287 I lost 3 days to design a portable address resolver, and then decided
4288 that the prototype mainly have to run on my box.
4289 Addressing portability too early may be like optimizing too early :-/
4291 - use gras_init in the Tests instead of the crappy parse_log_opt
4292 (the latter function is removed)
4293 [Conditional execution]
4294 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4295 [Code reorganisation]
4296 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4297 Build examples against the RL library, since there is no way to disable
4298 its creation for now.
4300 For information, the beginning of coding on GRAS was back in june
4301 2003. I guess that every line has been rewritten at least twice since