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 ----------------------------------------------------------------------------
34 SimGrid (3.26) December 15. 2020
36 The Release release (the French lockdown was eased today).
38 Important user-visible changes:
39 - SimGrid now requires a compiler with C++14 support.
40 Sibling projects should upgrade their FindSimgrid.cmake
41 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
42 several users had difficulties to understand issues when using high bandwidth or
43 small latency events. The new value was already the default for SMPI and
44 should not cause too much performance hit. It can change some simulation
45 timings if simulation was skipping a lot of small events (which was actually
46 an issue per see). The old value can be restored with --cfg=surf/precision:1e-5.
47 - On some network models (L07, CM02, LV08), an implicit loopback link is created
48 for local communication. Default bandwidth was 500MBps and latency was 15µs,
49 which was rather bad. Default values have been changed to 10GBps and 0 latency.
50 Options network/loopback-lat and network/loopback-bw have been added to adapt
51 these values if needed. Network and routing models already providing separate
52 loopbacks (clusters and such) are not impacted by this change.
53 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
54 metadata that are added to each MPI message by MPI libraries (tag,
55 src/dest, and communicator information). This may change simulation times
56 for very small message sizes, but should be more realistic.
59 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
60 - New plugins: wifi_energy and link_load. See the associated examples.
61 - New examples: wifi networks, and communication suspend.
64 - LTO is now enabled for Intel/clang compilers.
65 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
66 Setting it to "auto" will use all available cores, while setting it to n will
67 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
68 - Remove obsolete runtime option 'contexts/parallel-threshold'.
69 - Runtime option 'tracing/msg/process' renamed to 'tracing/actor'. The old name
70 has been kept for compatibility.
71 - Finally remove obsolete snake_case() aliases for runtime options.
72 - Further improve the documentation.
75 - the on_exit() of each actor is also executed when the simulation deadlocks.
76 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
77 An example is provided for s4u::Comm but it should work with Exec and Io.
80 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
81 https://framagit.org/simgrid/SMPI-proxy-apps
82 - MPI/IO: supports names starting with "./"
83 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
85 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
86 building shared libraries, this may cause issues, so environment variable
87 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
88 - most temporary files should now be created in /tmp dir (or equivalent).
89 If this one does not allow execution of code (noexec flag), this may cause issues.
90 Please use another tmp directory (using TMPDIR or equivalent system variable)
94 - A large refactoring is ongoing, but this should be transparent for users so far.
97 - Drop xbt_str_split().
99 C binding and interface:
100 - The return type of the sg_actor_on_exit() callbacks is now 'void'
101 (int value was previously ignored)
102 - Many C functions were renamed to follow a common naming style
103 (sg_object_get_something and sg_object_set_something). As usual, the deprecated
104 names are kept until version 3.30.
105 - Many MSG tests were converted to the new S4U's interface in C, that
106 was extended for that.
108 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
109 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
110 - FG#41: Add sg_actor_create C interface
111 - FG#43: xbt::random needs some care
112 - FG#48: The Impossible Did Happen (yet again)
113 - FG#50: Suspending an actor executed at the current timestamp fails
114 - FG#52: Zero-seconds timeout : "That's in the past already"
115 - FG#53: Crash while using ns-3 network model
116 - FG#54: How to suspend a comm?
117 - FG!22: Stochastic Profiles
118 - FG!24: Documentation and fix for xbt/random
119 - FG!35: Add a modeling hint for parallel links in doc
120 - FG!36: [xbt/random] Read/Write the state of the RNG
121 - FG!42: Link cumulated load plugin
122 - FG!44: Test condition variable wait_until+notify_all
123 - GF#18137: Allow different stack sizes?
124 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
125 - GH#139: Allow pthread creation in SMPI
126 - GH#336: Packet-level simulation using SMPI?
127 - GH#345: Error in the LMM while migrating a VM
128 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
129 - GH#352: pip install / python setup.py install fail to find pybind11
130 - GH!337: Fix link_energy plugin for wifi platforms
131 - GH!339: Add Mailbox set_receiver method to python binding
132 - GH!344: Cast hugepages macros parameters to int64
134 ----------------------------------------------------------------------------
136 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
138 The Palindrom Day Release.
140 Important user-visible changes:
141 - Improve the Python usability (stability and documentation).
142 - A nasty synchronization bug was ironed out, see also below.
143 - Python's doc was organized as a separate tree, now integrated with C++.
144 - C bindings of S4U were not part of the doc.
145 - The C++ doc was also improved as methods are now split by theme.
146 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
147 - This is mandatory to use the Java bindings.
148 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
149 - SimDAG++: Automatic dependencies on S4U activities (experimental)
150 - Some features are already implemented but not all of them
151 - Cannot block an activity until it's scheduled on a resource
152 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
153 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
156 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
157 - Drop support for Viva/Triva (old visualization tools), hence removed
158 TRACE_get_node_types() and TRACE_get_edge_types() functions.
161 - Require pybind11 version 2.4.
162 - Greatly improve locking during context switches with Python.
163 - Declaring Python callbacks to this_actor::on_exit is now working.
164 - No more tests are failing currently, even on win32.
167 - Actor: Merge signals on_migration_start/end into on_host_change
168 - Actor: Rename migrate() into set_host()
169 - Disk: Allow users to get the read and write nominal bandwidth values
170 - Exec: Implement wait_for(timeout)
171 - Io: Implement wait_for(timeout)
174 - Parse errors now raise a simgrid::ParseError that you may want to catch.
177 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
178 been renamed to fix a spelling error. As usual, the previous definitions are
179 kept with a deprecation warning. However, to avoid runtime errors, it is now
180 forbidden to override these deprecated methods in a derived class. Please use
181 the new names immediately if you need to override them.
184 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
185 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
186 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
187 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
188 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
189 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
190 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
191 - New module: random, providing classical random numbers generators.
194 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
195 allocations' size above which they are considered "shared" by default (as if
196 they were performed through SMPI_SHARED_MALLOC macros).
197 Default = 0 = disabled feature.
198 Note : malloc, calloc and free are now overridden by smpicc/cxx by default.
199 This can cause some troubles if codes are already overriding these. If this
200 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
201 help, but will make this feature unusable.
202 - Input check is now performed much more consistently and outputs precise
203 warnings for each error found
204 - New supported calls : MPI_File_set_view, MPI_File_get_view
205 - MPI I/O now only supports disks, as storage is discontinued.
206 - MPI I/O allows opening files without specifying absolute path
207 - Note : SMPI configuration options are now initialized by smpi_init_options()
208 call, which should be called if SMPI is not being used through smpirun.
210 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
211 - FG#9: Python bindings crashing
212 - FG#39: Missing s4u::Comm::wait_any_for example
213 - FG#42: Add support for ThreadSanitizer (TSan)
214 - FG!19: Removing RngStream
215 - FG!20: A module for RNG calls
216 - FG!21: Choice between ad-hoc and standard distributions implementations
218 - FG!25: Fix link in Mutex doc
219 - FG!26: Fix links in SMPI interface doc
220 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
221 - GH#305: Unscheduled tasks are still excuted
222 - GH#313: smpirun: manual outdated w.r.t. --help
223 - GH#321: [S4U] Get task remaining work ratio
224 - GH#323: Crash when an actor turn off his physical host
225 - GH#335: Missing links on dragonfly example svg
226 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
227 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
229 ----------------------------------------------------------------------------
231 SimGrid (3.24) October 9. 2019
233 The Clean Disk Release.
236 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
237 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
238 thus becomes the default resource to execute I/O operations. Examples
239 and the file system plugin have been modified to use disks instead of
240 storage elements. Storage can still be used but is doomed to disappear.
241 Users are thus adviced to update their simulators accordingly.
242 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
243 for consistency with pthread_barrier_wait()
244 - Host::get_englobing_zone() returns the englobing netzone
245 - Actor::on_destruction is now called in the destructor
246 Actor::on_termination new signal called when the actor terminates
248 - Global signals are now part of the Engine:
249 - on_platform_creation: after config settings, before the XML parsing
250 - on_platform_created: right after the XML parsing
251 - on_time_advance: each time the clock advances
252 - on_simulation_end: after simulation, before cleanups
253 - on_deadlock: as the name implies.
255 - sg_{actor,host,link}_{data,data_set}() now all exist.
256 Use them to attach user data to the object and retrieve it.
259 - Introduce an experimental Wifi model. It sounds reasonable
260 according to the state of the art, but it still has to be properly
261 validated, at least against ns-3 if not against reality.
262 - Improved the usability of ns-3. Several bugs were ironed out.
263 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
264 It is now expressed as 'idle:epsilon:allCores' to properly model the
265 consumption of non-whole tasks on mono-core hosts. More info in the doc.
268 - convert a new set of functions to the S4U C interface and move the old MSG
269 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
272 - Fortran bindings for DVFS have been removed.
273 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
274 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
275 - SMPI can now be selected by cmake's find_module(MPI) with
276 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
277 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
278 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
279 - trace-call-location can be used with TI traces, and replayed, and is compatible
280 with smpi/comp-adjustment-file.
281 - sleep events are now correctly traced/replayed.
282 - Default for trace-call-location is now to use file names and not full paths.
283 To revert to previous behavior (in case of collision of filenames), option
284 "smpi/trace-call-use-absolute-path" can be set to yes.
287 - Use the included xxHash as an hash implem when C++14 is usable.
288 - Option model-checker/hash was removed. This is always activated now.
289 - New option smpi/buffering controls the MPI buffering in MC mode.
290 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
291 This is useful to check for MPI compliance.
294 - New section on plugins: how to define a new one, and existing ones.
297 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
298 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
301 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
302 and <mount> tags. The rationale is finish to decouple what should be handle
303 by SimGrid kernel (i.e., resources and activities) from plugins built in
304 user space (e.g., the file system plugin). This new tag currently accepts
305 only three attributes (id, read_bw, and write_bw). All the other information
306 that was declared with the storage related tags now has to be expressed as
307 properties. An example of platform using this new tag is available at
308 examples/platforms/hosts_with_disks.xml
311 - 'expect signal' can now accept more than one potential signal.
313 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
314 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
315 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
316 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
317 - FG#31: per-actor data
318 - FG#34: SG_BARRIER_SERIAL_THREAD?
319 - FG#35: model-checker does not like buster-produced binaries
320 - FG!13: MC: complete workaround in the error msg seen on modern systems
321 - FG!15: execute_flops now logs compute
322 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
323 - FG!17: ns-3: unblock the right number of communications + others issues
324 - FG!18: Improving the performance of the ns-3 bindings
325 - GH#189: Energy consumption of parallel tasks
326 - GH#219: Error in the throughput of TCP transfer
327 - GH!330: Minor polishing of English: Configuring_Simgrid
328 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
329 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
330 - GH!334: Rework energy plugin, again
332 ----------------------------------------------------------------------------
334 SimGrid (3.23.2) July 8. 2019
337 - Nicer introduction page.
338 - Migrate the "Deploy your application" page to the new doc.
339 - Move Java as a subtree of MSG.
342 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
345 - Drop xbt_dynar_sort_strings().
348 - Really fix FG#26: Turning off a link should raise NetworkFailureException
349 - FG#27: Wrong exception thrown to wait_any when link is turned off
350 - GH#328: Java: Canceling multiple tasks in a single vm/host
352 ----------------------------------------------------------------------------
354 SimGrid (3.23) June 25. 2019
356 The Exotic Solstice Release.
359 - SunOS and Haiku OS support. Because exotic platforms are fun.
360 - Stop setting random seed with srand48() at initialization.
361 - Use addr2line as a fallback for stacktraces when backtrace is not available.
362 - Build option -Denable_documentation is now OFF by default.
363 - Network model 'NS3' was renamed into 'ns-3'.
366 - Simgrid can now hopefully be installed with pip.
369 - wait_any can now be used for asynchronous executions too.
372 - New log appenders: stdout and stderr. Use stdout for xbt_help.
373 - Drop xbt_dict_dump.
376 - 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
377 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
378 - MPI interface is now const correct for input parameters
379 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
380 - Fortran bindings for async collectives.
381 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
384 - Remove option 'model-check/record': Paths are recorded in any cases now.
385 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
386 always sparse. That's an immense gain in memory, and can even be
387 faster because of cache effects. So there is no need to clutter the
388 code to allow the user to go for the unefficient mode.
391 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
392 models proved to be more accurate than these old experiments.
394 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
395 - FG#1: Broken link in error messages
396 - FG#2: missing installation documentation
397 - FG#3: missing documentation in smpirun
398 - FG#6: Python bindings not available on PyPI
399 - FG#7: simple cmake call requires doxygen
400 - FG#8: make python bindings an optional dependency
401 - FG#10: Can not use MSG_process_set_data from SMPI any more
402 - FG#11: Auto-restart actors forget their on_exit behavior
403 - FG#12: -Denable_lto=OFF doesn't disable LTO
404 - FG#13: Installs unstripped file 'bin/graphicator'
405 - FG#14: Installs the empty directory 'doc/simgrid/html'
406 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
407 - FG#17: Dead link in doc (pls_ns3)
408 - FG#20: 'tesh --help' should return 0
409 - FG#21: Documentation link on http://simgrid.org/ broken
410 - FG#22: Debian installation instruction are broken
411 - FG#26: Turning off a link should raise NetworkFailureException exceptions
412 - GH#133: Java: a process can run on a VM even if its host is off
413 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
414 - GH#326: Valgrind-detected error for join() when energy plugin is activated
416 ----------------------------------------------------------------------------
418 SimGrid (3.22) April 1. 2019
420 The Easter Christmas Release (this one is somewhat late).
423 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
424 - Not complete yet: asynchronous activities (amongst others) are still missing
425 - Still ongoing: the interface may change in the future. We need more
426 testers! Please report any glitches.
427 - No new project using Java should start now. Please switch to Python.
430 - Some of the internal cleanups may lead to speed improvements:
431 - The hard limitation on the amount of simulated actors with
432 Java+Mac was removed. Now, the available memory is the only limit.
433 - Our refcounting was tidyied, leading to 10% speedups in some cases.
434 - We are still working on making our code robust to the actor kills
435 and hosts' churn. Things are improving, but it's not perfect yet.
436 - Replaced our own code to display a backtrace (that was forking addr2line)
437 with the Boost.Stacktrace library.
438 You won't see your backtraces without this optional dependency.
439 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
440 - Stop setting random seed with srand() at initialization.
443 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
444 XML file version remains 4.2 since old files are still compatible.
447 - Process termination which was broken at version 3.21 has been repaired.
448 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
449 - Hide the examples into examples/deprecated. New users should use Python.
452 - Drop MSG_process_create_from_stdfunc() from the API.
453 This C++ function was a pimple in the C API, made necessary at some
454 point by the Java bindings. This is fixed now.
455 - Hide the examples into examples/deprecated. New users should use S4U.
456 - MSG_process_create and MSG_process_attach now crash if the host used
457 to run this process is off.
458 - Fix the protype of MSG_process_on_exit()
459 Now use (int,void*) callbacks instead of (void*,void*) ones.
460 The implementation was ways too messy to actually work, I guess.
463 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
464 too often and provide more accurate timings.
465 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
466 - MPI_Alltoallw support
467 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
468 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
469 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
470 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
472 - Basic implementation of generalized requests (SMPI doesn't
473 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
476 - Drop sg_cmdline. Please use xbt_cmdline instead.
477 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
478 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
479 OsSemaphore is implemented in a portable way with C++11 threads.
480 This should allow much more threads to be created at the same time,
481 allowing Mac and Java users to simulate many more actors.
482 - Implement the 'thread' factory with std::thread instead of xbt ones.
483 It is not possible to set the stack size with threads anymore, but
484 -fsplit-stack is the way to go nowadays when using threads.
485 - Drop the xbt_os_thread_t module (now unused)
486 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
487 - Drop xbt_str_join_array().
488 - Drop cunit, use Catch2 instead.
491 - Many cleanups in the kernel::activity namespace. This was long
492 overdue, and shall open the path to many future endeavors.
495 - #132: Java: a process can not shut down its own host
496 - #220: S4U: detached send causes exception if sender terminates soon after sending
497 - #261: Document the parameters of parallel execution's constructor
498 - #300: [s4u] BarrierPtr is missing
499 - #314: SMPI args internal cleanup
500 - #316: Fix a bug related to the CPU utilization of multi-core VM
501 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
502 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
503 - #325: Turning off a host has different behavior on sleeping actors and computing actors
505 ----------------------------------------------------------------------------
507 SimGrid (3.21) October 3. 2018
509 The Restarting Documentation (TRD) Release.
512 - Convert the user manual to Sphinx for improved usability.
513 Unstable version now hosted on https://simgrid.frama.io/simgrid
514 Many glitches remain; stay tuned.
515 - Restrict the Doxygen reference API to the installed header files.
516 (documenting all internals this way is overkill + doxygen scales badly)
517 - New tutorial on S4U (highly inspired from the old MSG one, but with
518 a git repository to fork as a starting point, and a docker image)
519 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
520 - The Developper manual is still to be converted (not compiled until then)
521 - Some parts are still missing in this conversion (such as the
522 platform and deployment sections) while others were blindly converted
523 and would need more love (such as the configuration flags).
524 Things will certainly further improve in the future releases.
527 - Fully reimplement auto_restart mechanism that was utterly broken
528 (fix #22, #131 and #281 that were all very related)
529 - Implement semaphores (fix #303 and #291)
530 - When creating an actor from a function and its parameters,
531 move-only parameters are not allowed anymore, as it would prevent
532 the actor to be restartable if its parameters are consumed on run.
533 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
534 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
535 IO operations that can be started, waited for, or canceled as a regular
537 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
541 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
542 - New host variable: 'core_count'
545 - Remove xbt_os_thread_specific features
546 - Remove portability wrapper to condition variables
547 - Remove xbt_os_thread_yield()
550 - MPICH collective selector now mimics MPICH 3.3b
551 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
552 - MPI_Init(NULL,NULL) is now allowed
553 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
554 - Command-line arguments should now be handled properly in Fortran simulations
555 - MPI Attributes and related callbacks should work in Fortran
556 - Apps using C + Fortran should now work
557 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
558 - Support MPI_Error_String
561 - Due to an internal bug, Msg.run() must now be your last line.
562 We hope to fix it in a future release, and we are sorry for the inconvenience.
565 - #22: Process autorestart seem to only work with CAS01 cpus
566 - #56: Feature request: dynamic SMPI replay
567 - #93: simgrid should not eat --help
568 - #111: How to change the loopback link characteristics?
569 - #116: Platform cluster radical
570 - #117: Platform cluster/host order
571 - #129: Replay_multiple on steroid
572 - #131: Java : autostart does not work
573 - #143: Setting a breakpoint at a given time
574 - #235: Network energy models should be integrated
575 - #264: Add ptask L07 resource tracing
576 - #271: Dynamic and manual replay of SMPI traces
577 - #279: Breakpoints option uses time deltas instead of absolute time
578 - #281: Daemonized actor no longer daemonized after an auto-restart
579 - #282: TIT convertor erase traces when using absolute path in the trace list file
580 - #285: segfault when a s4u actor kills itself with msg process tracing activated
581 - #286: Pajé traces are not exposing the number of cores
582 - #287: Command-line parsing should stop when encountering '--'
583 - #288: MPI_Init(NULL, NULL)
584 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
585 - #290: Method to check if a mailbox has a message ready to be consumed
586 - #291: [s4u] semaphores do not seem available yet
587 - #294: Alltoall collective implementation of mpich
588 - #295: Starting a Host as OFF is not supported
589 - #296: DTD too permissive
590 - #299: Add a s4u barrier example
592 ----------------------------------------------------------------------------
594 SimGrid (3.20) Released June 24. 2018
596 The proxy snake_case() release.
599 - The whole API is now using snake_case() instead of mixing it with camelCase.
600 Compatibility wrappers in place for one year.
601 - We took the opportunity to increase the consistency of the API.
602 Please report any remaining glitches.
603 - New function simgrid::s4u::Host::get_actor_count:
604 Returns the number of actors running on a specific host.
607 - SMPI is now tested with ~45 proxy apps from various sources, with none or
608 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
609 - Replay: The replay file has been re-written in C++.
610 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
611 supported. They are stored in the trace and used when replayed.
612 - Basic support of MPI_Cancel. Robustness not guaranteed.
613 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
614 (only for MPI_COMM_TYPE_SHARED).
615 - New option: smpi/privatize-libs, to add external shared libs to be privatized
616 by SMPI. They will be copied locally and loaded separately by each process.
617 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
618 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
619 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
620 with internal index optind with multiple processes. Only works if MPI_Init has
622 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
623 syntax without preprocessing tricks.
626 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
627 behavior tracing is no longer limited to MSG
628 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
629 this means that traces now need to be updated or re-obtained; both Irecv
630 and Isend lines in your traces have now in total 5 mandatory fields:
631 <rankid> <command> <to/from rankid> <tag> <size>
632 To update your traces, it suffices to add a 0 for the tag here.
633 - TIT now also supports waiting for a distinct request via MPI_Wait.
634 Wait/Test now wait for a specific request, not just the last one that was
635 issued. This unfortunately means another update, because we need to
636 identify which request you want to wait for. We do this via the
637 triplet (sender, receiver, tag), which needs to be added:
638 <rankid> <command> <sender> <receiver> <tag>
639 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
640 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
643 - The deprecation of MSG is ongoing (but this should not impact you).
644 Many MSG functions are now simple wrappers on the C API of S4U. If
645 you wish to convert your code to S4U, find the S4U counterparts of
646 your MSG calls in src/msg/msg_legacy.cpp.
647 - MSG can still be used, but won't evolve anymore.
650 - Allow to run the Link energy plugin from the command line with
651 --cfg=plugin:link_energy
652 - Rename Energy plugin into host_energy
653 - Rename Load plugin into host_load
656 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
657 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
658 - Improve performance when handling timeouts of simix synchros.
661 - Config: the C API is now deprecated (will be removed in 3.23), and
662 the C++ API has been slightly improved.
665 - Fix several build issues on OSX.
666 - Move simgrid_config.h to simgrid/config.h (old header still working)
669 - #143: Setting a breakpoint at a given time
670 - #150: Inconsistent event names in SMPI replay
671 - #258: daemonized actors hang after all non-daemonized actors have completed
672 - #267: Linker error on unit_tmgr
673 - #269: SMPI: tracing of MPI_Wait/all/any broken
674 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
676 ----------------------------------------------------------------------------
678 SimGrid (3.19.1) Released March 22 2018
680 The Fixed ABI Release.
682 As you may know, we are currently refactoring SimGrid in deep.
683 Upcoming SimGrid4 will be really different from SimGrid3: modular,
684 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
686 Our goal is to smooth this transition, with backward compatibility
687 and automatic update paths, while still progressing toward SimGrid4.
689 SimGrid remains open during works: The last pure SimGrid3 release was
690 v3.12 while all subsequent versions are usable alpha versions of
691 SimGrid4: Existing interfaces remain unchanged, but the new S4U
692 interface is budding and the internals are deeply reorganized.
694 Since 2015, we work hard to reduce the changes to public APIs. When
695 we need to rename a public library symbol in S4U, we let your
696 compiler issue an explicative warning when you use the deprecated
697 function. These messages remain for four releases, i.e. for one full
698 year, before turning into an error. Starting with v3.15, your can
699 also adapt to API changes with the SIMGRID_VERSION macro, that is
700 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
702 Starting with this v3.19.1, our commitment to reduce the changes to
703 the public interfaces is extended from the API to the ABI: a program
704 using only MSG or SimDag and compiled against a given version of
705 simgrid can probably be used with a later version of SimGrid without
706 recompilation. We will do our best... but don't expect too much of
707 it, that's a really difficult goal during such profund refactoring.
709 The difference between v3.19 and v3.19.1 is that the former was
710 accidentally breaking the ABI of MSG, while the later is restoring
713 S4U and kernel APIs will still evolve until SimGrid4, with one-year
714 deprecation warnings as currently. In fact, cleaning up these
715 interfaces and converting them to snake_case() is one release goal of
716 v3.20. But don't worry, we are working to smooth this upgrade path.
718 In summary, new projects should start with S4U to benefit of the
719 future, but old MSG projects should still be usable with no change.
721 ----------------------------------------------------------------------------
723 SimGrid (3.19) Released March 20 2018
725 The Moscovitly-cold Spring release.
727 SMPI user-visible improvements:
728 - Support of MPI_Type_create_subarray
729 - Unimplemented MPI calls now abort (instead of issuing a warning)
730 Some harmless calls around errhandlers still only issue a warning.
731 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
732 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
733 is now possible (mmap privatization was refactored to this end).
734 Adding new SMPI instances should also be technically doable now.
736 SMPI internal cleanups:
737 - Removed index notion from SMPI -- no more "getPid() - 1"!
738 In tracing files, actors are called rank-<PROCESS-ID> but these are
739 NOT the real ranks (tracing several communicators is not supported yet).
740 - Deprecate smpi_process_index() (will be removed in 3.22)
741 - Better testing on classical proxy apps, through an external project:
742 https://github.com/simgrid/SMPI-proxy-apps/
745 - Use a graphical TOC to make it easier to find the documentation you need
746 - Revamp the MSG tutorial
749 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
750 - Remove parameter of MSG_process_killall().
751 Resetting the PID was bogus anyway (several actors could have the same PID).
754 - Execution->setHost() can be called after start(), to migrate it.
755 - Comm::test_any() is now implemented.
756 - s4u::Actor now has onCreation() and onDestruction() signals.
757 - Install some kernel header files for the users' plugins and more.
758 Warning, their API is really not stable yet.
759 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
760 include/simgrid/kernel/routing/*Zone.hpp
763 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
765 Build System and other cleanups:
766 - Remove unused run-time parameter "tracing/onelink-only".
767 - SimGrid now works with intel compilers (MC still buggy)
768 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
769 - tesh kills subprocesses on timeout (but not on windows)
772 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
775 - #194: Feature request: simgrid::s4u::Comm::test_any()
776 - #245: migrating an actor does not migrate its execution
777 - #253: Feature Request: expose clusters as objects
778 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
779 - #255: Tesh broken on Windows
780 - #256: Modernize FindSimGrid.cmake
781 - #257: Fix (ab)use of CMake install
783 ----------------------------------------------------------------------------
785 SimGrid (3.18) Released December 24 2017
787 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
789 MSG IS NOW DEPRECATED. Please use S4U for new projects.
790 - Support for MSG should not be removed from SimGrid before 2020,
791 but future evolutions will be limited to the integration of
792 user-provided patches.
795 - s4u::onDeadlock() signal, run before stopping the simulation.
796 - s4u::Exec: asynchronous executions (abstraction of background
798 - s4u::Host->getLoad() returns the achieved speed in flops/s
801 - Link::name() is deprecated and will be removed in v3.21, use
802 Link::getCname() instead.
803 - Mailbox::getName() changed to return a std::string, use
804 Mailbox::getCname() instead to get a char*.
805 - Storage::getName() changed to return a std::string, use
806 Storage::getCname() instead to get a char*.
807 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
808 to have both allocation and deallocation of the map in user space.
811 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
812 The amount of flops remaining to do is a vector, not a scalar.
813 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
814 name implies on both sequential and parallel tasks.
815 - Both changes fix GitHub's #223 using PR #237 as a basis.
816 Thanks Michael Mercier.
817 - Most examples were converted to S4U and hidden elsewhere at tests.
818 Rationale: we still want MSG to work; we want newcomers to use S4U.
821 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
822 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
825 - Switch to the faster dlopen privatization mechanism by default
826 - Documentation improvments
829 - Live migration is getting moved to a plugin. Dirty page tracking is
830 the first part of this plugin. This imply that VM migration is now
831 only possible if one this function is called:
832 - C/MSG: MSG_vm_live_migration_plugin_init()
833 - C/C++: sg_vm_live_migration_plugin_init()
834 - Java: Msg.liveMigrationInit()
835 For C and C++, "simgrid/plugins/live_migration.h" also has to be
839 - Define class simgrid::xbt::Path to manage file names.
840 - Removed unused functions:
841 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
842 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
843 xbt_graph_export_graphviz()
844 - xbt/str.h: xbt_str_join()
845 - Remove unused datatypes:
846 - xbt/heap.h: use std::priority_queue or boost::heap instead
847 - xbt/swag.h: use boost::intrusive::list instead
850 - New link_energy plugin for the consumption of the links.
851 - All of the operations on files and storage contents have been
852 packaged into a plugin (src/plugins/file_system). The current
853 public interface can be found in
854 include/simgrid/plugins/file_system.h
855 To use these functions you now have to initialize the plugin by
856 calling MSG_storage_file_system_init() just after calling
857 MSG_init() or sg_storage_file_system_init() just after creating
861 - Remove the undocumented/untested tag <include>
864 - Remove viva specific tracing as the tool is no longer maintained
867 - #248: Kill of finished processes leads to segfault
868 - #240: xbt_cond_wait_timeout should gracefully return for C apps
869 - #239: please implement signal s4u::onDeadlock()
870 - #230: segfaults when exit() without run()
871 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
872 - #223: MSG_task_get_flops_amount() not working with parallel tasks
873 - #222: Actor::kill() doesn't really kill and segfaults
874 - #221: odd LMM warning when killing an actor
875 - #120: Memory leak when the processes are forcefully killed
877 ----------------------------------------------------------------------------
879 SimGrid (3.17) Released October 8 2017
881 The Drained Leaks release: (almost) no known leaks despite the tests.
883 Even very long simulations will preserve your memory: our testsuite
884 stresses SimGrid for over 45mn on fast machines for a coverage of
885 over 80%, and there is only one single known leak, of about 4kb.
889 - Comm.detach(): start and forget about asynchronous emission. A cleanup
890 handler may be given to free resources if the comm cannot be completed.
891 - this_actor::send(mailbox) is now mailbox->put()
892 - New: simgrid::s4u::Comm::wait_all()
893 - New: Host.execute() for remote executions.
896 - Unused option network/sender-gap is removed.
899 - Deprecate MSG_task_isend_with_matching(): This unused feature
900 really complicates our internals. Will be removed in v3.20.
903 - Improved context termination. It is now possible to free resources, even
904 when a process is forcibly killed.
905 - ContextBoost: add support for Boost versions above 1.61.
908 - Bring back run-time option --cfg=exception/cutpath to remove exception
910 - Removed unused functions:
911 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
912 xbt_str_rtrim(), xbt_str_trim().
913 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
916 - Removed header files obsolete since SimGrid 3.12:
917 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
919 - #3: SD_exit should be made optional
920 - #120: Memory leak when the processes are forcefully killed
921 - #159: Threading test regression in Actor refcounting
922 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
923 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
924 - #186: Actor::killAll() segfaults if some process is blocked on wait()
925 - #191: VM migration and pstate
926 - #192: Updating the energy consumptions of all hosts crashes with VMs
927 - #195: All actors have PID=0 in the logs
928 - #204: Sometimes segfault with thread contexts and mmap privatization
929 - #222: Actor::kill() doesn't really kill and segfaults
930 - #225: Actor::kill() doesn't really kill when victims are doing a join()
932 ----------------------------------------------------------------------------
934 SimGrid (3.16) Released June 22. 2017.
936 The Blooming Spring Release: developments are budding.
939 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
940 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
941 - Java JAR file should be finally fully working on Mac OSX too.
943 XML platforms: Switch to platform v4.1 format.
944 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
945 - <zone> can be used as a synonym for the now deprecated <as>
946 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
947 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
948 - <actor> can be used as a synonym for the now deprecated <process>
949 - state_file and avail_file periodicity is now easier to express
950 (check the documentation) the old behavior should still work.
951 * Storage: not backward compatible, but it seems that nobody ever used it
952 - Remove attribute 'content_type' of <storage_type>: was never used
953 - Make attribute 'model' of <storage_type> optional: for future usage
954 - Remove Bconnection model property: was never (in)validated. Replaced by
955 the maximum of the read and write bandwidth as a resource constraint.
958 - New and Backwards Compatibility break:
959 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
960 When one wants to get the list of tasks whose states have changed during a
961 simulation round, s/he has to allocate and free a dynar and use it as
962 argument to this function. The former SD_simulate (double how_long)
966 - Allow multicore VMs, with the correct sharing (unless you overcommit)
967 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
970 - New function to update the consumption of all hosts at once.
971 - Fix the model for multi-core hosts, linear in the amount of busy
972 cores with an abnormality for pIdle (see paper).
973 BUG: Still not satisfactory for ptask on mono-cores.
976 - The netzone are now available from the MSG API.
977 The old names still work, but are now deprecated.
980 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
981 - New option: smpi/keep-temps to not cleanup temp files
982 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
983 mallocs mode is used, allows to change the size of the fake file used
984 (default 1MB), to potentially limit the number of mappings for large runs.
985 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
986 - Fortran ifort and flang compilers support
987 - New RMA calls supported (experimental):
988 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
989 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
990 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
991 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
992 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
993 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
996 - New: this_actor::isMaestro() returns whether we are in kernel mode.
997 - Behind the scene, ActivityImpl are now automatically refcounted.
998 Removing this nasty bottleneck will greatly simplify our internals.
1001 - Replay: New function xbt_replay_action_get():
1002 Retrieve the function previously associated to an event type.
1003 - DROPPED FUNCTION: xbt_str_varsubst()
1004 - DROPPED MODULE: strbuff. We don't need it anymore.
1005 - DROPPED MODULE: matrix. We don't need it anymore.
1006 - DROPPED MODULE: lib. We don't need it anymore.
1008 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1010 SimGrid (3.15) Released March 22. 2017
1012 The Spring Release: continuous integration servers become green
1013 We fixed even the transient bugs on all target architectures:
1014 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
1015 Windows, FreeBSD, NetBSD.
1017 - Rename NetCards to NetPoints.
1018 This was intended to help NS3 users, but that's not a netcard.
1019 That's a point in the routing algorithm, let's avoid wrong simplifications.
1022 - New: MSG_process_yield(). Stop and yield to other processes.
1023 - New: MSG_process_daemon(). Daemon processes are automatically killed
1024 when the last non-daemon process terminates
1025 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
1026 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
1027 to make room for the upcoming network energy plugin.
1028 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
1031 - Ensure that an actor can kill itself with Process::exit()
1032 - Kill the obscure NativeException. Nobody want to survive the issues
1033 it denotes, so use JniException that is a RuntimeException (not to
1034 be caught explicitly).
1035 - Partial bug fix in initialization. SimGrid flags on command line were
1036 consumed at C level but stayed in the original Java String[] args.
1037 This could mess users' args[i] if SG flags were not put at the end of
1039 The SimGrid flags are now removed from the Java arguments. However,
1040 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
1041 if args.length is greater than the number of YOUR OWN ARGUMENTS.
1042 It might be if you have --log or --cfg flags in the command line.
1043 - Fix numerous memleaks all around the place. In particular, around VMs.
1047 - s4u::onPlatformCreated: right before the simulation starts
1048 - s4u::onSimulationEnd: right after the main simulation loop
1049 - s4u::onTimeAdvance: right after a clock change (time is discrete)
1050 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1051 event from the availability_file changes the avail speed.
1052 - Links are now usable from s4u
1053 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1054 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1055 - New examples: The conversion of MSG examples to S4U has begun
1056 - Actors: create, kill, migration, and suspend
1057 - Applications: master-worker and token-ring
1058 - Action replay: communications and storage
1059 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1062 - Backwards Compatibility breaks
1063 - The SD_route_get_list and SD_route_get_size functions have been removed.
1064 They are replaced by the unique
1065 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1066 The route from src to dst is built in the links dynar whose size can
1067 be obtained with xbt_dynar_length.
1068 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1069 sg_host_route_bandwidth and sg_host_route_latency.
1070 Macros ensure the backwards compatibility, but you should fix your code
1073 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1074 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1075 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1076 We don't support 100% of it, but it should be enough. Contact us if not.
1077 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1078 naive and probably inaccurate)
1079 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1080 mapped onto a unique small file using some system magic.
1081 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1082 - Fortran: cleanups, fixes, support of user-added operations
1083 - RMA: MPI_Accumulate are applied in correct order.
1084 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1085 - Better support for MPI_IN_PLACE
1086 - Support for MPI_Win attrs and keyvals.
1087 - Support MPI_Comm_group_create, MPI_Type_size_x
1088 - MPI ops should be properly applied to all allowed types
1089 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1090 algorithm by a non failing one from mpich.
1093 - Kill the fifo data container: we don't use it anymore.
1096 - A <cluster> can now be created with different speed values to represent pstates.
1097 see examples/platforms/energy_cluster.xml for an example.
1099 SimGrid (3.14.159) Released December 28. 2016
1101 The previous release was only a crude approximation of a Pi release;
1102 this one should be a bit better.
1104 - Revert a last minute change that broke on Mac OSX.
1105 - Fix the changelog and NEWS files.
1106 - Improve the documentation, in particular of the routing module.
1107 - Integrate some patches flying around in Debian.
1109 SimGrid (3.14) Released December 24. 2016
1112 * The whole documentation was reworked and reorganized. There is
1113 still much room for improvement here, but we're on it.
1116 * We now need python3 for our testing infrastructure.
1117 * Model checking needs libevent
1120 * Now works on FreeBSD
1123 * Privatization now works on FreeBSD
1125 * Privatization is now activated by default in smpirun
1126 You can turn it off with -no-privatize if something goes wrong.
1128 * Call-location tracing for SMPI.
1129 You can add the exact location (filename / linenumber) of an MPI call to
1130 your trace files and slow down or speed up the simulation between two
1131 consecutive calls by using an adjustment file (see the documentation).
1133 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1134 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1135 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1136 values were ignored in some cases. The timings of these functions can now
1137 be significantly different.
1139 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1140 smpi/running-power is renamed to smpi/host-speed
1142 * smpi/grow-injected-times option to enable or disable multiplication of the
1143 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1144 make simulation less precise (but also much faster).
1146 * smpirun script should be (much) faster for large deployments.
1148 * SMPI tracing: fixed issue with poor matching of send/receives.
1150 * Replay: Fix broken waitall
1152 New functions and features
1153 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1155 Dropped / renamed functions and features
1156 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1157 - MSG_mailbox_is_empty() -> Mailbox::empty()
1158 - MSG_mailbox_front() -> Mailbox::front()
1159 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1160 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1161 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1162 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1163 * Don't pass the free_f parameter to property related functions:
1164 - MSG_host_set_property_value()
1165 - MSG_as_router_set_property_value()
1166 - MSG_storage_set_property_value()
1167 * VM properties. Since msg_vm_t are msg_host_t, just use
1168 MSG_host_get_property_value() and friends
1169 * VM I/O related things:
1170 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1171 - Unimplemented save/restore methods
1172 * MSG_as_router_get_property_value() was redundant with
1173 MSG_environment_as_get_property_value().
1174 - Removed MSG_as_router_*propert*() functions
1175 - Added MSG_environment_as_set_property_value() for consistency
1176 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1177 Well, they are still there for now, but deprecated with a warning.
1178 Please switch to xbt_dict_new_homogeneous() before this is removed
1180 * Task affinity. Its intended behavior (that was very badly tested
1181 and probably not really working) was deceiving what most users
1182 would have hoped here.
1183 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1184 * Option network/coordinates is now useless and should be dropped.
1187 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1190 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1191 * Added Dragonfly topology support
1194 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1195 EiBps, EBps, Eibps, Ebps for bandwidth.
1196 They may become useful to some lucky ones.
1199 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1200 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1201 * ex/async_waitAll: new example, on asynchronous communications
1204 * Memory usage should be decreased for simulations with a large number
1205 of processes. This also helps for SMPI.
1207 SimGrid (3.13) stable; urgency=low
1209 The Half Release, a.k.a. the Zealous Easter Trim.
1211 - We removed half of the lines, that were mostly experimental cruft.
1212 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1213 (not counting blanks and comments -- according to openhub.net)
1214 - The internals are now compiled in C++ (and will soon be clean C++)
1215 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1216 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1218 * Backwards Compatibility breaks
1219 - Removed Lua simulation bindings (switch to C or Java for that).
1220 Lua can still be used to describe platforms
1221 - Removed Java kernel plug-ins.
1222 Will be reintroduced after the ongoing major internals reorg.
1224 - the following functions were removed.
1225 They were too specific and should be reimplemented in a generic
1226 way, with filter function.
1227 - MSG_task_listen_from_host
1228 - MSG_mailbox_get_count_host_waiting_tasks
1229 - MSG_mailbox_put_with_timeout was removed.
1230 Please use MSG_task_send_with_timeout instead.
1232 - the SD_application_reinit function was removed. It has been a noop for a while.
1233 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1234 replaced by a more flexible mechanism.
1235 - The following functions thus do not exist anymore
1236 - SD_workstation_get_access_mode
1237 - SD_workstation_set_access_mode
1238 - SD_workstation_get_current_task
1239 - Basic estimation functions have been removed but can easily be replaced
1240 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1241 - SD_workstation_get_computation_time => amount / sg_host_speed()
1243 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1244 Use VM.getSpeed()*load/100 for the legacy behavior.
1246 - option enable_tracing was removed. It was not doing anything for a while.
1247 - In the ModelChecker:
1248 - the model-checker now ptraces the model-checked process which means
1249 you cannot use a debugger on the latter anymore (we might make this
1250 optional in the feature);
1251 - removed soft-dirty page tracking;
1252 - remove model-checked side snapshot management,
1253 MC_snapshot() and MC_compare_snapshot();
1254 - keep the MC_cut() function as a stub (it was not really working
1255 in the previous release).
1258 * All options are consistently in kebab-case. Old names are kept as alias.
1261 * Switch to platform v4 format.
1262 - Rename from 'power' to 'speed' the attributes describing the amount of
1263 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1264 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1265 - In <host> and <link>, attributes availability and state are gone.
1266 It was redundant with state and availability traces, and with peak values.
1267 - In <cluster>, attributes availability_file and state_file are gone.
1268 It was too complex and unused.
1269 - Kill <gpu>. Was not doing anything.
1270 - The DOCTYPE points to the right URL:
1271 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1272 (the file at this address now documents the changelog since its v1)
1273 - A warning is emitted for unit-less values (they are still accepted).
1274 - speed. Default: 'f' or 'flops'. Also defined:
1275 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1276 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1277 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1278 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1279 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1280 - latency. Default: 's' second. Also defined:
1281 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1283 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1284 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1287 * s4u::Host is now the preferred public interface to the Host features.
1288 sg_host_* functions are C bindings to the exact same behavior
1289 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1292 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1293 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1294 They were deprecated since a few years, and probably did not even compile anymore.
1297 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1298 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1299 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1300 encouraged to update their codes. A list of the modified functions can be found at the end of
1301 include/simgrid/simdag.h
1304 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1305 the lambda or closure passed as a parameter will run in kernel mode.
1306 All the callback functions should be rewritten to that interface at some point.
1309 * Reorganizing and cleaning the internals all around the place.
1312 * Remove old default barrier/bcast buggy algorithms (see #18407)
1313 * Various bug fixes to handle more codes
1314 * Remove the need for the --foreground option of smpirun (it is still
1315 accepted for backward compatibility).
1318 * Kill the setset data container: MC don't use it anymore.
1319 * Kill the queue data container: it made more sense with GRAS.
1320 * Kill the xbt_peer_t data type: it's useless without GRAS.
1321 * Kill rm_cb feature of config sets: it was never useful.
1322 * Kill graphxml parsing feature. It was not used.
1323 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1325 - xbt_dynar_sort_strings(), when the content is char*
1326 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1327 They throw exceptions on invalid input;
1328 * C++ support for declaring CLI flags (simgrid::config::Flag);
1329 * class for abstracting different signal backends (simgrid::xbt::signal).
1330 with no external dependencies (we need very simple signals).
1333 * refactoring and cleanup of the code;
1334 * ongoing process to cleanly separate the model-checking algorithms
1335 from the code model-checking support.
1337 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1339 SimGrid (3.12) stable; urgency=low
1341 The Facelift Release.
1344 * Require g++ v4.7 at least to not speak prehistorical C++.
1345 * Require Boost 1.48 (for signal2 component).
1346 * Java must be version 7 at least when activated.
1347 * Builds on Windows again (including Java bindings).
1348 * Tracing is now always enabled (no way to turn it out)
1349 * Remove GTNetS. It was not working anyway.
1350 * Various cleanups in the cmake scripts.
1351 * Move headers around to sort them out on installed systems:
1352 - instr/instr.h -> simgrid/instr.h
1353 - instr/jedule/* -> simgrid/jedule
1354 - simdag/datatypes.h was removed
1355 - simdag/simdag.h -> simgrid/simdag.h
1356 - msg/datatypes.h was removed
1357 - msg/msg.h -> simgrid/msg.h
1360 * Interface improvement:
1361 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1362 - Sanitize the interface in MSG_task_ module:
1363 - Merge two functions that were close enough but misleading:
1364 set_compute_duration(t) -> set_flops_amount(t)
1365 get_remaining_computation(t) -> get_flops_amount(t)
1366 - set_data_size(t) -> set_bytes_amount(t)
1367 get_data_size(t) -> get_bytes_amount(t)
1368 - Massive cleanups in the functions related to the energy
1369 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1370 - New: MSG_host_get_pstate()
1372 - msg/energy/onoff: switching hosts on and off
1374 * Interface improvement:
1375 - Rename Host.isAvail() to Host.isOn()
1376 - Rename Process.currentProcess() to Process.getCurrentProcess()
1377 - Rename Task.setDataSize() to Task.setBytesAmount()
1378 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1380 - #18874: Actually allows the GC to reclaim tasks
1384 - SIMIX_process_throw: raises an exception in a remote process
1385 * Refactoring: Separate sub-modules
1386 - libsmx: the public interface, as libc in a real system
1387 - popping: the strange dance that converts a user request into a kernel handling
1388 - smx_context_*: the virtualization mechanisms that embed the user code
1389 - smx_*: the handling of each simcalls
1390 * Interface cleanup:
1391 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1392 * Rename smx_action_t into smx_synchro_t, making explicit that these
1393 things are used to synchronize processes with their environment.
1394 For example, a communication is a sort of synchronization involving
1395 the communicating processes (that may block until the exchange) and
1396 the platform. The same can be said from computations, etc.
1398 - #18888: segfault when a process ends before its kill_time
1402 - 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.
1403 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1404 - MPI_Type_set_name, MPI_Type_get_name
1405 - MPI_*_c2f and MPI_*_f2c functions
1406 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1407 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1408 - Activate a lot of new tests from the mpich 3 testsuite
1410 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1411 - InfiniBand network model added: Based on the works of Jerome Vienne
1412 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1413 - When smpi/display_timing is set, also display global simulation time and application times
1414 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1415 * Collective communications
1416 - 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
1417 - 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/).
1418 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1420 - Replay now uses algorithms from wanted collective selector
1421 - Replay can be used with SMP-aware algorithms
1422 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1423 - 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.
1425 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1426 - Correctly use loopback on fat-tree clusters
1427 - Asynchronous small messages shouldn't trigger deadlocks anymore
1428 * Energy/DVFS cleanup and improvement
1429 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1430 - new: smpi_get_host_pstate()
1433 - "Full" network optimization flag was broken since Surf++
1434 - Better handling of precision flags in maxmin
1435 - Fix bug causing sometimes "Impossible" errors
1436 - Properly pass cluster properties to included hosts
1437 * Improvement of the Energy plugin.
1438 - Always update the consumption before returning that value
1439 - New property: watt_off to denote the dissipation when the host is off
1440 - New functions getWattMinAt and getWattMaxAt to retrieve the
1441 dissipation of pstates that we are not currently at.
1442 * Java: class NetworkLink renamed to Link
1443 * New function: simcall_process_get_kill_time()
1444 * Massive rename s/workstation/host/
1445 - That's intrusive, but that's good for the project consistency. Sorry.
1446 - Change config option "workstation/model" into "host/model"
1449 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1450 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1451 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1452 - xbt_dirname and xbt_basename for non-POSIX systems
1454 * The model checker now runs as a separate process.
1455 * The model checker runs must now be launched with the new simgrid-mc program.
1456 * Record/Replay: the MC can display a textual representation of a path in the
1457 execution graph. It can then be replayed outside of the model checker.
1459 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1461 SimGrid (3.11) stable; urgency=low
1466 * Normalizing pointers addresses tool for better diff between logs
1469 * Add cloud examples using new VMs
1470 - examples/msg/cloud/two_tasks_vm.tesh
1471 - examples/msg/cloud/simple_vm.tesh
1472 - examples/java/cloud/cloud.tesh
1473 - examples/java/cloud/migration/migration.tesh
1474 * Add java surf examples:
1475 - examples/java/surfPlugin/surf_plugin.tesh
1476 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1477 - examples/java/surfCpuModel/surf_cpu_model.tesh
1478 * Add SMPI+MSG example:
1479 - examples/smpi/smpi_msg_masterslave/
1484 - msg task destroy cancel test
1485 - msg_host on/off test
1486 * Move all tests in testsuite to teshsuite (adding tesh files)
1487 * Restructure teshsuites
1488 - one folder for each kind of test
1489 * Restructure AddTests.cmake
1491 - structure the order of tests (with sections)
1494 * Add virtual machine
1495 - creation of a VM on a PM
1496 - migration of a VM from a PM to another PM
1498 - MSG_process_join(msg_process_t process, double timeout)
1499 - msg_bar_t MSG_barrier_init(unsigned int count)
1500 - int MSG_barrier_wait(msg_bar_t barrier)
1501 - void MSG_barrier_destroy(msg_bar_t barrier)
1502 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1503 * New option "msg/debug_multiple_use" to help debugging when a task is used
1506 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1508 - Increase code coverage in test suites
1512 * Protect context stacks against stack overflow. The number of protected memory
1513 pages allocated on the top of each stack (1 by default) can be configured
1514 with the new command line option --cfg=contexts/guard_size:<value>.
1515 * Simcalls are now generated by a python script that
1516 - generates files included by SimGrid
1517 - checks that all the functions exist, and proposes prototypes
1519 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1520 host_set_data, host_get_data
1522 - simcall_process_join(smx_process_t process, double timeout)
1523 * Fix bug where sleeping processing could not be suspended.
1526 * Translate surf models from C to C++
1527 - Generic classes for all models: Model, Resource, Action
1528 - A generic interface for each kind of model (CPU, Network, Storage
1529 Workstation, WorkstationVM)
1531 * Translate surf routings from C to C++
1532 * Add callbacks using sigc++ or boost::signals2
1533 - Add callback functions for resource creation/destruction
1534 - Add callback functions for action state change
1535 - Handle Energy as a plugin
1536 * Replace swag by boost::intrusive
1537 * Add new routing models for clusters. For documentation, see
1538 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1539 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1540 parameters for cluster tag
1541 - Fat trees, with topology="FAT_TREE" and
1542 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1543 - see examples/platforms/torus_cluster.xml and
1544 examples/platforms/fat_tree_cluster.xml
1545 * More documentation
1548 * Hostfiles support host:nb_processes construct to deploy several processes on
1550 * Collective communication algorithms should not crash if used with
1551 improper number of nodes and report the error.
1552 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1553 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1554 MPI_Dims_create, MPI_Cart_sub are supported.
1555 * New interface to use SMPI programmatically (still depends on MSG for
1556 some parts, see examples/smpi/smpi_msg_masterslave):
1557 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1561 * Global variables privatization in MPI executables is now performed at runtime
1562 with the option smpi/privatize_global_variables (default:no).
1563 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1564 dynamic libraries loaded by the application are not privatized (static
1565 linking with these libraries is advised in this case)
1568 * Options defined in XML work correctly now.
1571 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1572 copied into simgrid.jar or not (ON by default). Use this option if you want
1573 to reduce the size of the installed simgrid.jar, *and* the native libraries
1574 are kept installed elsewhere.
1575 * Surf binding with SWIG (code generated in maintainer mode only):
1576 - plugin to handle callbacks
1577 - CPU model only for the moment
1580 * Supernovae build mode is definitively removed. It was used to improve
1581 inlining and inter-module optimizations. It is nowadays superseded by
1582 link-time optimizations commonly available in compilers.
1583 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1584 * Add boost dependency for surf++
1585 * Add new macro for tests
1586 - ADD_TESH(name <tesh_args>)
1587 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1591 - xbt_bar_t XBT_barrier_init(unsigned int count)
1592 - int XBT_barrier_wait(xbt_bar_t barrier)
1593 - void XBT_barrier_destroy(xbt_bar_t barrier)
1594 * Make the xbt_os_time module public
1596 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1598 SimGrid (3.10) stable; urgency=low
1600 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1603 * Reintegrate Java to the main archive as desynchronizing these
1604 package is not acceptable anymore (Java is now considered stable)
1605 * Add explicit synchronization facilities through semaphores
1606 * Bug fix: Task.setDataSize() only changed the C world, not the value
1607 cached in the Java world
1610 * Dramatically change the way files are handled. API and internals changed, but
1611 this part of MSG was not considered as production grade either.
1612 * Add explicit synchronization facilities through semaphores
1613 * Add a new function MSG_host_get_process_list()
1614 * Preliminary DVFS support (see examples/msg/energy/ for details)
1617 * SMPI is now included directly in the libsimgrid as the windows
1618 linker doesn't force us on splitting it anymore.
1619 * Improvements of the SMPI replay tool:
1620 - Most of the collective communications are now rooted in the same process as
1621 in the original application.
1622 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1623 used until now). Multiple data types can now be used in a trace.
1624 - The replay tool now supports traces produce either by TAU or a modified
1626 - Bug Fix: the compute part of the reduce action is now taken into account.
1627 - Gatherv collective is now supported
1628 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1629 of smpirun outputs time independent traces for the current run. One file
1630 is created per process. If too many processes are simulated, this behavior
1631 can be changed to one file for all processes by using the
1632 tracing/smpi/format/ti_one_file flag
1633 * smpirun generates the host file if needed (with given host count and platform)
1634 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1635 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1636 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1637 --cfg=smpi/coll_selector:(mpich/ompi)
1638 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1640 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1641 happening inside a collective SMPI call.
1642 * Fix the behavior of complex data types handling.
1643 * Make MPI_Wtime another synchronization point to take computations into
1645 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1646 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1647 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1648 (no privatization of global variables yet)
1649 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1652 * Verification of liveness properties is now available for SMPI applications
1653 (in addition to MSG applications)
1654 * Bugged examples using SMPI in examples/smpi/mc/
1655 * Add --cfg=model-check/visited option. Allows the verification of infinite
1656 programs. Detection of loops in the execution thanks to the system state
1657 comparison and reduction of the state space to explore. Can be combined with
1658 DPOR for safety properties.
1661 * Allow to change SimGrid configuration (see --help) within the code
1662 thanks to SD_config() as it can be done in MSG.
1663 * Add a new function SD_task_set_amount() upon user request.
1666 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1667 * Remove rule based routing (no more PCRE dependency)
1668 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1669 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1670 * Add a loopback_bw and loopback_lat options to cluster tag.
1671 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1672 routing. Now peers are encapsulated in an AS and have their own private
1673 router but this is transparent.
1676 * Our own implementation of getline is renamed xbt_getline, and gets
1677 used even if the OS provide a getline(). This should reduce the
1678 configuration complexity by using the same code on all platforms.
1679 * New type: xbt_cfg_elm_boolean.
1680 * Allow to use yes/no for boolean configuration options in the command line.
1681 * Allow to disable SimGrid cleanups at exit from command line option.
1682 There are situations where one may want a simulation to end with an exit.
1683 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1684 annoying when scripting around the simulator. Adding a
1685 --cfg=clean_atexit:no allows to circumvent this issue.
1688 * Lots of memory leaks were corrected in this release.
1689 There is no known memory leaks anymore, in all of our 600+ tests.
1690 * New command line option --version, to get SimGrid version information.
1691 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1693 * Supernovae builds are deprecated, and expected to be removed in the next
1696 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1698 SimGrid (3.9) stable; urgency=low
1700 The Grasgory release: GRAS is really dead now.
1702 * Complete overhaul of the internal host structures scheme.
1705 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1706 was considered as experimental and badly maintained since then.
1707 * Keeping it was thus a trap to our potential users, that could take
1708 it instead of MSG or SMPI by mistake despite is pity state.
1709 * GRAS seems to have very few users (if any), and no one volunteered
1710 to maintain it further. It also induces a lot of XBT code (for
1711 portability sake), that must be maintained too.
1712 * For all these reasons, we killed GRAS. If someone wants to revive it
1713 in the future, don't cry, our git history still remembers of GRAS.
1716 * Major overhaul. Merge our documentation again as time proved that
1717 splitting it was really not helping our users.
1718 * Further improve the developer documentation to help newcomers
1719 hacking on SimGrid itself. The user documentation (and in
1720 particular, the beginner documentation) is still in a sorry state.
1723 * Now works on Windows too!
1724 * Much more extensive test suite, from MPICH
1727 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1728 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1729 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1730 * Bug fix: let task be scheduled when the last dependency to be solved is
1731 a control dependency.
1732 * Remove SD_load_environment_script function.
1733 Use the C sg_platf function if you want to declare a platform
1737 * New function: MSG_process_get_number()
1738 * Old function documented: MSG_config()
1739 * Remove MSG_load_platform_script function
1740 Use the C sg_platf function if you want to declare a platform
1744 * Change the default value of the TCP_gamma constant (maximal size of TCP
1745 congestion window) to a more realistic 4MiB value. If you notice changes in
1746 your simulation results, you can fall back to the previous 20k tiny window
1747 by adding --cfg=network/TCP_gamma:20000 on command line.
1748 * (Hopefully) fix a bug wrt periodic availability and state traces
1749 * Bug fix: use default values at start when first event in availability/state
1750 trace is not at time 0.
1753 * remove the "new_" part of function name sg_platf_new_trace_connect
1754 (resulting in sg_platf_trace_connect), since it does not create
1758 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1759 use case where it's really mandatory, and maintaining it was a pain in
1761 * New: xbt_fifo_search(), search an item with a user-provided
1762 comparison function instead of dumb pointer comparison.
1765 * Fix the lua deployment:
1766 Use `simgrid.init_application()` before deployment instead of
1767 `simgrid.msg_register_application()` after.
1770 * Transfer the tracing files into the corresponding modules.
1772 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1774 SimGrid (3.8.1) stable; urgency=low
1776 The "we are told that some people want to also *install* the simgrid
1779 * Add missing file "tesh.1" to the archive.
1781 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1783 SimGrid (3.8) stable; urgency=low
1785 The Psssshiiiit release: SimGrid jumps into the Cloud.
1788 * Add an experimental interface to manipulate VMs. They are mainly
1789 process groups with very few intrinsic semantic, but they should
1790 allow you to build the semantic you want easily.
1791 * New function: MSG_host_set_property_value()
1792 * New function: MSG_process_on_exit(). To clean memory in all cases.
1793 * Bug fixes that made the host (and link) failures unusable.
1794 * Add a way to auto-restart process when the host in which they are
1795 executing comes back (ON_FAILURE="RESTART" on deployment file,
1796 MSG_process_auto_restart_set).
1797 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1798 please stop using the old ones, they are DEPRECATED.
1800 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1801 Please use MSG_init() instead. (reducing the amount of entry
1802 points in the library helps us).
1803 * Make it impossible to link against the wrong version of the lib
1804 * Deprecate MSG_clean(). No need to call it anymore.
1805 * Function MSG_get_host_number() is not deprecated anymore.
1808 * Split the doc into a user guide and a reference guide.
1809 * Start a developper guide to help people hacking on SimGrid.
1812 * Enable tracing by default. This modules rocks you should use it.
1813 * Remove option custom_flags. Now use environment variables CFLAGS
1815 * Use default cmake things to detect lua instead of home grown ones.
1816 * New option "enable_mallocators" to disable mallocators, for debugging
1817 purpose ("on" by default).
1820 * Bug fixes around the resource failures: don't let the processes
1821 survive the host they are running onto.
1822 * Add an interface to auto-restart processes when the host in which they are
1823 executing comes back.
1824 * Ensures that SIMIX_clean is called automatically. It's not part of
1825 the public interface anymore (bindings should be updated).
1828 * Bug fix for when SD_Simulate is called with a positive value: be careful
1829 when comparing doubles. Sometimes they are different for non significant
1831 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1832 parallel task whose initial work is distributed among host according
1833 to the Amdahl's law. Such tasks are created with a parameter alpha
1834 that corresponds to the non-parallelizable part of the computation.
1835 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1836 between two sets of workstations assuming a 1D block distribution (each
1837 workstation owns a similar share of data) on both sides.
1839 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1840 Data redistribution will be automatically scheduled once parent and child
1841 are both scheduled. The filling of computation_amount and
1842 communication_amount structures is now done seamlessly thanks to the chosen
1844 * New function SD_workstation_dump to display various information
1845 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1846 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1847 bandwidth on the route onto which the task is scheduled and the amount of
1849 To divide the nominal bandwidth by 2, the rate then has to be:
1850 rate = bandwidth/(2*amount)
1851 * Compute tasks that have failed can now be rescheduled and executed again
1852 (from their beginning)
1853 * Increasing source code coverage (src/simdag is now covered at 95.8%
1857 * Re-implement time-independent trace replay using SMPI (at the
1858 smpi_smp_* level) instead of MSG. This should replace
1859 examples/msg/actions/actions.c
1860 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1861 hindexed and structs)
1862 * Implement the exchange of non-contiguous data.
1863 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1864 * Correct behavior of smpi/sender_gap and set its default value to 0
1865 * Add option to asynchronously send small messages to allow better
1866 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1867 specifies the size in bytes under which messages will be asynchronously sent.
1868 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1869 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1872 * New C interface to define a platform: XML is now optional.
1873 For more info, please check include/simgrid/platf.h.
1874 * New interface to define random platforms from the C:
1875 For more info, please check include/simgrid/platf_generator.h and
1876 examples/msg/masterslave/masterslave_platfgen.c
1877 * Export a sg_cmdline dynar containing all the arguments we got from
1881 * Two new tracing options for adding comments to trace file so you
1882 can track your experiments (see --help-tracing for details).
1883 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
1884 * Adding the SimGrid version that generated the trace file as a comment.
1885 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1886 * Fix to avoid key clashes on Paje links
1887 * Other minor fixes related to the Paje specification
1890 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1891 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1892 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1893 * Greatly improve the robustness of mmalloc to user errors (such as
1894 using an area after freeing it, or freeing it twice)
1896 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1898 SimGrid-java (3.8.1) stable; urgency=low
1900 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1901 Warning: all methods are visible, but only some of them are
1902 implemented so far. Check the source (src/jtrace.c)
1903 for further information.
1904 * New module: org.simgrid.msg.File (SimGrid File management functions)
1905 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1906 * Change the meaning of Process.restart: now restart the process from
1907 the beginning, like MSG_process_restart in C.
1908 * Add Process.setAutoRestart: handling of process restart when failed
1910 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1911 you to retrieve the properties of the processes/hosts
1912 * Deprecate Msg.clean(): you can just forget about it now.
1913 * New function Process.getCount(), that only works when compiling
1914 with the not yet released version 3.9 of the C library.
1916 * New context factory based on Coroutines. It mandates a modified JVM
1917 but then, the simulations run about five times faster, and there is
1918 no limit to the amount of processes (beside of the available memory).
1920 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1922 SimGrid (3.7.1) stable; urgency=low
1925 * Restore the prototype of MSG_process_create_with_environment() to
1926 the pre-3.7 situation by removing the kill_time argument.
1927 * Add a MSG_process_set_kill_time() function instead.
1930 * Fix weird behaviors when dealing with parallel tasks.
1933 * Simgrid is now built as a dll.
1934 * Simgrid-java now works on Windows.
1935 * Simgrid-Java is now included into Windows package.
1938 * First pre-build package for MacOSX.
1941 * Fix compilation when using MSG_USE_DEPRECATED.
1942 * Fix some compilation issues on Macs and Windows.
1943 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1944 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1946 We discovered that the Lua console is broken, but we are missing the
1947 manpower to fix it right now. The problem existed in 3.7 too, so we
1948 are not blocking the release for that. Sorry if you depended on this
1949 feature, any help would be really welcome.
1951 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1954 SimGrid-java (3.7.1) stable; urgency=low
1956 The "Java aint got to be bloated and slow" release
1959 * Various internal cleanups and performance improvement
1960 Simulations are expected to run up to twice faster or so
1961 * Make Process.kill(process) an instance method, not a static one
1962 * User processes are not java.lang.Thread subclasses.
1963 This breaks the compatibility (sorry), but previous API was
1964 brain-dead, making it impossible to have non-trivial
1965 initializations in the process constructor.
1966 * Require a full constructor per Process sub-class.
1967 Kinda breaks the compatibility (sorry), but this allows a much more
1968 efficient way to launch the processes at simulation startup.
1969 * Do not embeed our version of semaphores, java 1.5 can be considered
1970 as sufficiently prevalent for us to not dupplicate its features.
1974 * Add examples for almost every part of the API
1975 We spotted and fixed a lot of bugs in the process
1977 * New module: asynchronous communication API
1978 * New function: Process.sleep()
1979 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1980 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1981 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1983 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1985 SimGrid (3.7) stable; urgency=low
1987 The "spring cleanups (before the next Big Project kicks in)" release.
1990 * We can specify the SMPI latency/bandwidth factor with command line
1991 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1992 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1993 You can also use the "config tag" from platform file by adding this line
1994 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1995 (see "example/platforms/tag_config.xml" to use "config tag").
1996 Note that the command line supersedes the platform file configuration.
1997 * Change the correction factors used in LMM model, according to
1998 the latest experiments described in INRIA RR-7821.
1999 Accuracy should be improved this way.
2000 * Use the partial invalidation optimization by default for the
2001 network too. Should produce the exact same results, only faster.
2002 * Major cleanup in surf to merge models and split some optimization
2003 mechanisms from the core of the models. As a result you can now
2004 specify which model to use (e.g., --cfg=network/model:LV08
2005 --cfg=cpu/model:Cas01) and which optimization mode to use
2006 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
2007 Incompatible combinations should err at initialization. See
2008 --help-models for the list of all models and optimization modes.
2009 * The CLM03 workstation model was dropped for simplicity because it
2010 used the deprecated CM02 network model. Use default instead.
2011 * Rename the TCP_gamma configuration option to network/TCP_gamma
2012 * Rename the coordinates configuration option to
2013 network/coordinates, and document it
2014 * Use now crosstraffic keyword instead of the terribly misleading
2015 fullduplex keyword. It is activated by default now in the current
2016 default model, use --cfg=network/crosstraffic:0 to turn it off.
2017 * Ongoing refactoring the model parsing to make XML files optional
2018 See include/simgrid/platf.h for details (still to be completed)
2021 * Major overhaul of the documentation. Almost instructive now :/
2022 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
2023 functions and friends. This interface was considered as
2024 deprecated since over 2 years, it's time to inform our users that it is.
2025 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
2026 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
2027 use these (crappy) functions in your code.
2028 These functions will be removed soon. Stop using them now.
2029 * Deprecate MSG_get_host_{table,number}
2030 Implement MSG_hosts_as_dynar() instead.
2031 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
2032 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
2035 * Stabilize the parallel execution mode of user contexts
2036 * Introduce configuration variables to control parallel execution:
2037 - contexts/synchro: Synchronization mode to use when running
2038 contexts in parallel (either futex, posix or busy_wait)
2039 - contexts/parallel_threshold: Minimal number of user contexts
2040 that must be part of a scheduling round to switch to parallel
2041 execution mode (raw contexts only)
2042 * Fix bugs that prevented to use suspend/resume along with
2043 synchronization structures.
2044 * Fix bugs in process termination that lead to invalid memory access
2045 in very specific conditions.
2048 * Introduce a parallel mode for the models (controlled by surf/nthreads
2049 configuration item). In our tests, running the models in parallel
2050 never lead to any speedups because they are so fast that the gain
2051 of computing each model in parallel does not amortizes the
2052 synchronization costs, even when ultra fast futexes are used.
2053 This is released anyway because YMMV.
2056 * Performance boost by using a swag internally to compute the set of
2057 tasks that are finished and should constitute the return value of
2061 * Enable it by default now that it is considered rather stable.
2064 * Documentation of the tracing functions.
2065 * Performance gains when tracing categorized/uncategorized resource
2066 utilization by avoiding calls to get route when updating resource
2067 variables. LMM constraints are being used instead.
2068 * API changed to set task categories. Use MSG_task_set_category instead
2069 of TRACE_msg_set_task_category, and SD_task_set_category instead
2070 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2071 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2072 * Examples for link user variables added (see at examples/msg/tracing/)
2073 * Deprecated function TRACE_msg_set_process_category completely removed
2074 * Trace header updated according to the latest Paje file format
2075 * Tracing network lazy updates, no longer obligate users to use full updates
2076 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2077 * Experimental: let user code declare/set/push/pop application states for hosts
2078 * API changed to allow the manual creation of graph configuration files
2079 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2082 * Improve the API of Lua MSG bindings, using the Lua spirit.
2083 * Each simulated process now lives in its own Lua world (globals are
2084 automatically duplicated). It helps writing simulators. Will allow
2085 to run Splay programs within SimGrid in the future.
2086 * Add a Chord example in Lua, equivalent to the MSG one.
2089 * Start the implementation of a solution to express temporal
2090 properties, not only local assertions. This is still an
2091 experimental work in progress, stay clear from it to be safe.
2095 - Add new runtime parameters --help-logs and --help-log-categories
2096 to display information about supported logging parameters and
2098 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2099 don't exists anymore.
2100 * Mallocators: allow value NULL for the reset function.
2102 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2103 create homogeneous dictionaries, where all the elements share the
2104 same free function. Non homogeneous dictionaries will be
2105 deprecated in the next release.
2106 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2107 - Multi-level dictionaries are deprecated.
2109 - new function xbt_dynar_search_or_negative() that is useful when
2110 you have less than 2 million elements in your dynar and don't
2111 want of the extra complexity of catching exceptions when the
2112 element is not found.
2114 - Make xbt_os_thread module (for thread portability) public.
2115 Documentation is still to come, sorry.
2117 - Cleanups and simplifications to make it maintainable again.
2118 - Exotic features (such as memalign and valloc) were removed.
2119 - The metadata were extended and improved so that the
2120 model-checker becomes able to explore and inspect the heaps.
2121 - This may induce a performance drop when enable_model-checking is
2122 ON in cmake (even if it's not used in the simulation), but it is
2123 necessary at this point to get MC working.
2125 Turn model-checking OFF if simulation performance matters to you.
2126 Not enabling it at runtime is not enough, disable it in cmake.
2128 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2131 SimGrid (3.6.2) stable; urgency=low
2133 The "Not coding new stuff allows to polish old things" release.
2136 * New bindings to the NS3 packet level simulator (experimental)
2137 * Use the raw (efficient) execution contexts instead of the sysv
2138 (portable) ones when possible.
2139 * libpcre is now mandatory in any cases since not using it led to
2140 severe performance loss and possibly other issues
2141 * Update the XML platforms:
2142 - G5K: include the latest machine in Nancy
2143 - GridPP and LCG: new platforms
2144 * Documentation was partially updated, at least (more to come)
2146 Bug fixes, cosmetics and small improvements
2147 * Free terminated processes before the end of the simulation to avoid
2148 exhausting the memory of users having very dynamic amount of
2150 * Bug fix and cosmetics about canceling non-running tasks
2151 * Bug fix about the dot loader's issues when using libcgraph
2154 * Create an installer for windows with nsis (amd64 and win32)
2155 - Add an hello world project to illustrate simgrid project creation.
2156 - Embed libpcre into the Simgrid installer to avoid
2157 its compilation burden
2158 * The raw execution contexts should work on Apple now
2159 * Port to Windows 64 bits
2160 - Sysv contexts now have an implementation for this arch
2161 - GRAS communication features now support this arch
2162 * Drop support for borland compiler on windows
2163 - this code was not maintained, and we kinda depend on gcc nowadays
2164 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2165 * Fix portability issue on unstable ubuntu: linker became picky on
2168 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2170 SimGrid (3.6.1) stable; urgency=low
2172 The "Oops, we broke Macs too" release
2175 * Fixed contexts detection so that raw ones are used when possible
2176 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2177 leads to a strange error, with user code segfaulting sometimes when
2178 the generated code is not perfectly aligned (which is not
2179 controllable from the user side, depends on the amount of code)
2182 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2184 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2186 SimGrid-java (3.6) unstable; urgency=low
2189 * Split of every thing from simgrid v3.5 into a separate package.
2191 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2193 SimGrid (3.6) stable; urgency=medium
2195 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2198 * Bindings now constitute their own package, separated from the main one.
2199 Rationale: reduce our maintenance nightmare by reducing the module coupling
2200 They will soon be released on their own on gforge.
2202 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2203 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2205 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2206 * It's not quite deprecated for now because we have no replacement,
2207 but it may soon become the case.
2210 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2211 * Fortran: New user-level cache variable to store the rank of the running
2212 process. This improves performance by an order of magnitude.
2213 * C: New coccinelle script to automatically locate and modify global and
2214 local static variables.
2215 * Improved SMPI network model with a sender-side gap to account for multiple
2219 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2220 only say if a communication is finished, no matter whether it succeeded or
2221 failed. You can call MSG_comm_get_status() to know the status of a finished
2223 * New function MSG_task_dsend() to send a task and detach it. When a
2224 communication is detached, you are never notified of its success or failure
2225 and the memory is released automatically once it is finished. This function
2226 is useful when you don't care about the end nor the success of a
2228 * Change the prototypes of action replay. Sorry for inconvenience,
2229 but this is really more efficient this way (and to adapt your code,
2230 you just have to fix the initialization, that shouldn't be too long)
2231 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2232 ever managed to do anything useful with it.
2233 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2234 takes a dynar of communications. It returns immediately and gives the
2235 index of a finished communication (if any).
2236 * New example: a basic implementation of the Chord P2P algorithm.
2239 * New model for multi-core CPUs. You can now use the core attribute to
2240 precise the number of cores of a host. This is a basic model. Every
2241 process running on the host receives at most the power provided in
2242 the DTD (throughput<=power). Total throughput of process cannot exceed
2244 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2245 router linked by an up-link and a down-link (possibly asymmetrical).
2246 This kind of pattern allows to easily build last-mile model style platforms.
2247 Aggregating such patterns in a rule-based AS is thus the technique of
2248 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2249 * New model for Vivaldi routing. We transformed the Vivaldi network model
2250 into a Vivaldi routing model (based on the rule-based model). This allows to
2251 combine Vivaldi based latencies with last-mile platforms.
2254 * Added a check for NaN of IEEE754 infinite in the double entries of
2256 * Introduce a new context factory "raw", highly inspirated from the
2257 ucontext factory, but using manually crafted functions in assembly to
2258 do the work in an efficient manner.
2259 * Allow to change the used context factory at run time, not only at
2260 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2261 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2262 contexts at runtime (only with raw contexts or ucontexts).
2263 * Completely rewrote this module to allow parallel execution of user
2264 processes. Use --cfg=contexts/nthreads:N to execute user processes
2265 with N parallel threads (the default is 1, meaning no parallelism).
2266 * Allow to decide dynamically between sequential and parallel modes.
2267 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2268 processes in parallel only when their number is greater than or equal to P
2270 * Added a check for NaN of IEEE754 infinite in the double entries of
2274 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2275 won't output the state of processes when interrupted with Ctrl-C
2276 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2277 NULL-terminated array. This may solve backward compatibility issues
2278 due to the change to return type of SD_simulate. See also:
2279 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2280 * Add new macros with variable number of arguments.
2281 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2282 - in xbt/asserts.h: xbt_assert
2283 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2284 - in xbt/ex.h: THROWF and RETHROWF.
2285 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2287 * Change xbt_die() to accept a format string with arguments, just like printf.
2288 * New data structure: xbt_lib_t, like a dict but more general and with better
2292 * New configuration options
2293 Options triva/categorized and triva/uncategorized can be used to generate
2294 graph configuration files for Triva visualization tool.
2295 * Configuration option tracing/platform is renamed to tracing/categorized
2296 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2297 * New TRACE_user_link_variable interface:
2298 User provides the name of the link and the tracing variable to attach to it
2299 * the declaration of tracing categories must be done after the environment
2301 * simpler tracing interface, just one way to declare categories
2302 TRACE_category or TRACE_category_with_color, it is up to you
2303 * links in the trace file are again identified by their names
2304 * trace contains the full platform hierarchy exactly as declared using the ASes
2305 * Options tracing/msg/[task|process]:1 groups the process by hosts
2306 for both cases, tasks and processes must have names that are unique during the simulation
2307 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2308 * The experimental option tracing/msg/volume is deprecated
2309 its functionality may be reincorporated if needed
2311 The tracing generates a trace file with unordered timestamped events,
2312 because of the way the core simulator (surf) works. A script available
2313 at the tools directory (fix-paje-trace.sh) can be used to put the events
2314 in order. We have changed the tracing so it can generate ordered timestamped
2315 events in the final trace, but depending on the simulator (and how much time
2316 is simulated) that can lead to a huge memory utilization. It is deactivated
2317 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2319 Build Infrastructure
2320 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2321 - We are trying hard to keep the API stable, but it may happen that
2322 some things change (we're a research project after all, not a
2323 nuclear plant operating system). If such things should happen, you
2324 could rely on that macro to adapt.
2325 - current value: 30600 for 3.06.00, aka 3.6
2326 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2327 a number that can be compared with SIMGRID_VERSION.
2328 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2329 assertions and verbose/debug logging events are disabled at compile time.
2331 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2333 SimGrid (3.5) stable; urgency=medium
2336 * New feature to any SimGrid-based simulator: Model-Checking
2337 Check SIN#1 for more details.
2340 * New Model SMPI (three-interval linear regression for correction factors)
2341 See RR-7426, available at http://hal.inria.fr/inria-00527150
2342 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2343 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2344 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2345 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2346 * See also src/smpi/README
2351 - Tracing API changes: TRACE_start and TRACE_end should not be called
2352 by user-code. They are automatically called by simulators created
2353 with SimDAG, MSG and SMPI if the toolkit is compiled with
2354 tracing_enabled=ON. Categories declaration and utilization remain the
2355 same for MSG and SimDag.
2356 - A function was added to the tracing API to declare categories with
2358 - TRACE_category_with_color (char *category, char *color)
2359 where color must be in the following format
2360 "%f %f %f", red, green, blue
2361 and red, green, blue are float values in the interval [0, 1]
2362 - User can specify NULL as color parameter, or continue calling
2363 TRACE_category (cat)
2364 On that case, the tracing system will define random colors
2365 - The following command-line options are supported:
2366 --cfg=tracing/filename:msg.trace
2367 --cfg=tracing:1 (activate tracing, needed to use others)
2368 --cfg=tracing/platform:1 (categorized resource use)
2369 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2370 --cfg=tracing/msg/task:1 (task creation)
2371 --cfg=tracing/msg/process:1 (process creation, migration)
2372 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2373 --cfg=tracing/smpi:1 (SMPI interface tracing)
2374 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2375 - examples of examples/msg/tracing updated
2377 - DAXLoader and DOTLoader functions can generate tasks with categories
2378 - A new function to attribute a category to SD tasks:
2379 TRACE_sd_set_task_category (SD_task_t task, char *category)
2380 * Tracing the MPI interface implemented by SMPI
2381 - Collective operations are traced with states
2382 - Point-to-Point operations are traced with states/links
2383 - Tracing activated by a parameter "-trace filename" passed
2384 to smpirun during execution (considering that simgrid
2385 is compiled with tracing enabled)
2386 - To run the simulation with gdb, the simulator
2387 accepts --cfg=tracing/smpi:1 to trace SMPI
2388 - tesh files to check if smpi tracing is ok
2389 - See examples/smpi/NAS/DT-trace
2390 * GTNetS tracing re-worked
2391 - adaptation to the tracing system of GTNets to cope
2392 with modifications regarding the fullduplex mode
2393 - new tesh files to check if gtnets tracing is ok
2396 * Asynchronous communications through the functions:
2397 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2398 * New function: MSG_load_platform_script()
2399 to make possible using a lua script instead of XML files to set up platforms
2400 * New function: MSG_set_function
2401 to associate functions to processes, used when bypassing the parser
2402 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2404 Platforms: Add some more examples in examples/platforms
2405 * Grid'5000: see www.grid5000.fr
2406 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2409 * Change the XML format. This is a very important modification. SimGrid
2410 3.5 introduces a new hierarchical format based on the notion of
2411 Autonomous Systems. Compatibility with old format is ensured
2412 through the perl script provided in the install bin directory
2413 bin/simgrid_update_xml.
2414 It is now possible to build platforms with specific routing
2415 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2416 connect several platforms together. We will try to provide soon
2417 set of realistic platforms exploiting these properties (have a
2418 look at examples/platforms/ for the moment).
2419 * Take the opportunity of the XML format change to be a good XML citizen:
2420 rename link:ctn to link_ctn and similar changes (also dealt with by
2422 * Add a new routing scheme (rule-based) using regular expressions. It
2423 enables to have an extremely low memory footprint when the
2424 underlying routing is simple and can be compactly described. You
2425 need to have libpcre4-dev (perl regular expressions) installed if
2426 you want to use this routing scheme.
2427 * Revive the cluster TAG and allow to easily and efficiently (both in
2428 term of memory and speed) connect clusters together. Have a look
2429 at teshsuite/simdag/platforms/ to see how this can be done. With
2430 this tag, you can create clusters with thousands of tasks at no
2431 cost (have a look at examples/platforms/).
2432 Note: clusters are implemented as ASes, so there is no need for an
2433 enclosing AS tag if you have only one cluster in your platform.
2434 * Add new generic functions in the public interface that allows the user
2435 to call SURF 'create_resource' methods from your code (same
2436 functionality as the XML bypass mechanism but with a much lighter
2438 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2439 that uses a piecewise linear approximation to produce better
2440 results when exchanging small messages.
2441 * Add a new parameter to handle correctly full duplex link and account
2442 for interferences between uplink and downlink communications
2443 (activate with --cfg=fullduplex:1).
2446 * Rename the SD_READY (all dependencies are satisfied and task is
2447 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2448 dependencies are satisfied) state.
2449 This prevents a confusion between the notion of "ready to schedule"
2450 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2451 simulated" (SD_RUNNABLE) used by the simulation kernel.
2452 * Change the way a task is considered as ready. Instead of removing
2453 dependencies when a task is done, a counter is decreased. This way,
2454 it is always possible to reach ancestors thanks to the
2455 SD_taks_get_parents function (even after the end of the simulation.)
2456 * Change the return type of SD_Simulate from (SD_task_t*) into
2457 xbt_dynar_t. This function was in handling a dynar internally and
2458 converted it into a NULL terminated array for historical reasons.
2459 * New function SD_dotload(char*) to load a DAG described in dot
2460 format. This loader and the corresponding examples require the
2461 installation of the graphviz library.
2462 * Fix a bug in the management of tasks of size 0 in the surf network
2463 models. This problem was only visible with SIMDAG and you should
2464 thus disregard results produced with earlier versions if you
2465 relied on this feature (some tasks were blocked because of this).
2466 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2467 even though your applications were doing only point-to-point
2468 communications and sequential computations. Now you can really use any
2469 model you want (of course, if you create real parallel tasks, which are
2470 not implemented in most models beside ptaskL07, this will abort).
2471 * Add an example that schedules a DAX on an heterogeneous platform
2472 using a Min-Min strategy.
2473 * New function SD_workstation_get_current_task() that returns the kind
2474 of task currently running on a workstation in the sequential access
2476 * Raise some warnings when unexecuted tasks remains at the end of the
2477 simulation. This is usually caused by cycles in the DAG.
2480 * New function: SIMIX_process_set_function() called by MSG_set_function
2481 * Change the underlying waiting queue in semaphores so that a process
2482 can wait on several of them simultaneously (as in waitany).
2483 * Fix the way to handle tokens in semaphores so that all access patterns
2484 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2485 * kill the dirty pimple SIMIX_message_sizes_output()
2486 Please use (proper) visualization instead
2489 * New data container: setset (set of sets of elements)
2490 * New module: mmalloc (mapped malloc, allowing to have several
2491 independent segments of malloc)
2492 * New function: xbt_dict_cursor_set_data()
2493 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2494 * New function: xbt_dynar_is_empty()
2495 * New function: xbt_fifo_get_last_item()
2496 * Fix xbt_dynar_shrink(): use the right element size.
2497 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2498 uninitialized areas during expand.
2499 * Fix semaphores: previous implementation was severely broken.
2500 * Use library init/fini functions for our initialization.
2501 - you can use logs and other feature as soon as you want in your
2502 code (even before the xbt_init / MSG_init)
2503 - xbt_exit is now a no-op and produce a warning when used.
2506 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2507 benefit from the latest improvement to the simulation kernel.
2508 * Kill measurement sockets for now. If you rely on them, sorry. This
2509 release is not for you. This feature will be reintroduced in the
2510 future, but we cannot delay the release any further.
2511 * New function: gras_msgtype_get_name().
2512 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2513 * Fix (at last) the pmm example: it should not randomly fail anymore.
2515 Build chain: bug fixes and overall polishing
2516 * Cmake is now stable enough. Hence, we killed the autotools.
2517 * Port to windows ( TM :)
2518 * Fix the 'make install' target.
2519 No need to use 'make install-simgrid' anymore
2520 * Introduce a 'make dist' target compiling a *source* archive
2521 'make package' compiles a binary archive
2522 * Compile java files only on need
2523 * Add --cd and --setenv command line options to tesh
2524 * Out of source builds are not fully supported yet, but we are close
2525 * Enable supernovae and optimization flags by default for our users
2528 * Add layer to set up environment directly from lua, without XML.
2529 * The effect of gras_stub_generator can be achieved through
2530 lua too (check examples/gras/console/ping_generator.lua)
2532 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2534 SimGrid (3.4.1) stable; urgency=low
2536 The "Polishing easter eggs is probably a good idea" release.
2537 This is a bug fixes release only.
2541 * Fix a bug preventing the tasks from begin garbage collected.
2544 * Fix a bug occurring when a host involved in a communication fails.
2545 This was not detected properly by the other peer involved in the
2546 communication. Now, it's reported as a network error.
2549 * Warn the user about loop dependencies in data flow of DAX files
2550 * Obey the control-flow dependencies of DAX files
2553 * Add option "enable_smpi" allowing to not compile SMPI.
2554 Probably useful for the (Mac) users experiencing a build error here
2555 * Improve the detection of lua5.1 and ruby1.8
2557 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2559 SimGrid (3.4) stable; urgency=low
2561 The "Easter in Cargese" release. Also known as (major changes):
2563 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2564 ~> bindings were greatly improved
2565 ~> new tracing infrastructure for better visualization introduced
2567 * the "Welcome to configury modernity" release.
2568 ~> we switched from autotools to cmake, and improved our cdash
2571 A more detailled list of changes follow (full detail in svn log).
2573 Java Bindings: Various Cleanups
2574 * (install java-gcj-compat-dev on debian-like to use them)
2575 * Remove put/get: no need to export deprecated interface in Java
2576 Use send/receive instead.
2577 * Cleanup the examples and add a README per directory
2578 * Remove example autoDestination (that's the only way to go now)
2579 * Remove example explicitDestination (was a plain copy of basic)
2580 * Make JniException a runtime exception, so that there is no need to
2581 declare the fact that you may encounter such a beast. I guess that
2582 nobody will ever want to survive such error.
2583 * Create specific errors for each MSG case of failure:
2584 host failure, transfer failure, timeout, task cancelled
2585 * Cleanup the exceptions that may get thrown by each function
2586 * Other internal cleanups in Java bindings. Performance still bad :/
2587 Ruby and Lua Bindings: create them
2588 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2589 * That's new and great, you should try them out.
2590 Same functionalities than Java bindings, only even less polished
2592 * Kill the useless "rate" argument of SD_task_get_execution_time()
2593 Everyone used to provide -1 as a value, it was not used, and the
2594 semantic of a possible use wasn't even clear.
2595 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2596 either as comm costs or compute costs to mean that there is no
2597 such thing for that specific task.
2598 * Add a SD_task_set_name() function
2599 * Fix SD_task_unschedule() on typed tasks
2600 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2601 * In DAX loader, accept useless 'level' attributes to <job> since
2602 LIGO DAGs have them (seem to be to ease graphical representation).
2604 * Add an example masterslave_mailbox.c using send/receive and not
2605 the deprecated put/get interface.
2606 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2607 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2609 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2610 (declare MSG_USE_DEPRECATED to still have the old name)
2611 * Add a MSG_task_set_data() function
2612 * About trace replay (see examples/msg/actions):
2614 - Allow to work with splitted trace files for each process
2615 Give the specific trace file as argument of each process,
2616 and call MSG_action_trace_run(NULL)
2617 You can still have one merged file for all processes.
2618 - Fix implementation of collective operations
2619 * Allow task_execute() on 0-sized tasks (closes #10063)
2621 * This is the first release of SimGrid where SMPI is not considered
2622 beta anymore (even if some corners should still be improved)
2623 * Port over the new SIMIX_network submodule (internal refactoring)
2624 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2625 * Implement more missing elements of the standard:
2627 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2629 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2630 MPI_Scan() MPI_Get_processor_name()
2631 - Added implementation of missing case for Alltoall (warning: it's
2632 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2633 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2634 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2635 SMPI_MPI_Allgatherv()
2636 * Bug fixes include:
2637 - MPI_Waitsome() was broken
2638 - Allow relative includes in smpicc
2639 - Command line cfg argument 'reference_speed' was ignored...
2640 - Some functions did not properly lead to auto-benching of user code
2641 - smpicc passes -O2 by default (just like openmpi one)
2643 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2644 * Bug fixes about timeouts during communications
2645 * add SIMIX_message_sizes_output() as a pimple to write to file the
2646 amount of messages per size. Use gnuplot to get histogram.
2647 Pimple because that's the only user-visible function of simix,
2648 defined directly in xbt.h (irk, sorry)
2650 - Add a SIMIX_sem_get_capacity() function
2651 - Fix interactions with processe resume/suspende
2652 - release_forever() was stupidly broken
2653 - Fix SIMIX_display_process_status() for processes in a semaphore
2654 - Make SIMIX_sem_block_onto() user-visible
2655 * Refactoring context stuff:
2656 - Use pseudo-OOP for better modularity
2657 - reimplement SIMIX_process_kill() without process_schedule() so
2658 that the latter can take as invariant that it is called from
2660 - Merge context_start into context_new for sake of simplicity
2662 * Add a Vivaldi network model, coded live during SUD'10 ;)
2663 * Rename configuration variables to start a hierarchy:
2664 o cpu_model -> cpu/model
2665 o network_model -> network/model
2666 o workstation_model -> workstation/model
2667 * New configuration variables:
2668 o network/bandwidth_factor: correction to bandwidth
2669 o network/latency_factor: correction to latency
2670 o netwotk/weight_S: correction to the weight of competing streams
2671 * Add a long description to the models, that users can see with such
2672 argument on the command line: --cfg=cpu/model:help
2673 * --help-models display the long description of all known models
2675 * config: add the ability to set a default value after registration
2676 Does not override any previously set value (e.g. from cmd line)
2677 * dict: allow to have integer key and data.
2678 When so, you need to use the following functions
2679 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2680 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2681 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2682 In contrary to regular dicts, the key is not malloced before copy.
2683 Mixing scalar and regular elements in the same dict is not tested
2685 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2686 Tracing for Visualization:
2687 * SimGrid is now instrumented in order to generate a trace file for
2688 visualization analysis: to use it, need to compile SimGrid with the
2689 "tracing" option enabled, and instrument the program using SimGrid with
2690 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2691 (among other functions).
2692 * The instrumentation only traces the platform utilization for now
2693 * Documentation to use the tracing functions and how to analyze the
2694 traces with the Triva tool is written.
2695 * More information about: SimGrid FAQ (in the section Tracing Simulations
2698 * We moved to cmake as default build system. Autotools support will
2699 be dropped soon. Check the FAQ for more info about how to use it.
2700 * Greatly improved our cdash/ctest interactions
2701 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2702 * Added memory checking tests with valgrind; lot of memleak fixing.
2703 This may be the first release of simgrid with so few memory issues
2704 * Added code coverage tests.
2705 Our coverage is still improvable, but at least we see it on cdash.
2707 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2709 SimGrid (3.3.4) stable; urgency=low
2711 The "Desktop Grid needs love too" release (also called Xmas release).
2713 Models improvements:
2714 * Major speedup in the maxmin system solving by using lazy evaluation
2715 Instead of solving completely the maxmin system at each iteration,
2716 only invalidate (and recompute) the modified parts.
2717 This new feature is enabled in default models but you can try to
2718 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2719 * Cas01 IMproved as default CPU model
2720 This CPU model is the same Cas01 model, but it uses the
2721 maxmin-selective-update flag and a heap structure to manage
2722 actions on SURF kernel.
2723 It reduces the complexity to find the next action to finish and,
2724 consequently, it's faster than the old Cas01.
2725 This is the new default CPU model (Cas01).
2726 * Rename the old Cas01 model to Cas01_fullupdate
2727 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2728 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2729 * CpuTI (CPU Trace Integration)
2730 A new CPU model whose objective is simulate faster when using
2731 availability trace files.
2732 Instead of using a full featured, over engineered maxmin system for
2733 CPU modeling, this model does the pre-integration of traces files
2734 to calculate the amount of CPU power available, and so, executes
2735 faster than the old CPU models.
2736 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2737 * Use LV08 as default network model since it gives better accuracy
2738 for small messages and shouldn't change things for big ones.
2739 Use --cfg=network_model:CM02 to get the previous behavior.
2742 ******************************************
2743 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2744 ******************************************
2745 * The new CPU model may changes simulations!
2746 The point is that events occurring at the exact same timestamp
2747 are not scheduled in the same order with the old and new
2748 version. This may be enough to completely change the execution
2749 of simulations in some cases.
2750 * The new network model will change simulations!
2751 This new model is more realistic than the previous one, so you
2752 should consider redoing your old experiments with this model.
2753 Sorry for the inconvenience.
2756 * Introduce the supernovae compilation mode
2757 When compiled that way, the whole SimGrid (or almost) is put in a
2758 single compilation unit and compiled in one shoot.
2759 This is to help gcc which has difficulties to inline stuff from one
2761 The speedup seem to be above 15%, althrough more tests are needed on
2762 amd64 to confirm that gain.
2765 * Port of MSG's mailbox on top of SIMIX network
2766 The put/get mechanism was greatly simplified on the way.
2769 * New SIMIX network module. Provides:
2770 - Mailbox: rendez-vous mechanism to find with who you want to speak
2771 - Synchronous send/recv: easier and hopefully faster since the
2772 logic is handled in the maestro process directly now
2773 - Asynchronous send/recv: you dreamt of it? It's here now
2774 Too bad that nobody cared enough to propagate the change to MSG.
2775 * Add semaphores as SIMIX synchronization mechanism.
2778 * new function SD_daxload(char*) to load a DAX file
2779 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2780 * Introduce typed tasks. Specify its kind and cost at creation.
2781 At scheduling, just give where it should be placed, and the cost
2782 for each involved resource is automatically computed.
2783 Existing constructors so far (more to come of course):
2784 - SD_task_create_comm_e2e() for end-to-end communication
2785 - SD_task_create_comp_seq() for sequential computation
2786 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2787 * new function SD_task_dump() for debugging display
2788 * new function SD_task_dotty(task,FILE*) writing to file the info
2789 about the task in dotty format
2790 * SD_task_dependency_exists() can now cope with having one of its
2791 arguments NULL. If so, it tests whether the other argument has any
2793 * Add getters on list of preceding/following tasks:
2794 SD_task_get_parents(task) and SD_task_get_children(task)
2795 * Add getters on amount of workstations and list:
2796 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2797 * Add getter on task kind: SD_task_get_kind(task)
2798 * Update the start_time and finish_time of tasks on completion/failure
2799 * Bugfix: Remove task from state swags when destroyed
2802 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2805 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2806 dynar with one separate thread per value of the dynar.
2807 * Change the prototype of xbt_thread_create(), sorry.
2808 Added a boolean parameter indicating whether we want to join this
2809 thread (used in SG only for now)
2810 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2813 * GTNetS wrappers should now be usable again (and betterly tested too)
2814 * Fix a major regression from 3.2 where the timeout provided to
2815 MSG_task_put_with_timeout() was used as absolute time before which
2816 the comm should be done.
2817 * Start to fix the <cluster> tag.
2818 - Internal links should be good now (beside of the loopback, which
2819 use the private link instead)
2820 - paths to the external world is still rather broken
2821 - the <route:multi> tag is just broken. Actually that's brain-dead.
2822 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2823 to make it less stupid
2824 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2825 * Fix a source-level compatibility glitch from 3.2: after defining
2826 MSG_USE_DEPRECATED, you can use the old name
2827 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2828 * Allow to compile from the SVN with automake 1.11
2829 * Fix some problems when using the "start_time" tag in deployment XMLs.
2830 * Fix #8569: XBT/synchro.h has redundant declarations
2831 * Fix #8563: MSG return values and exceptions
2832 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2833 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2834 Thanks to Robson Peixoto.
2835 * Fix a vicious bug in dictionaries inducing that some elements were
2836 not freed on xbt_dict_free()
2838 Portability report of this version:
2839 * Main portability targets:
2840 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2841 on (amd64/i386/ia64)
2842 - mac leopard on i386
2843 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2844 but nothing critical.
2845 * Other platforms: windows, AIX and others were not tested for this release
2847 Timing report of this version:
2848 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2849 scenario parameters). From 8h to a few seconds in desktop grid settings.
2850 * Supernovae brings about 25% speedup on i386.
2852 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2854 SimGrid (3.3.3) stable; urgency=low
2856 The "Need for Speed" release.
2858 The timings done to validate the 3.3.2 were faulty.
2859 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2861 The problem was a conversion from a manually handled vector to
2862 xbt_dynar_t on the critical path.
2863 xbt_dynar_foreach calls functions, inducing stack management crap.
2865 We inlined these functions and xbt_dynar_foreach is now breath taking.
2866 We also inlined xbt_swag_belong on the way.
2868 Here are some approximate speedup measurements (on master/slaves
2869 simulations lasting between 10s and 20s each):
2870 3.3.1 -> 3.3.2: about same performance
2871 3.3.2 -> 3.3.3: 40% speedup
2872 3.3.1 -> 3.3.3: 40% speedup
2873 3.3.1 with inline patch -> 3.3.3: 30% speedup
2875 Our reading is that the refactoring which occurred in 3.3.2 made us
2876 suffer much more from the xbt_dynar_foreach low performance, but
2877 once we solved this, this refactoring proved to be very performance
2878 effective. From the 40% speedup, somehow, 10% are due to the
2879 inlining and 30% to the refactoring.
2881 That's a pitty that gcc cannot inline functions placed in other files
2882 alone. We have to choose between:
2883 - break the encapsulation (by putting private data structures and
2884 accessors in headers files to help gcc)
2885 - live with low performance
2886 - switch to a decent compiler such as icc (not quite possible).
2888 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2890 SimGrid (3.3.2) stable; urgency=low
2892 The "Simplicity does not preceed complexity, but follows it" release.
2894 The main contributors of this release were (lexical order):
2895 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2898 * Extract the routing logic into its own object.
2899 (was dupplicated in network.c and workstation_LV07.c;
2900 Allows to implement other ways of storing that info)
2901 => kill now useless network_card concept
2902 - Use dynar to represent routes (instead of void** + int*)
2903 - kill link_set (use surf_network_model->resource_set instead)
2904 - Add a command-line option to choose the routing schema to use
2905 - Add three new models:
2906 * Floyd (shortest path computed at initialization)
2907 * Dijikstra (shortest path recomputed all the time)
2908 * Cached Dijikstra (shortest path computed on need)
2909 All these models where contributed by Silas De Munck, and are
2910 described in his ICCS09 paper.
2912 * Simplify model declaration
2913 (less redirections, less function to write when defining a model)
2914 - Factorize stuff between models:
2917 surf_model_resource_set(model)
2918 surf_model_resource_by_name(model, name)
2919 - Unify the types of models in s_surf_model_t (using an union)
2920 - Embeed fields of common_public directly into s_surf_model_t
2921 - Rename model methods:
2922 action_free ~> action_unref
2923 action_change_state ~> action_state_set
2924 action_get_state ~> action_state_get
2925 - Change model methods into functions:
2926 (model)->common_public->action_use ~> surf_action_ref
2928 * Implement a generic resource; use it as ancestor to specific ones
2929 (allows to kill duplicated code in models)
2930 Drawback: timer command don't need no name nor properties;
2931 workstation_CLM03 don't need no properties
2932 (but I guess we can live with those few bytes wasted)
2934 * Improve the action object model
2935 - implement a constructor avoiding dupplicated code about field
2936 initialization in generic_action part.
2938 * Kill the SDP model: it has an external dependency, is deprecated
2939 in flavor of modern lmm models, and didn't compile since a while
2942 * Relocation of the context module from XBT to SIMIX.
2943 (the context were decoupled from the simix processes, duplicating a lot of code)
2944 => a lot of code was factorized
2945 - less overhead is introduced during scheduling
2946 - simpler API for the context factory
2947 - the logic for process creation,destruction and manipulation was simplified
2948 * Simplification of the s_smx_process_t data structure.
2949 => accesing the simix level data associated to a process is faster now,
2950 and the code is a lot more readable.
2953 * Implement some more MPI primitives:
2954 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2955 -implementation: Bcast: flat or 2-ary tree (default),
2956 Barrier: 4-ary tree,
2958 Allreduce: Reduce then Bcast
2959 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2960 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2961 Alltoallv: flat tree, like ompi
2963 * Add support for optimized collectives (Bcast is now binomial by default)
2964 * Port smpirun and smpicc to OS X
2967 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2968 Shout out if you used it.
2971 * Display the list of still queued messages in SG mode when existing
2975 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2976 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2977 * Add xbt_str_from_file(FILE*)
2978 * Add xbt_dict_get_key achieving a linear reverse search
2979 * Remove the context module
2981 Portability report of this version:
2982 * Main portability targets:
2983 - Linux(debian)/x86/context
2984 - Linux(debian)/x86/pthreads
2985 - Linux(debian)/amd64/context
2986 - Linux(debian)/amd64/pthreads
2987 On these, we still have the eratic breakages of gras/pmm and
2988 amok/saturate_sg reported in previous version. We still think
2989 that the tests are the cause of the fault, not the tested code.
2991 - Mac OSX Leopard/x86/context
2992 Still false negative in tesh autotesting.
2993 Smpi still fails, but this time because readlink does not accept -f
2994 Everything seems to work properly beside of that.
2997 - AIX version 5.3 (only tested contexts this time)
2998 Smpi still fails there because mktemp is not installed.
2999 Everything seems to work properly beside of that.
3001 I managed to compile it for the first time, but several breakages.
3002 Won't delay the release for this exotic platform.
3004 * Windows: it's still lagging behind. If you want to help, please
3007 Timing report of this version:
3008 This version seem to be more than 5% faster than 3.3.1 (on linux
3009 64bits with contextes). The gain is less than expected, we are
3010 investigating this for next release.
3012 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
3014 SimGrid (3.3.1) stable; urgency=low
3017 * Implement a --cfg-help to show existing configuration variables
3018 * Build chain do not require doxygen in maintainer mode
3021 * fix a bug on struct sizeof computation, which prevented the
3022 exchange of arrays of structs in some conditions
3023 - added a regression test about this in datadesc_usage
3024 * Allow the exchange of 0-long dynamic vectors.
3025 - for that, use -1 as indicator of dynamic size instead of 0
3026 - This implied to change any size from unsigned long to long,
3027 reducing a bit communication abilities, but I guess that with
3028 64bits being quite common, this is more than enough.
3029 - This also induce a protocol change, thus bumping network protocol
3030 version from 0 to 1 (if we have external users, we have to get
3031 clean on that point too ;)
3032 - added two regression tests about this in datadesc_usage
3033 * Be more verbose when propagating local exceptions
3034 This helps debugging.
3035 * Display the status of simulated processes when receiving SIGINT in
3039 * Allow to control the simulation from a trace file.
3040 New functions MSG_action_register() and MSG_action_trace_run()
3041 The first one allows to associate a function execution to each
3042 kind of action while the second one parses a trace file and
3043 triggers the corresponding actions within the system.
3044 For now, only a toy example is provided in examples/msg/actions
3045 * Add an example of process migration in examples/msg/migration
3046 * Fix a bug in task exchange which broke MSG_task_get_sender()
3047 Add a teshsuite regression test for that.
3048 [Bug: if MSG_task_get_sender() is called after sender exit,
3050 * Fix a bug which prevented suspend/resume to work properly
3051 * Display the status of simulated processes when receiving SIGINT
3052 This fixes a regression of v3.3. due to the introduction of SIMIX
3053 * Bug fixing in failure management:
3054 - trace could not start by a failure at time 0
3055 - failure during communications were not working
3058 * Add SIMIX_process_set_name() to change the name of the current
3059 process in the log messages.
3060 * Store smx_hosts in a dict since we only retrieve them by name
3061 * Move the configuration infrastructure to surf
3064 * Move the configuration infrastructure to surf
3067 * Massive internal cleanups:
3068 - Store internal structures on processes instead of hosts (allows
3069 to have more than one process per host, in addition of being more
3071 - Cleanup the initialization/finalization process
3072 - Kill a whole bunch of unneeded synchronization:
3073 processes run in exclusive manner within the simulator
3074 - Move queues from global tables to process data fields
3076 - now accept -platform and -hostfile arguments
3077 - Pass the right rank value to processes according to the hostfile
3078 * Compile the examples by default, and use them as regression tests
3079 * Implement MPI_Wtime()
3080 * Change the reference speed to a command line option
3083 * TCP_gamma can now be specified as command line option using
3084 --cfg=TCP_gamma:10000000.0
3085 * Change the --surf-path cmd line option into --cfg=path:
3088 * Also include strbuff from xbt.h public header
3089 * xbt_ex_display(): do not free the exception after displaying
3090 This allows to do more with the given exception afterward.
3091 Users should call xbt_ex_free() themselves.
3095 Portability report of this version:
3096 * Main portability targets:
3097 - Linux(debian)/x86/context
3098 - Linux(debian)/x86/pthreads
3099 - Linux(debian)/amd64/context
3100 - Linux(debian)/amd64/pthreads
3101 These targets fail about 1/10 of times on gras/pmm, but we believe
3102 that this is because of the test, not because of simgrid.
3103 amok/saturate_sg fails even more rarely, and the test may not be
3106 - Mac OSX Leopard/x86/context
3107 The test suite still spits tons of errors because some obscure
3108 force prevents us from removing the temporary directories
3109 arguing that they still contain some metadata I've never heard of.
3110 Smpi fails because seq is not installed.
3111 Everything seems to work properly beside of that.
3114 - AIX version 5.3 (both contexts and pthread)
3115 Smpi still fails there because mktemp is not installed.
3116 XML inclusions seems rosty on AIX.
3118 * Windows: it's still lagging behind. If you want to help, please
3121 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3123 SimGrid (3.3) stable; urgency=high
3127 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3128 [Malek Cherier & Mt]
3130 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3133 Until now, GRAS were using MSG as an interface to SURF. It was
3134 quite difficult because both interface have several differences
3135 (MSG channels vs GRAS sockets were the most notable point).
3137 This also opens the gate to SMPI (which should occur soon) and speed
3138 up simulations by to 40% (even if it were not the main goal).
3140 **************************************
3141 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3142 **************************************
3143 The point is that events occurring at the exact same timestamp are
3144 not scheduled in the same order with the old and new version. This
3145 may be enough to completely change the execution of simulations in
3146 some cases. Sorry for the inconvenience.
3148 * Cleanup and upgrade the XML format to push further scalability
3149 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3151 * Improve the testing infrastructure with tesh. Now a very large part of
3152 the code is tested not only by being run but also by checking that the
3153 output match an expected output [Mt].
3155 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3156 is really less memory-demanding, which should allow you to use
3157 larger files in SimGrid [AL].
3159 * Inform valgrind about our contextes, so that it becomes usable
3160 with the default (and more effecient) version of SimGrid
3161 [contributed by Sékou Diakite, many thanks]
3164 * Introduce a listener thread in charge of receiving incoming
3165 messages from the network. It allows to overlap communication and
3166 computation but most notably, it removes some stupid deadlocks due
3167 to the fact that so far, a process could not send and receive at
3168 the same time. This made most non trivial communication schema
3170 * Convert the PIDs from long int to int to match the MSG ones (and
3171 linux ones too) [Mt]
3172 * New function: gras_agent_spawn() to launch a new process on
3173 current host. Only working in simulation for now. [Mt]
3174 * New function: gras_os_hostport() returning a constant form (ie,
3175 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3178 * Make the backtrace of exceptions more human readable [Mt]
3179 * New module: xbt/str [Mt]
3180 a ton of string utility functions (split, join, printf to a newly
3181 allocated buffer, trim, etc)
3182 * New module: xbt/hash [Mt]
3183 SHA1 hashing algorithm (more to come if needed)
3184 * New module: xbt/synchro [Mt]
3185 synchronization tools (mutex and conditions) working the same way
3186 in simulation and in real life (mainly useful for GRAS, but not
3188 * New module: xbt/queue [Mt]
3189 classical producer/consumer synchronization scheme
3190 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3191 (using the classical functions will get serialized) [Mt]
3192 * Make dictionary internal table dynamic. No need to specify its size
3193 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3195 * Make sure the log channels are organized as a tree under windows
3196 (because of ANSI C compatibility issue, any channel were child of
3197 root directly) [Mt].
3200 * Cleaned many thing in surf and fixed a few bugs [AL].
3201 * Add a nice command line configuration mechanism to compose models [AL].
3202 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3203 the previous one (KCCFLN05). It relies on something that looks like
3204 a max-min sharing mechanism but cannot be written as such. A new solver
3205 was thus designed [AL].
3206 * Add a new solver to lmm. Based on Lagrange optimization and
3207 gradient-based descent, it enables to efficiently maximise systems s.a
3209 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3211 This solver enables to propose two new network models for TCP Reno and
3212 TCP Vegas based on Low's work. These models still need to be fully
3213 tested though [Pedro Velho].
3216 * Bug fix in SD_simulate. Now the time bound given as argument is
3218 * Use the new parallel task model (ptask_L07) as default.
3219 * Use the SURF command line configuration mechanism.
3220 * 0-size tasks (for synchronization) should now work.
3222 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3224 SimGrid (3.2) stable; urgency=high
3228 We still experience issues on this platform, but we believe that at
3229 least MSG is usable.
3231 GRAS API BREAKAGE (for simplification purpose, sorry):
3232 * the gras_msgtype_by_name is not used anymore. Instead of
3233 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3234 you can write (and must)
3235 gras_msg_send(toserver, "request", &request);
3236 - If you still want to pass a gras_msgtype_t to the function (to cache
3237 the type and avoid the lookup time), use the gras_msg_send_() variant.
3238 - Impacted functions:
3239 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3240 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3241 * The callbacks are now expected to return 0 when everything went well
3242 (just like the main() function)
3244 GRAS new features and improvements:
3245 * New module mechanism where user code can use per process globals [Mt]
3246 This is similar to gras_userdata_*() functions, but for libraries. It
3247 factorize some code developped over and over in the examples and AMOK.
3248 It has still to be documented and used (only amok/peermanagement is
3250 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3251 loss when they were small enough to fit into the buffer and sent quickly
3252 enough so that they can all get received in one shoot.
3253 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3254 if not found. Use the *_or_null() variant for the old semantic.
3255 * In gras_msg_handle, do not discard messages without callback.
3256 They are probably messages to be explicitly awaited later (ie, proofs of
3257 mis-synchronization in userland since they are sent before being awaited)
3258 No big deal usually.
3259 * gras_socket_meas_send/recv: semantic changed!
3260 The numerical arguments used to be (1) the total amount of data to send
3261 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3262 messages. This was need for the fool willing to send more than MAXINT
3263 bytes on quite fat pipes.
3266 * Do really rename the hostmanagement module to peermanagement. [Mt]
3267 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3268 but this is rather new and this was documented in the module
3269 documentation (poor excuses, I admit)
3270 * Bandwidth measurement semantic changed! This follows the changes to
3271 gras_socket_meas_send/recv explained above.
3274 * A sequential mode has been added to the workstations. When a workstation
3275 is in sequential mode, it can execute only one task, and the other tasks
3276 are waiting in a FIFO. [Christophe Thiery]
3279 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3280 model for SIMDAG. [Christophe Thiery]
3281 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3282 equal to 0, which caused some bad behaviors in
3283 saturated_constraint_set_update. I now use a threshold mechanism like in
3287 * When running manually src/testall, you select specific units [Mt]
3288 testall is the result of our cunit mechanism, and should replace all
3289 the scripty thingy around since bash don't run easily on billware.
3291 * A mallocator system has been added. [Christophe Thiery]
3292 Mallocators allow you to recycle your unused objects instead of freeing them
3293 and allocating new ones.
3295 Documentation update:
3296 * FAQ reworking + New FAQs:
3297 - "Valgrind spits tons of errors!" [Mt]
3298 - "How to repport bugs" [Mt]
3299 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3300 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3301 - Communication time measurement within MSG [AL]
3302 - I experience weird communication times when I change the latency [AL]
3303 * GRAS tutorial [Mt]
3305 - an introduction to the framework and to the used communication model
3306 - an initiatic tour introducing the most proheminent features:
3308 . Lesson 0: Installing GRAS
3309 . Lesson 1: Setting up your own project
3310 o Part 2: Message passing
3311 . Lesson 2: Exchanging simple messages
3312 . Lesson 3: Passing arguments to the processes (in SG)
3313 . Lesson 4: Attaching callbacks to messages
3314 . Lesson 5: Using globals in processes
3315 . Lesson 6: Logging information properly
3316 . Lesson 7: Using internal timers
3317 . Lesson 8: Handling errors through exceptions
3318 . Lesson 9: Exchanging simple data
3319 . Lesson 10: Remote Procedure Calling (RPC)
3320 . Lesson 11: Explicitly waiting for messages
3321 . Recapping of message passing features in GRAS
3322 - A HOWTO section containing:
3323 o HOWTO design a GRAS application
3324 More are due, of course. They will come latter. In the meanwhile, you can
3325 check the examples which are still here.
3327 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3329 SimGrid (3.1) stable; urgency=high
3333 There was a stack corruption somewhere, visible only when optimizing
3334 with these versions. [Vince]
3337 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3338 module that enabled to play with Directed Acyclic Graphs. It is built
3339 directly on top of SURF and provides an API rather close to the old
3340 SG. Some old codes using SG are currently under rewrite to check that
3341 all needful functions are provided. [Christophe Thiery]
3344 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3345 extension of the classical CLM03 model that gracefully handles
3346 failures. This is now the default model for MSG and GRAS. It doesn't
3347 handle parallel tasks yet though. [AL]
3348 * Bug fix: Weights were not correctly set in the network part.
3349 WARNING: This may have resulted in incorrect results with simulations
3350 where there are more than one flow on a given link. [AL]
3353 * After a (long ?) discussion on simgrid-devel, we have decided that the
3354 convention we had on units was stupid. That is why it has been decided
3355 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3356 WARNING: This means that all previous platform files will not work as
3357 such with this version! A warning is issued to ask users to update
3359 A conversion script can be found in the contrib module of the CVS, under
3360 the name contrib/platform_generation/surfxml_update.pl [MQ]
3363 * Bug fix: Processes were started in reverse order, wrt deployment file.
3364 WARNING: if your code relies on this bug, please fix it. [AL]
3365 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3366 being executed on two different locations. [AL]
3367 * Bug fix: Failures are now better supported thanks to Derrick's tests
3368 (there was many failure situations I hadn't thought of and that weren't
3369 correctly handled). [AL]
3370 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3374 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3375 Exception occurring on server-side are propagated back to client (!).
3377 API CHANGE: the callback changed their prototype. Change:
3378 int my_handler(gras_socket_t expeditor, void *payload_data) {
3380 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3381 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3383 * New! function: gras_msg_handleall to deal with all messages arriving
3384 within a given period.
3385 * New! function: gras_socket_server_range to get a server socket in a
3386 range of port numbers (ease to avoid port number conflicts) [MQ]
3387 * New! gras processes display their backtrace when they get a SIGUSR1
3388 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3389 Sweet to debug RL processes [MQ]
3393 - Do not force experiment sizes to be expressed in kb, or it becomes
3394 impossible to measure the latency this way (needs one byte-long tests)
3395 WARNING: this changes the amok_bw_* function semantic. [MQ]
3396 - Implements the link saturation stuff. [MQ]
3397 * Peer management module:
3398 New! module factorizing code that we wrote over and over [MQ].
3401 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3402 - Test units are placed directly into the library code, they get extracted
3403 automatically and placed into the src/testall binary.
3404 - Convert most of the XBT tests to this system.
3405 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3406 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3407 live in the same C code. This required to change a little bit the API
3408 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3409 * New module: sparse graph structure with basic algorithms (this is work
3410 in progress and the API is not considered to be frozen yet). [AL]
3411 * Display more information on backtraces: source line & function names are
3412 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3413 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3414 * Reworked a little bit some #include statements to load only required
3415 headers. Some user code that relied on SimGrid to include stdlib or
3416 stdio may need to include it by themselves. [AL]
3417 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3419 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3420 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3421 but I guess that almost nobody used those parts. [MQ]
3423 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3425 SimGrid (3.0.1) stable; urgency=low
3428 * Unfortunately, I had missed 5 misnamed functions:
3429 xbt_fifo_item_t xbt_fifo_newitem(void);
3430 void xbt_fifo_freeitem(xbt_fifo_item_t);
3431 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3432 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3433 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3434 They're now deprecated. Please use their new versions:
3435 xbt_fifo_item_t xbt_fifo_new_item(void);
3436 void xbt_fifo_free_item(xbt_fifo_item_t);
3437 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3438 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3439 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3441 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3442 * Documentation: xbt_log module unmercifully reworked [MQ]
3443 * Bugfix: there was a problem with the ending of contexts with
3444 the pthread backend. It caused some weird deadlock or behavior
3445 depending on the pthread implementation. [AL]
3446 * Bugfix: get the exceptions raised in the simulator repport where
3447 and why they come from when they are not caught in time [AL, MQ]
3450 * Bugfix: Do repport the error when two non-connected hosts try to
3451 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3454 * Add additionnal checkings on communications. Assert that two
3455 communicating hosts are connected by a set of links... [AL]
3458 * Add additionnal checkings on channel values in communication [AL]
3459 * New: MSG_task_get_source to see on which host a task was generated [HC]
3460 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3461 the number of tasks waiting to be received on channel and sent
3463 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3464 waits for the first task coming from a given host.. [AL]
3466 GRAS new functionnalities: [MQ]
3467 * Enhance the parsing macro to allow the size of multidimentional objects
3468 to be given thru annotations.
3469 * New example (and documentation): Matrix Multiplication a la RPC
3470 (as when I was young!) and fix a bunch of bugs found on the way.
3472 GRAS performance improvements: [MQ]
3474 * Reduce the amount of cbps creation/destruction by making it static to
3475 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3478 * Change libdata to a set so that we can search for stuff by ID (and thus
3479 reduce the insane amount of dict lookups)
3482 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3483 RL communication on top of SG since it's so uneffective.
3484 It may also be used for inter-thread communication in RL, one day.
3485 * Use gras_datadesc_copy() to exchange messages on top of SG
3487 - improve message exchange performance on top of SG
3488 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3489 * Don't exchange on the network the size of the used part of buffer,
3490 instead, specify the possible buffer size to read().
3492 - reduces the amount of read/write calls (one pair per exchange)
3493 - reduces the amount of exchanged data (the size)
3494 - allows to retrieve all arrived data on receiver side, if we don't need
3495 it right now (subsequent read will peek the buffer)
3496 - allows the receiver to proceed with the beginning of the stream before
3497 everything is arrived
3498 - make it possible to build an iov transport (using readv/writev)
3500 - take care of the data with non-stable storage (like stacked data),
3502 * If possible, TCP send uses vector I/O (when writev() is here)
3503 - Don't use it for receive since we send data sizes and data on the
3504 same stream, so we wouldn't be able to chain large amount of chunks
3505 before having to flush the stuff to read the size.
3506 * Rework the transport plugin mechanism to simplify it and reduce the
3507 amount of pointer dereferencement when searching for the right function
3510 * I guess that now, we do almost as few system calls as possible while
3511 doing as few data copy as possible.
3513 To improve it further, we could try to send all the sizes first and then
3514 all the data (to use iov on receiving size), but it's only a partial
3515 solution: when you have 2 dimensional data, the sizes of the second
3516 dimension is data of the first dimension, so you need 3 streams.
3518 I'm not sure the potential performance gains justify the coding burden.
3520 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3522 SimGrid (3.00) stable; urgency=high
3525 * New! Give the possibility to hijack the surf parser and thus bypass
3526 MSG_create_environment and MSG_launch_application. Have a look at
3527 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3529 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3531 SimGrid (2.96) unstable; urgency=low
3536 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3537 This deprecates the xbt_error_t mechanisms.
3538 It modifies (simplifies) all XBT and GRAS API.
3539 MSG API keeps unchanged (exceptions raised by XBT are caught from
3540 within MSG and masked with existing error handling facilities)
3543 * New! Add a FATPIPE model. [AL]
3544 * New! Add a parallel task model. [AL]
3545 * New! Add automatically a loopback interface (in the default
3546 network model) if none was precised.
3549 * Bugfix: MSG_process_resume now works with the current running process.
3551 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3552 * Modification of MSG_task_get_compute_duration. Once a task has been
3553 processed, the value returned by this function is now equal to 0. [AL]
3554 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3555 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3556 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3557 of a m_task. For now, MSG_task_cancel only works with computation
3559 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3560 of the processor (in Mflop/s) regardless of the current load on the
3562 * API Change: use proper naming convention for MSG_getClock and
3563 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3565 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3566 This function changes the priority of a computation task. This priority
3567 doesn't affect the transfer rate. A priority of 2 will make a task
3568 receive two times more cpu power than the other ones. This function
3569 has been added to suit the needs of Nguyen The Loc and hasn't been that
3570 much tested yet. So if it fails, please report it and send me your code.
3572 * API Change: removed all functions and types that were marked "deprecated"
3573 since many months. Renamed MSG_global_init_args to MSG_global_init.
3575 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3577 SimGrid (2.95) unstable; urgency=low
3580 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3581 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3582 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3585 * Bugfix: Allow absolute paths for platform description files [MQ]
3586 * Bugfix: do free the variables after use. Leads to drastic performance
3588 * Implement max_duration (ie, timeouts) on resources [AL]
3591 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3593 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3594 the needed support (timeouts) [AL]
3597 * Implement measurement sockets. You can now get the bandwidth between two
3598 hosts thanks to AMOK (see below). [MQ]
3599 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3600 dynar over the network (yeah) [MQ]
3601 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3602 * simple/static token-ring example. [Alexandre Colucci and MQ]
3603 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3604 in place before [MQ]
3607 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3608 type parsing. "short" and "long" modifiers were also ignored; other
3609 modifier (such as reference level) are still ignored. [MQ]
3610 * Embeed the buffer size within the buffer itself on SG. [MQ]
3611 That way, send() are atomic and cannot get intermixed anymore (at least
3612 the ones which are less than 100k; bigger messages still have the issue)
3613 * Array size pushed by the field, not by the field type (or each
3614 and every long int will push stuff to the cbps) [MQ]
3615 * use select() to sleep since it allows to portably sleep less than one
3618 GRAS (minor cleanups)
3619 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3622 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3623 * type category 'ignored' killed as it was never used and were difficult
3625 * whether a type can cycle or not is now a flag, leaving room for more
3626 flags (as "ignored", if we feel the need one day ;)
3627 * Rename raw sockets to measurement sockets since "raw" has another
3628 meaning in networking community.
3631 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3632 and offers features not belonging to GRAS but that most applications
3633 need. One day, it may be a set of plugins loadable at runtime.
3634 * New module: bandwidth
3635 bandwidth measurement between arbitrary nodes running this module. [MQ]
3637 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3639 SimGrid (2.94) unstable; urgency=low
3641 The first beta release of SimGrid 3 !
3645 * Update the main page and the FAQ. Adding references to gforge.
3648 * Add a gras_os_getpid function.
3651 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3652 * Extend the logs so that they also print PID, hostname, date, ... if
3654 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3655 and kill the old version which were in testsuite/
3656 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3660 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3661 for users. Instead, each programming environment (GRAS, MSG,...) use it
3662 when needed to provide such a feature to users.
3663 Don't shortcut the mechanism or you will also shortcut the virtualization
3664 you need on the simulator.
3668 * Cleanups in configury with regard to compile optimization/warning flags.
3669 Also add -fno-loop-optimize to any powerpc since it's the optimization
3670 killing gcc (< 3.4.0).
3671 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3672 needed by MSG examples complications
3673 * Borrow configury beautifications from PHP
3676 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3677 warning (thanks loris for stumbling into this one).
3678 * Bugfix: stop loading private headers (gras_config.h) from the public
3682 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3683 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3684 RL would imply a DNS resolver, which is *hard* to do in a portable way
3685 (and therefore delayed).
3686 * Implement a real timer mechanism and use it in timing macros. This allows
3687 to avoid rounding errors and get a 0.000005 sec precision in timing
3688 macros. While I was at it, various cleanups:
3689 - allow to declare more than one timed section per file (fix a stupid bug)
3690 - move some private declaration to the right place
3691 - merge conditional execution and timing macros into emulation module
3692 - document the module
3693 - make sure the module cleanups its mess on gras_exit
3694 * Documentation improvements:
3695 - (new) how to compile applications using GRAS
3696 - (new) emulation support (timing macros)
3698 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3700 SimGrid (2.93) unstable; urgency=low
3702 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3705 - Use Paje properly where used. Still to be sanitized properly.
3706 - Portability fix: Add an implementation of the contexts using pthread
3710 - Add xbt_procname(): returns the name of the current process.
3711 Use it to show the current process's name in all logging.
3713 - fix detection of older flex version and the reaction, since we do
3714 depend on modern ones (we use lex_destroy)
3715 - Better separation of SG and RL in the libs: remove all simulation code
3716 from libgras. As a result, this lib is now only 200k when stripped.
3717 Some of the xbt modules may also be duplicated (two sets and such) and
3718 should be cleaned/killed before SG3.
3719 - Insist on using xlC on AIX because of weird problems involving gcc there.
3720 - Cleanup the make remote stuff. This is now done by scripts
3721 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3722 mainly for our private use, but we're working on changing them to user
3725 - Bugfix: flush the socket on close only if there is some *output*.
3726 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3728 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3731 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3733 SimGrid (2.92) unstable; urgency=low
3735 Alpha 3 on the path to SimGrid 3
3739 - New! First try of benchmarking macros.
3740 - New! First try so that gras_stub_generator generate deployment and
3741 remote compilation helpers.
3743 - Bugfix: Initialization fix in msg_test.
3747 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3749 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3751 - API Change: gras_os_sleep to take the amount of seconds as a double.
3752 Accepting an int was error prone since it was the only location where
3753 seconds were coded as such. It leaded to damn rounding errors.
3754 - Bugfix: Hard to belive that timers ever worked before this.
3756 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3758 SimGrid (2.91) unstable; urgency=low
3760 Alpha 2 on the path to SimGrid 3
3764 - Bug fix in the lmm_solver.
3766 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3767 through the function MSG_paje_output.
3768 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3769 - It is possible to bound the rate of a communication in MSG with
3770 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3771 to write it in the changelog).
3772 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3777 - Various cleanups to the autotools stuff
3778 - Begin to move Gras examples to examples/gras/
3779 - Let make distcheck work again (yeah!)
3781 - documentation overhauled using doxygen.
3782 gtk-doc-tools is dead in SimGrid now.
3783 - Automatically extract all existing logging categories, and add the list
3784 to the documentation (long standing one, to say the less)
3786 - Cleanup the known architecture table. Reorder the entries to group what
3787 should be, and use a more consistent naming scheme.
3788 (some of the test dataset are still to be regenerated)
3789 - New! Allow library to register globals on each process just as userdata
3791 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3792 lookup time (for now).
3793 Use it in msg and trp.
3794 This cleans a lot the internals and helps enforcing privacy of the
3795 headers between the gras components.
3796 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3797 - Bugfix: gras_os_time was delirious in RL.
3798 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3800 - Reenable GRAS now that it works.
3802 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3804 SimGrid (2.90) unstable; urgency=low
3806 Alpha 1 on the path to SimGrid 3
3808 * It is a long time since the last release of SimGrid. I'm sorry about
3809 that but as I had told you, I was rewriting a lot of things. I apologize
3810 to those who had been reporting bugs to me and that I had not answered.
3811 If your bug is still in the new version, please tell me. Here is a
3812 summary of the main changes.
3814 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3815 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3816 portability, speed, and a lot more but you'll figure it out later.
3817 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3818 I wanted to release the new MSG as soon as possible and I have
3819 broken GRAS, which is the reason why, for now, only MSG is fully
3820 functional. A laconic description of these projects is available
3821 in the documentation.
3823 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3824 kernel optimized for our needs (called SURF but only the developers
3825 should use it). Hence, MSG is now roughly 30 times faster and I think
3826 that by rewriting a little bit MSG, I could event speed it up a little
3827 bit more. Beside the gain in speed, it is also much easier to encode a
3828 new platform model with SURF than it was with SG. More to come...
3830 * REVOLUTION 3: I have tried to change a little as possible the API of
3831 MSG but a few things really had to disappear. The main differences
3832 with the previous version are:
3833 1) no more m_links_t and the corresponding functions. Platforms are
3834 directly read from an XML description and cannot be hard-coded
3835 anymore. The same format is used for application deployment
3836 description. The new format is described in the documentation.
3837 Have a look in tools/platform_generation. There is a tiny script
3838 that converts from the old platform format to the new one. Concerning
3839 the application deployment format, parsing the old one is tricky.
3840 I think most of you should however be able to convert your files. If
3841 it is really an issue, I can write a C code that does the conversion.
3843 2) the toolbox tbx does not exist anymore. We now have a library
3844 with much more data-structures but without the hash-tables (we have
3845 dictionaries that are much faster).
3847 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3849 *****************************************************************************
3850 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3851 * don't feel like converting the oldies. (Mt) *
3852 *****************************************************************************
3855 Version 2.90: "the long awaited one"
3856 - Finished rewriting and debugging MSG. Rewrote the documentation.
3857 - disable GRAS for now since it needs to be ported to the newest SG
3860 - Finish the port to windows (using mingw32 for cross-compile)
3863 - Main loop and datastructures of SURF. A cpu resource object is
3864 functional. Surf can thus be used to create cpu's with variable
3865 performance on which you can execute some actions.
3867 2004-11-15 Martin Quinson
3868 - Port to ARM. Simply added the alignment and size descriptions. Should
3869 work, but the ARM machines are so slow that I didn't had the opportunity
3870 to 'make check' over there yet.
3872 2004-11-15 Arnaud Legrand
3873 - Trace manager now written. It uses a heap structure and is therefore
3874 expected to be efficient. It may however be speeded up (particularly
3875 when many events occur at the same date) by using red and black
3876 trees. One day maybe...
3877 - Max-min linear system solver written. It uses a sparse matrix
3878 structure taking advantage of its expected use. Most operations are
3879 O(1) and free/calloc are called as few as possible. The computation of
3880 the minimum could however be improved by using a red and black tree
3883 2004-11-03 Arnaud Legrand
3884 - Rename every gras_* function that was in xbt/ to its xbt_
3886 - Add a heap and a doubly-linked list to xbt
3887 - Added a dichotomy to the dictionaries. make check works as well before
3888 so I assume that the patch is correct. I do not know however if things
3889 run effectively faster than before now. :)
3891 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3892 SimGrid, and the version number is bumped to 2.x
3894 2004-10-29 Martin Quinson
3895 - Introduction of the remote errors.
3896 They are the result of a RMI/RPC on the remote machine.
3897 ErrCodes being scalar values, you can't get the host on which those
3898 errors did happen. Extending the error mechanism as in Gnome is possible.
3899 No idea yet whether it is a good idea.
3901 2004-10-28 Martin Quinson
3902 - Interface revolution: the Starred Structure Eradication.
3903 I used to do typedef struct {} toto_t; and then handle *toto_t.
3904 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3905 - ???_t is a valid type (builded with typedef)
3906 - s_toto_t is a structure (access to fields with .)
3907 - s_toto is a structure needing 'struct' keyword to be used
3908 - e_toto_t is an enum
3909 - toto_t is an 'object' (struct*)
3911 typedef struct s_toto {} s_toto_t, *toto_t;
3912 typedef enum {} e_toto_t;
3913 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3914 s_toto_t) is private.
3916 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3917 so that it can set the variable to NULL. It was so for dicts and sets,
3918 it changed for dynars.
3920 - Fix a bunch of memleaks in dict_remove
3921 - Fix a bug in sg/server_socket opening: it failed all the time.
3923 2004-10-07 Martin Quinson
3924 - Speed up dynar lookup operation a bit.
3926 gras_dynar_get is dead.
3928 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3929 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3930 (which gives you the address of the stored data).
3932 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3933 copy of the data using an affectation to do the job and not a memcpy.
3935 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3936 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3938 It does not really speedup the dynar test because they are
3939 setting elements all the time (and look them seldom). But the dict does
3940 far more lookup than setting.
3942 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3944 2004-10-05 Martin Quinson
3945 - Allow to (en/dis)able the cycle detection at run time.
3947 Whether we should check for cycle or not is now a property of each
3948 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3949 datadesc_cycle_unset allow to remove this property when previously set.
3951 Note that the cycle detection is off by default since it impacts the
3952 performance. Watch the data you feed GRAS with ;)
3954 This property is hereditary. Any element embedded in a structure having it
3955 set have it set for the time of this data exchange.
3957 You should set it both on sender and receiver side. If you don't set it on
3958 sender side, it will enter an endless loop. If you forget on receiver
3959 side, the cycles won't be recreated after communication.
3961 - Header reorganization.
3962 Kill gras_private.h, each submodule must load the headers it needs.
3964 2004-10-04 Martin Quinson
3965 - Interface revolution: do not try to survive to malloc failure.
3967 Now, gras_malloc and friends call gras_abort() on failure.
3968 As a conclusion, malloc_error is not a valid error anymore, and all
3969 functions for which it was the only gras_error_t return value are
3970 changed. They now return void, or there result directly.
3971 This simplify the API a lot.
3973 2004-09-29 Martin Quinson
3974 - Re-enable raw sockets.
3975 Created by gras_socket_{client,server}_ext;
3976 Used with gras_raw_{send,recv}
3979 It should allow to kill the last bits of gras first version soon.
3981 This is not completely satisfactory yet (duplicate code with
3982 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3985 - Simplify transport plugin (internal) interface by not passing any
3986 argument to _server and _client, but embedding them in the socket
3989 2004-09-28 Martin Quinson
3990 - Finish the port to AIX.
3991 autoconf was my problem (segfault within the malloc replacement
3992 function. No idea why)
3994 2004-09-16 Martin Quinson
3995 - Fix some size_t madness on 64bit architectures.
3997 2004-09-08 Martin Quinson
3998 - Reduce the number of system headers loaded, overload some more system
3999 calls (such as malloc to cast the result of the system one, and work
4001 - Fix and reintroduce the config support
4003 2004-09-07 Martin Quinson
4004 - Source code reorganization to allow Arnaud to surf all over there.
4005 - Allow to document the logging categories.
4006 - Remove all uppercase from logging categories and useless cleanup in names.
4008 2004-08-18 Martin Quinson
4009 Version 0.6.2 (protocol not changed; API changed)
4010 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
4011 gras_error_t), and NULL when not found. Functions expecting a msgtype
4012 as argument (msg_wait; msg_send) deal with NULL argument by providing a
4013 hopefully usefull message.
4014 - Portability to prehistoric sparcs again
4016 2004-08-17 Martin Quinson
4017 Version 0.6.1 (protocol not changed; ABI not changed)
4018 - prealloc some buffers to speed things up
4020 2004-08-11 Martin Quinson
4021 Version 0.6 (protocol not changed; ABI expended)
4022 - The parsing macro can deal with the references, provided that you add
4023 the relevant annotations (using GRAS_ANNOTE(size,field_name))
4025 2004-08-09 Martin Quinson
4026 Version 0.5 (protocol not changed; ABI changed)
4027 - Allow to off turn the cycle detection code in data exchange at
4028 compilation time. It should be at run time, but I'm short of time (and
4029 the config stuff is still broken). That way, we keep dict out of the
4030 critical path, which is good because the performance is poor:
4031 - search not dichotomial yet
4032 - dynar give no way to access their content and memcpy everytime
4033 - In composed data description (struct, ref and so on), stop foolness of
4034 keeping the subtype's ID, but store the type itself. This keeps sets out
4035 of the critical path, which is good since they rely on dynar and
4036 dictionaries. The only loose of that is that we cannot detect the
4037 redeclaration of a structure/union with another content (but I'm not sure
4038 the code detected well this error before anyway). We still can detect
4039 the redefinition discrepancy for the other types.
4040 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
4041 it breaks the code because of type-punning used all over the place).
4042 This breaks on all non-gcc architectures (for now).
4044 All those changes (plus the buffer of last time) allow me to gain 2 order
4045 of magnitude on cruel tests consisting of 800000 array of integers on two
4046 level of a hierarchical structure (200 secondes -> 4 secondes)
4049 - the selector of reference must now return the type it points to, not
4050 the ID of this type.
4052 2004-08-06 Martin Quinson
4053 Version 0.4 (protocol changed; ABI not changed)
4054 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4055 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4056 - Implement a buffer transport, and use it by default (it relies on tcp in
4057 real life and on sg in simulation).
4058 That's a bit hackish since I had a new field to the structure to store
4059 its data without interfering with the subtype ones. Inheritance
4060 is tricky in C. And that's a kind of reverse inheritance with one class
4061 derivating two classes. Or maybe a game with java interfaces. Anyway,
4062 that's damn hard in C (at least).
4063 Moreover, I got tired while trying to ensure plugin separation and
4064 genericity in SG mode. MSG wants me to do weird things, so let's go for
4065 cruel hacks (temporarily of course ;).
4066 See comment in transport_private.h:71
4067 - do not include all the _interface headers in private but in the files
4068 which really need them (to cut the compilation time when they are
4071 2004-07-26 Martin Quinson
4072 Version 0.3 (protocol not changed; ABI changed)
4073 - Major overhault of the datadesc interface to simplify it:
4074 - shorted the function names:
4075 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4076 - add a trivial way to push/pop integers into the cbps without malloc.
4077 This allows to make really generic sub_type description, which simply
4078 pop their size of the stack.
4079 - add a function gras_datadesc_ref_pop_arr() which does what users want
4080 most of the time: Declare a dynamic array (which pops its size of the
4081 stack) and declare a reference to it. Poor name, but anyway.
4082 - kill the post-send callback, add a post-receive one
4084 2004-07-23 Martin Quinson
4085 Version 0.2 (protocol changed; ABI changed)
4086 - add some testing for cpbs in the test cases, and fix some more bugs.
4087 This invalidate again the little64 data file, since I cannot regenerate
4089 - remove an awfull optimization in the logging stuff, allowing me to:
4090 - understand it again
4091 - learn gcc how to check that the argument match the provided format
4092 - fix all errors revealed by gcc after that
4093 - internal keys of dict are not \0 terminated. Deal with it properly in
4094 loggings instead of segfaulting when the user want to see the logs :-/
4096 2004-07-22 Martin Quinson
4097 - Fix some stupid bug preventing cbps (callback postit) from working
4099 2004-07-21 Martin Quinson
4100 - Some documentation cleanups
4101 - remove the useless last argument of msgtype_declare
4102 - rename the Virtu functions to fit into the 'os' namespace
4103 - move headers src/include -> src/include/gras/ and stop fooling with
4104 gras -> . symbolic link
4105 - make distcheck is now successful
4107 2004-07-19 Martin Quinson
4109 - Build shared library also
4110 - Install html doc to the right location
4111 - stop removing maintainer files in make clean
4112 - build tests only on make check
4114 2004-07-13 Martin Quinson
4116 - No major issue in previous version => change versioning schema
4117 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4118 corresponding dataset.
4120 2004-07-11 Martin Quinson
4122 - Get it working with any kind of structure (we can compute the padding
4123 bytes remotely for all the architectures I have access to)
4124 - Implement the structure parsing macro (still not quite robust/complete)
4125 - Improvement to the remote testing toysuite
4127 2004-07-10 Martin Quinson
4128 [autoconf mechanism]
4129 - get ride of a bunch of deprecated macros
4130 - actually run the test for two-compliment, not only compile it :-/
4131 - test whether the structures get packed (and bail out if yes. Damn.
4132 Alignment is a serious matter)
4133 - test whether the structures get compacted (but respecting the alignment
4134 constraints of each types)
4135 - test whether the array fields of structures can straddle alignment boundaries
4137 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4138 (took me 2 hours to find that bug, looking at the wrong place)
4139 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4140 before sending/receiving and are used to compute the offsets of fields
4141 - Given that padding size depend even on compiler options, keep track of
4142 alignment and aligned_size only for the current architecture. Not a big
4143 deal since we send structure fields one after the other (seems
4145 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4146 program, let it work on linux/gcc/little32. portability todo.
4148 2004-07-08 Martin Quinson
4149 - import and improve remote compilation support from FAST
4150 - make sure make check works on half a dozen of machines out there
4152 2004-07-07 Martin Quinson
4153 Let's say it's version 0.0.3 ;)
4154 - Implement conversions (yuhu!)
4155 - Let it work on solaris (beside conversion, of course)
4156 - Stupid me, using rand() to generate the conversion datatests in not wise.
4158 2004-07-06 Martin Quinson
4159 - Let make dist work, since I'm gonna need it to compile on remote hosts
4160 - Let Tests/datadesc_usage write the architecture on which the file was
4161 generated as first byte.
4162 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4163 architecture descriptions.
4164 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4165 on those architectures.
4166 - Optimization: send/recv array of scalar in one shoot
4168 2004-07-05 Martin Quinson
4169 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4171 - Plug a whole bunch of memleaks
4172 - each process now have to call gras_{init,exit}. One day, their log
4173 settings will be separated
4174 - Continue the code factorisation between SG, RL and common in Transport.
4176 2004-07-04 Martin Quinson
4178 - Redistribution between SG and RL.
4179 We wanna have to accept in SG, so move accepted related parts of RL in
4180 the common part. (more precisely, the dynar of all known sockets is no
4181 more a static in transport.c, but part of the process_data)
4183 [gras_stub_generator]
4184 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4186 2004-07-03 Martin Quinson
4187 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4188 them all up in one shot)
4190 - Fix the ugly case of reference to dynamic array.
4191 - New (semi-public) function gras_datadesc_size to allow the messaging
4192 layer to malloc the needed space for the buffer.
4194 - gras_socket_close now expect the socket to close (and not its address to
4195 put NULL in it after it). This is because the socket passed to handlers
4196 is one of their argument (=> not writable).
4198 - propagate the interface cleanup from last week in datadesc, ie remove a
4199 superfluous level of indirection. User pass adress of variable
4200 containing data (both when sending and receiving), and not of a variable
4201 being a pointer to the data. Let's say that I like it better ;)
4202 The price for that is constructs like "int msg=*(int*)payload" in
4203 handlers, but it's a fine price, IMHO.
4205 - Let it work in RL (yuhu)
4207 2004-06-21 Martin Quinson
4209 - porting SG plugin and SG select to new standards (works almost).
4210 - plug memleaks and fix bugs around.
4213 - cleanup the prototype of data recv and force users to specify when they
4214 want to handle references to objects. Test case working even for cycles.
4215 - plug memleaks. Valgrind is perfectly ok with this.
4217 2004-06-12 Martin Quinson
4219 - cleanup the separation between plugin and main code in plugin creation
4221 2004-06-11 Martin Quinson
4223 - Reput hook for raw sockets, needed for BW experiments
4224 - kill a few lines of dead code
4225 [Data description] Interface cleanup
4226 - gras_datadesc_by_name returns the searched type or NULL.
4227 That way, no variable is needed to use a type desc once, which makes
4229 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4230 two parameters were strings (field name, type name), leading to
4232 [Dicos] Interface cleanup
4233 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4234 This is consistant with the dynar API.
4236 2004-04-21 Martin Quinson
4238 - Porting to new standards.
4240 - interface cleanup.
4241 There is no bag anymore, no need to take extra provision to mask the
4242 pointers behind "ID".
4243 Better splitup of functions between files create/exchange/convert.
4244 This is still a bit artificial since convert and receive are so
4245 interleaved, but anyway.
4247 - add a queued message list to procdata (the ones not matching criteria
4249 - factorize some more code between SG and RL wrt procdata
4251 - use gras_exit in example to track memleaks
4252 - get rid of gs_example now that GS is properly integrated into gras
4253 - update run_test to integrate the latest tests (datadesc)
4255 - rename WARNINGn macros to WARNn since it prooved error-prone
4257 2004-04-19 Martin Quinson
4259 - register init/exit functions within gras module mechanism
4260 - send/receive function.
4261 Convertion is not implemented, but short-cutted if not needed.
4262 struct/array elements are sent one by one (instead of block-wise), but
4263 nobody really cares (yet). Get a prototype before optimizing.
4264 - tests (using a file socket) for DD send/receive on:
4265 - base types: int, float
4266 - array: fixed size, string (ie ref to dynamic string)
4267 - structure: homogeneous, heterogeneous
4268 - chained list, graph with cycle
4269 Believe it or not, valgrind is not too unhappy with the results. The
4270 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4274 - Bugfix when using a filename explicitly (instead of '-')
4276 2004-04-09 Martin Quinson
4278 - factorize more code between RL and SG in socket creation
4279 - Complete the implementation and tests of:
4281 o file (only in RL, and mainly for debugging)
4283 I lost 3 days to design a portable address resolver, and then decided
4284 that the prototype mainly have to run on my box.
4285 Addressing portability too early may be like optimizing too early :-/
4287 - use gras_init in the Tests instead of the crappy parse_log_opt
4288 (the latter function is removed)
4289 [Conditional execution]
4290 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4291 [Code reorganisation]
4292 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4293 Build examples against the RL library, since there is no way to disable
4294 its creation for now.
4296 For information, the beginning of coding on GRAS was back in june
4297 2003. I guess that every line has been rewritten at least twice since