1 ----------------------------------------------------------------------------
3 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
5 The Palindrom Day Release.
7 Important user-visible changes:
8 - Improve the Python usability (stability and documentation).
9 - A nasty synchronization bug was ironed out, see also below.
10 - Python's doc was organized as a separate tree, now integrated with C++.
11 - C bindings of S4U were not part of the doc.
12 - The C++ doc was also improved as methods are now split by theme.
13 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
14 - This is mandatory to use the Java bindings.
15 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
16 - SimDAG++: Automatic dependencies on S4U activities (experimental)
17 - Some features are already implemented but not all of them
18 - Cannot block an activity until it's scheduled on a resource
19 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
20 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
25 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
26 - Drop support for Viva/Triva (old visualization tools), hence removed
27 TRACE_get_node_types() and TRACE_get_edge_types() functions.
30 - Require pybind11 version 2.4.
31 - Greatly improve locking during context switches with Python.
32 - Declaring Python callbacks to this_actor::on_exit is now working.
33 - No more tests are failing currently, even on win32.
36 - Actor: Merge signals on_migration_start/end into on_host_change
37 - Actor: Rename migrate() into set_host()
38 - Disk: Allow users to get the read and write nominal bandwidth values
39 - Exec: Implement wait_for(timeout)
40 - Io: Implement wait_for(timeout)
43 - Parse errors now raise a simgrid::ParseError that you may want to catch.
46 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
47 been renamed to fix a spelling error. As usual, the previous definitions are
48 kept with a deprecation warning. However, to avoid runtime errors, it is now
49 forbidden to override these deprecated methods in a derived class. Please use
50 the new names immediately if you need to override them.
53 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
54 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
55 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
56 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
57 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
58 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
59 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
60 - New module: random, providing classical random numbers generators.
63 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
64 allocations' size above which they are considered "shared" by default (as if
65 they were performed through SMPI_SHARED_MALLOC macros).
66 Default = 0 = disabled feature.
67 Note : malloc, calloc and free are now overriden by smpicc/cxx by default.
68 This can cause some troubles if codes are already overriding these. If this
69 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
70 help, but will make this feature unusable.
71 - Input check is now performed much more consistently and outputs precise
72 warnings for each error found
73 - New supported calls : MPI_File_set_view, MPI_File_get_view
74 - MPI I/O now only supports disks, as storage is discontinued.
75 - MPI I/O allows opening files without specifying absolute path
76 - Note : SMPI configuration options are now initialized by smpi_init_options()
77 call, which should be called if SMPI is not being used through smpirun.
79 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
80 - FG#9: Python bindings crashing
81 - FG#39: Missing s4u::Comm::wait_any_for example
82 - FG#42: Add support for ThreadSanitizer (TSan)
83 - FG!19: Removing RngStream
84 - FG!20: A module for RNG calls
85 - FG!21: Choice between ad-hoc and standard distributions implementations
87 - FG!25: Fix link in Mutex doc
88 - FG!26: Fix links in SMPI interface doc
89 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
90 - GH#305: Unscheduled tasks are still excuted
91 - GH#313: smpirun: manual outdated w.r.t. --help
92 - GH#321: [S4U] Get task remaining work ratio
93 - GH#323: Crash when an actor turn off his physical host
94 - GH#335: Missing links on dragonfly example svg
95 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
96 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
98 ----------------------------------------------------------------------------
100 SimGrid (3.24) October 9. 2019
102 The Clean Disk Release.
105 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
106 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
107 thus becomes the default resource to execute I/O operations. Examples
108 and the file system plugin have been modified to use disks instead of
109 storage elements. Storage can still be used but is doomed to disappear.
110 Users are thus adviced to update their simulators accordingly.
111 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
112 for consistency with pthread_barrier_wait()
113 - Host::get_englobing_zone() returns the englobing netzone
114 - Actor::on_destruction is now called in the destructor
115 Actor::on_termination new signal called when the actor terminates
117 - Global signals are now part of the Engine:
118 - on_platform_creation: after config settings, before the XML parsing
119 - on_platform_created: right after the XML parsing
120 - on_time_advance: each time the clock advances
121 - on_simulation_end: after simulation, before cleanups
122 - on_deadlock: as the name implies.
124 - sg_{actor,host,link}_{data,data_set}() now all exist.
125 Use them to attach user data to the object and retrieve it.
128 - Introduce an experimental Wifi model. It sounds reasonable
129 according to the state of the art, but it still has to be properly
130 validated, at least against ns-3 if not against reality.
131 - Improved the usability of ns-3. Several bugs were ironed out.
132 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
133 It is now expressed as 'idle:epsilon:allCores' to properly model the
134 consumption of non-whole tasks on mono-core hosts. More info in the doc.
137 - convert a new set of functions to the S4U C interface and move the old MSG
138 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
141 - Fortran bindings for DVFS have been removed.
142 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
143 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
144 - SMPI can now be selected by cmake's find_module(MPI) with
145 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
146 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
147 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
148 - trace-call-location can be used with TI traces, and replayed, and is compatible
149 with smpi/comp-adjustment-file.
150 - sleep events are now correctly traced/replayed.
151 - Default for trace-call-location is now to use file names and not full paths.
152 To revert to previous behavior (in case of collision of filenames), option
153 "smpi/trace-call-use-absolute-path" can be set to yes.
156 - Use the included xxHash as an hash implem when C++14 is usable.
157 - Option model-checker/hash was removed. This is always activated now.
158 - New option smpi/buffering controls the MPI buffering in MC mode.
159 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
160 This is useful to check for MPI compliance.
163 - New section on plugins: how to define a new one, and existing ones.
166 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
167 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
170 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
171 and <mount> tags. The rationale is finish to decouple what should be handle
172 by SimGrid kernel (i.e., resources and activities) from plugins built in
173 user space (e.g., the file system plugin). This new tag currently accepts
174 only three attributes (id, read_bw, and write_bw). All the other information
175 that was declared with the storage related tags now has to be expressed as
176 properties. An example of platform using this new tag is available at
177 examples/platforms/hosts_with_disks.xml
180 - 'expect signal' can now accept more than one potential signal.
182 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
183 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
184 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
185 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
186 - FG#31: per-actor data
187 - FG#34: SG_BARRIER_SERIAL_THREAD?
188 - FG#35: model-checker does not like buster-produced binaries
189 - FG!13: MC: complete workaround in the error msg seen on modern systems
190 - FG!15: execute_flops now logs compute
191 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
192 - FG!17: ns-3: unblock the right number of communications + others issues
193 - FG!18: Improving the performance of the ns-3 bindings
194 - GH#189: Energy consumption of parallel tasks
195 - GH#219: Error in the throughput of TCP transfer
196 - GH!330: Minor polishing of English: Configuring_Simgrid
197 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
198 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
199 - GH!334: Rework energy plugin, again
201 ----------------------------------------------------------------------------
203 SimGrid (3.23.2) July 8. 2019
206 - Nicer introduction page.
207 - Migrate the "Deploy your application" page to the new doc.
208 - Move Java as a subtree of MSG.
211 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
214 - Drop xbt_dynar_sort_strings().
217 - Really fix FG#26: Turning off a link should raise NetworkFailureException
218 - FG#27: Wrong exception thrown to wait_any when link is turned off
219 - GH#328: Java: Canceling multiple tasks in a single vm/host
221 ----------------------------------------------------------------------------
223 SimGrid (3.23) June 25. 2019
225 The Exotic Solstice Release.
228 - SunOS and Haiku OS support. Because exotic platforms are fun.
229 - Stop setting random seed with srand48() at initialization.
230 - Use addr2line as a fallback for stacktraces when backtrace is not available.
231 - Build option -Denable_documentation is now OFF by default.
232 - Network model 'NS3' was renamed into 'ns-3'.
235 - Simgrid can now hopefully be installed with pip.
238 - wait_any can now be used for asynchronous executions too.
241 - New log appenders: stdout and stderr. Use stdout for xbt_help.
242 - Drop xbt_dict_dump.
245 - 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
246 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
247 - MPI interface is now const correct for input parameters
248 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
249 - Fortran bindings for async collectives.
250 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
253 - Remove option 'model-check/record': Paths are recorded in any cases now.
254 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
255 always sparse. That's an immense gain in memory, and can even be
256 faster because of cache effects. So there is no need to clutter the
257 code to allow the user to go for the unefficient mode.
260 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
261 models proved to be more accurate than these old experiments.
263 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
264 - FG#1: Broken link in error messages
265 - FG#2: missing installation documentation
266 - FG#3: missing documentation in smpirun
267 - FG#6: Python bindings not available on PyPI
268 - FG#7: simple cmake call requires doxygen
269 - FG#8: make python bindings an optional dependency
270 - FG#10: Can not use MSG_process_set_data from SMPI any more
271 - FG#11: Auto-restart actors forget their on_exit behavior
272 - FG#12: -Denable_lto=OFF doesn't disable LTO
273 - FG#13: Installs unstripped file 'bin/graphicator'
274 - FG#14: Installs the empty directory 'doc/simgrid/html'
275 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
276 - FG#17: Dead link in doc (pls_ns3)
277 - FG#20: 'tesh --help' should return 0
278 - FG#21: Documentation link on http://simgrid.org/ broken
279 - FG#22: Debian installation instruction are broken
280 - FG#26: Turning off a link should raise NetworkFailureException exceptions
281 - GH#133: Java: a process can run on a VM even if its host is off
282 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
283 - GH#326: Valgrind-detected error for join() when energy plugin is activated
285 ----------------------------------------------------------------------------
287 SimGrid (3.22) April 1. 2019
289 The Easter Christmas Release (this one is somewhat late).
292 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
293 - Not complete yet: asynchronous activities (amongst others) are still missing
294 - Still ongoing: the interface may change in the future. We need more
295 testers! Please report any glitches.
296 - No new project using Java should start now. Please switch to Python.
299 - Some of the internal cleanups may lead to speed improvements:
300 - The hard limitation on the amount of simulated actors with
301 Java+Mac was removed. Now, the available memory is the only limit.
302 - Our refcounting was tidyied, leading to 10% speedups in some cases.
303 - We are still working on making our code robust to the actor kills
304 and hosts' churn. Things are improving, but it's not perfect yet.
305 - Replaced our own code to display a backtrace (that was forking addr2line)
306 with the Boost.Stacktrace library.
307 You won't see your backtraces without this optional dependency.
308 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
309 - Stop setting random seed with srand() at initialization.
312 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
313 XML file version remains 4.2 since old files are still compatible.
316 - Process termination which was broken at version 3.21 has been repaired.
317 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
318 - Hide the examples into examples/deprecated. New users should use Python.
321 - Drop MSG_process_create_from_stdfunc() from the API.
322 This C++ function was a pimple in the C API, made necessary at some
323 point by the Java bindings. This is fixed now.
324 - Hide the examples into examples/deprecated. New users should use S4U.
325 - MSG_process_create and MSG_process_attach now crash if the host used
326 to run this process is off.
327 - Fix the protype of MSG_process_on_exit()
328 Now use (int,void*) callbacks instead of (void*,void*) ones.
329 The implementation was ways too messy to actually work, I guess.
332 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
333 too often and provide more accurate timings.
334 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
335 - MPI_Alltoallw support
336 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
337 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
338 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
339 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
341 - Basic implementation of generalized requests (SMPI doesn't
342 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
345 - Drop sg_cmdline. Please use xbt_cmdline instead.
346 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
347 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
348 OsSemaphore is implemented in a portable way with C++11 threads.
349 This should allow much more threads to be created at the same time,
350 allowing Mac and Java users to simulate many more actors.
351 - Implement the 'thread' factory with std::thread instead of xbt ones.
352 It is not possible to set the stack size with threads anymore, but
353 -fsplit-stack is the way to go nowadays when using threads.
354 - Drop the xbt_os_thread_t module (now unused)
355 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
356 - Drop xbt_str_join_array().
357 - Drop cunit, use Catch2 instead.
360 - Many cleanups in the kernel::activity namespace. This was long
361 overdue, and shall open the path to many future endeavors.
364 - #132: Java: a process can not shut down its own host
365 - #220: S4U: detached send causes exception if sender terminates soon after sending
366 - #261: Document the parameters of parallel execution's constructor
367 - #300: [s4u] BarrierPtr is missing
368 - #314: SMPI args internal cleanup
369 - #316: Fix a bug related to the CPU utilization of multi-core VM
370 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
371 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
372 - #325: Turning off a host has different behavior on sleeping actors and computing actors
374 ----------------------------------------------------------------------------
376 SimGrid (3.21) October 3. 2018
378 The Restarting Documentation (TRD) Release.
381 - Convert the user manual to Sphinx for improved usability.
382 Unstable version now hosted on https://simgrid.frama.io/simgrid
383 Many glitches remain; stay tuned.
384 - Restrict the Doxygen reference API to the installed header files.
385 (documenting all internals this way is overkill + doxygen scales badly)
386 - New tutorial on S4U (highly inspired from the old MSG one, but with
387 a git repository to fork as a starting point, and a docker image)
388 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
389 - The Developper manual is still to be converted (not compiled until then)
390 - Some parts are still missing in this conversion (such as the
391 platform and deployment sections) while others were blindly converted
392 and would need more love (such as the configuration flags).
393 Things will certainly further improve in the future releases.
396 - Fully reimplement auto_restart mechanism that was utterly broken
397 (fix #22, #131 and #281 that were all very related)
398 - Implement semaphores (fix #303 and #291)
399 - When creating an actor from a function and its parameters,
400 move-only parameters are not allowed anymore, as it would prevent
401 the actor to be restartable if its parameters are consumed on run.
402 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
403 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
404 IO operations that can be started, waited for, or canceled as a regular
406 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
410 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
411 - New host variable: 'core_count'
414 - Remove xbt_os_thread_specific features
415 - Remove portability wrapper to condition variables
416 - Remove xbt_os_thread_yield()
419 - MPICH collective selector now mimics MPICH 3.3b
420 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
421 - MPI_Init(NULL,NULL) is now allowed
422 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
423 - Command-line arguments should now be handled properly in Fortran simulations
424 - MPI Attributes and related callbacks should work in Fortran
425 - Apps using C + Fortran should now work
426 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
427 - Support MPI_Error_String
430 - Due to an internal bug, Msg.run() must now be your last line.
431 We hope to fix it in a future release, and we are sorry for the inconvenience.
434 - #22: Process autorestart seem to only work with CAS01 cpus
435 - #56: Feature request: dynamic SMPI replay
436 - #93: simgrid should not eat --help
437 - #111: How to change the loopback link characteristics?
438 - #116: Platform cluster radical
439 - #117: Platform cluster/host order
440 - #129: Replay_multiple on steroid
441 - #131: Java : autostart does not work
442 - #143: Setting a breakpoint at a given time
443 - #235: Network energy models should be integrated
444 - #264: Add ptask L07 resource tracing
445 - #271: Dynamic and manual replay of SMPI traces
446 - #279: Breakpoints option uses time deltas instead of absolute time
447 - #281: Daemonized actor no longer daemonized after an auto-restart
448 - #282: TIT convertor erase traces when using absolute path in the trace list file
449 - #285: segfault when a s4u actor kills itself with msg process tracing activated
450 - #286: Pajé traces are not exposing the number of cores
451 - #287: Command-line parsing should stop when encountering '--'
452 - #288: MPI_Init(NULL, NULL)
453 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
454 - #290: Method to check if a mailbox has a message ready to be consumed
455 - #291: [s4u] semaphores do not seem available yet
456 - #294: Alltoall collective implementation of mpich
457 - #295: Starting a Host as OFF is not supported
458 - #296: DTD too permissive
459 - #299: Add a s4u barrier example
461 ----------------------------------------------------------------------------
463 SimGrid (3.20) Released June 24. 2018
465 The proxy snake_case() release.
468 - The whole API is now using snake_case() instead of mixing it with camelCase.
469 Compatibility wrappers in place for one year.
470 - We took the opportunity to increase the consistency of the API.
471 Please report any remaining glitches.
472 - New function simgrid::s4u::Host::get_actor_count:
473 Returns the number of actors running on a specific host.
476 - SMPI is now tested with ~45 proxy apps from various sources, with none or
477 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
478 - Replay: The replay file has been re-written in C++.
479 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
480 supported. They are stored in the trace and used when replayed.
481 - Basic support of MPI_Cancel. Robustness not guaranteed.
482 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
483 (only for MPI_COMM_TYPE_SHARED).
484 - New option: smpi/privatize-libs, to add external shared libs to be privatized
485 by SMPI. They will be copied locally and loaded separately by each process.
486 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
487 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
488 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
489 with internal index optind with multiple processes. Only works if MPI_Init has
491 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
492 syntax without preprocessing tricks.
495 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
496 behavior tracing is no longer limited to MSG
497 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
498 this means that traces now need to be updated or re-obtained; both Irecv
499 and Isend lines in your traces have now in total 5 mandatory fields:
500 <rankid> <command> <to/from rankid> <tag> <size>
501 To update your traces, it suffices to add a 0 for the tag here.
502 - TIT now also supports waiting for a distinct request via MPI_Wait.
503 Wait/Test now wait for a specific request, not just the last one that was
504 issued. This unfortunately means another update, because we need to
505 identify which request you want to wait for. We do this via the
506 triplet (sender, receiver, tag), which needs to be added:
507 <rankid> <command> <sender> <receiver> <tag>
508 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
509 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
512 - The deprecation of MSG is ongoing (but this should not impact you).
513 Many MSG functions are now simple wrappers on the C API of S4U. If
514 you wish to convert your code to S4U, find the S4U counterparts of
515 your MSG calls in src/msg/msg_legacy.cpp.
516 - MSG can still be used, but won't evolve anymore.
519 - Allow to run the Link energy plugin from the command line with
520 --cfg=plugin:link_energy
521 - Rename Energy plugin into host_energy
522 - Rename Load plugin into host_load
525 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
526 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
527 - Improve performance when handling timeouts of simix synchros.
530 - Config: the C API is now deprecated (will be removed in 3.23), and
531 the C++ API has been slightly improved.
534 - Fix several build issues on OSX.
535 - Move simgrid_config.h to simgrid/config.h (old header still working)
538 - #143: Setting a breakpoint at a given time
539 - #150: Inconsistent event names in SMPI replay
540 - #258: daemonized actors hang after all non-daemonized actors have completed
541 - #267: Linker error on unit_tmgr
542 - #269: SMPI: tracing of MPI_Wait/all/any broken
543 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
545 ----------------------------------------------------------------------------
547 SimGrid (3.19.1) Released March 22 2018
549 The Fixed ABI Release.
551 As you may know, we are currently refactoring SimGrid in deep.
552 Upcoming SimGrid4 will be really different from SimGrid3: modular,
553 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
555 Our goal is to smooth this transition, with backward compatibility
556 and automatic update paths, while still progressing toward SimGrid4.
558 SimGrid remains open during works: The last pure SimGrid3 release was
559 v3.12 while all subsequent versions are usable alpha versions of
560 SimGrid4: Existing interfaces remain unchanged, but the new S4U
561 interface is budding and the internals are deeply reorganized.
563 Since 2015, we work hard to reduce the changes to public APIs. When
564 we need to rename a public library symbol in S4U, we let your
565 compiler issue an explicative warning when you use the deprecated
566 function. These messages remain for four releases, i.e. for one full
567 year, before turning into an error. Starting with v3.15, your can
568 also adapt to API changes with the SIMGRID_VERSION macro, that is
569 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
571 Starting with this v3.19.1, our commitment to reduce the changes to
572 the public interfaces is extended from the API to the ABI: a program
573 using only MSG or SimDag and compiled against a given version of
574 simgrid can probably be used with a later version of SimGrid without
575 recompilation. We will do our best... but don't expect too much of
576 it, that's a really difficult goal during such profund refactoring.
578 The difference between v3.19 and v3.19.1 is that the former was
579 accidentally breaking the ABI of MSG, while the later is restoring
582 S4U and kernel APIs will still evolve until SimGrid4, with one-year
583 deprecation warnings as currently. In fact, cleaning up these
584 interfaces and converting them to snake_case() is one release goal of
585 v3.20. But don't worry, we are working to smooth this upgrade path.
587 In summary, new projects should start with S4U to benefit of the
588 future, but old MSG projects should still be usable with no change.
590 ----------------------------------------------------------------------------
592 SimGrid (3.19) Released March 20 2018
594 The Moscovitly-cold Spring release.
596 SMPI user-visible improvements:
597 - Support of MPI_Type_create_subarray
598 - Unimplemented MPI calls now abort (instead of issuing a warning)
599 Some harmless calls around errhandlers still only issue a warning.
600 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
601 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
602 is now possible (mmap privatization was refactored to this end).
603 Adding new SMPI instances should also be technically doable now.
605 SMPI internal cleanups:
606 - Removed index notion from SMPI -- no more "getPid() - 1"!
607 In tracing files, actors are called rank-<PROCESS-ID> but these are
608 NOT the real ranks (tracing several communicators is not supported yet).
609 - Deprecate smpi_process_index() (will be removed in 3.22)
610 - Better testing on classical proxy apps, through an external project:
611 https://github.com/simgrid/SMPI-proxy-apps/
614 - Use a graphical TOC to make it easier to find the documentation you need
615 - Revamp the MSG tutorial
618 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
619 - Remove parameter of MSG_process_killall().
620 Resetting the PID was bogus anyway (several actors could have the same PID).
623 - Execution->setHost() can be called after start(), to migrate it.
624 - Comm::test_any() is now implemented.
625 - s4u::Actor now has onCreation() and onDestruction() signals.
626 - Install some kernel header files for the users' plugins and more.
627 Warning, their API is really not stable yet.
628 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
629 include/simgrid/kernel/routing/*Zone.hpp
632 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
634 Build System and other cleanups:
635 - Remove unused run-time parameter "tracing/onelink-only".
636 - SimGrid now works with intel compilers (MC still buggy)
637 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
638 - tesh kills subprocesses on timeout (but not on windows)
641 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
644 - #194: Feature request: simgrid::s4u::Comm::test_any()
645 - #245: migrating an actor does not migrate its execution
646 - #253: Feature Request: expose clusters as objects
647 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
648 - #255: Tesh broken on Windows
649 - #256: Modernize FindSimGrid.cmake
650 - #257: Fix (ab)use of CMake install
652 ----------------------------------------------------------------------------
654 SimGrid (3.18) Released December 24 2017
656 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
658 MSG IS NOW DEPRECATED. Please use S4U for new projects.
659 - Support for MSG should not be removed from SimGrid before 2020,
660 but future evolutions will be limited to the integration of
661 user-provided patches.
664 - s4u::onDeadlock() signal, run before stopping the simulation.
665 - s4u::Exec: asynchronous executions (abstraction of background
667 - s4u::Host->getLoad() returns the achieved speed in flops/s
670 - Link::name() is deprecated and will be removed in v3.21, use
671 Link::getCname() instead.
672 - Mailbox::getName() changed to return a std::string, use
673 Mailbox::getCname() instead to get a char*.
674 - Storage::getName() changed to return a std::string, use
675 Storage::getCname() instead to get a char*.
676 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
677 to have both allocation and deallocation of the map in user space.
680 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
681 The amount of flops remaining to do is a vector, not a scalar.
682 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
683 name implies on both sequential and parallel tasks.
684 - Both changes fix GitHub's #223 using PR #237 as a basis.
685 Thanks Michael Mercier.
686 - Most examples were converted to S4U and hidden elsewhere at tests.
687 Rationale: we still want MSG to work; we want newcomers to use S4U.
690 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
691 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
694 - Switch to the faster dlopen privatization mechanism by default
695 - Documentation improvments
698 - Live migration is getting moved to a plugin. Dirty page tracking is
699 the first part of this plugin. This imply that VM migration is now
700 only possible if one this function is called:
701 - C/MSG: MSG_vm_live_migration_plugin_init()
702 - C/C++: sg_vm_live_migration_plugin_init()
703 - Java: Msg.liveMigrationInit()
704 For C and C++, "simgrid/plugins/live_migration.h" also has to be
708 - Define class simgrid::xbt::Path to manage file names.
709 - Removed unused functions:
710 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
711 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
712 xbt_graph_export_graphviz()
713 - xbt/str.h: xbt_str_join()
714 - Remove unused datatypes:
715 - xbt/heap.h: use std::priority_queue or boost::heap instead
716 - xbt/swag.h: use boost::intrusive::list instead
719 - New link_energy plugin for the consumption of the links.
720 - All of the operations on files and storage contents have been
721 packaged into a plugin (src/plugins/file_system). The current
722 public interface can be found in
723 include/simgrid/plugins/file_system.h
724 To use these functions you now have to initialize the plugin by
725 calling MSG_storage_file_system_init() just after calling
726 MSG_init() or sg_storage_file_system_init() just after creating
730 - Remove the undocumented/untested tag <include>
733 - Remove viva specific tracing as the tool is no longer maintained
736 - #248: Kill of finished processes leads to segfault
737 - #240: xbt_cond_wait_timeout should gracefully return for C apps
738 - #239: please implement signal s4u::onDeadlock()
739 - #230: segfaults when exit() without run()
740 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
741 - #223: MSG_task_get_flops_amount() not working with parallel tasks
742 - #222: Actor::kill() doesn't really kill and segfaults
743 - #221: odd LMM warning when killing an actor
744 - #120: Memory leak when the processes are forcefully killed
746 ----------------------------------------------------------------------------
748 SimGrid (3.17) Released October 8 2017
750 The Drained Leaks release: (almost) no known leaks despite the tests.
752 Even very long simulations will preserve your memory: our testsuite
753 stresses SimGrid for over 45mn on fast machines for a coverage of
754 over 80%, and there is only one single known leak, of about 4kb.
758 - Comm.detach(): start and forget about asynchronous emission. A cleanup
759 handler may be given to free resources if the comm cannot be completed.
760 - this_actor::send(mailbox) is now mailbox->put()
761 - New: simgrid::s4u::Comm::wait_all()
762 - New: Host.execute() for remote executions.
765 - Unused option network/sender-gap is removed.
768 - Deprecate MSG_task_isend_with_matching(): This unused feature
769 really complicates our internals. Will be removed in v3.20.
772 - Improved context termination. It is now possible to free resources, even
773 when a process is forcibly killed.
774 - ContextBoost: add support for Boost versions above 1.61.
777 - Bring back run-time option --cfg=exception/cutpath to remove exception
779 - Removed unused functions:
780 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
781 xbt_str_rtrim(), xbt_str_trim().
782 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
785 - Removed header files obsolete since SimGrid 3.12:
786 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
788 - #3: SD_exit should be made optional
789 - #120: Memory leak when the processes are forcefully killed
790 - #159: Threading test regression in Actor refcounting
791 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
792 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
793 - #186: Actor::killAll() segfaults if some process is blocked on wait()
794 - #191: VM migration and pstate
795 - #192: Updating the energy consumptions of all hosts crashes with VMs
796 - #195: All actors have PID=0 in the logs
797 - #204: Sometimes segfault with thread contexts and mmap privatization
798 - #222: Actor::kill() doesn't really kill and segfaults
799 - #225: Actor::kill() doesn't really kill when victims are doing a join()
801 ----------------------------------------------------------------------------
803 SimGrid (3.16) Released June 22. 2017.
805 The Blooming Spring Release: developments are budding.
808 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
809 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
810 - Java JAR file should be finally fully working on Mac OSX too.
812 XML platforms: Switch to platform v4.1 format.
813 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
814 - <zone> can be used as a synonym for the now deprecated <as>
815 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
816 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
817 - <actor> can be used as a synonym for the now deprecated <process>
818 - state_file and avail_file periodicity is now easier to express
819 (check the documentation) the old behavior should still work.
820 * Storage: not backward compatible, but it seems that nobody ever used it
821 - Remove attribute 'content_type' of <storage_type>: was never used
822 - Make attribute 'model' of <storage_type> optional: for future usage
823 - Remove Bconnection model property: was never (in)validated. Replaced by
824 the maximum of the read and write bandwidth as a resource constraint.
827 - New and Backwards Compatibility break:
828 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
829 When one wants to get the list of tasks whose states have changed during a
830 simulation round, s/he has to allocate and free a dynar and use it as
831 argument to this function. The former SD_simulate (double how_long)
835 - Allow multicore VMs, with the correct sharing (unless you overcommit)
836 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
839 - New function to update the consumption of all hosts at once.
840 - Fix the model for multi-core hosts, linear in the amount of busy
841 cores with an abnormality for pIdle (see paper).
842 BUG: Still not satisfactory for ptask on mono-cores.
845 - The netzone are now available from the MSG API.
846 The old names still work, but are now deprecated.
849 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
850 - New option: smpi/keep-temps to not cleanup temp files
851 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
852 mallocs mode is used, allows to change the size of the fake file used
853 (default 1MB), to potentially limit the number of mappings for large runs.
854 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
855 - Fortran ifort and flang compilers support
856 - New RMA calls supported (experimental):
857 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
858 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
859 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
860 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
861 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
862 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
865 - New: this_actor::isMaestro() returns whether we are in kernel mode.
866 - Behind the scene, ActivityImpl are now automatically refcounted.
867 Removing this nasty bottleneck will greatly simplify our internals.
870 - Replay: New function xbt_replay_action_get():
871 Retrieve the function previously associated to an event type.
872 - DROPPED FUNCTION: xbt_str_varsubst()
873 - DROPPED MODULE: strbuff. We don't need it anymore.
874 - DROPPED MODULE: matrix. We don't need it anymore.
875 - DROPPED MODULE: lib. We don't need it anymore.
877 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
879 SimGrid (3.15) Released March 22. 2017
881 The Spring Release: continuous integration servers become green
882 We fixed even the transient bugs on all target architectures:
883 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
884 Windows, FreeBSD, NetBSD.
886 - Rename NetCards to NetPoints.
887 This was intended to help NS3 users, but that's not a netcard.
888 That's a point in the routing algorithm, let's avoid wrong simplifications.
891 - New: MSG_process_yield(). Stop and yield to other processes.
892 - New: MSG_process_daemon(). Daemon processes are automatically killed
893 when the last non-daemon process terminates
894 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
895 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
896 to make room for the upcoming network energy plugin.
897 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
900 - Ensure that an actor can kill itself with Process::exit()
901 - Kill the obscure NativeException. Nobody want to survive the issues
902 it denotes, so use JniException that is a RuntimeException (not to
903 be caught explicitly).
904 - Partial bug fix in initialization. SimGrid flags on command line were
905 consumed at C level but stayed in the original Java String[] args.
906 This could mess users' args[i] if SG flags were not put at the end of
908 The SimGrid flags are now removed from the Java arguments. However,
909 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
910 if args.length is greater than the number of YOUR OWN ARGUMENTS.
911 It might be if you have --log or --cfg flags in the command line.
912 - Fix numerous memleaks all around the place. In particular, around VMs.
916 - s4u::onPlatformCreated: right before the simulation starts
917 - s4u::onSimulationEnd: right after the main simulation loop
918 - s4u::onTimeAdvance: right after a clock change (time is discrete)
919 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
920 event from the availability_file changes the avail speed.
921 - Links are now usable from s4u
922 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
923 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
924 - New examples: The conversion of MSG examples to S4U has begun
925 - Actors: create, kill, migration, and suspend
926 - Applications: master-worker and token-ring
927 - Action replay: communications and storage
928 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
931 - Backwards Compatibility breaks
932 - The SD_route_get_list and SD_route_get_size functions have been removed.
933 They are replaced by the unique
934 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
935 The route from src to dst is built in the links dynar whose size can
936 be obtained with xbt_dynar_length.
937 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
938 sg_host_route_bandwidth and sg_host_route_latency.
939 Macros ensure the backwards compatibility, but you should fix your code
942 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
943 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
944 - Bump our claim of support from MPI 1.1 to MPI 2.2.
945 We don't support 100% of it, but it should be enough. Contact us if not.
946 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
947 naive and probably inaccurate)
948 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
949 mapped onto a unique small file using some system magic.
950 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
951 - Fortran: cleanups, fixes, support of user-added operations
952 - RMA: MPI_Accumulate are applied in correct order.
953 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
954 - Better support for MPI_IN_PLACE
955 - Support for MPI_Win attrs and keyvals.
956 - Support MPI_Comm_group_create, MPI_Type_size_x
957 - MPI ops should be properly applied to all allowed types
958 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
959 algorithm by a non failing one from mpich.
962 - Kill the fifo data container: we don't use it anymore.
965 - A <cluster> can now be created with different speed values to represent pstates.
966 see examples/platforms/energy_cluster.xml for an example.
968 SimGrid (3.14.159) Released December 28. 2016
970 The previous release was only a crude approximation of a Pi release;
971 this one should be a bit better.
973 - Revert a last minute change that broke on Mac OSX.
974 - Fix the changelog and NEWS files.
975 - Improve the documentation, in particular of the routing module.
976 - Integrate some patches flying around in Debian.
978 SimGrid (3.14) Released December 24. 2016
981 * The whole documentation was reworked and reorganized. There is
982 still much room for improvement here, but we're on it.
985 * We now need python3 for our testing infrastructure.
986 * Model checking needs libevent
989 * Now works on FreeBSD
992 * Privatization now works on FreeBSD
994 * Privatization is now activated by default in smpirun
995 You can turn it off with -no-privatize if something goes wrong.
997 * Call-location tracing for SMPI.
998 You can add the exact location (filename / linenumber) of an MPI call to
999 your trace files and slow down or speed up the simulation between two
1000 consecutive calls by using an adjustment file (see the documentation).
1002 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1003 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1004 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1005 values were ignored in some cases. The timings of these functions can now
1006 be significantly different.
1008 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1009 smpi/running-power is renamed to smpi/host-speed
1011 * smpi/grow-injected-times option to enable or disable multiplication of the
1012 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1013 make simulation less precise (but also much faster).
1015 * smpirun script should be (much) faster for large deployments.
1017 * SMPI tracing: fixed issue with poor matching of send/receives.
1019 * Replay: Fix broken waitall
1021 New functions and features
1022 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1024 Dropped / renamed functions and features
1025 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1026 - MSG_mailbox_is_empty() -> Mailbox::empty()
1027 - MSG_mailbox_front() -> Mailbox::front()
1028 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1029 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1030 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1031 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1032 * Don't pass the free_f parameter to property related functions:
1033 - MSG_host_set_property_value()
1034 - MSG_as_router_set_property_value()
1035 - MSG_storage_set_property_value()
1036 * VM properties. Since msg_vm_t are msg_host_t, just use
1037 MSG_host_get_property_value() and friends
1038 * VM I/O related things:
1039 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1040 - Unimplemented save/restore methods
1041 * MSG_as_router_get_property_value() was redundant with
1042 MSG_environment_as_get_property_value().
1043 - Removed MSG_as_router_*propert*() functions
1044 - Added MSG_environment_as_set_property_value() for consistency
1045 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1046 Well, they are still there for now, but deprecated with a warning.
1047 Please switch to xbt_dict_new_homogeneous() before this is removed
1049 * Task affinity. Its intended behavior (that was very badly tested
1050 and probably not really working) was deceiving what most users
1051 would have hoped here.
1052 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1053 * Option network/coordinates is now useless and should be dropped.
1056 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1059 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1060 * Added Dragonfly topology support
1063 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1064 EiBps, EBps, Eibps, Ebps for bandwidth.
1065 They may become useful to some lucky ones.
1068 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1069 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1070 * ex/async_waitAll: new example, on asynchronous communications
1073 * Memory usage should be decreased for simulations with a large number
1074 of processes. This also helps for SMPI.
1076 SimGrid (3.13) stable; urgency=low
1078 The Half Release, a.k.a. the Zealous Easter Trim.
1080 - We removed half of the lines, that were mostly experimental cruft.
1081 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1082 (not counting blanks and comments -- according to openhub.net)
1083 - The internals are now compiled in C++ (and will soon be clean C++)
1084 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1085 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1087 * Backwards Compatibility breaks
1088 - Removed Lua simulation bindings (switch to C or Java for that).
1089 Lua can still be used to describe platforms
1090 - Removed Java kernel plug-ins.
1091 Will be reintroduced after the ongoing major internals reorg.
1093 - the following functions were removed.
1094 They were too specific and should be reimplemented in a generic
1095 way, with filter function.
1096 - MSG_task_listen_from_host
1097 - MSG_mailbox_get_count_host_waiting_tasks
1098 - MSG_mailbox_put_with_timeout was removed.
1099 Please use MSG_task_send_with_timeout instead.
1101 - the SD_application_reinit function was removed. It has been a noop for a while.
1102 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1103 replaced by a more flexible mechanism.
1104 - The following functions thus do not exist anymore
1105 - SD_workstation_get_access_mode
1106 - SD_workstation_set_access_mode
1107 - SD_workstation_get_current_task
1108 - Basic estimation functions have been removed but can easily be replaced
1109 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1110 - SD_workstation_get_computation_time => amount / sg_host_speed()
1112 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1113 Use VM.getSpeed()*load/100 for the legacy behavior.
1115 - option enable_tracing was removed. It was not doing anything for a while.
1116 - In the ModelChecker:
1117 - the model-checker now ptraces the model-checked process which means
1118 you cannot use a debugger on the latter anymore (we might make this
1119 optional in the feature);
1120 - removed soft-dirty page tracking;
1121 - remove model-checked side snapshot management,
1122 MC_snapshot() and MC_compare_snapshot();
1123 - keep the MC_cut() function as a stub (it was not really working
1124 in the previous release).
1127 * All options are consistently in kebab-case. Old names are kept as alias.
1130 * Switch to platform v4 format.
1131 - Rename from 'power' to 'speed' the attributes describing the amount of
1132 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1133 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1134 - In <host> and <link>, attributes availability and state are gone.
1135 It was redundant with state and availability traces, and with peak values.
1136 - In <cluster>, attributes availability_file and state_file are gone.
1137 It was too complex and unused.
1138 - Kill <gpu>. Was not doing anything.
1139 - The DOCTYPE points to the right URL:
1140 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1141 (the file at this address now documents the changelog since its v1)
1142 - A warning is emitted for unit-less values (they are still accepted).
1143 - speed. Default: 'f' or 'flops'. Also defined:
1144 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1145 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1146 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1147 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1148 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1149 - latency. Default: 's' second. Also defined:
1150 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1152 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1153 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1156 * s4u::Host is now the preferred public interface to the Host features.
1157 sg_host_* functions are C bindings to the exact same behavior
1158 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1161 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1162 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1163 They were deprecated since a few years, and probably did not even compile anymore.
1166 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1167 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1168 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1169 encouraged to update their codes. A list of the modified functions can be found at the end of
1170 include/simgrid/simdag.h
1173 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1174 the lambda or closure passed as a parameter will run in kernel mode.
1175 All the callback functions should be rewritten to that interface at some point.
1178 * Reorganizing and cleaning the internals all around the place.
1181 * Remove old default barrier/bcast buggy algorithms (see #18407)
1182 * Various bug fixes to handle more codes
1183 * Remove the need for the --foreground option of smpirun (it is still
1184 accepted for backward compatibility).
1187 * Kill the setset data container: MC don't use it anymore.
1188 * Kill the queue data container: it made more sense with GRAS.
1189 * Kill the xbt_peer_t data type: it's useless without GRAS.
1190 * Kill rm_cb feature of config sets: it was never useful.
1191 * Kill graphxml parsing feature. It was not used.
1192 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1194 - xbt_dynar_sort_strings(), when the content is char*
1195 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1196 They throw exceptions on invalid input;
1197 * C++ support for declaring CLI flags (simgrid::config::Flag);
1198 * class for abstracting different signal backends (simgrid::xbt::signal).
1199 with no external dependencies (we need very simple signals).
1202 * refactoring and cleanup of the code;
1203 * ongoing process to cleanly separate the model-checking algorithms
1204 from the code model-checking support.
1206 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1208 SimGrid (3.12) stable; urgency=low
1210 The Facelift Release.
1213 * Require g++ v4.7 at least to not speak prehistorical C++.
1214 * Require Boost 1.48 (for signal2 component).
1215 * Java must be version 7 at least when activated.
1216 * Builds on Windows again (including Java bindings).
1217 * Tracing is now always enabled (no way to turn it out)
1218 * Remove GTNetS. It was not working anyway.
1219 * Various cleanups in the cmake scripts.
1220 * Move headers around to sort them out on installed systems:
1221 - instr/instr.h -> simgrid/instr.h
1222 - instr/jedule/* -> simgrid/jedule
1223 - simdag/datatypes.h was removed
1224 - simdag/simdag.h -> simgrid/simdag.h
1225 - msg/datatypes.h was removed
1226 - msg/msg.h -> simgrid/msg.h
1229 * Interface improvement:
1230 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1231 - Sanitize the interface in MSG_task_ module:
1232 - Merge two functions that were close enough but misleading:
1233 set_compute_duration(t) -> set_flops_amount(t)
1234 get_remaining_computation(t) -> get_flops_amount(t)
1235 - set_data_size(t) -> set_bytes_amount(t)
1236 get_data_size(t) -> get_bytes_amount(t)
1237 - Massive cleanups in the functions related to the energy
1238 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1239 - New: MSG_host_get_pstate()
1241 - msg/energy/onoff: switching hosts on and off
1243 * Interface improvement:
1244 - Rename Host.isAvail() to Host.isOn()
1245 - Rename Process.currentProcess() to Process.getCurrentProcess()
1246 - Rename Task.setDataSize() to Task.setBytesAmount()
1247 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1249 - #18874: Actually allows the GC to reclaim tasks
1253 - SIMIX_process_throw: raises an exception in a remote process
1254 * Refactoring: Separate sub-modules
1255 - libsmx: the public interface, as libc in a real system
1256 - popping: the strange dance that converts a user request into a kernel handling
1257 - smx_context_*: the virtualization mechanisms that embed the user code
1258 - smx_*: the handling of each simcalls
1259 * Interface cleanup:
1260 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1261 * Rename smx_action_t into smx_synchro_t, making explicit that these
1262 things are used to synchronize processes with their environment.
1263 For example, a communication is a sort of synchronization involving
1264 the communicating processes (that may block until the exchange) and
1265 the platform. The same can be said from computations, etc.
1267 - #18888: segfault when a process ends before its kill_time
1271 - 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.
1272 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1273 - MPI_Type_set_name, MPI_Type_get_name
1274 - MPI_*_c2f and MPI_*_f2c functions
1275 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1276 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1277 - Activate a lot of new tests from the mpich 3 testsuite
1279 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1280 - InfiniBand network model added: Based on the works of Jerome Vienne
1281 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1282 - When smpi/display_timing is set, also display global simulation time and application times
1283 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1284 * Collective communications
1285 - 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
1286 - 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/).
1287 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1289 - Replay now uses algorithms from wanted collective selector
1290 - Replay can be used with SMP-aware algorithms
1291 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1292 - 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.
1294 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1295 - Correctly use loopback on fat-tree clusters
1296 - Asynchronous small messages shouldn't trigger deadlocks anymore
1297 * Energy/DVFS cleanup and improvement
1298 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1299 - new: smpi_get_host_pstate()
1302 - "Full" network optimization flag was broken since Surf++
1303 - Better handling of precision flags in maxmin
1304 - Fix bug causing sometimes "Impossible" errors
1305 - Properly pass cluster properties to included hosts
1306 * Improvement of the Energy plugin.
1307 - Always update the consumption before returning that value
1308 - New property: watt_off to denote the dissipation when the host is off
1309 - New functions getWattMinAt and getWattMaxAt to retrieve the
1310 dissipation of pstates that we are not currently at.
1311 * Java: class NetworkLink renamed to Link
1312 * New function: simcall_process_get_kill_time()
1313 * Massive rename s/workstation/host/
1314 - That's intrusive, but that's good for the project consistency. Sorry.
1315 - Change config option "workstation/model" into "host/model"
1318 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1319 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1320 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1321 - xbt_dirname and xbt_basename for non-POSIX systems
1323 * The model checker now runs as a separate process.
1324 * The model checker runs must now be launched with the new simgrid-mc program.
1325 * Record/Replay: the MC can display a textual representation of a path in the
1326 execution graph. It can then be replayed outside of the model checker.
1328 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1330 SimGrid (3.11) stable; urgency=low
1335 * Normalizing pointers addresses tool for better diff between logs
1338 * Add cloud examples using new VMs
1339 - examples/msg/cloud/two_tasks_vm.tesh
1340 - examples/msg/cloud/simple_vm.tesh
1341 - examples/java/cloud/cloud.tesh
1342 - examples/java/cloud/migration/migration.tesh
1343 * Add java surf examples:
1344 - examples/java/surfPlugin/surf_plugin.tesh
1345 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1346 - examples/java/surfCpuModel/surf_cpu_model.tesh
1347 * Add SMPI+MSG example:
1348 - examples/smpi/smpi_msg_masterslave/
1353 - msg task destroy cancel test
1354 - msg_host on/off test
1355 * Move all tests in testsuite to teshsuite (adding tesh files)
1356 * Restructure teshsuites
1357 - one folder for each kind of test
1358 * Restructure AddTests.cmake
1360 - structure the order of tests (with sections)
1363 * Add virtual machine
1364 - creation of a VM on a PM
1365 - migration of a VM from a PM to another PM
1367 - MSG_process_join(msg_process_t process, double timeout)
1368 - msg_bar_t MSG_barrier_init(unsigned int count)
1369 - int MSG_barrier_wait(msg_bar_t barrier)
1370 - void MSG_barrier_destroy(msg_bar_t barrier)
1371 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1372 * New option "msg/debug_multiple_use" to help debugging when a task is used
1375 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1377 - Increase code coverage in test suites
1381 * Protect context stacks against stack overflow. The number of protected memory
1382 pages allocated on the top of each stack (1 by default) can be configured
1383 with the new command line option --cfg=contexts/guard_size:<value>.
1384 * Simcalls are now generated by a python script that
1385 - generates files included by SimGrid
1386 - checks that all the functions exist, and proposes prototypes
1388 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1389 host_set_data, host_get_data
1391 - simcall_process_join(smx_process_t process, double timeout)
1392 * Fix bug where sleeping processing could not be suspended.
1395 * Translate surf models from C to C++
1396 - Generic classes for all models: Model, Resource, Action
1397 - A generic interface for each kind of model (CPU, Network, Storage
1398 Workstation, WorkstationVM)
1400 * Translate surf routings from C to C++
1401 * Add callbacks using sigc++ or boost::signals2
1402 - Add callback functions for resource creation/destruction
1403 - Add callback functions for action state change
1404 - Handle Energy as a plugin
1405 * Replace swag by boost::intrusive
1406 * Add new routing models for clusters. For documentation, see
1407 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1408 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1409 parameters for cluster tag
1410 - Fat trees, with topology="FAT_TREE" and
1411 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1412 - see examples/platforms/torus_cluster.xml and
1413 examples/platforms/fat_tree_cluster.xml
1414 * More documentation
1417 * Hostfiles support host:nb_processes construct to deploy several processes on
1419 * Collective communication algorithms should not crash if used with
1420 improper number of nodes and report the error.
1421 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1422 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1423 MPI_Dims_create, MPI_Cart_sub are supported.
1424 * New interface to use SMPI programmatically (still depends on MSG for
1425 some parts, see examples/smpi/smpi_msg_masterslave):
1426 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1430 * Global variables privatization in MPI executables is now performed at runtime
1431 with the option smpi/privatize_global_variables (default:no).
1432 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1433 dynamic libraries loaded by the application are not privatized (static
1434 linking with these libraries is advised in this case)
1437 * Options defined in XML work correctly now.
1440 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1441 copied into simgrid.jar or not (ON by default). Use this option if you want
1442 to reduce the size of the installed simgrid.jar, *and* the native libraries
1443 are kept installed elsewhere.
1444 * Surf binding with SWIG (code generated in maintainer mode only):
1445 - plugin to handle callbacks
1446 - CPU model only for the moment
1449 * Supernovae build mode is definitively removed. It was used to improve
1450 inlining and inter-module optimizations. It is nowadays superseded by
1451 link-time optimizations commonly available in compilers.
1452 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1453 * Add boost dependency for surf++
1454 * Add new macro for tests
1455 - ADD_TESH(name <tesh_args>)
1456 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1460 - xbt_bar_t XBT_barrier_init(unsigned int count)
1461 - int XBT_barrier_wait(xbt_bar_t barrier)
1462 - void XBT_barrier_destroy(xbt_bar_t barrier)
1463 * Make the xbt_os_time module public
1465 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1467 SimGrid (3.10) stable; urgency=low
1469 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1472 * Reintegrate Java to the main archive as desynchronizing these
1473 package is not acceptable anymore (Java is now considered stable)
1474 * Add explicit synchronization facilities through semaphores
1475 * Bug fix: Task.setDataSize() only changed the C world, not the value
1476 cached in the Java world
1479 * Dramatically change the way files are handled. API and internals changed, but
1480 this part of MSG was not considered as production grade either.
1481 * Add explicit synchronization facilities through semaphores
1482 * Add a new function MSG_host_get_process_list()
1483 * Preliminary DVFS support (see examples/msg/energy/ for details)
1486 * SMPI is now included directly in the libsimgrid as the windows
1487 linker doesn't force us on splitting it anymore.
1488 * Improvements of the SMPI replay tool:
1489 - Most of the collective communications are now rooted in the same process as
1490 in the original application.
1491 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1492 used until now). Multiple data types can now be used in a trace.
1493 - The replay tool now supports traces produce either by TAU or a modified
1495 - Bug Fix: the compute part of the reduce action is now taken into account.
1496 - Gatherv collective is now supported
1497 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1498 of smpirun outputs time independent traces for the current run. One file
1499 is created per process. If too many processes are simulated, this behavior
1500 can be changed to one file for all processes by using the
1501 tracing/smpi/format/ti_one_file flag
1502 * smpirun generates the host file if needed (with given host count and platform)
1503 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1504 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1505 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1506 --cfg=smpi/coll_selector:(mpich/ompi)
1507 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1509 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1510 happening inside a collective SMPI call.
1511 * Fix the behavior of complex data types handling.
1512 * Make MPI_Wtime another synchronization point to take computations into
1514 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1515 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1516 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1517 (no privatization of global variables yet)
1518 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1521 * Verification of liveness properties is now available for SMPI applications
1522 (in addition to MSG applications)
1523 * Bugged examples using SMPI in examples/smpi/mc/
1524 * Add --cfg=model-check/visited option. Allows the verification of infinite
1525 programs. Detection of loops in the execution thanks to the system state
1526 comparison and reduction of the state space to explore. Can be combined with
1527 DPOR for safety properties.
1530 * Allow to change SimGrid configuration (see --help) within the code
1531 thanks to SD_config() as it can be done in MSG.
1532 * Add a new function SD_task_set_amount() upon user request.
1535 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1536 * Remove rule based routing (no more PCRE dependency)
1537 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1538 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1539 * Add a loopback_bw and loopback_lat options to cluster tag.
1540 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1541 routing. Now peers are encapsulated in an AS and have their own private
1542 router but this is transparent.
1545 * Our own implementation of getline is renamed xbt_getline, and gets
1546 used even if the OS provide a getline(). This should reduce the
1547 configuration complexity by using the same code on all platforms.
1548 * New type: xbt_cfg_elm_boolean.
1549 * Allow to use yes/no for boolean configuration options in the command line.
1550 * Allow to disable SimGrid cleanups at exit from command line option.
1551 There are situations where one may want a simulation to end with an exit.
1552 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1553 annoying when scripting around the simulator. Adding a
1554 --cfg=clean_atexit:no allows to circumvent this issue.
1557 * Lots of memory leaks were corrected in this release.
1558 There is no known memory leaks anymore, in all of our 600+ tests.
1559 * New command line option --version, to get SimGrid version information.
1560 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1562 * Supernovae builds are deprecated, and expected to be removed in the next
1565 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1567 SimGrid (3.9) stable; urgency=low
1569 The Grasgory release: GRAS is really dead now.
1571 * Complete overhaul of the internal host structures scheme.
1574 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1575 was considered as experimental and badly maintained since then.
1576 * Keeping it was thus a trap to our potential users, that could take
1577 it instead of MSG or SMPI by mistake despite is pity state.
1578 * GRAS seems to have very few users (if any), and no one volunteered
1579 to maintain it further. It also induces a lot of XBT code (for
1580 portability sake), that must be maintained too.
1581 * For all these reasons, we killed GRAS. If someone wants to revive it
1582 in the future, don't cry, our git history still remembers of GRAS.
1585 * Major overhaul. Merge our documentation again as time proved that
1586 splitting it was really not helping our users.
1587 * Further improve the developer documentation to help newcomers
1588 hacking on SimGrid itself. The user documentation (and in
1589 particular, the beginner documentation) is still in a sorry state.
1592 * Now works on Windows too!
1593 * Much more extensive test suite, from MPICH
1596 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1597 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1598 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1599 * Bug fix: let task be scheduled when the last dependency to be solved is
1600 a control dependency.
1601 * Remove SD_load_environment_script function.
1602 Use the C sg_platf function if you want to declare a platform
1606 * New function: MSG_process_get_number()
1607 * Old function documented: MSG_config()
1608 * Remove MSG_load_platform_script function
1609 Use the C sg_platf function if you want to declare a platform
1613 * Change the default value of the TCP_gamma constant (maximal size of TCP
1614 congestion window) to a more realistic 4MiB value. If you notice changes in
1615 your simulation results, you can fall back to the previous 20k tiny window
1616 by adding --cfg=network/TCP_gamma:20000 on command line.
1617 * (Hopefully) fix a bug wrt periodic availability and state traces
1618 * Bug fix: use default values at start when first event in availability/state
1619 trace is not at time 0.
1622 * remove the "new_" part of function name sg_platf_new_trace_connect
1623 (resulting in sg_platf_trace_connect), since it does not create
1627 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1628 use case where it's really mandatory, and maintaining it was a pain in
1630 * New: xbt_fifo_search(), search an item with a user-provided
1631 comparison function instead of dumb pointer comparison.
1634 * Fix the lua deployment:
1635 Use `simgrid.init_application()` before deployment instead of
1636 `simgrid.msg_register_application()` after.
1639 * Transfer the tracing files into the corresponding modules.
1641 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1643 SimGrid (3.8.1) stable; urgency=low
1645 The "we are told that some people want to also *install* the simgrid
1648 * Add missing file "tesh.1" to the archive.
1650 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1652 SimGrid (3.8) stable; urgency=low
1654 The Psssshiiiit release: SimGrid jumps into the Cloud.
1657 * Add an experimental interface to manipulate VMs. They are mainly
1658 process groups with very few intrinsic semantic, but they should
1659 allow you to build the semantic you want easily.
1660 * New function: MSG_host_set_property_value()
1661 * New function: MSG_process_on_exit(). To clean memory in all cases.
1662 * Bug fixes that made the host (and link) failures unusable.
1663 * Add a way to auto-restart process when the host in which they are
1664 executing comes back (ON_FAILURE="RESTART" on deployment file,
1665 MSG_process_auto_restart_set).
1666 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1667 please stop using the old ones, they are DEPRECATED.
1669 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1670 Please use MSG_init() instead. (reducing the amount of entry
1671 points in the library helps us).
1672 * Make it impossible to link against the wrong version of the lib
1673 * Deprecate MSG_clean(). No need to call it anymore.
1674 * Function MSG_get_host_number() is not deprecated anymore.
1677 * Split the doc into a user guide and a reference guide.
1678 * Start a developper guide to help people hacking on SimGrid.
1681 * Enable tracing by default. This modules rocks you should use it.
1682 * Remove option custom_flags. Now use environment variables CFLAGS
1684 * Use default cmake things to detect lua instead of home grown ones.
1685 * New option "enable_mallocators" to disable mallocators, for debugging
1686 purpose ("on" by default).
1689 * Bug fixes around the resource failures: don't let the processes
1690 survive the host they are running onto.
1691 * Add an interface to auto-restart processes when the host in which they are
1692 executing comes back.
1693 * Ensures that SIMIX_clean is called automatically. It's not part of
1694 the public interface anymore (bindings should be updated).
1697 * Bug fix for when SD_Simulate is called with a positive value: be careful
1698 when comparing doubles. Sometimes they are different for non significant
1700 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1701 parallel task whose initial work is distributed among host according
1702 to the Amdahl's law. Such tasks are created with a parameter alpha
1703 that corresponds to the non-parallelizable part of the computation.
1704 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1705 between two sets of workstations assuming a 1D block distribution (each
1706 workstation owns a similar share of data) on both sides.
1708 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1709 Data redistribution will be automatically scheduled once parent and child
1710 are both scheduled. The filling of computation_amount and
1711 communication_amount structures is now done seamlessly thanks to the chosen
1713 * New function SD_workstation_dump to display various information
1714 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1715 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1716 bandwidth on the route onto which the task is scheduled and the amount of
1718 To divide the nominal bandwidth by 2, the rate then has to be:
1719 rate = bandwidth/(2*amount)
1720 * Compute tasks that have failed can now be rescheduled and executed again
1721 (from their beginning)
1722 * Increasing source code coverage (src/simdag is now covered at 95.8%
1726 * Re-implement time-independent trace replay using SMPI (at the
1727 smpi_smp_* level) instead of MSG. This should replace
1728 examples/msg/actions/actions.c
1729 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1730 hindexed and structs)
1731 * Implement the exchange of non-contiguous data.
1732 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1733 * Correct behavior of smpi/sender_gap and set its default value to 0
1734 * Add option to asynchronously send small messages to allow better
1735 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1736 specifies the size in bytes under which messages will be asynchronously sent.
1737 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1738 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1741 * New C interface to define a platform: XML is now optional.
1742 For more info, please check include/simgrid/platf.h.
1743 * New interface to define random platforms from the C:
1744 For more info, please check include/simgrid/platf_generator.h and
1745 examples/msg/masterslave/masterslave_platfgen.c
1746 * Export a sg_cmdline dynar containing all the arguments we got from
1750 * Two new tracing options for adding comments to trace file so you
1751 can track your experiments (see --help-tracing for details).
1752 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1753 * Adding the SimGrid version that generated the trace file as a comment.
1754 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1755 * Fix to avoid key clashes on Paje links
1756 * Other minor fixes related to the Paje specification
1759 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1760 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1761 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1762 * Greatly improve the robustness of mmalloc to user errors (such as
1763 using an area after freeing it, or freeing it twice)
1765 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1767 SimGrid-java (3.8.1) stable; urgency=low
1769 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1770 Warning: all methods are visible, but only some of them are
1771 implemented so far. Check the source (src/jtrace.c)
1772 for further information.
1773 * New module: org.simgrid.msg.File (SimGrid File management functions)
1774 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1775 * Change the meaning of Process.restart: now restart the process from
1776 the beginning, like MSG_process_restart in C.
1777 * Add Process.setAutoRestart: handling of process restart when failed
1779 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1780 you to retrieve the properties of the processes/hosts
1781 * Deprecate Msg.clean(): you can just forget about it now.
1782 * New function Process.getCount(), that only works when compiling
1783 with the not yet released version 3.9 of the C library.
1785 * New context factory based on Coroutines. It mandates a modified JVM
1786 but then, the simulations run about five times faster, and there is
1787 no limit to the amount of processes (beside of the available memory).
1789 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1791 SimGrid (3.7.1) stable; urgency=low
1794 * Restore the prototype of MSG_process_create_with_environment() to
1795 the pre-3.7 situation by removing the kill_time argument.
1796 * Add a MSG_process_set_kill_time() function instead.
1799 * Fix weird behaviors when dealing with parallel tasks.
1802 * Simgrid is now built as a dll.
1803 * Simgrid-java now works on Windows.
1804 * Simgrid-Java is now included into Windows package.
1807 * First pre-build package for MacOSX.
1810 * Fix compilation when using MSG_USE_DEPRECATED.
1811 * Fix some compilation issues on Macs and Windows.
1812 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1813 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1815 We discovered that the Lua console is broken, but we are missing the
1816 manpower to fix it right now. The problem existed in 3.7 too, so we
1817 are not blocking the release for that. Sorry if you depended on this
1818 feature, any help would be really welcome.
1820 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1823 SimGrid-java (3.7.1) stable; urgency=low
1825 The "Java aint got to be bloated and slow" release
1828 * Various internal cleanups and performance improvement
1829 Simulations are expected to run up to twice faster or so
1830 * Make Process.kill(process) an instance method, not a static one
1831 * User processes are not java.lang.Thread subclasses.
1832 This breaks the compatibility (sorry), but previous API was
1833 brain-dead, making it impossible to have non-trivial
1834 initializations in the process constructor.
1835 * Require a full constructor per Process sub-class.
1836 Kinda breaks the compatibility (sorry), but this allows a much more
1837 efficient way to launch the processes at simulation startup.
1838 * Do not embeed our version of semaphores, java 1.5 can be considered
1839 as sufficiently prevalent for us to not dupplicate its features.
1843 * Add examples for almost every part of the API
1844 We spotted and fixed a lot of bugs in the process
1846 * New module: asynchronous communication API
1847 * New function: Process.sleep()
1848 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1849 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1850 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1852 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1854 SimGrid (3.7) stable; urgency=low
1856 The "spring cleanups (before the next Big Project kicks in)" release.
1859 * We can specify the SMPI latency/bandwidth factor with command line
1860 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1861 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1862 You can also use the "config tag" from platform file by adding this line
1863 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1864 (see "example/platforms/tag_config.xml" to use "config tag").
1865 Note that the command line supersedes the platform file configuration.
1866 * Change the correction factors used in LMM model, according to
1867 the latest experiments described in INRIA RR-7821.
1868 Accuracy should be improved this way.
1869 * Use the partial invalidation optimization by default for the
1870 network too. Should produce the exact same results, only faster.
1871 * Major cleanup in surf to merge models and split some optimization
1872 mechanisms from the core of the models. As a result you can now
1873 specify which model to use (e.g., --cfg=network/model:LV08
1874 --cfg=cpu/model:Cas01) and which optimization mode to use
1875 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1876 Incompatible combinations should err at initialization. See
1877 --help-models for the list of all models and optimization modes.
1878 * The CLM03 workstation model was dropped for simplicity because it
1879 used the deprecated CM02 network model. Use default instead.
1880 * Rename the TCP_gamma configuration option to network/TCP_gamma
1881 * Rename the coordinates configuration option to
1882 network/coordinates, and document it
1883 * Use now crosstraffic keyword instead of the terribly misleading
1884 fullduplex keyword. It is activated by default now in the current
1885 default model, use --cfg=network/crosstraffic:0 to turn it off.
1886 * Ongoing refactoring the model parsing to make XML files optional
1887 See include/simgrid/platf.h for details (still to be completed)
1890 * Major overhaul of the documentation. Almost instructive now :/
1891 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1892 functions and friends. This interface was considered as
1893 deprecated since over 2 years, it's time to inform our users that it is.
1894 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1895 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1896 use these (crappy) functions in your code.
1897 These functions will be removed soon. Stop using them now.
1898 * Deprecate MSG_get_host_{table,number}
1899 Implement MSG_hosts_as_dynar() instead.
1900 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1901 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1904 * Stabilize the parallel execution mode of user contexts
1905 * Introduce configuration variables to control parallel execution:
1906 - contexts/synchro: Synchronization mode to use when running
1907 contexts in parallel (either futex, posix or busy_wait)
1908 - contexts/parallel_threshold: Minimal number of user contexts
1909 that must be part of a scheduling round to switch to parallel
1910 execution mode (raw contexts only)
1911 * Fix bugs that prevented to use suspend/resume along with
1912 synchronization structures.
1913 * Fix bugs in process termination that lead to invalid memory access
1914 in very specific conditions.
1917 * Introduce a parallel mode for the models (controlled by surf/nthreads
1918 configuration item). In our tests, running the models in parallel
1919 never lead to any speedups because they are so fast that the gain
1920 of computing each model in parallel does not amortizes the
1921 synchronization costs, even when ultra fast futexes are used.
1922 This is released anyway because YMMV.
1925 * Performance boost by using a swag internally to compute the set of
1926 tasks that are finished and should constitute the return value of
1930 * Enable it by default now that it is considered rather stable.
1933 * Documentation of the tracing functions.
1934 * Performance gains when tracing categorized/uncategorized resource
1935 utilization by avoiding calls to get route when updating resource
1936 variables. LMM constraints are being used instead.
1937 * API changed to set task categories. Use MSG_task_set_category instead
1938 of TRACE_msg_set_task_category, and SD_task_set_category instead
1939 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1940 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1941 * Examples for link user variables added (see at examples/msg/tracing/)
1942 * Deprecated function TRACE_msg_set_process_category completely removed
1943 * Trace header updated according to the latest Paje file format
1944 * Tracing network lazy updates, no longer obligate users to use full updates
1945 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1946 * Experimental: let user code declare/set/push/pop application states for hosts
1947 * API changed to allow the manual creation of graph configuration files
1948 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1951 * Improve the API of Lua MSG bindings, using the Lua spirit.
1952 * Each simulated process now lives in its own Lua world (globals are
1953 automatically duplicated). It helps writing simulators. Will allow
1954 to run Splay programs within SimGrid in the future.
1955 * Add a Chord example in Lua, equivalent to the MSG one.
1958 * Start the implementation of a solution to express temporal
1959 properties, not only local assertions. This is still an
1960 experimental work in progress, stay clear from it to be safe.
1964 - Add new runtime parameters --help-logs and --help-log-categories
1965 to display information about supported logging parameters and
1967 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1968 don't exists anymore.
1969 * Mallocators: allow value NULL for the reset function.
1971 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1972 create homogeneous dictionaries, where all the elements share the
1973 same free function. Non homogeneous dictionaries will be
1974 deprecated in the next release.
1975 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1976 - Multi-level dictionaries are deprecated.
1978 - new function xbt_dynar_search_or_negative() that is useful when
1979 you have less than 2 million elements in your dynar and don't
1980 want of the extra complexity of catching exceptions when the
1981 element is not found.
1983 - Make xbt_os_thread module (for thread portability) public.
1984 Documentation is still to come, sorry.
1986 - Cleanups and simplifications to make it maintainable again.
1987 - Exotic features (such as memalign and valloc) were removed.
1988 - The metadata were extended and improved so that the
1989 model-checker becomes able to explore and inspect the heaps.
1990 - This may induce a performance drop when enable_model-checking is
1991 ON in cmake (even if it's not used in the simulation), but it is
1992 necessary at this point to get MC working.
1994 Turn model-checking OFF if simulation performance matters to you.
1995 Not enabling it at runtime is not enough, disable it in cmake.
1997 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2000 SimGrid (3.6.2) stable; urgency=low
2002 The "Not coding new stuff allows to polish old things" release.
2005 * New bindings to the NS3 packet level simulator (experimental)
2006 * Use the raw (efficient) execution contexts instead of the sysv
2007 (portable) ones when possible.
2008 * libpcre is now mandatory in any cases since not using it led to
2009 severe performance loss and possibly other issues
2010 * Update the XML platforms:
2011 - G5K: include the latest machine in Nancy
2012 - GridPP and LCG: new platforms
2013 * Documentation was partially updated, at least (more to come)
2015 Bug fixes, cosmetics and small improvements
2016 * Free terminated processes before the end of the simulation to avoid
2017 exhausting the memory of users having very dynamic amount of
2019 * Bug fix and cosmetics about canceling non-running tasks
2020 * Bug fix about the dot loader's issues when using libcgraph
2023 * Create an installer for windows with nsis (amd64 and win32)
2024 - Add an hello world project to illustrate simgrid project creation.
2025 - Embed libpcre into the Simgrid installer to avoid
2026 its compilation burden
2027 * The raw execution contexts should work on Apple now
2028 * Port to Windows 64 bits
2029 - Sysv contexts now have an implementation for this arch
2030 - GRAS communication features now support this arch
2031 * Drop support for borland compiler on windows
2032 - this code was not maintained, and we kinda depend on gcc nowadays
2033 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2034 * Fix portability issue on unstable ubuntu: linker became picky on
2037 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2039 SimGrid (3.6.1) stable; urgency=low
2041 The "Oops, we broke Macs too" release
2044 * Fixed contexts detection so that raw ones are used when possible
2045 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2046 leads to a strange error, with user code segfaulting sometimes when
2047 the generated code is not perfectly aligned (which is not
2048 controllable from the user side, depends on the amount of code)
2051 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2053 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2055 SimGrid-java (3.6) unstable; urgency=low
2058 * Split of every thing from simgrid v3.5 into a separate package.
2060 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2062 SimGrid (3.6) stable; urgency=medium
2064 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2067 * Bindings now constitute their own package, separated from the main one.
2068 Rationale: reduce our maintenance nightmare by reducing the module coupling
2069 They will soon be released on their own on gforge.
2071 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2072 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2074 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2075 * It's not quite deprecated for now because we have no replacement,
2076 but it may soon become the case.
2079 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2080 * Fortran: New user-level cache variable to store the rank of the running
2081 process. This improves performance by an order of magnitude.
2082 * C: New coccinelle script to automatically locate and modify global and
2083 local static variables.
2084 * Improved SMPI network model with a sender-side gap to account for multiple
2088 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2089 only say if a communication is finished, no matter whether it succeeded or
2090 failed. You can call MSG_comm_get_status() to know the status of a finished
2092 * New function MSG_task_dsend() to send a task and detach it. When a
2093 communication is detached, you are never notified of its success or failure
2094 and the memory is released automatically once it is finished. This function
2095 is useful when you don't care about the end nor the success of a
2097 * Change the prototypes of action replay. Sorry for inconvenience,
2098 but this is really more efficient this way (and to adapt your code,
2099 you just have to fix the initialization, that shouldn't be too long)
2100 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2101 ever managed to do anything useful with it.
2102 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2103 takes a dynar of communications. It returns immediately and gives the
2104 index of a finished communication (if any).
2105 * New example: a basic implementation of the Chord P2P algorithm.
2108 * New model for multi-core CPUs. You can now use the core attribute to
2109 precise the number of cores of a host. This is a basic model. Every
2110 process running on the host receives at most the power provided in
2111 the DTD (throughput<=power). Total throughput of process cannot exceed
2113 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2114 router linked by an up-link and a down-link (possibly asymmetrical).
2115 This kind of pattern allows to easily build last-mile model style platforms.
2116 Aggregating such patterns in a rule-based AS is thus the technique of
2117 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2118 * New model for Vivaldi routing. We transformed the Vivaldi network model
2119 into a Vivaldi routing model (based on the rule-based model). This allows to
2120 combine Vivaldi based latencies with last-mile platforms.
2123 * Added a check for NaN of IEEE754 infinite in the double entries of
2125 * Introduce a new context factory "raw", highly inspirated from the
2126 ucontext factory, but using manually crafted functions in assembly to
2127 do the work in an efficient manner.
2128 * Allow to change the used context factory at run time, not only at
2129 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2130 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2131 contexts at runtime (only with raw contexts or ucontexts).
2132 * Completely rewrote this module to allow parallel execution of user
2133 processes. Use --cfg=contexts/nthreads:N to execute user processes
2134 with N parallel threads (the default is 1, meaning no parallelism).
2135 * Allow to decide dynamically between sequential and parallel modes.
2136 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2137 processes in parallel only when their number is greater than or equal to P
2139 * Added a check for NaN of IEEE754 infinite in the double entries of
2143 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2144 won't output the state of processes when interrupted with Ctrl-C
2145 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2146 NULL-terminated array. This may solve backward compatibility issues
2147 due to the change to return type of SD_simulate. See also:
2148 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2149 * Add new macros with variable number of arguments.
2150 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2151 - in xbt/asserts.h: xbt_assert
2152 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2153 - in xbt/ex.h: THROWF and RETHROWF.
2154 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2156 * Change xbt_die() to accept a format string with arguments, just like printf.
2157 * New data structure: xbt_lib_t, like a dict but more general and with better
2161 * New configuration options
2162 Options triva/categorized and triva/uncategorized can be used to generate
2163 graph configuration files for Triva visualization tool.
2164 * Configuration option tracing/platform is renamed to tracing/categorized
2165 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2166 * New TRACE_user_link_variable interface:
2167 User provides the name of the link and the tracing variable to attach to it
2168 * the declaration of tracing categories must be done after the environment
2170 * simpler tracing interface, just one way to declare categories
2171 TRACE_category or TRACE_category_with_color, it is up to you
2172 * links in the trace file are again identified by their names
2173 * trace contains the full platform hierarchy exactly as declared using the ASes
2174 * Options tracing/msg/[task|process]:1 groups the process by hosts
2175 for both cases, tasks and processes must have names that are unique during the simulation
2176 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2177 * The experimental option tracing/msg/volume is deprecated
2178 its functionality may be reincorporated if needed
2180 The tracing generates a trace file with unordered timestamped events,
2181 because of the way the core simulator (surf) works. A script available
2182 at the tools directory (fix-paje-trace.sh) can be used to put the events
2183 in order. We have changed the tracing so it can generate ordered timestamped
2184 events in the final trace, but depending on the simulator (and how much time
2185 is simulated) that can lead to a huge memory utilization. It is deactivated
2186 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2188 Build Infrastructure
2189 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2190 - We are trying hard to keep the API stable, but it may happen that
2191 some things change (we're a research project after all, not a
2192 nuclear plant operating system). If such things should happen, you
2193 could rely on that macro to adapt.
2194 - current value: 30600 for 3.06.00, aka 3.6
2195 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2196 a number that can be compared with SIMGRID_VERSION.
2197 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2198 assertions and verbose/debug logging events are disabled at compile time.
2200 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2202 SimGrid (3.5) stable; urgency=medium
2205 * New feature to any SimGrid-based simulator: Model-Checking
2206 Check SIN#1 for more details.
2209 * New Model SMPI (three-interval linear regression for correction factors)
2210 See RR-7426, available at http://hal.inria.fr/inria-00527150
2211 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2212 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2213 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2214 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2215 * See also src/smpi/README
2220 - Tracing API changes: TRACE_start and TRACE_end should not be called
2221 by user-code. They are automatically called by simulators created
2222 with SimDAG, MSG and SMPI if the toolkit is compiled with
2223 tracing_enabled=ON. Categories declaration and utilization remain the
2224 same for MSG and SimDag.
2225 - A function was added to the tracing API to declare categories with
2227 - TRACE_category_with_color (char *category, char *color)
2228 where color must be in the following format
2229 "%f %f %f", red, green, blue
2230 and red, green, blue are float values in the interval [0, 1]
2231 - User can specify NULL as color parameter, or continue calling
2232 TRACE_category (cat)
2233 On that case, the tracing system will define random colors
2234 - The following command-line options are supported:
2235 --cfg=tracing/filename:msg.trace
2236 --cfg=tracing:1 (activate tracing, needed to use others)
2237 --cfg=tracing/platform:1 (categorized resource use)
2238 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2239 --cfg=tracing/msg/task:1 (task creation)
2240 --cfg=tracing/msg/process:1 (process creation, migration)
2241 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2242 --cfg=tracing/smpi:1 (SMPI interface tracing)
2243 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2244 - examples of examples/msg/tracing updated
2246 - DAXLoader and DOTLoader functions can generate tasks with categories
2247 - A new function to attribute a category to SD tasks:
2248 TRACE_sd_set_task_category (SD_task_t task, char *category)
2249 * Tracing the MPI interface implemented by SMPI
2250 - Collective operations are traced with states
2251 - Point-to-Point operations are traced with states/links
2252 - Tracing activated by a parameter "-trace filename" passed
2253 to smpirun during execution (considering that simgrid
2254 is compiled with tracing enabled)
2255 - To run the simulation with gdb, the simulator
2256 accepts --cfg=tracing/smpi:1 to trace SMPI
2257 - tesh files to check if smpi tracing is ok
2258 - See examples/smpi/NAS/DT-trace
2259 * GTNetS tracing re-worked
2260 - adaptation to the tracing system of GTNets to cope
2261 with modifications regarding the fullduplex mode
2262 - new tesh files to check if gtnets tracing is ok
2265 * Asynchronous communications through the functions:
2266 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2267 * New function: MSG_load_platform_script()
2268 to make possible using a lua script instead of XML files to set up platforms
2269 * New function: MSG_set_function
2270 to associate functions to processes, used when bypassing the parser
2271 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2273 Platforms: Add some more examples in examples/platforms
2274 * Grid'5000: see www.grid5000.fr
2275 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2278 * Change the XML format. This is a very important modification. SimGrid
2279 3.5 introduces a new hierarchical format based on the notion of
2280 Autonomous Systems. Compatibility with old format is ensured
2281 through the perl script provided in the install bin directory
2282 bin/simgrid_update_xml.
2283 It is now possible to build platforms with specific routing
2284 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2285 connect several platforms together. We will try to provide soon
2286 set of realistic platforms exploiting these properties (have a
2287 look at examples/platforms/ for the moment).
2288 * Take the opportunity of the XML format change to be a good XML citizen:
2289 rename link:ctn to link_ctn and similar changes (also dealt with by
2291 * Add a new routing scheme (rule-based) using regular expressions. It
2292 enables to have an extremely low memory footprint when the
2293 underlying routing is simple and can be compactly described. You
2294 need to have libpcre4-dev (perl regular expressions) installed if
2295 you want to use this routing scheme.
2296 * Revive the cluster TAG and allow to easily and efficiently (both in
2297 term of memory and speed) connect clusters together. Have a look
2298 at teshsuite/simdag/platforms/ to see how this can be done. With
2299 this tag, you can create clusters with thousands of tasks at no
2300 cost (have a look at examples/platforms/).
2301 Note: clusters are implemented as ASes, so there is no need for an
2302 enclosing AS tag if you have only one cluster in your platform.
2303 * Add new generic functions in the public interface that allows the user
2304 to call SURF 'create_resource' methods from your code (same
2305 functionality as the XML bypass mechanism but with a much lighter
2307 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2308 that uses a piecewise linear approximation to produce better
2309 results when exchanging small messages.
2310 * Add a new parameter to handle correctly full duplex link and account
2311 for interferences between uplink and downlink communications
2312 (activate with --cfg=fullduplex:1).
2315 * Rename the SD_READY (all dependencies are satisfied and task is
2316 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2317 dependencies are satisfied) state.
2318 This prevents a confusion between the notion of "ready to schedule"
2319 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2320 simulated" (SD_RUNNABLE) used by the simulation kernel.
2321 * Change the way a task is considered as ready. Instead of removing
2322 dependencies when a task is done, a counter is decreased. This way,
2323 it is always possible to reach ancestors thanks to the
2324 SD_taks_get_parents function (even after the end of the simulation.)
2325 * Change the return type of SD_Simulate from (SD_task_t*) into
2326 xbt_dynar_t. This function was in handling a dynar internally and
2327 converted it into a NULL terminated array for historical reasons.
2328 * New function SD_dotload(char*) to load a DAG described in dot
2329 format. This loader and the corresponding examples require the
2330 installation of the graphviz library.
2331 * Fix a bug in the management of tasks of size 0 in the surf network
2332 models. This problem was only visible with SIMDAG and you should
2333 thus disregard results produced with earlier versions if you
2334 relied on this feature (some tasks were blocked because of this).
2335 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2336 even though your applications were doing only point-to-point
2337 communications and sequential computations. Now you can really use any
2338 model you want (of course, if you create real parallel tasks, which are
2339 not implemented in most models beside ptaskL07, this will abort).
2340 * Add an example that schedules a DAX on an heterogeneous platform
2341 using a Min-Min strategy.
2342 * New function SD_workstation_get_current_task() that returns the kind
2343 of task currently running on a workstation in the sequential access
2345 * Raise some warnings when unexecuted tasks remains at the end of the
2346 simulation. This is usually caused by cycles in the DAG.
2349 * New function: SIMIX_process_set_function() called by MSG_set_function
2350 * Change the underlying waiting queue in semaphores so that a process
2351 can wait on several of them simultaneously (as in waitany).
2352 * Fix the way to handle tokens in semaphores so that all access patterns
2353 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2354 * kill the dirty pimple SIMIX_message_sizes_output()
2355 Please use (proper) visualization instead
2358 * New data container: setset (set of sets of elements)
2359 * New module: mmalloc (mapped malloc, allowing to have several
2360 independent segments of malloc)
2361 * New function: xbt_dict_cursor_set_data()
2362 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2363 * New function: xbt_dynar_is_empty()
2364 * New function: xbt_fifo_get_last_item()
2365 * Fix xbt_dynar_shrink(): use the right element size.
2366 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2367 uninitialized areas during expand.
2368 * Fix semaphores: previous implementation was severely broken.
2369 * Use library init/fini functions for our initialization.
2370 - you can use logs and other feature as soon as you want in your
2371 code (even before the xbt_init / MSG_init)
2372 - xbt_exit is now a no-op and produce a warning when used.
2375 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2376 benefit from the latest improvement to the simulation kernel.
2377 * Kill measurement sockets for now. If you rely on them, sorry. This
2378 release is not for you. This feature will be reintroduced in the
2379 future, but we cannot delay the release any further.
2380 * New function: gras_msgtype_get_name().
2381 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2382 * Fix (at last) the pmm example: it should not randomly fail anymore.
2384 Build chain: bug fixes and overall polishing
2385 * Cmake is now stable enough. Hence, we killed the autotools.
2386 * Port to windows ( TM :)
2387 * Fix the 'make install' target.
2388 No need to use 'make install-simgrid' anymore
2389 * Introduce a 'make dist' target compiling a *source* archive
2390 'make package' compiles a binary archive
2391 * Compile java files only on need
2392 * Add --cd and --setenv command line options to tesh
2393 * Out of source builds are not fully supported yet, but we are close
2394 * Enable supernovae and optimization flags by default for our users
2397 * Add layer to set up environment directly from lua, without XML.
2398 * The effect of gras_stub_generator can be achieved through
2399 lua too (check examples/gras/console/ping_generator.lua)
2401 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2403 SimGrid (3.4.1) stable; urgency=low
2405 The "Polishing easter eggs is probably a good idea" release.
2406 This is a bug fixes release only.
2410 * Fix a bug preventing the tasks from begin garbage collected.
2413 * Fix a bug occurring when a host involved in a communication fails.
2414 This was not detected properly by the other peer involved in the
2415 communication. Now, it's reported as a network error.
2418 * Warn the user about loop dependencies in data flow of DAX files
2419 * Obey the control-flow dependencies of DAX files
2422 * Add option "enable_smpi" allowing to not compile SMPI.
2423 Probably useful for the (Mac) users experiencing a build error here
2424 * Improve the detection of lua5.1 and ruby1.8
2426 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2428 SimGrid (3.4) stable; urgency=low
2430 The "Easter in Cargese" release. Also known as (major changes):
2432 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2433 ~> bindings were greatly improved
2434 ~> new tracing infrastructure for better visualization introduced
2436 * the "Welcome to configury modernity" release.
2437 ~> we switched from autotools to cmake, and improved our cdash
2440 A more detailled list of changes follow (full detail in svn log).
2442 Java Bindings: Various Cleanups
2443 * (install java-gcj-compat-dev on debian-like to use them)
2444 * Remove put/get: no need to export deprecated interface in Java
2445 Use send/receive instead.
2446 * Cleanup the examples and add a README per directory
2447 * Remove example autoDestination (that's the only way to go now)
2448 * Remove example explicitDestination (was a plain copy of basic)
2449 * Make JniException a runtime exception, so that there is no need to
2450 declare the fact that you may encounter such a beast. I guess that
2451 nobody will ever want to survive such error.
2452 * Create specific errors for each MSG case of failure:
2453 host failure, transfer failure, timeout, task cancelled
2454 * Cleanup the exceptions that may get thrown by each function
2455 * Other internal cleanups in Java bindings. Performance still bad :/
2456 Ruby and Lua Bindings: create them
2457 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2458 * That's new and great, you should try them out.
2459 Same functionalities than Java bindings, only even less polished
2461 * Kill the useless "rate" argument of SD_task_get_execution_time()
2462 Everyone used to provide -1 as a value, it was not used, and the
2463 semantic of a possible use wasn't even clear.
2464 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2465 either as comm costs or compute costs to mean that there is no
2466 such thing for that specific task.
2467 * Add a SD_task_set_name() function
2468 * Fix SD_task_unschedule() on typed tasks
2469 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2470 * In DAX loader, accept useless 'level' attributes to <job> since
2471 LIGO DAGs have them (seem to be to ease graphical representation).
2473 * Add an example masterslave_mailbox.c using send/receive and not
2474 the deprecated put/get interface.
2475 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2476 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2478 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2479 (declare MSG_USE_DEPRECATED to still have the old name)
2480 * Add a MSG_task_set_data() function
2481 * About trace replay (see examples/msg/actions):
2483 - Allow to work with splitted trace files for each process
2484 Give the specific trace file as argument of each process,
2485 and call MSG_action_trace_run(NULL)
2486 You can still have one merged file for all processes.
2487 - Fix implementation of collective operations
2488 * Allow task_execute() on 0-sized tasks (closes #10063)
2490 * This is the first release of SimGrid where SMPI is not considered
2491 beta anymore (even if some corners should still be improved)
2492 * Port over the new SIMIX_network submodule (internal refactoring)
2493 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2494 * Implement more missing elements of the standard:
2496 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2498 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2499 MPI_Scan() MPI_Get_processor_name()
2500 - Added implementation of missing case for Alltoall (warning: it's
2501 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2502 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2503 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2504 SMPI_MPI_Allgatherv()
2505 * Bug fixes include:
2506 - MPI_Waitsome() was broken
2507 - Allow relative includes in smpicc
2508 - Command line cfg argument 'reference_speed' was ignored...
2509 - Some functions did not properly lead to auto-benching of user code
2510 - smpicc passes -O2 by default (just like openmpi one)
2512 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2513 * Bug fixes about timeouts during communications
2514 * add SIMIX_message_sizes_output() as a pimple to write to file the
2515 amount of messages per size. Use gnuplot to get histogram.
2516 Pimple because that's the only user-visible function of simix,
2517 defined directly in xbt.h (irk, sorry)
2519 - Add a SIMIX_sem_get_capacity() function
2520 - Fix interactions with processe resume/suspende
2521 - release_forever() was stupidly broken
2522 - Fix SIMIX_display_process_status() for processes in a semaphore
2523 - Make SIMIX_sem_block_onto() user-visible
2524 * Refactoring context stuff:
2525 - Use pseudo-OOP for better modularity
2526 - reimplement SIMIX_process_kill() without process_schedule() so
2527 that the latter can take as invariant that it is called from
2529 - Merge context_start into context_new for sake of simplicity
2531 * Add a Vivaldi network model, coded live during SUD'10 ;)
2532 * Rename configuration variables to start a hierarchy:
2533 o cpu_model -> cpu/model
2534 o network_model -> network/model
2535 o workstation_model -> workstation/model
2536 * New configuration variables:
2537 o network/bandwidth_factor: correction to bandwidth
2538 o network/latency_factor: correction to latency
2539 o netwotk/weight_S: correction to the weight of competing streams
2540 * Add a long description to the models, that users can see with such
2541 argument on the command line: --cfg=cpu/model:help
2542 * --help-models display the long description of all known models
2544 * config: add the ability to set a default value after registration
2545 Does not override any previously set value (e.g. from cmd line)
2546 * dict: allow to have integer key and data.
2547 When so, you need to use the following functions
2548 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2549 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2550 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2551 In contrary to regular dicts, the key is not malloced before copy.
2552 Mixing scalar and regular elements in the same dict is not tested
2554 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2555 Tracing for Visualization:
2556 * SimGrid is now instrumented in order to generate a trace file for
2557 visualization analysis: to use it, need to compile SimGrid with the
2558 "tracing" option enabled, and instrument the program using SimGrid with
2559 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2560 (among other functions).
2561 * The instrumentation only traces the platform utilization for now
2562 * Documentation to use the tracing functions and how to analyze the
2563 traces with the Triva tool is written.
2564 * More information about: SimGrid FAQ (in the section Tracing Simulations
2567 * We moved to cmake as default build system. Autotools support will
2568 be dropped soon. Check the FAQ for more info about how to use it.
2569 * Greatly improved our cdash/ctest interactions
2570 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2571 * Added memory checking tests with valgrind; lot of memleak fixing.
2572 This may be the first release of simgrid with so few memory issues
2573 * Added code coverage tests.
2574 Our coverage is still improvable, but at least we see it on cdash.
2576 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2578 SimGrid (3.3.4) stable; urgency=low
2580 The "Desktop Grid needs love too" release (also called Xmas release).
2582 Models improvements:
2583 * Major speedup in the maxmin system solving by using lazy evaluation
2584 Instead of solving completely the maxmin system at each iteration,
2585 only invalidate (and recompute) the modified parts.
2586 This new feature is enabled in default models but you can try to
2587 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2588 * Cas01 IMproved as default CPU model
2589 This CPU model is the same Cas01 model, but it uses the
2590 maxmin-selective-update flag and a heap structure to manage
2591 actions on SURF kernel.
2592 It reduces the complexity to find the next action to finish and,
2593 consequently, it's faster than the old Cas01.
2594 This is the new default CPU model (Cas01).
2595 * Rename the old Cas01 model to Cas01_fullupdate
2596 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2597 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2598 * CpuTI (CPU Trace Integration)
2599 A new CPU model whose objective is simulate faster when using
2600 availability trace files.
2601 Instead of using a full featured, over engineered maxmin system for
2602 CPU modeling, this model does the pre-integration of traces files
2603 to calculate the amount of CPU power available, and so, executes
2604 faster than the old CPU models.
2605 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2606 * Use LV08 as default network model since it gives better accuracy
2607 for small messages and shouldn't change things for big ones.
2608 Use --cfg=network_model:CM02 to get the previous behavior.
2611 ******************************************
2612 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2613 ******************************************
2614 * The new CPU model may changes simulations!
2615 The point is that events occurring at the exact same timestamp
2616 are not scheduled in the same order with the old and new
2617 version. This may be enough to completely change the execution
2618 of simulations in some cases.
2619 * The new network model will change simulations!
2620 This new model is more realistic than the previous one, so you
2621 should consider redoing your old experiments with this model.
2622 Sorry for the inconvenience.
2625 * Introduce the supernovae compilation mode
2626 When compiled that way, the whole SimGrid (or almost) is put in a
2627 single compilation unit and compiled in one shoot.
2628 This is to help gcc which has difficulties to inline stuff from one
2630 The speedup seem to be above 15%, althrough more tests are needed on
2631 amd64 to confirm that gain.
2634 * Port of MSG's mailbox on top of SIMIX network
2635 The put/get mechanism was greatly simplified on the way.
2638 * New SIMIX network module. Provides:
2639 - Mailbox: rendez-vous mechanism to find with who you want to speak
2640 - Synchronous send/recv: easier and hopefully faster since the
2641 logic is handled in the maestro process directly now
2642 - Asynchronous send/recv: you dreamt of it? It's here now
2643 Too bad that nobody cared enough to propagate the change to MSG.
2644 * Add semaphores as SIMIX synchronization mechanism.
2647 * new function SD_daxload(char*) to load a DAX file
2648 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2649 * Introduce typed tasks. Specify its kind and cost at creation.
2650 At scheduling, just give where it should be placed, and the cost
2651 for each involved resource is automatically computed.
2652 Existing constructors so far (more to come of course):
2653 - SD_task_create_comm_e2e() for end-to-end communication
2654 - SD_task_create_comp_seq() for sequential computation
2655 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2656 * new function SD_task_dump() for debugging display
2657 * new function SD_task_dotty(task,FILE*) writing to file the info
2658 about the task in dotty format
2659 * SD_task_dependency_exists() can now cope with having one of its
2660 arguments NULL. If so, it tests whether the other argument has any
2662 * Add getters on list of preceding/following tasks:
2663 SD_task_get_parents(task) and SD_task_get_children(task)
2664 * Add getters on amount of workstations and list:
2665 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2666 * Add getter on task kind: SD_task_get_kind(task)
2667 * Update the start_time and finish_time of tasks on completion/failure
2668 * Bugfix: Remove task from state swags when destroyed
2671 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2674 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2675 dynar with one separate thread per value of the dynar.
2676 * Change the prototype of xbt_thread_create(), sorry.
2677 Added a boolean parameter indicating whether we want to join this
2678 thread (used in SG only for now)
2679 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2682 * GTNetS wrappers should now be usable again (and betterly tested too)
2683 * Fix a major regression from 3.2 where the timeout provided to
2684 MSG_task_put_with_timeout() was used as absolute time before which
2685 the comm should be done.
2686 * Start to fix the <cluster> tag.
2687 - Internal links should be good now (beside of the loopback, which
2688 use the private link instead)
2689 - paths to the external world is still rather broken
2690 - the <route:multi> tag is just broken. Actually that's brain-dead.
2691 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2692 to make it less stupid
2693 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2694 * Fix a source-level compatibility glitch from 3.2: after defining
2695 MSG_USE_DEPRECATED, you can use the old name
2696 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2697 * Allow to compile from the SVN with automake 1.11
2698 * Fix some problems when using the "start_time" tag in deployment XMLs.
2699 * Fix #8569: XBT/synchro.h has redundant declarations
2700 * Fix #8563: MSG return values and exceptions
2701 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2702 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2703 Thanks to Robson Peixoto.
2704 * Fix a vicious bug in dictionaries inducing that some elements were
2705 not freed on xbt_dict_free()
2707 Portability report of this version:
2708 * Main portability targets:
2709 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2710 on (amd64/i386/ia64)
2711 - mac leopard on i386
2712 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2713 but nothing critical.
2714 * Other platforms: windows, AIX and others were not tested for this release
2716 Timing report of this version:
2717 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2718 scenario parameters). From 8h to a few seconds in desktop grid settings.
2719 * Supernovae brings about 25% speedup on i386.
2721 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2723 SimGrid (3.3.3) stable; urgency=low
2725 The "Need for Speed" release.
2727 The timings done to validate the 3.3.2 were faulty.
2728 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2730 The problem was a conversion from a manually handled vector to
2731 xbt_dynar_t on the critical path.
2732 xbt_dynar_foreach calls functions, inducing stack management crap.
2734 We inlined these functions and xbt_dynar_foreach is now breath taking.
2735 We also inlined xbt_swag_belong on the way.
2737 Here are some approximate speedup measurements (on master/slaves
2738 simulations lasting between 10s and 20s each):
2739 3.3.1 -> 3.3.2: about same performance
2740 3.3.2 -> 3.3.3: 40% speedup
2741 3.3.1 -> 3.3.3: 40% speedup
2742 3.3.1 with inline patch -> 3.3.3: 30% speedup
2744 Our reading is that the refactoring which occurred in 3.3.2 made us
2745 suffer much more from the xbt_dynar_foreach low performance, but
2746 once we solved this, this refactoring proved to be very performance
2747 effective. From the 40% speedup, somehow, 10% are due to the
2748 inlining and 30% to the refactoring.
2750 That's a pitty that gcc cannot inline functions placed in other files
2751 alone. We have to choose between:
2752 - break the encapsulation (by putting private data structures and
2753 accessors in headers files to help gcc)
2754 - live with low performance
2755 - switch to a decent compiler such as icc (not quite possible).
2757 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2759 SimGrid (3.3.2) stable; urgency=low
2761 The "Simplicity does not preceed complexity, but follows it" release.
2763 The main contributors of this release were (lexical order):
2764 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2767 * Extract the routing logic into its own object.
2768 (was dupplicated in network.c and workstation_LV07.c;
2769 Allows to implement other ways of storing that info)
2770 => kill now useless network_card concept
2771 - Use dynar to represent routes (instead of void** + int*)
2772 - kill link_set (use surf_network_model->resource_set instead)
2773 - Add a command-line option to choose the routing schema to use
2774 - Add three new models:
2775 * Floyd (shortest path computed at initialization)
2776 * Dijikstra (shortest path recomputed all the time)
2777 * Cached Dijikstra (shortest path computed on need)
2778 All these models where contributed by Silas De Munck, and are
2779 described in his ICCS09 paper.
2781 * Simplify model declaration
2782 (less redirections, less function to write when defining a model)
2783 - Factorize stuff between models:
2786 surf_model_resource_set(model)
2787 surf_model_resource_by_name(model, name)
2788 - Unify the types of models in s_surf_model_t (using an union)
2789 - Embeed fields of common_public directly into s_surf_model_t
2790 - Rename model methods:
2791 action_free ~> action_unref
2792 action_change_state ~> action_state_set
2793 action_get_state ~> action_state_get
2794 - Change model methods into functions:
2795 (model)->common_public->action_use ~> surf_action_ref
2797 * Implement a generic resource; use it as ancestor to specific ones
2798 (allows to kill duplicated code in models)
2799 Drawback: timer command don't need no name nor properties;
2800 workstation_CLM03 don't need no properties
2801 (but I guess we can live with those few bytes wasted)
2803 * Improve the action object model
2804 - implement a constructor avoiding dupplicated code about field
2805 initialization in generic_action part.
2807 * Kill the SDP model: it has an external dependency, is deprecated
2808 in flavor of modern lmm models, and didn't compile since a while
2811 * Relocation of the context module from XBT to SIMIX.
2812 (the context were decoupled from the simix processes, duplicating a lot of code)
2813 => a lot of code was factorized
2814 - less overhead is introduced during scheduling
2815 - simpler API for the context factory
2816 - the logic for process creation,destruction and manipulation was simplified
2817 * Simplification of the s_smx_process_t data structure.
2818 => accesing the simix level data associated to a process is faster now,
2819 and the code is a lot more readable.
2822 * Implement some more MPI primitives:
2823 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2824 -implementation: Bcast: flat or 2-ary tree (default),
2825 Barrier: 4-ary tree,
2827 Allreduce: Reduce then Bcast
2828 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2829 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2830 Alltoallv: flat tree, like ompi
2832 * Add support for optimized collectives (Bcast is now binomial by default)
2833 * Port smpirun and smpicc to OS X
2836 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2837 Shout out if you used it.
2840 * Display the list of still queued messages in SG mode when existing
2844 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2845 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2846 * Add xbt_str_from_file(FILE*)
2847 * Add xbt_dict_get_key achieving a linear reverse search
2848 * Remove the context module
2850 Portability report of this version:
2851 * Main portability targets:
2852 - Linux(debian)/x86/context
2853 - Linux(debian)/x86/pthreads
2854 - Linux(debian)/amd64/context
2855 - Linux(debian)/amd64/pthreads
2856 On these, we still have the eratic breakages of gras/pmm and
2857 amok/saturate_sg reported in previous version. We still think
2858 that the tests are the cause of the fault, not the tested code.
2860 - Mac OSX Leopard/x86/context
2861 Still false negative in tesh autotesting.
2862 Smpi still fails, but this time because readlink does not accept -f
2863 Everything seems to work properly beside of that.
2866 - AIX version 5.3 (only tested contexts this time)
2867 Smpi still fails there because mktemp is not installed.
2868 Everything seems to work properly beside of that.
2870 I managed to compile it for the first time, but several breakages.
2871 Won't delay the release for this exotic platform.
2873 * Windows: it's still lagging behind. If you want to help, please
2876 Timing report of this version:
2877 This version seem to be more than 5% faster than 3.3.1 (on linux
2878 64bits with contextes). The gain is less than expected, we are
2879 investigating this for next release.
2881 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2883 SimGrid (3.3.1) stable; urgency=low
2886 * Implement a --cfg-help to show existing configuration variables
2887 * Build chain do not require doxygen in maintainer mode
2890 * fix a bug on struct sizeof computation, which prevented the
2891 exchange of arrays of structs in some conditions
2892 - added a regression test about this in datadesc_usage
2893 * Allow the exchange of 0-long dynamic vectors.
2894 - for that, use -1 as indicator of dynamic size instead of 0
2895 - This implied to change any size from unsigned long to long,
2896 reducing a bit communication abilities, but I guess that with
2897 64bits being quite common, this is more than enough.
2898 - This also induce a protocol change, thus bumping network protocol
2899 version from 0 to 1 (if we have external users, we have to get
2900 clean on that point too ;)
2901 - added two regression tests about this in datadesc_usage
2902 * Be more verbose when propagating local exceptions
2903 This helps debugging.
2904 * Display the status of simulated processes when receiving SIGINT in
2908 * Allow to control the simulation from a trace file.
2909 New functions MSG_action_register() and MSG_action_trace_run()
2910 The first one allows to associate a function execution to each
2911 kind of action while the second one parses a trace file and
2912 triggers the corresponding actions within the system.
2913 For now, only a toy example is provided in examples/msg/actions
2914 * Add an example of process migration in examples/msg/migration
2915 * Fix a bug in task exchange which broke MSG_task_get_sender()
2916 Add a teshsuite regression test for that.
2917 [Bug: if MSG_task_get_sender() is called after sender exit,
2919 * Fix a bug which prevented suspend/resume to work properly
2920 * Display the status of simulated processes when receiving SIGINT
2921 This fixes a regression of v3.3. due to the introduction of SIMIX
2922 * Bug fixing in failure management:
2923 - trace could not start by a failure at time 0
2924 - failure during communications were not working
2927 * Add SIMIX_process_set_name() to change the name of the current
2928 process in the log messages.
2929 * Store smx_hosts in a dict since we only retrieve them by name
2930 * Move the configuration infrastructure to surf
2933 * Move the configuration infrastructure to surf
2936 * Massive internal cleanups:
2937 - Store internal structures on processes instead of hosts (allows
2938 to have more than one process per host, in addition of being more
2940 - Cleanup the initialization/finalization process
2941 - Kill a whole bunch of unneeded synchronization:
2942 processes run in exclusive manner within the simulator
2943 - Move queues from global tables to process data fields
2945 - now accept -platform and -hostfile arguments
2946 - Pass the right rank value to processes according to the hostfile
2947 * Compile the examples by default, and use them as regression tests
2948 * Implement MPI_Wtime()
2949 * Change the reference speed to a command line option
2952 * TCP_gamma can now be specified as command line option using
2953 --cfg=TCP_gamma:10000000.0
2954 * Change the --surf-path cmd line option into --cfg=path:
2957 * Also include strbuff from xbt.h public header
2958 * xbt_ex_display(): do not free the exception after displaying
2959 This allows to do more with the given exception afterward.
2960 Users should call xbt_ex_free() themselves.
2964 Portability report of this version:
2965 * Main portability targets:
2966 - Linux(debian)/x86/context
2967 - Linux(debian)/x86/pthreads
2968 - Linux(debian)/amd64/context
2969 - Linux(debian)/amd64/pthreads
2970 These targets fail about 1/10 of times on gras/pmm, but we believe
2971 that this is because of the test, not because of simgrid.
2972 amok/saturate_sg fails even more rarely, and the test may not be
2975 - Mac OSX Leopard/x86/context
2976 The test suite still spits tons of errors because some obscure
2977 force prevents us from removing the temporary directories
2978 arguing that they still contain some metadata I've never heard of.
2979 Smpi fails because seq is not installed.
2980 Everything seems to work properly beside of that.
2983 - AIX version 5.3 (both contexts and pthread)
2984 Smpi still fails there because mktemp is not installed.
2985 XML inclusions seems rosty on AIX.
2987 * Windows: it's still lagging behind. If you want to help, please
2990 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2992 SimGrid (3.3) stable; urgency=high
2996 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2997 [Malek Cherier & Mt]
2999 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3002 Until now, GRAS were using MSG as an interface to SURF. It was
3003 quite difficult because both interface have several differences
3004 (MSG channels vs GRAS sockets were the most notable point).
3006 This also opens the gate to SMPI (which should occur soon) and speed
3007 up simulations by to 40% (even if it were not the main goal).
3009 **************************************
3010 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3011 **************************************
3012 The point is that events occurring at the exact same timestamp are
3013 not scheduled in the same order with the old and new version. This
3014 may be enough to completely change the execution of simulations in
3015 some cases. Sorry for the inconvenience.
3017 * Cleanup and upgrade the XML format to push further scalability
3018 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3020 * Improve the testing infrastructure with tesh. Now a very large part of
3021 the code is tested not only by being run but also by checking that the
3022 output match an expected output [Mt].
3024 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3025 is really less memory-demanding, which should allow you to use
3026 larger files in SimGrid [AL].
3028 * Inform valgrind about our contextes, so that it becomes usable
3029 with the default (and more effecient) version of SimGrid
3030 [contributed by Sékou Diakite, many thanks]
3033 * Introduce a listener thread in charge of receiving incoming
3034 messages from the network. It allows to overlap communication and
3035 computation but most notably, it removes some stupid deadlocks due
3036 to the fact that so far, a process could not send and receive at
3037 the same time. This made most non trivial communication schema
3039 * Convert the PIDs from long int to int to match the MSG ones (and
3040 linux ones too) [Mt]
3041 * New function: gras_agent_spawn() to launch a new process on
3042 current host. Only working in simulation for now. [Mt]
3043 * New function: gras_os_hostport() returning a constant form (ie,
3044 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3047 * Make the backtrace of exceptions more human readable [Mt]
3048 * New module: xbt/str [Mt]
3049 a ton of string utility functions (split, join, printf to a newly
3050 allocated buffer, trim, etc)
3051 * New module: xbt/hash [Mt]
3052 SHA1 hashing algorithm (more to come if needed)
3053 * New module: xbt/synchro [Mt]
3054 synchronization tools (mutex and conditions) working the same way
3055 in simulation and in real life (mainly useful for GRAS, but not
3057 * New module: xbt/queue [Mt]
3058 classical producer/consumer synchronization scheme
3059 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3060 (using the classical functions will get serialized) [Mt]
3061 * Make dictionary internal table dynamic. No need to specify its size
3062 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3064 * Make sure the log channels are organized as a tree under windows
3065 (because of ANSI C compatibility issue, any channel were child of
3066 root directly) [Mt].
3069 * Cleaned many thing in surf and fixed a few bugs [AL].
3070 * Add a nice command line configuration mechanism to compose models [AL].
3071 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3072 the previous one (KCCFLN05). It relies on something that looks like
3073 a max-min sharing mechanism but cannot be written as such. A new solver
3074 was thus designed [AL].
3075 * Add a new solver to lmm. Based on Lagrange optimization and
3076 gradient-based descent, it enables to efficiently maximise systems s.a
3078 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3080 This solver enables to propose two new network models for TCP Reno and
3081 TCP Vegas based on Low's work. These models still need to be fully
3082 tested though [Pedro Velho].
3085 * Bug fix in SD_simulate. Now the time bound given as argument is
3087 * Use the new parallel task model (ptask_L07) as default.
3088 * Use the SURF command line configuration mechanism.
3089 * 0-size tasks (for synchronization) should now work.
3091 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3093 SimGrid (3.2) stable; urgency=high
3097 We still experience issues on this platform, but we believe that at
3098 least MSG is usable.
3100 GRAS API BREAKAGE (for simplification purpose, sorry):
3101 * the gras_msgtype_by_name is not used anymore. Instead of
3102 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3103 you can write (and must)
3104 gras_msg_send(toserver, "request", &request);
3105 - If you still want to pass a gras_msgtype_t to the function (to cache
3106 the type and avoid the lookup time), use the gras_msg_send_() variant.
3107 - Impacted functions:
3108 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3109 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3110 * The callbacks are now expected to return 0 when everything went well
3111 (just like the main() function)
3113 GRAS new features and improvements:
3114 * New module mechanism where user code can use per process globals [Mt]
3115 This is similar to gras_userdata_*() functions, but for libraries. It
3116 factorize some code developped over and over in the examples and AMOK.
3117 It has still to be documented and used (only amok/peermanagement is
3119 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3120 loss when they were small enough to fit into the buffer and sent quickly
3121 enough so that they can all get received in one shoot.
3122 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3123 if not found. Use the *_or_null() variant for the old semantic.
3124 * In gras_msg_handle, do not discard messages without callback.
3125 They are probably messages to be explicitly awaited later (ie, proofs of
3126 mis-synchronization in userland since they are sent before being awaited)
3127 No big deal usually.
3128 * gras_socket_meas_send/recv: semantic changed!
3129 The numerical arguments used to be (1) the total amount of data to send
3130 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3131 messages. This was need for the fool willing to send more than MAXINT
3132 bytes on quite fat pipes.
3135 * Do really rename the hostmanagement module to peermanagement. [Mt]
3136 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3137 but this is rather new and this was documented in the module
3138 documentation (poor excuses, I admit)
3139 * Bandwidth measurement semantic changed! This follows the changes to
3140 gras_socket_meas_send/recv explained above.
3143 * A sequential mode has been added to the workstations. When a workstation
3144 is in sequential mode, it can execute only one task, and the other tasks
3145 are waiting in a FIFO. [Christophe Thiery]
3148 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3149 model for SIMDAG. [Christophe Thiery]
3150 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3151 equal to 0, which caused some bad behaviors in
3152 saturated_constraint_set_update. I now use a threshold mechanism like in
3156 * When running manually src/testall, you select specific units [Mt]
3157 testall is the result of our cunit mechanism, and should replace all
3158 the scripty thingy around since bash don't run easily on billware.
3160 * A mallocator system has been added. [Christophe Thiery]
3161 Mallocators allow you to recycle your unused objects instead of freeing them
3162 and allocating new ones.
3164 Documentation update:
3165 * FAQ reworking + New FAQs:
3166 - "Valgrind spits tons of errors!" [Mt]
3167 - "How to repport bugs" [Mt]
3168 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3169 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3170 - Communication time measurement within MSG [AL]
3171 - I experience weird communication times when I change the latency [AL]
3172 * GRAS tutorial [Mt]
3174 - an introduction to the framework and to the used communication model
3175 - an initiatic tour introducing the most proheminent features:
3177 . Lesson 0: Installing GRAS
3178 . Lesson 1: Setting up your own project
3179 o Part 2: Message passing
3180 . Lesson 2: Exchanging simple messages
3181 . Lesson 3: Passing arguments to the processes (in SG)
3182 . Lesson 4: Attaching callbacks to messages
3183 . Lesson 5: Using globals in processes
3184 . Lesson 6: Logging information properly
3185 . Lesson 7: Using internal timers
3186 . Lesson 8: Handling errors through exceptions
3187 . Lesson 9: Exchanging simple data
3188 . Lesson 10: Remote Procedure Calling (RPC)
3189 . Lesson 11: Explicitly waiting for messages
3190 . Recapping of message passing features in GRAS
3191 - A HOWTO section containing:
3192 o HOWTO design a GRAS application
3193 More are due, of course. They will come latter. In the meanwhile, you can
3194 check the examples which are still here.
3196 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3198 SimGrid (3.1) stable; urgency=high
3202 There was a stack corruption somewhere, visible only when optimizing
3203 with these versions. [Vince]
3206 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3207 module that enabled to play with Directed Acyclic Graphs. It is built
3208 directly on top of SURF and provides an API rather close to the old
3209 SG. Some old codes using SG are currently under rewrite to check that
3210 all needful functions are provided. [Christophe Thiery]
3213 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3214 extension of the classical CLM03 model that gracefully handles
3215 failures. This is now the default model for MSG and GRAS. It doesn't
3216 handle parallel tasks yet though. [AL]
3217 * Bug fix: Weights were not correctly set in the network part.
3218 WARNING: This may have resulted in incorrect results with simulations
3219 where there are more than one flow on a given link. [AL]
3222 * After a (long ?) discussion on simgrid-devel, we have decided that the
3223 convention we had on units was stupid. That is why it has been decided
3224 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3225 WARNING: This means that all previous platform files will not work as
3226 such with this version! A warning is issued to ask users to update
3228 A conversion script can be found in the contrib module of the CVS, under
3229 the name contrib/platform_generation/surfxml_update.pl [MQ]
3232 * Bug fix: Processes were started in reverse order, wrt deployment file.
3233 WARNING: if your code relies on this bug, please fix it. [AL]
3234 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3235 being executed on two different locations. [AL]
3236 * Bug fix: Failures are now better supported thanks to Derrick's tests
3237 (there was many failure situations I hadn't thought of and that weren't
3238 correctly handled). [AL]
3239 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3243 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3244 Exception occurring on server-side are propagated back to client (!).
3246 API CHANGE: the callback changed their prototype. Change:
3247 int my_handler(gras_socket_t expeditor, void *payload_data) {
3249 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3250 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3252 * New! function: gras_msg_handleall to deal with all messages arriving
3253 within a given period.
3254 * New! function: gras_socket_server_range to get a server socket in a
3255 range of port numbers (ease to avoid port number conflicts) [MQ]
3256 * New! gras processes display their backtrace when they get a SIGUSR1
3257 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3258 Sweet to debug RL processes [MQ]
3262 - Do not force experiment sizes to be expressed in kb, or it becomes
3263 impossible to measure the latency this way (needs one byte-long tests)
3264 WARNING: this changes the amok_bw_* function semantic. [MQ]
3265 - Implements the link saturation stuff. [MQ]
3266 * Peer management module:
3267 New! module factorizing code that we wrote over and over [MQ].
3270 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3271 - Test units are placed directly into the library code, they get extracted
3272 automatically and placed into the src/testall binary.
3273 - Convert most of the XBT tests to this system.
3274 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3275 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3276 live in the same C code. This required to change a little bit the API
3277 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3278 * New module: sparse graph structure with basic algorithms (this is work
3279 in progress and the API is not considered to be frozen yet). [AL]
3280 * Display more information on backtraces: source line & function names are
3281 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3282 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3283 * Reworked a little bit some #include statements to load only required
3284 headers. Some user code that relied on SimGrid to include stdlib or
3285 stdio may need to include it by themselves. [AL]
3286 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3288 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3289 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3290 but I guess that almost nobody used those parts. [MQ]
3292 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3294 SimGrid (3.0.1) stable; urgency=low
3297 * Unfortunately, I had missed 5 misnamed functions:
3298 xbt_fifo_item_t xbt_fifo_newitem(void);
3299 void xbt_fifo_freeitem(xbt_fifo_item_t);
3300 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3301 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3302 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3303 They're now deprecated. Please use their new versions:
3304 xbt_fifo_item_t xbt_fifo_new_item(void);
3305 void xbt_fifo_free_item(xbt_fifo_item_t);
3306 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3307 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3308 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3310 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3311 * Documentation: xbt_log module unmercifully reworked [MQ]
3312 * Bugfix: there was a problem with the ending of contexts with
3313 the pthread backend. It caused some weird deadlock or behavior
3314 depending on the pthread implementation. [AL]
3315 * Bugfix: get the exceptions raised in the simulator repport where
3316 and why they come from when they are not caught in time [AL, MQ]
3319 * Bugfix: Do repport the error when two non-connected hosts try to
3320 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3323 * Add additionnal checkings on communications. Assert that two
3324 communicating hosts are connected by a set of links... [AL]
3327 * Add additionnal checkings on channel values in communication [AL]
3328 * New: MSG_task_get_source to see on which host a task was generated [HC]
3329 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3330 the number of tasks waiting to be received on channel and sent
3332 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3333 waits for the first task coming from a given host.. [AL]
3335 GRAS new functionnalities: [MQ]
3336 * Enhance the parsing macro to allow the size of multidimentional objects
3337 to be given thru annotations.
3338 * New example (and documentation): Matrix Multiplication a la RPC
3339 (as when I was young!) and fix a bunch of bugs found on the way.
3341 GRAS performance improvements: [MQ]
3343 * Reduce the amount of cbps creation/destruction by making it static to
3344 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3347 * Change libdata to a set so that we can search for stuff by ID (and thus
3348 reduce the insane amount of dict lookups)
3351 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3352 RL communication on top of SG since it's so uneffective.
3353 It may also be used for inter-thread communication in RL, one day.
3354 * Use gras_datadesc_copy() to exchange messages on top of SG
3356 - improve message exchange performance on top of SG
3357 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3358 * Don't exchange on the network the size of the used part of buffer,
3359 instead, specify the possible buffer size to read().
3361 - reduces the amount of read/write calls (one pair per exchange)
3362 - reduces the amount of exchanged data (the size)
3363 - allows to retrieve all arrived data on receiver side, if we don't need
3364 it right now (subsequent read will peek the buffer)
3365 - allows the receiver to proceed with the beginning of the stream before
3366 everything is arrived
3367 - make it possible to build an iov transport (using readv/writev)
3369 - take care of the data with non-stable storage (like stacked data),
3371 * If possible, TCP send uses vector I/O (when writev() is here)
3372 - Don't use it for receive since we send data sizes and data on the
3373 same stream, so we wouldn't be able to chain large amount of chunks
3374 before having to flush the stuff to read the size.
3375 * Rework the transport plugin mechanism to simplify it and reduce the
3376 amount of pointer dereferencement when searching for the right function
3379 * I guess that now, we do almost as few system calls as possible while
3380 doing as few data copy as possible.
3382 To improve it further, we could try to send all the sizes first and then
3383 all the data (to use iov on receiving size), but it's only a partial
3384 solution: when you have 2 dimensional data, the sizes of the second
3385 dimension is data of the first dimension, so you need 3 streams.
3387 I'm not sure the potential performance gains justify the coding burden.
3389 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3391 SimGrid (3.00) stable; urgency=high
3394 * New! Give the possibility to hijack the surf parser and thus bypass
3395 MSG_create_environment and MSG_launch_application. Have a look at
3396 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3398 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3400 SimGrid (2.96) unstable; urgency=low
3405 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3406 This deprecates the xbt_error_t mechanisms.
3407 It modifies (simplifies) all XBT and GRAS API.
3408 MSG API keeps unchanged (exceptions raised by XBT are caught from
3409 within MSG and masked with existing error handling facilities)
3412 * New! Add a FATPIPE model. [AL]
3413 * New! Add a parallel task model. [AL]
3414 * New! Add automatically a loopback interface (in the default
3415 network model) if none was precised.
3418 * Bugfix: MSG_process_resume now works with the current running process.
3420 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3421 * Modification of MSG_task_get_compute_duration. Once a task has been
3422 processed, the value returned by this function is now equal to 0. [AL]
3423 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3424 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3425 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3426 of a m_task. For now, MSG_task_cancel only works with computation
3428 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3429 of the processor (in Mflop/s) regardless of the current load on the
3431 * API Change: use proper naming convention for MSG_getClock and
3432 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3434 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3435 This function changes the priority of a computation task. This priority
3436 doesn't affect the transfer rate. A priority of 2 will make a task
3437 receive two times more cpu power than the other ones. This function
3438 has been added to suit the needs of Nguyen The Loc and hasn't been that
3439 much tested yet. So if it fails, please report it and send me your code.
3441 * API Change: removed all functions and types that were marked "deprecated"
3442 since many months. Renamed MSG_global_init_args to MSG_global_init.
3444 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3446 SimGrid (2.95) unstable; urgency=low
3449 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3450 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3451 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3454 * Bugfix: Allow absolute paths for platform description files [MQ]
3455 * Bugfix: do free the variables after use. Leads to drastic performance
3457 * Implement max_duration (ie, timeouts) on resources [AL]
3460 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3462 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3463 the needed support (timeouts) [AL]
3466 * Implement measurement sockets. You can now get the bandwidth between two
3467 hosts thanks to AMOK (see below). [MQ]
3468 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3469 dynar over the network (yeah) [MQ]
3470 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3471 * simple/static token-ring example. [Alexandre Colucci and MQ]
3472 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3473 in place before [MQ]
3476 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3477 type parsing. "short" and "long" modifiers were also ignored; other
3478 modifier (such as reference level) are still ignored. [MQ]
3479 * Embeed the buffer size within the buffer itself on SG. [MQ]
3480 That way, send() are atomic and cannot get intermixed anymore (at least
3481 the ones which are less than 100k; bigger messages still have the issue)
3482 * Array size pushed by the field, not by the field type (or each
3483 and every long int will push stuff to the cbps) [MQ]
3484 * use select() to sleep since it allows to portably sleep less than one
3487 GRAS (minor cleanups)
3488 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3491 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3492 * type category 'ignored' killed as it was never used and were difficult
3494 * whether a type can cycle or not is now a flag, leaving room for more
3495 flags (as "ignored", if we feel the need one day ;)
3496 * Rename raw sockets to measurement sockets since "raw" has another
3497 meaning in networking community.
3500 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3501 and offers features not belonging to GRAS but that most applications
3502 need. One day, it may be a set of plugins loadable at runtime.
3503 * New module: bandwidth
3504 bandwidth measurement between arbitrary nodes running this module. [MQ]
3506 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3508 SimGrid (2.94) unstable; urgency=low
3510 The first beta release of SimGrid 3 !
3514 * Update the main page and the FAQ. Adding references to gforge.
3517 * Add a gras_os_getpid function.
3520 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3521 * Extend the logs so that they also print PID, hostname, date, ... if
3523 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3524 and kill the old version which were in testsuite/
3525 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3529 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3530 for users. Instead, each programming environment (GRAS, MSG,...) use it
3531 when needed to provide such a feature to users.
3532 Don't shortcut the mechanism or you will also shortcut the virtualization
3533 you need on the simulator.
3537 * Cleanups in configury with regard to compile optimization/warning flags.
3538 Also add -fno-loop-optimize to any powerpc since it's the optimization
3539 killing gcc (< 3.4.0).
3540 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3541 needed by MSG examples complications
3542 * Borrow configury beautifications from PHP
3545 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3546 warning (thanks loris for stumbling into this one).
3547 * Bugfix: stop loading private headers (gras_config.h) from the public
3551 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3552 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3553 RL would imply a DNS resolver, which is *hard* to do in a portable way
3554 (and therefore delayed).
3555 * Implement a real timer mechanism and use it in timing macros. This allows
3556 to avoid rounding errors and get a 0.000005 sec precision in timing
3557 macros. While I was at it, various cleanups:
3558 - allow to declare more than one timed section per file (fix a stupid bug)
3559 - move some private declaration to the right place
3560 - merge conditional execution and timing macros into emulation module
3561 - document the module
3562 - make sure the module cleanups its mess on gras_exit
3563 * Documentation improvements:
3564 - (new) how to compile applications using GRAS
3565 - (new) emulation support (timing macros)
3567 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3569 SimGrid (2.93) unstable; urgency=low
3571 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3574 - Use Paje properly where used. Still to be sanitized properly.
3575 - Portability fix: Add an implementation of the contexts using pthread
3579 - Add xbt_procname(): returns the name of the current process.
3580 Use it to show the current process's name in all logging.
3582 - fix detection of older flex version and the reaction, since we do
3583 depend on modern ones (we use lex_destroy)
3584 - Better separation of SG and RL in the libs: remove all simulation code
3585 from libgras. As a result, this lib is now only 200k when stripped.
3586 Some of the xbt modules may also be duplicated (two sets and such) and
3587 should be cleaned/killed before SG3.
3588 - Insist on using xlC on AIX because of weird problems involving gcc there.
3589 - Cleanup the make remote stuff. This is now done by scripts
3590 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3591 mainly for our private use, but we're working on changing them to user
3594 - Bugfix: flush the socket on close only if there is some *output*.
3595 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3597 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3600 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3602 SimGrid (2.92) unstable; urgency=low
3604 Alpha 3 on the path to SimGrid 3
3608 - New! First try of benchmarking macros.
3609 - New! First try so that gras_stub_generator generate deployment and
3610 remote compilation helpers.
3612 - Bugfix: Initialization fix in msg_test.
3616 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3618 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3620 - API Change: gras_os_sleep to take the amount of seconds as a double.
3621 Accepting an int was error prone since it was the only location where
3622 seconds were coded as such. It leaded to damn rounding errors.
3623 - Bugfix: Hard to belive that timers ever worked before this.
3625 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3627 SimGrid (2.91) unstable; urgency=low
3629 Alpha 2 on the path to SimGrid 3
3633 - Bug fix in the lmm_solver.
3635 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3636 through the function MSG_paje_output.
3637 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3638 - It is possible to bound the rate of a communication in MSG with
3639 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3640 to write it in the changelog).
3641 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3646 - Various cleanups to the autotools stuff
3647 - Begin to move Gras examples to examples/gras/
3648 - Let make distcheck work again (yeah!)
3650 - documentation overhauled using doxygen.
3651 gtk-doc-tools is dead in SimGrid now.
3652 - Automatically extract all existing logging categories, and add the list
3653 to the documentation (long standing one, to say the less)
3655 - Cleanup the known architecture table. Reorder the entries to group what
3656 should be, and use a more consistent naming scheme.
3657 (some of the test dataset are still to be regenerated)
3658 - New! Allow library to register globals on each process just as userdata
3660 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3661 lookup time (for now).
3662 Use it in msg and trp.
3663 This cleans a lot the internals and helps enforcing privacy of the
3664 headers between the gras components.
3665 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3666 - Bugfix: gras_os_time was delirious in RL.
3667 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3669 - Reenable GRAS now that it works.
3671 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3673 SimGrid (2.90) unstable; urgency=low
3675 Alpha 1 on the path to SimGrid 3
3677 * It is a long time since the last release of SimGrid. I'm sorry about
3678 that but as I had told you, I was rewriting a lot of things. I apologize
3679 to those who had been reporting bugs to me and that I had not answered.
3680 If your bug is still in the new version, please tell me. Here is a
3681 summary of the main changes.
3683 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3684 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3685 portability, speed, and a lot more but you'll figure it out later.
3686 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3687 I wanted to release the new MSG as soon as possible and I have
3688 broken GRAS, which is the reason why, for now, only MSG is fully
3689 functional. A laconic description of these projects is available
3690 in the documentation.
3692 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3693 kernel optimized for our needs (called SURF but only the developers
3694 should use it). Hence, MSG is now roughly 30 times faster and I think
3695 that by rewriting a little bit MSG, I could event speed it up a little
3696 bit more. Beside the gain in speed, it is also much easier to encode a
3697 new platform model with SURF than it was with SG. More to come...
3699 * REVOLUTION 3: I have tried to change a little as possible the API of
3700 MSG but a few things really had to disappear. The main differences
3701 with the previous version are:
3702 1) no more m_links_t and the corresponding functions. Platforms are
3703 directly read from a XML description and cannot be hard-coded
3704 anymore. The same format is used for application deployment
3705 description. The new format is described in the documentation.
3706 Have a look in tools/platform_generation. There is a tiny script
3707 that converts from the old platform format to the new one. Concerning
3708 the application deployment format, parsing the old one is tricky.
3709 I think most of you should however be able to convert your files. If
3710 it is really an issue, I can write a C code that does the conversion.
3712 2) the toolbox tbx does not exist anymore. We now have a library
3713 with much more data-structures but without the hash-tables (we have
3714 dictionaries that are much faster).
3716 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3718 *****************************************************************************
3719 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3720 * don't feel like converting the oldies. (Mt) *
3721 *****************************************************************************
3724 Version 2.90: "the long awaited one"
3725 - Finished rewriting and debugging MSG. Rewrote the documentation.
3726 - disable GRAS for now since it needs to be ported to the newest SG
3729 - Finish the port to windows (using mingw32 for cross-compile)
3732 - Main loop and datastructures of SURF. A cpu resource object is
3733 functional. Surf can thus be used to create cpu's with variable
3734 performance on which you can execute some actions.
3736 2004-11-15 Martin Quinson
3737 - Port to ARM. Simply added the alignment and size descriptions. Should
3738 work, but the ARM machines are so slow that I didn't had the opportunity
3739 to 'make check' over there yet.
3741 2004-11-15 Arnaud Legrand
3742 - Trace manager now written. It uses a heap structure and is therefore
3743 expected to be efficient. It may however be speeded up (particularly
3744 when many events occur at the same date) by using red and black
3745 trees. One day maybe...
3746 - Max-min linear system solver written. It uses a sparse matrix
3747 structure taking advantage of its expected use. Most operations are
3748 O(1) and free/calloc are called as few as possible. The computation of
3749 the minimum could however be improved by using a red and black tree
3752 2004-11-03 Arnaud Legrand
3753 - Rename every gras_* function that was in xbt/ to its xbt_
3755 - Add a heap and a doubly-linked list to xbt
3756 - Added a dichotomy to the dictionaries. make check works as well before
3757 so I assume that the patch is correct. I do not know however if things
3758 run effectively faster than before now. :)
3760 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3761 SimGrid, and the version number is bumped to 2.x
3763 2004-10-29 Martin Quinson
3764 - Introduction of the remote errors.
3765 They are the result of a RMI/RPC on the remote machine.
3766 ErrCodes being scalar values, you can't get the host on which those
3767 errors did happen. Extending the error mechanism as in Gnome is possible.
3768 No idea yet whether it is a good idea.
3770 2004-10-28 Martin Quinson
3771 - Interface revolution: the Starred Structure Eradication.
3772 I used to do typedef struct {} toto_t; and then handle *toto_t.
3773 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3774 - ???_t is a valid type (builded with typedef)
3775 - s_toto_t is a structure (access to fields with .)
3776 - s_toto is a structure needing 'struct' keyword to be used
3777 - e_toto_t is an enum
3778 - toto_t is an 'object' (struct*)
3780 typedef struct s_toto {} s_toto_t, *toto_t;
3781 typedef enum {} e_toto_t;
3782 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3783 s_toto_t) is private.
3785 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3786 so that it can set the variable to NULL. It was so for dicts and sets,
3787 it changed for dynars.
3789 - Fix a bunch of memleaks in dict_remove
3790 - Fix a bug in sg/server_socket opening: it failed all the time.
3792 2004-10-07 Martin Quinson
3793 - Speed up dynar lookup operation a bit.
3795 gras_dynar_get is dead.
3797 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3798 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3799 (which gives you the address of the stored data).
3801 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3802 copy of the data using an affectation to do the job and not a memcpy.
3804 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3805 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3807 It does not really speedup the dynar test because they are
3808 setting elements all the time (and look them seldom). But the dict does
3809 far more lookup than setting.
3811 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3813 2004-10-05 Martin Quinson
3814 - Allow to (en/dis)able the cycle detection at run time.
3816 Whether we should check for cycle or not is now a property of each
3817 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3818 datadesc_cycle_unset allow to remove this property when previously set.
3820 Note that the cycle detection is off by default since it impacts the
3821 performance. Watch the data you feed GRAS with ;)
3823 This property is hereditary. Any element embedded in a structure having it
3824 set have it set for the time of this data exchange.
3826 You should set it both on sender and receiver side. If you don't set it on
3827 sender side, it will enter an endless loop. If you forget on receiver
3828 side, the cycles won't be recreated after communication.
3830 - Header reorganization.
3831 Kill gras_private.h, each submodule must load the headers it needs.
3833 2004-10-04 Martin Quinson
3834 - Interface revolution: do not try to survive to malloc failure.
3836 Now, gras_malloc and friends call gras_abort() on failure.
3837 As a conclusion, malloc_error is not a valid error anymore, and all
3838 functions for which it was the only gras_error_t return value are
3839 changed. They now return void, or there result directly.
3840 This simplify the API a lot.
3842 2004-09-29 Martin Quinson
3843 - Re-enable raw sockets.
3844 Created by gras_socket_{client,server}_ext;
3845 Used with gras_raw_{send,recv}
3848 It should allow to kill the last bits of gras first version soon.
3850 This is not completely satisfactory yet (duplicate code with
3851 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3854 - Simplify transport plugin (internal) interface by not passing any
3855 argument to _server and _client, but embedding them in the socket
3858 2004-09-28 Martin Quinson
3859 - Finish the port to AIX.
3860 autoconf was my problem (segfault within the malloc replacement
3861 function. No idea why)
3863 2004-09-16 Martin Quinson
3864 - Fix some size_t madness on 64bit architectures.
3866 2004-09-08 Martin Quinson
3867 - Reduce the number of system headers loaded, overload some more system
3868 calls (such as malloc to cast the result of the system one, and work
3870 - Fix and reintroduce the config support
3872 2004-09-07 Martin Quinson
3873 - Source code reorganization to allow Arnaud to surf all over there.
3874 - Allow to document the logging categories.
3875 - Remove all uppercase from logging categories and useless cleanup in names.
3877 2004-08-18 Martin Quinson
3878 Version 0.6.2 (protocol not changed; API changed)
3879 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3880 gras_error_t), and NULL when not found. Functions expecting a msgtype
3881 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3882 hopefully usefull message.
3883 - Portability to prehistoric sparcs again
3885 2004-08-17 Martin Quinson
3886 Version 0.6.1 (protocol not changed; ABI not changed)
3887 - prealloc some buffers to speed things up
3889 2004-08-11 Martin Quinson
3890 Version 0.6 (protocol not changed; ABI expended)
3891 - The parsing macro can deal with the references, provided that you add
3892 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3894 2004-08-09 Martin Quinson
3895 Version 0.5 (protocol not changed; ABI changed)
3896 - Allow to off turn the cycle detection code in data exchange at
3897 compilation time. It should be at run time, but I'm short of time (and
3898 the config stuff is still broken). That way, we keep dict out of the
3899 critical path, which is good because the performance is poor:
3900 - search not dichotomial yet
3901 - dynar give no way to access their content and memcpy everytime
3902 - In composed data description (struct, ref and so on), stop foolness of
3903 keeping the subtype's ID, but store the type itself. This keeps sets out
3904 of the critical path, which is good since they rely on dynar and
3905 dictionaries. The only loose of that is that we cannot detect the
3906 redeclaration of a structure/union with another content (but I'm not sure
3907 the code detected well this error before anyway). We still can detect
3908 the redefinition discrepancy for the other types.
3909 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3910 it breaks the code because of type-punning used all over the place).
3911 This breaks on all non-gcc architectures (for now).
3913 All those changes (plus the buffer of last time) allow me to gain 2 order
3914 of magnitude on cruel tests consisting of 800000 array of integers on two
3915 level of a hierarchical structure (200 secondes -> 4 secondes)
3918 - the selector of reference must now return the type it points to, not
3919 the ID of this type.
3921 2004-08-06 Martin Quinson
3922 Version 0.4 (protocol changed; ABI not changed)
3923 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3924 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3925 - Implement a buffer transport, and use it by default (it relies on tcp in
3926 real life and on sg in simulation).
3927 That's a bit hackish since I had a new field to the structure to store
3928 its data without interfering with the subtype ones. Inheritance
3929 is tricky in C. And that's a kind of reverse inheritance with one class
3930 derivating two classes. Or maybe a game with java interfaces. Anyway,
3931 that's damn hard in C (at least).
3932 Moreover, I got tired while trying to ensure plugin separation and
3933 genericity in SG mode. MSG wants me to do weird things, so let's go for
3934 cruel hacks (temporarily of course ;).
3935 See comment in transport_private.h:71
3936 - do not include all the _interface headers in private but in the files
3937 which really need them (to cut the compilation time when they are
3940 2004-07-26 Martin Quinson
3941 Version 0.3 (protocol not changed; ABI changed)
3942 - Major overhault of the datadesc interface to simplify it:
3943 - shorted the function names:
3944 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3945 - add a trivial way to push/pop integers into the cbps without malloc.
3946 This allows to make really generic sub_type description, which simply
3947 pop their size of the stack.
3948 - add a function gras_datadesc_ref_pop_arr() which does what users want
3949 most of the time: Declare a dynamic array (which pops its size of the
3950 stack) and declare a reference to it. Poor name, but anyway.
3951 - kill the post-send callback, add a post-receive one
3953 2004-07-23 Martin Quinson
3954 Version 0.2 (protocol changed; ABI changed)
3955 - add some testing for cpbs in the test cases, and fix some more bugs.
3956 This invalidate again the little64 data file, since I cannot regenerate
3958 - remove an awfull optimization in the logging stuff, allowing me to:
3959 - understand it again
3960 - learn gcc how to check that the argument match the provided format
3961 - fix all errors revealed by gcc after that
3962 - internal keys of dict are not \0 terminated. Deal with it properly in
3963 loggings instead of segfaulting when the user want to see the logs :-/
3965 2004-07-22 Martin Quinson
3966 - Fix some stupid bug preventing cbps (callback postit) from working
3968 2004-07-21 Martin Quinson
3969 - Some documentation cleanups
3970 - remove the useless last argument of msgtype_declare
3971 - rename the Virtu functions to fit into the 'os' namespace
3972 - move headers src/include -> src/include/gras/ and stop fooling with
3973 gras -> . symbolic link
3974 - make distcheck is now successful
3976 2004-07-19 Martin Quinson
3978 - Build shared library also
3979 - Install html doc to the right location
3980 - stop removing maintainer files in make clean
3981 - build tests only on make check
3983 2004-07-13 Martin Quinson
3985 - No major issue in previous version => change versioning schema
3986 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3987 corresponding dataset.
3989 2004-07-11 Martin Quinson
3991 - Get it working with any kind of structure (we can compute the padding
3992 bytes remotely for all the architectures I have access to)
3993 - Implement the structure parsing macro (still not quite robust/complete)
3994 - Improvement to the remote testing toysuite
3996 2004-07-10 Martin Quinson
3997 [autoconf mechanism]
3998 - get ride of a bunch of deprecated macros
3999 - actually run the test for two-compliment, not only compile it :-/
4000 - test whether the structures get packed (and bail out if yes. Damn.
4001 Alignment is a serious matter)
4002 - test whether the structures get compacted (but respecting the alignment
4003 constraints of each types)
4004 - test whether the array fields of structures can straddle alignment boundaries
4006 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4007 (took me 2 hours to find that bug, looking at the wrong place)
4008 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4009 before sending/receiving and are used to compute the offsets of fields
4010 - Given that padding size depend even on compiler options, keep track of
4011 alignment and aligned_size only for the current architecture. Not a big
4012 deal since we send structure fields one after the other (seems
4014 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4015 program, let it work on linux/gcc/little32. portability todo.
4017 2004-07-08 Martin Quinson
4018 - import and improve remote compilation support from FAST
4019 - make sure make check works on half a dozen of machines out there
4021 2004-07-07 Martin Quinson
4022 Let's say it's version 0.0.3 ;)
4023 - Implement conversions (yuhu!)
4024 - Let it work on solaris (beside conversion, of course)
4025 - Stupid me, using rand() to generate the conversion datatests in not wise.
4027 2004-07-06 Martin Quinson
4028 - Let make dist work, since I'm gonna need it to compile on remote hosts
4029 - Let Tests/datadesc_usage write the architecture on which the file was
4030 generated as first byte.
4031 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4032 architecture descriptions.
4033 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4034 on those architectures.
4035 - Optimization: send/recv array of scalar in one shoot
4037 2004-07-05 Martin Quinson
4038 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4040 - Plug a whole bunch of memleaks
4041 - each process now have to call gras_{init,exit}. One day, their log
4042 settings will be separated
4043 - Continue the code factorisation between SG, RL and common in Transport.
4045 2004-07-04 Martin Quinson
4047 - Redistribution between SG and RL.
4048 We wanna have to accept in SG, so move accepted related parts of RL in
4049 the common part. (more precisely, the dynar of all known sockets is no
4050 more a static in transport.c, but part of the process_data)
4052 [gras_stub_generator]
4053 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4055 2004-07-03 Martin Quinson
4056 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4057 them all up in one shot)
4059 - Fix the ugly case of reference to dynamic array.
4060 - New (semi-public) function gras_datadesc_size to allow the messaging
4061 layer to malloc the needed space for the buffer.
4063 - gras_socket_close now expect the socket to close (and not its address to
4064 put NULL in it after it). This is because the socket passed to handlers
4065 is one of their argument (=> not writable).
4067 - propagate the interface cleanup from last week in datadesc, ie remove a
4068 superfluous level of indirection. User pass adress of variable
4069 containing data (both when sending and receiving), and not of a variable
4070 being a pointer to the data. Let's say that I like it better ;)
4071 The price for that is constructs like "int msg=*(int*)payload" in
4072 handlers, but it's a fine price, IMHO.
4074 - Let it work in RL (yuhu)
4076 2004-06-21 Martin Quinson
4078 - porting SG plugin and SG select to new standards (works almost).
4079 - plug memleaks and fix bugs around.
4082 - cleanup the prototype of data recv and force users to specify when they
4083 want to handle references to objects. Test case working even for cycles.
4084 - plug memleaks. Valgrind is perfectly ok with this.
4086 2004-06-12 Martin Quinson
4088 - cleanup the separation between plugin and main code in plugin creation
4090 2004-06-11 Martin Quinson
4092 - Reput hook for raw sockets, needed for BW experiments
4093 - kill a few lines of dead code
4094 [Data description] Interface cleanup
4095 - gras_datadesc_by_name returns the searched type or NULL.
4096 That way, no variable is needed to use a type desc once, which makes
4098 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4099 two parameters were strings (field name, type name), leading to
4101 [Dicos] Interface cleanup
4102 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4103 This is consistant with the dynar API.
4105 2004-04-21 Martin Quinson
4107 - Porting to new standards.
4109 - interface cleanup.
4110 There is no bag anymore, no need to take extra provision to mask the
4111 pointers behind "ID".
4112 Better splitup of functions between files create/exchange/convert.
4113 This is still a bit artificial since convert and receive are so
4114 interleaved, but anyway.
4116 - add a queued message list to procdata (the ones not matching criteria
4118 - factorize some more code between SG and RL wrt procdata
4120 - use gras_exit in example to track memleaks
4121 - get rid of gs_example now that GS is properly integrated into gras
4122 - update run_test to integrate the latest tests (datadesc)
4124 - rename WARNINGn macros to WARNn since it prooved error-prone
4126 2004-04-19 Martin Quinson
4128 - register init/exit functions within gras module mechanism
4129 - send/receive function.
4130 Convertion is not implemented, but short-cutted if not needed.
4131 struct/array elements are sent one by one (instead of block-wise), but
4132 nobody really cares (yet). Get a prototype before optimizing.
4133 - tests (using a file socket) for DD send/receive on:
4134 - base types: int, float
4135 - array: fixed size, string (ie ref to dynamic string)
4136 - structure: homogeneous, heterogeneous
4137 - chained list, graph with cycle
4138 Believe it or not, valgrind is not too unhappy with the results. The
4139 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4143 - Bugfix when using a filename explicitly (instead of '-')
4145 2004-04-09 Martin Quinson
4147 - factorize more code between RL and SG in socket creation
4148 - Complete the implementation and tests of:
4150 o file (only in RL, and mainly for debugging)
4152 I lost 3 days to design a portable address resolver, and then decided
4153 that the prototype mainly have to run on my box.
4154 Addressing portability too early may be like optimizing too early :-/
4156 - use gras_init in the Tests instead of the crappy parse_log_opt
4157 (the latter function is removed)
4158 [Conditional execution]
4159 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4160 [Code reorganisation]
4161 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4162 Build examples against the RL library, since there is no way to disable
4163 its creation for now.
4165 For information, the beginning of coding on GRAS was back in june
4166 2003. I guess that every line has been rewritten at least twice since