1 ----------------------------------------------------------------------------
3 SimGrid (3.24.1) NOT RELEASED YET (v3.25 expected December 22. 2029, 04:19 UTC)
5 Important user-visible changes:
6 - Improve the Python usability (stability and documentation).
7 - A nasty synchronization bug was ironed out, see also below.
8 - Python's doc was organized as a separate tree, now integrated with C++.
9 - C bindings of S4U were not part of the doc.
10 - The C++ doc was also improved as methods are now split by theme.
11 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
12 - This is mandatory to use the Java bindings.
13 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
14 - SimDAG++: Automatic dependencies on S4U activities (experimental)
15 - Some features are already implemented but not all of them
16 - Cannot block an activity until it's scheduled on a resource
17 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
18 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
23 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
24 - Drop support for Viva/Triva (old visualization tools), hence removed
25 TRACE_get_node_types() and TRACE_get_edge_types() functions.
28 - Require pybind11 version 2.4.
29 - Greatly improve locking during context switches with Python.
30 - Declaring Python callbacks to this_actor::on_exit is now working.
31 - No more tests are failing currently, even on win32.
34 - Actor: Merge signals on_migration_start/end into on_host_change
35 - Actor: Rename migrate() into set_host()
36 - Disk: Allow users to get the read and write nominal bandwidth values
37 - Exec: Implement wait_for(timeout)
38 - Io: Implement wait_for(timeout)
41 - Parse errors now raise a simgrid::ParseError that you may want to catch.
44 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
45 been renamed to fix a spelling error. As usual, the previous definitions are
46 kept with a deprecation warning. However, to avoid runtime errors, it is now
47 forbidden to override these deprecated methods in a derived class. Please use
48 the new names immediately if you need to override them.
51 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
52 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
53 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
54 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
55 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
56 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
57 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
58 - New module: random, providing classical random numbers generators.
61 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
62 allocations' size above which they are considered "shared" by default (as if
63 they were performed through SMPI_SHARED_MALLOC macros).
64 Default = 0 = disabled feature.
65 Note : malloc, calloc and free are now overriden by smpicc/cxx by default.
66 This can cause some troubles if codes are already overriding these. If this
67 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
68 help, but will make this feature unusable.
69 - Input check is now performed much more consistently and outputs precise
70 warnings for each error found
71 - New supported calls : MPI_File_set_view, MPI_File_get_view
72 - MPI I/O now only supports disks, as storage is discontinued.
73 - MPI I/O allows opening files without specifying absolute path
74 - Note : SMPI configuration options are now initialized by smpi_init_options()
75 call, which should be called if SMPI is not being used through smpirun.
77 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
78 - FG#9: Python bindings crashing
79 - FG#39: Missing s4u::Comm::wait_any_for example
80 - FG#42: Add support for ThreadSanitizer (TSan)
81 - FG!19: Removing RngStream
82 - FG!20: A module for RNG calls
83 - FG!21: Choice between ad-hoc and standard distributions implementations
85 - FG!25: Fix link in Mutex doc
86 - FG!26: Fix links in SMPI interface doc
87 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
88 - GH#305: Unscheduled tasks are still excuted
89 - GH#313: smpirun: manual outdated w.r.t. --help
90 - GH#321: [S4U] Get task remaining work ratio
91 - GH#323: Crash when an actor turn off his physical host
92 - GH#335: Missing links on dragonfly example svg
93 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
94 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
96 ----------------------------------------------------------------------------
98 SimGrid (3.24) October 9. 2019
100 The Clean Disk Release.
103 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
104 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
105 thus becomes the default resource to execute I/O operations. Examples
106 and the file system plugin have been modified to use disks instead of
107 storage elements. Storage can still be used but is doomed to disappear.
108 Users are thus adviced to update their simulators accordingly.
109 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
110 for consistency with pthread_barrier_wait()
111 - Host::get_englobing_zone() returns the englobing netzone
112 - Actor::on_destruction is now called in the destructor
113 Actor::on_termination new signal called when the actor terminates
115 - Global signals are now part of the Engine:
116 - on_platform_creation: after config settings, before the XML parsing
117 - on_platform_created: right after the XML parsing
118 - on_time_advance: each time the clock advances
119 - on_simulation_end: after simulation, before cleanups
120 - on_deadlock: as the name implies.
122 - sg_{actor,host,link}_{data,data_set}() now all exist.
123 Use them to attach user data to the object and retrieve it.
126 - Introduce an experimental Wifi model. It sounds reasonable
127 according to the state of the art, but it still has to be properly
128 validated, at least against ns-3 if not against reality.
129 - Improved the usability of ns-3. Several bugs were ironed out.
130 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
131 It is now expressed as 'idle:epsilon:allCores' to properly model the
132 consumption of non-whole tasks on mono-core hosts. More info in the doc.
135 - convert a new set of functions to the S4U C interface and move the old MSG
136 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
139 - Fortran bindings for DVFS have been removed.
140 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
141 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
142 - SMPI can now be selected by cmake's find_module(MPI) with
143 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
144 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
145 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
146 - trace-call-location can be used with TI traces, and replayed, and is compatible
147 with smpi/comp-adjustment-file.
148 - sleep events are now correctly traced/replayed.
149 - Default for trace-call-location is now to use file names and not full paths.
150 To revert to previous behavior (in case of collision of filenames), option
151 "smpi/trace-call-use-absolute-path" can be set to yes.
154 - Use the included xxHash as an hash implem when C++14 is usable.
155 - Option model-checker/hash was removed. This is always activated now.
156 - New option smpi/buffering controls the MPI buffering in MC mode.
157 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
158 This is useful to check for MPI compliance.
161 - New section on plugins: how to define a new one, and existing ones.
164 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
165 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
168 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
169 and <mount> tags. The rationale is finish to decouple what should be handle
170 by SimGrid kernel (i.e., resources and activities) from plugins built in
171 user space (e.g., the file system plugin). This new tag currently accepts
172 only three attributes (id, read_bw, and write_bw). All the other information
173 that was declared with the storage related tags now has to be expressed as
174 properties. An example of platform using this new tag is available at
175 examples/platforms/hosts_with_disks.xml
178 - 'expect signal' can now accept more than one potential signal.
180 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
181 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
182 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
183 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
184 - FG#31: per-actor data
185 - FG#34: SG_BARRIER_SERIAL_THREAD?
186 - FG#35: model-checker does not like buster-produced binaries
187 - FG!13: MC: complete workaround in the error msg seen on modern systems
188 - FG!15: execute_flops now logs compute
189 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
190 - FG!17: ns-3: unblock the right number of communications + others issues
191 - FG!18: Improving the performance of the ns-3 bindings
192 - GH#189: Energy consumption of parallel tasks
193 - GH#219: Error in the throughput of TCP transfer
194 - GH!330: Minor polishing of English: Configuring_Simgrid
195 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
196 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
197 - GH!334: Rework energy plugin, again
199 ----------------------------------------------------------------------------
201 SimGrid (3.23.2) July 8. 2019
204 - Nicer introduction page.
205 - Migrate the "Deploy your application" page to the new doc.
206 - Move Java as a subtree of MSG.
209 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
212 - Drop xbt_dynar_sort_strings().
215 - Really fix FG#26: Turning off a link should raise NetworkFailureException
216 - FG#27: Wrong exception thrown to wait_any when link is turned off
217 - GH#328: Java: Canceling multiple tasks in a single vm/host
219 ----------------------------------------------------------------------------
221 SimGrid (3.23) June 25. 2019
223 The Exotic Solstice Release.
226 - SunOS and Haiku OS support. Because exotic platforms are fun.
227 - Stop setting random seed with srand48() at initialization.
228 - Use addr2line as a fallback for stacktraces when backtrace is not available.
229 - Build option -Denable_documentation is now OFF by default.
230 - Network model 'NS3' was renamed into 'ns-3'.
233 - Simgrid can now hopefully be installed with pip.
236 - wait_any can now be used for asynchronous executions too.
239 - New log appenders: stdout and stderr. Use stdout for xbt_help.
240 - Drop xbt_dict_dump.
243 - 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
244 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
245 - MPI interface is now const correct for input parameters
246 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
247 - Fortran bindings for async collectives.
248 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
251 - Remove option 'model-check/record': Paths are recorded in any cases now.
252 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
253 always sparse. That's an immense gain in memory, and can even be
254 faster because of cache effects. So there is no need to clutter the
255 code to allow the user to go for the unefficient mode.
258 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
259 models proved to be more accurate than these old experiments.
261 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
262 - FG#1: Broken link in error messages
263 - FG#2: missing installation documentation
264 - FG#3: missing documentation in smpirun
265 - FG#6: Python bindings not available on PyPI
266 - FG#7: simple cmake call requires doxygen
267 - FG#8: make python bindings an optional dependency
268 - FG#10: Can not use MSG_process_set_data from SMPI any more
269 - FG#11: Auto-restart actors forget their on_exit behavior
270 - FG#12: -Denable_lto=OFF doesn't disable LTO
271 - FG#13: Installs unstripped file 'bin/graphicator'
272 - FG#14: Installs the empty directory 'doc/simgrid/html'
273 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
274 - FG#17: Dead link in doc (pls_ns3)
275 - FG#20: 'tesh --help' should return 0
276 - FG#21: Documentation link on http://simgrid.org/ broken
277 - FG#22: Debian installation instruction are broken
278 - FG#26: Turning off a link should raise NetworkFailureException exceptions
279 - GH#133: Java: a process can run on a VM even if its host is off
280 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
281 - GH#326: Valgrind-detected error for join() when energy plugin is activated
283 ----------------------------------------------------------------------------
285 SimGrid (3.22) April 1. 2019
287 The Easter Christmas Release (this one is somewhat late).
290 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
291 - Not complete yet: asynchronous activities (amongst others) are still missing
292 - Still ongoing: the interface may change in the future. We need more
293 testers! Please report any glitches.
294 - No new project using Java should start now. Please switch to Python.
297 - Some of the internal cleanups may lead to speed improvements:
298 - The hard limitation on the amount of simulated actors with
299 Java+Mac was removed. Now, the available memory is the only limit.
300 - Our refcounting was tidyied, leading to 10% speedups in some cases.
301 - We are still working on making our code robust to the actor kills
302 and hosts' churn. Things are improving, but it's not perfect yet.
303 - Replaced our own code to display a backtrace (that was forking addr2line)
304 with the Boost.Stacktrace library.
305 You won't see your backtraces without this optional dependency.
306 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
307 - Stop setting random seed with srand() at initialization.
310 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
311 XML file version remains 4.2 since old files are still compatible.
314 - Process termination which was broken at version 3.21 has been repaired.
315 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
316 - Hide the examples into examples/deprecated. New users should use Python.
319 - Drop MSG_process_create_from_stdfunc() from the API.
320 This C++ function was a pimple in the C API, made necessary at some
321 point by the Java bindings. This is fixed now.
322 - Hide the examples into examples/deprecated. New users should use S4U.
323 - MSG_process_create and MSG_process_attach now crash if the host used
324 to run this process is off.
325 - Fix the protype of MSG_process_on_exit()
326 Now use (int,void*) callbacks instead of (void*,void*) ones.
327 The implementation was ways too messy to actually work, I guess.
330 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
331 too often and provide more accurate timings.
332 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
333 - MPI_Alltoallw support
334 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
335 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
336 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
337 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
339 - Basic implementation of generalized requests (SMPI doesn't
340 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
343 - Drop sg_cmdline. Please use xbt_cmdline instead.
344 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
345 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
346 OsSemaphore is implemented in a portable way with C++11 threads.
347 This should allow much more threads to be created at the same time,
348 allowing Mac and Java users to simulate many more actors.
349 - Implement the 'thread' factory with std::thread instead of xbt ones.
350 It is not possible to set the stack size with threads anymore, but
351 -fsplit-stack is the way to go nowadays when using threads.
352 - Drop the xbt_os_thread_t module (now unused)
353 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
354 - Drop xbt_str_join_array().
355 - Drop cunit, use Catch2 instead.
358 - Many cleanups in the kernel::activity namespace. This was long
359 overdue, and shall open the path to many future endeavors.
362 - #132: Java: a process can not shut down its own host
363 - #220: S4U: detached send causes exception if sender terminates soon after sending
364 - #261: Document the parameters of parallel execution's constructor
365 - #300: [s4u] BarrierPtr is missing
366 - #314: SMPI args internal cleanup
367 - #316: Fix a bug related to the CPU utilization of multi-core VM
368 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
369 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
370 - #325: Turning off a host has different behavior on sleeping actors and computing actors
372 ----------------------------------------------------------------------------
374 SimGrid (3.21) October 3. 2018
376 The Restarting Documentation (TRD) Release.
379 - Convert the user manual to Sphinx for improved usability.
380 Unstable version now hosted on https://simgrid.frama.io/simgrid
381 Many glitches remain; stay tuned.
382 - Restrict the Doxygen reference API to the installed header files.
383 (documenting all internals this way is overkill + doxygen scales badly)
384 - New tutorial on S4U (highly inspired from the old MSG one, but with
385 a git repository to fork as a starting point, and a docker image)
386 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
387 - The Developper manual is still to be converted (not compiled until then)
388 - Some parts are still missing in this conversion (such as the
389 platform and deployment sections) while others were blindly converted
390 and would need more love (such as the configuration flags).
391 Things will certainly further improve in the future releases.
394 - Fully reimplement auto_restart mechanism that was utterly broken
395 (fix #22, #131 and #281 that were all very related)
396 - Implement semaphores (fix #303 and #291)
397 - When creating an actor from a function and its parameters,
398 move-only parameters are not allowed anymore, as it would prevent
399 the actor to be restartable if its parameters are consumed on run.
400 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
401 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
402 IO operations that can be started, waited for, or canceled as a regular
404 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
408 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
409 - New host variable: 'core_count'
412 - Remove xbt_os_thread_specific features
413 - Remove portability wrapper to condition variables
414 - Remove xbt_os_thread_yield()
417 - MPICH collective selector now mimics MPICH 3.3b
418 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
419 - MPI_Init(NULL,NULL) is now allowed
420 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
421 - Command-line arguments should now be handled properly in Fortran simulations
422 - MPI Attributes and related callbacks should work in Fortran
423 - Apps using C + Fortran should now work
424 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
425 - Support MPI_Error_String
428 - Due to an internal bug, Msg.run() must now be your last line.
429 We hope to fix it in a future release, and we are sorry for the inconvenience.
432 - #22: Process autorestart seem to only work with CAS01 cpus
433 - #56: Feature request: dynamic SMPI replay
434 - #93: simgrid should not eat --help
435 - #111: How to change the loopback link characteristics?
436 - #116: Platform cluster radical
437 - #117: Platform cluster/host order
438 - #129: Replay_multiple on steroid
439 - #131: Java : autostart does not work
440 - #143: Setting a breakpoint at a given time
441 - #235: Network energy models should be integrated
442 - #264: Add ptask L07 resource tracing
443 - #271: Dynamic and manual replay of SMPI traces
444 - #279: Breakpoints option uses time deltas instead of absolute time
445 - #281: Daemonized actor no longer daemonized after an auto-restart
446 - #282: TIT convertor erase traces when using absolute path in the trace list file
447 - #285: segfault when a s4u actor kills itself with msg process tracing activated
448 - #286: Pajé traces are not exposing the number of cores
449 - #287: Command-line parsing should stop when encountering '--'
450 - #288: MPI_Init(NULL, NULL)
451 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
452 - #290: Method to check if a mailbox has a message ready to be consumed
453 - #291: [s4u] semaphores do not seem available yet
454 - #294: Alltoall collective implementation of mpich
455 - #295: Starting a Host as OFF is not supported
456 - #296: DTD too permissive
457 - #299: Add a s4u barrier example
459 ----------------------------------------------------------------------------
461 SimGrid (3.20) Released June 24. 2018
463 The proxy snake_case() release.
466 - The whole API is now using snake_case() instead of mixing it with camelCase.
467 Compatibility wrappers in place for one year.
468 - We took the opportunity to increase the consistency of the API.
469 Please report any remaining glitches.
470 - New function simgrid::s4u::Host::get_actor_count:
471 Returns the number of actors running on a specific host.
474 - SMPI is now tested with ~45 proxy apps from various sources, with none or
475 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
476 - Replay: The replay file has been re-written in C++.
477 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
478 supported. They are stored in the trace and used when replayed.
479 - Basic support of MPI_Cancel. Robustness not guaranteed.
480 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
481 (only for MPI_COMM_TYPE_SHARED).
482 - New option: smpi/privatize-libs, to add external shared libs to be privatized
483 by SMPI. They will be copied locally and loaded separately by each process.
484 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
485 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
486 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
487 with internal index optind with multiple processes. Only works if MPI_Init has
489 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
490 syntax without preprocessing tricks.
493 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
494 behavior tracing is no longer limited to MSG
495 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
496 this means that traces now need to be updated or re-obtained; both Irecv
497 and Isend lines in your traces have now in total 5 mandatory fields:
498 <rankid> <command> <to/from rankid> <tag> <size>
499 To update your traces, it suffices to add a 0 for the tag here.
500 - TIT now also supports waiting for a distinct request via MPI_Wait.
501 Wait/Test now wait for a specific request, not just the last one that was
502 issued. This unfortunately means another update, because we need to
503 identify which request you want to wait for. We do this via the
504 triplet (sender, receiver, tag), which needs to be added:
505 <rankid> <command> <sender> <receiver> <tag>
506 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
507 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
510 - The deprecation of MSG is ongoing (but this should not impact you).
511 Many MSG functions are now simple wrappers on the C API of S4U. If
512 you wish to convert your code to S4U, find the S4U counterparts of
513 your MSG calls in src/msg/msg_legacy.cpp.
514 - MSG can still be used, but won't evolve anymore.
517 - Allow to run the Link energy plugin from the command line with
518 --cfg=plugin:link_energy
519 - Rename Energy plugin into host_energy
520 - Rename Load plugin into host_load
523 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
524 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
525 - Improve performance when handling timeouts of simix synchros.
528 - Config: the C API is now deprecated (will be removed in 3.23), and
529 the C++ API has been slightly improved.
532 - Fix several build issues on OSX.
533 - Move simgrid_config.h to simgrid/config.h (old header still working)
536 - #143: Setting a breakpoint at a given time
537 - #150: Inconsistent event names in SMPI replay
538 - #258: daemonized actors hang after all non-daemonized actors have completed
539 - #267: Linker error on unit_tmgr
540 - #269: SMPI: tracing of MPI_Wait/all/any broken
541 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
543 ----------------------------------------------------------------------------
545 SimGrid (3.19.1) Released March 22 2018
547 The Fixed ABI Release.
549 As you may know, we are currently refactoring SimGrid in deep.
550 Upcoming SimGrid4 will be really different from SimGrid3: modular,
551 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
553 Our goal is to smooth this transition, with backward compatibility
554 and automatic update paths, while still progressing toward SimGrid4.
556 SimGrid remains open during works: The last pure SimGrid3 release was
557 v3.12 while all subsequent versions are usable alpha versions of
558 SimGrid4: Existing interfaces remain unchanged, but the new S4U
559 interface is budding and the internals are deeply reorganized.
561 Since 2015, we work hard to reduce the changes to public APIs. When
562 we need to rename a public library symbol in S4U, we let your
563 compiler issue an explicative warning when you use the deprecated
564 function. These messages remain for four releases, i.e. for one full
565 year, before turning into an error. Starting with v3.15, your can
566 also adapt to API changes with the SIMGRID_VERSION macro, that is
567 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
569 Starting with this v3.19.1, our commitment to reduce the changes to
570 the public interfaces is extended from the API to the ABI: a program
571 using only MSG or SimDag and compiled against a given version of
572 simgrid can probably be used with a later version of SimGrid without
573 recompilation. We will do our best... but don't expect too much of
574 it, that's a really difficult goal during such profund refactoring.
576 The difference between v3.19 and v3.19.1 is that the former was
577 accidentally breaking the ABI of MSG, while the later is restoring
580 S4U and kernel APIs will still evolve until SimGrid4, with one-year
581 deprecation warnings as currently. In fact, cleaning up these
582 interfaces and converting them to snake_case() is one release goal of
583 v3.20. But don't worry, we are working to smooth this upgrade path.
585 In summary, new projects should start with S4U to benefit of the
586 future, but old MSG projects should still be usable with no change.
588 ----------------------------------------------------------------------------
590 SimGrid (3.19) Released March 20 2018
592 The Moscovitly-cold Spring release.
594 SMPI user-visible improvements:
595 - Support of MPI_Type_create_subarray
596 - Unimplemented MPI calls now abort (instead of issuing a warning)
597 Some harmless calls around errhandlers still only issue a warning.
598 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
599 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
600 is now possible (mmap privatization was refactored to this end).
601 Adding new SMPI instances should also be technically doable now.
603 SMPI internal cleanups:
604 - Removed index notion from SMPI -- no more "getPid() - 1"!
605 In tracing files, actors are called rank-<PROCESS-ID> but these are
606 NOT the real ranks (tracing several communicators is not supported yet).
607 - Deprecate smpi_process_index() (will be removed in 3.22)
608 - Better testing on classical proxy apps, through an external project:
609 https://github.com/simgrid/SMPI-proxy-apps/
612 - Use a graphical TOC to make it easier to find the documentation you need
613 - Revamp the MSG tutorial
616 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
617 - Remove parameter of MSG_process_killall().
618 Resetting the PID was bogus anyway (several actors could have the same PID).
621 - Execution->setHost() can be called after start(), to migrate it.
622 - Comm::test_any() is now implemented.
623 - s4u::Actor now has onCreation() and onDestruction() signals.
624 - Install some kernel header files for the users' plugins and more.
625 Warning, their API is really not stable yet.
626 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
627 include/simgrid/kernel/routing/*Zone.hpp
630 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
632 Build System and other cleanups:
633 - Remove unused run-time parameter "tracing/onelink-only".
634 - SimGrid now works with intel compilers (MC still buggy)
635 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
636 - tesh kills subprocesses on timeout (but not on windows)
639 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
642 - #194: Feature request: simgrid::s4u::Comm::test_any()
643 - #245: migrating an actor does not migrate its execution
644 - #253: Feature Request: expose clusters as objects
645 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
646 - #255: Tesh broken on Windows
647 - #256: Modernize FindSimGrid.cmake
648 - #257: Fix (ab)use of CMake install
650 ----------------------------------------------------------------------------
652 SimGrid (3.18) Released December 24 2017
654 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
656 MSG IS NOW DEPRECATED. Please use S4U for new projects.
657 - Support for MSG should not be removed from SimGrid before 2020,
658 but future evolutions will be limited to the integration of
659 user-provided patches.
662 - s4u::onDeadlock() signal, run before stopping the simulation.
663 - s4u::Exec: asynchronous executions (abstraction of background
665 - s4u::Host->getLoad() returns the achieved speed in flops/s
668 - Link::name() is deprecated and will be removed in v3.21, use
669 Link::getCname() instead.
670 - Mailbox::getName() changed to return a std::string, use
671 Mailbox::getCname() instead to get a char*.
672 - Storage::getName() changed to return a std::string, use
673 Storage::getCname() instead to get a char*.
674 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
675 to have both allocation and deallocation of the map in user space.
678 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
679 The amount of flops remaining to do is a vector, not a scalar.
680 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
681 name implies on both sequential and parallel tasks.
682 - Both changes fix GitHub's #223 using PR #237 as a basis.
683 Thanks Michael Mercier.
684 - Most examples were converted to S4U and hidden elsewhere at tests.
685 Rationale: we still want MSG to work; we want newcomers to use S4U.
688 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
689 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
692 - Switch to the faster dlopen privatization mechanism by default
693 - Documentation improvments
696 - Live migration is getting moved to a plugin. Dirty page tracking is
697 the first part of this plugin. This imply that VM migration is now
698 only possible if one this function is called:
699 - C/MSG: MSG_vm_live_migration_plugin_init()
700 - C/C++: sg_vm_live_migration_plugin_init()
701 - Java: Msg.liveMigrationInit()
702 For C and C++, "simgrid/plugins/live_migration.h" also has to be
706 - Define class simgrid::xbt::Path to manage file names.
707 - Removed unused functions:
708 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
709 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
710 xbt_graph_export_graphviz()
711 - xbt/str.h: xbt_str_join()
712 - Remove unused datatypes:
713 - xbt/heap.h: use std::priority_queue or boost::heap instead
714 - xbt/swag.h: use boost::intrusive::list instead
717 - New link_energy plugin for the consumption of the links.
718 - All of the operations on files and storage contents have been
719 packaged into a plugin (src/plugins/file_system). The current
720 public interface can be found in
721 include/simgrid/plugins/file_system.h
722 To use these functions you now have to initialize the plugin by
723 calling MSG_storage_file_system_init() just after calling
724 MSG_init() or sg_storage_file_system_init() just after creating
728 - Remove the undocumented/untested tag <include>
731 - Remove viva specific tracing as the tool is no longer maintained
734 - #248: Kill of finished processes leads to segfault
735 - #240: xbt_cond_wait_timeout should gracefully return for C apps
736 - #239: please implement signal s4u::onDeadlock()
737 - #230: segfaults when exit() without run()
738 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
739 - #223: MSG_task_get_flops_amount() not working with parallel tasks
740 - #222: Actor::kill() doesn't really kill and segfaults
741 - #221: odd LMM warning when killing an actor
742 - #120: Memory leak when the processes are forcefully killed
744 ----------------------------------------------------------------------------
746 SimGrid (3.17) Released October 8 2017
748 The Drained Leaks release: (almost) no known leaks despite the tests.
750 Even very long simulations will preserve your memory: our testsuite
751 stresses SimGrid for over 45mn on fast machines for a coverage of
752 over 80%, and there is only one single known leak, of about 4kb.
756 - Comm.detach(): start and forget about asynchronous emission. A cleanup
757 handler may be given to free resources if the comm cannot be completed.
758 - this_actor::send(mailbox) is now mailbox->put()
759 - New: simgrid::s4u::Comm::wait_all()
760 - New: Host.execute() for remote executions.
763 - Unused option network/sender-gap is removed.
766 - Deprecate MSG_task_isend_with_matching(): This unused feature
767 really complicates our internals. Will be removed in v3.20.
770 - Improved context termination. It is now possible to free resources, even
771 when a process is forcibly killed.
772 - ContextBoost: add support for Boost versions above 1.61.
775 - Bring back run-time option --cfg=exception/cutpath to remove exception
777 - Removed unused functions:
778 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
779 xbt_str_rtrim(), xbt_str_trim().
780 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
783 - Removed header files obsolete since SimGrid 3.12:
784 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
786 - #3: SD_exit should be made optional
787 - #120: Memory leak when the processes are forcefully killed
788 - #159: Threading test regression in Actor refcounting
789 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
790 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
791 - #186: Actor::killAll() segfaults if some process is blocked on wait()
792 - #191: VM migration and pstate
793 - #192: Updating the energy consumptions of all hosts crashes with VMs
794 - #195: All actors have PID=0 in the logs
795 - #204: Sometimes segfault with thread contexts and mmap privatization
796 - #222: Actor::kill() doesn't really kill and segfaults
797 - #225: Actor::kill() doesn't really kill when victims are doing a join()
799 ----------------------------------------------------------------------------
801 SimGrid (3.16) Released June 22. 2017.
803 The Blooming Spring Release: developments are budding.
806 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
807 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
808 - Java JAR file should be finally fully working on Mac OSX too.
810 XML platforms: Switch to platform v4.1 format.
811 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
812 - <zone> can be used as a synonym for the now deprecated <as>
813 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
814 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
815 - <actor> can be used as a synonym for the now deprecated <process>
816 - state_file and avail_file periodicity is now easier to express
817 (check the documentation) the old behavior should still work.
818 * Storage: not backward compatible, but it seems that nobody ever used it
819 - Remove attribute 'content_type' of <storage_type>: was never used
820 - Make attribute 'model' of <storage_type> optional: for future usage
821 - Remove Bconnection model property: was never (in)validated. Replaced by
822 the maximum of the read and write bandwidth as a resource constraint.
825 - New and Backwards Compatibility break:
826 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
827 When one wants to get the list of tasks whose states have changed during a
828 simulation round, s/he has to allocate and free a dynar and use it as
829 argument to this function. The former SD_simulate (double how_long)
833 - Allow multicore VMs, with the correct sharing (unless you overcommit)
834 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
837 - New function to update the consumption of all hosts at once.
838 - Fix the model for multi-core hosts, linear in the amount of busy
839 cores with an abnormality for pIdle (see paper).
840 BUG: Still not satisfactory for ptask on mono-cores.
843 - The netzone are now available from the MSG API.
844 The old names still work, but are now deprecated.
847 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
848 - New option: smpi/keep-temps to not cleanup temp files
849 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
850 mallocs mode is used, allows to change the size of the fake file used
851 (default 1MB), to potentially limit the number of mappings for large runs.
852 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
853 - Fortran ifort and flang compilers support
854 - New RMA calls supported (experimental):
855 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
856 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
857 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
858 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
859 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
860 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
863 - New: this_actor::isMaestro() returns whether we are in kernel mode.
864 - Behind the scene, ActivityImpl are now automatically refcounted.
865 Removing this nasty bottleneck will greatly simplify our internals.
868 - Replay: New function xbt_replay_action_get():
869 Retrieve the function previously associated to an event type.
870 - DROPPED FUNCTION: xbt_str_varsubst()
871 - DROPPED MODULE: strbuff. We don't need it anymore.
872 - DROPPED MODULE: matrix. We don't need it anymore.
873 - DROPPED MODULE: lib. We don't need it anymore.
875 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
877 SimGrid (3.15) Released March 22. 2017
879 The Spring Release: continuous integration servers become green
880 We fixed even the transient bugs on all target architectures:
881 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
882 Windows, FreeBSD, NetBSD.
884 - Rename NetCards to NetPoints.
885 This was intended to help NS3 users, but that's not a netcard.
886 That's a point in the routing algorithm, let's avoid wrong simplifications.
889 - New: MSG_process_yield(). Stop and yield to other processes.
890 - New: MSG_process_daemon(). Daemon processes are automatically killed
891 when the last non-daemon process terminates
892 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
893 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
894 to make room for the upcoming network energy plugin.
895 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
898 - Ensure that an actor can kill itself with Process::exit()
899 - Kill the obscure NativeException. Nobody want to survive the issues
900 it denotes, so use JniException that is a RuntimeException (not to
901 be caught explicitly).
902 - Partial bug fix in initialization. SimGrid flags on command line were
903 consumed at C level but stayed in the original Java String[] args.
904 This could mess users' args[i] if SG flags were not put at the end of
906 The SimGrid flags are now removed from the Java arguments. However,
907 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
908 if args.length is greater than the number of YOUR OWN ARGUMENTS.
909 It might be if you have --log or --cfg flags in the command line.
910 - Fix numerous memleaks all around the place. In particular, around VMs.
914 - s4u::onPlatformCreated: right before the simulation starts
915 - s4u::onSimulationEnd: right after the main simulation loop
916 - s4u::onTimeAdvance: right after a clock change (time is discrete)
917 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
918 event from the availability_file changes the avail speed.
919 - Links are now usable from s4u
920 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
921 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
922 - New examples: The conversion of MSG examples to S4U has begun
923 - Actors: create, kill, migration, and suspend
924 - Applications: master-worker and token-ring
925 - Action replay: communications and storage
926 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
929 - Backwards Compatibility breaks
930 - The SD_route_get_list and SD_route_get_size functions have been removed.
931 They are replaced by the unique
932 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
933 The route from src to dst is built in the links dynar whose size can
934 be obtained with xbt_dynar_length.
935 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
936 sg_host_route_bandwidth and sg_host_route_latency.
937 Macros ensure the backwards compatibility, but you should fix your code
940 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
941 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
942 - Bump our claim of support from MPI 1.1 to MPI 2.2.
943 We don't support 100% of it, but it should be enough. Contact us if not.
944 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
945 naive and probably inaccurate)
946 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
947 mapped onto a unique small file using some system magic.
948 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
949 - Fortran: cleanups, fixes, support of user-added operations
950 - RMA: MPI_Accumulate are applied in correct order.
951 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
952 - Better support for MPI_IN_PLACE
953 - Support for MPI_Win attrs and keyvals.
954 - Support MPI_Comm_group_create, MPI_Type_size_x
955 - MPI ops should be properly applied to all allowed types
956 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
957 algorithm by a non failing one from mpich.
960 - Kill the fifo data container: we don't use it anymore.
963 - A <cluster> can now be created with different speed values to represent pstates.
964 see examples/platforms/energy_cluster.xml for an example.
966 SimGrid (3.14.159) Released December 28. 2016
968 The previous release was only a crude approximation of a Pi release;
969 this one should be a bit better.
971 - Revert a last minute change that broke on Mac OSX.
972 - Fix the changelog and NEWS files.
973 - Improve the documentation, in particular of the routing module.
974 - Integrate some patches flying around in Debian.
976 SimGrid (3.14) Released December 24. 2016
979 * The whole documentation was reworked and reorganized. There is
980 still much room for improvement here, but we're on it.
983 * We now need python3 for our testing infrastructure.
984 * Model checking needs libevent
987 * Now works on FreeBSD
990 * Privatization now works on FreeBSD
992 * Privatization is now activated by default in smpirun
993 You can turn it off with -no-privatize if something goes wrong.
995 * Call-location tracing for SMPI.
996 You can add the exact location (filename / linenumber) of an MPI call to
997 your trace files and slow down or speed up the simulation between two
998 consecutive calls by using an adjustment file (see the documentation).
1000 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1001 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1002 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1003 values were ignored in some cases. The timings of these functions can now
1004 be significantly different.
1006 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1007 smpi/running-power is renamed to smpi/host-speed
1009 * smpi/grow-injected-times option to enable or disable multiplication of the
1010 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1011 make simulation less precise (but also much faster).
1013 * smpirun script should be (much) faster for large deployments.
1015 * SMPI tracing: fixed issue with poor matching of send/receives.
1017 * Replay: Fix broken waitall
1019 New functions and features
1020 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1022 Dropped / renamed functions and features
1023 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1024 - MSG_mailbox_is_empty() -> Mailbox::empty()
1025 - MSG_mailbox_front() -> Mailbox::front()
1026 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1027 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1028 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1029 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1030 * Don't pass the free_f parameter to property related functions:
1031 - MSG_host_set_property_value()
1032 - MSG_as_router_set_property_value()
1033 - MSG_storage_set_property_value()
1034 * VM properties. Since msg_vm_t are msg_host_t, just use
1035 MSG_host_get_property_value() and friends
1036 * VM I/O related things:
1037 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1038 - Unimplemented save/restore methods
1039 * MSG_as_router_get_property_value() was redundant with
1040 MSG_environment_as_get_property_value().
1041 - Removed MSG_as_router_*propert*() functions
1042 - Added MSG_environment_as_set_property_value() for consistency
1043 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1044 Well, they are still there for now, but deprecated with a warning.
1045 Please switch to xbt_dict_new_homogeneous() before this is removed
1047 * Task affinity. Its intended behavior (that was very badly tested
1048 and probably not really working) was deceiving what most users
1049 would have hoped here.
1050 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1051 * Option network/coordinates is now useless and should be dropped.
1054 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1057 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1058 * Added Dragonfly topology support
1061 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1062 EiBps, EBps, Eibps, Ebps for bandwidth.
1063 They may become useful to some lucky ones.
1066 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1067 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1068 * ex/async_waitAll: new example, on asynchronous communications
1071 * Memory usage should be decreased for simulations with a large number
1072 of processes. This also helps for SMPI.
1074 SimGrid (3.13) stable; urgency=low
1076 The Half Release, a.k.a. the Zealous Easter Trim.
1078 - We removed half of the lines, that were mostly experimental cruft.
1079 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1080 (not counting blanks and comments -- according to openhub.net)
1081 - The internals are now compiled in C++ (and will soon be clean C++)
1082 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1083 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1085 * Backwards Compatibility breaks
1086 - Removed Lua simulation bindings (switch to C or Java for that).
1087 Lua can still be used to describe platforms
1088 - Removed Java kernel plug-ins.
1089 Will be reintroduced after the ongoing major internals reorg.
1091 - the following functions were removed.
1092 They were too specific and should be reimplemented in a generic
1093 way, with filter function.
1094 - MSG_task_listen_from_host
1095 - MSG_mailbox_get_count_host_waiting_tasks
1096 - MSG_mailbox_put_with_timeout was removed.
1097 Please use MSG_task_send_with_timeout instead.
1099 - the SD_application_reinit function was removed. It has been a noop for a while.
1100 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1101 replaced by a more flexible mechanism.
1102 - The following functions thus do not exist anymore
1103 - SD_workstation_get_access_mode
1104 - SD_workstation_set_access_mode
1105 - SD_workstation_get_current_task
1106 - Basic estimation functions have been removed but can easily be replaced
1107 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1108 - SD_workstation_get_computation_time => amount / sg_host_speed()
1110 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1111 Use VM.getSpeed()*load/100 for the legacy behavior.
1113 - option enable_tracing was removed. It was not doing anything for a while.
1114 - In the ModelChecker:
1115 - the model-checker now ptraces the model-checked process which means
1116 you cannot use a debugger on the latter anymore (we might make this
1117 optional in the feature);
1118 - removed soft-dirty page tracking;
1119 - remove model-checked side snapshot management,
1120 MC_snapshot() and MC_compare_snapshot();
1121 - keep the MC_cut() function as a stub (it was not really working
1122 in the previous release).
1125 * All options are consistently in kebab-case. Old names are kept as alias.
1128 * Switch to platform v4 format.
1129 - Rename from 'power' to 'speed' the attributes describing the amount of
1130 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1131 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1132 - In <host> and <link>, attributes availability and state are gone.
1133 It was redundant with state and availability traces, and with peak values.
1134 - In <cluster>, attributes availability_file and state_file are gone.
1135 It was too complex and unused.
1136 - Kill <gpu>. Was not doing anything.
1137 - The DOCTYPE points to the right URL:
1138 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1139 (the file at this address now documents the changelog since its v1)
1140 - A warning is emitted for unit-less values (they are still accepted).
1141 - speed. Default: 'f' or 'flops'. Also defined:
1142 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1143 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1144 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1145 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1146 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1147 - latency. Default: 's' second. Also defined:
1148 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1150 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1151 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1154 * s4u::Host is now the preferred public interface to the Host features.
1155 sg_host_* functions are C bindings to the exact same behavior
1156 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1159 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1160 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1161 They were deprecated since a few years, and probably did not even compile anymore.
1164 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1165 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1166 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1167 encouraged to update their codes. A list of the modified functions can be found at the end of
1168 include/simgrid/simdag.h
1171 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1172 the lambda or closure passed as a parameter will run in kernel mode.
1173 All the callback functions should be rewritten to that interface at some point.
1176 * Reorganizing and cleaning the internals all around the place.
1179 * Remove old default barrier/bcast buggy algorithms (see #18407)
1180 * Various bug fixes to handle more codes
1181 * Remove the need for the --foreground option of smpirun (it is still
1182 accepted for backward compatibility).
1185 * Kill the setset data container: MC don't use it anymore.
1186 * Kill the queue data container: it made more sense with GRAS.
1187 * Kill the xbt_peer_t data type: it's useless without GRAS.
1188 * Kill rm_cb feature of config sets: it was never useful.
1189 * Kill graphxml parsing feature. It was not used.
1190 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1192 - xbt_dynar_sort_strings(), when the content is char*
1193 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1194 They throw exceptions on invalid input;
1195 * C++ support for declaring CLI flags (simgrid::config::Flag);
1196 * class for abstracting different signal backends (simgrid::xbt::signal).
1197 with no external dependencies (we need very simple signals).
1200 * refactoring and cleanup of the code;
1201 * ongoing process to cleanly separate the model-checking algorithms
1202 from the code model-checking support.
1204 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1206 SimGrid (3.12) stable; urgency=low
1208 The Facelift Release.
1211 * Require g++ v4.7 at least to not speak prehistorical C++.
1212 * Require Boost 1.48 (for signal2 component).
1213 * Java must be version 7 at least when activated.
1214 * Builds on Windows again (including Java bindings).
1215 * Tracing is now always enabled (no way to turn it out)
1216 * Remove GTNetS. It was not working anyway.
1217 * Various cleanups in the cmake scripts.
1218 * Move headers around to sort them out on installed systems:
1219 - instr/instr.h -> simgrid/instr.h
1220 - instr/jedule/* -> simgrid/jedule
1221 - simdag/datatypes.h was removed
1222 - simdag/simdag.h -> simgrid/simdag.h
1223 - msg/datatypes.h was removed
1224 - msg/msg.h -> simgrid/msg.h
1227 * Interface improvement:
1228 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1229 - Sanitize the interface in MSG_task_ module:
1230 - Merge two functions that were close enough but misleading:
1231 set_compute_duration(t) -> set_flops_amount(t)
1232 get_remaining_computation(t) -> get_flops_amount(t)
1233 - set_data_size(t) -> set_bytes_amount(t)
1234 get_data_size(t) -> get_bytes_amount(t)
1235 - Massive cleanups in the functions related to the energy
1236 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1237 - New: MSG_host_get_pstate()
1239 - msg/energy/onoff: switching hosts on and off
1241 * Interface improvement:
1242 - Rename Host.isAvail() to Host.isOn()
1243 - Rename Process.currentProcess() to Process.getCurrentProcess()
1244 - Rename Task.setDataSize() to Task.setBytesAmount()
1245 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1247 - #18874: Actually allows the GC to reclaim tasks
1251 - SIMIX_process_throw: raises an exception in a remote process
1252 * Refactoring: Separate sub-modules
1253 - libsmx: the public interface, as libc in a real system
1254 - popping: the strange dance that converts a user request into a kernel handling
1255 - smx_context_*: the virtualization mechanisms that embed the user code
1256 - smx_*: the handling of each simcalls
1257 * Interface cleanup:
1258 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1259 * Rename smx_action_t into smx_synchro_t, making explicit that these
1260 things are used to synchronize processes with their environment.
1261 For example, a communication is a sort of synchronization involving
1262 the communicating processes (that may block until the exchange) and
1263 the platform. The same can be said from computations, etc.
1265 - #18888: segfault when a process ends before its kill_time
1269 - 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.
1270 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1271 - MPI_Type_set_name, MPI_Type_get_name
1272 - MPI_*_c2f and MPI_*_f2c functions
1273 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1274 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1275 - Activate a lot of new tests from the mpich 3 testsuite
1277 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1278 - InfiniBand network model added: Based on the works of Jerome Vienne
1279 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1280 - When smpi/display_timing is set, also display global simulation time and application times
1281 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1282 * Collective communications
1283 - 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
1284 - 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/).
1285 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1287 - Replay now uses algorithms from wanted collective selector
1288 - Replay can be used with SMP-aware algorithms
1289 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1290 - 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.
1292 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1293 - Correctly use loopback on fat-tree clusters
1294 - Asynchronous small messages shouldn't trigger deadlocks anymore
1295 * Energy/DVFS cleanup and improvement
1296 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1297 - new: smpi_get_host_pstate()
1300 - "Full" network optimization flag was broken since Surf++
1301 - Better handling of precision flags in maxmin
1302 - Fix bug causing sometimes "Impossible" errors
1303 - Properly pass cluster properties to included hosts
1304 * Improvement of the Energy plugin.
1305 - Always update the consumption before returning that value
1306 - New property: watt_off to denote the dissipation when the host is off
1307 - New functions getWattMinAt and getWattMaxAt to retrieve the
1308 dissipation of pstates that we are not currently at.
1309 * Java: class NetworkLink renamed to Link
1310 * New function: simcall_process_get_kill_time()
1311 * Massive rename s/workstation/host/
1312 - That's intrusive, but that's good for the project consistency. Sorry.
1313 - Change config option "workstation/model" into "host/model"
1316 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1317 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1318 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1319 - xbt_dirname and xbt_basename for non-POSIX systems
1321 * The model checker now runs as a separate process.
1322 * The model checker runs must now be launched with the new simgrid-mc program.
1323 * Record/Replay: the MC can display a textual representation of a path in the
1324 execution graph. It can then be replayed outside of the model checker.
1326 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1328 SimGrid (3.11) stable; urgency=low
1333 * Normalizing pointers addresses tool for better diff between logs
1336 * Add cloud examples using new VMs
1337 - examples/msg/cloud/two_tasks_vm.tesh
1338 - examples/msg/cloud/simple_vm.tesh
1339 - examples/java/cloud/cloud.tesh
1340 - examples/java/cloud/migration/migration.tesh
1341 * Add java surf examples:
1342 - examples/java/surfPlugin/surf_plugin.tesh
1343 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1344 - examples/java/surfCpuModel/surf_cpu_model.tesh
1345 * Add SMPI+MSG example:
1346 - examples/smpi/smpi_msg_masterslave/
1351 - msg task destroy cancel test
1352 - msg_host on/off test
1353 * Move all tests in testsuite to teshsuite (adding tesh files)
1354 * Restructure teshsuites
1355 - one folder for each kind of test
1356 * Restructure AddTests.cmake
1358 - structure the order of tests (with sections)
1361 * Add virtual machine
1362 - creation of a VM on a PM
1363 - migration of a VM from a PM to another PM
1365 - MSG_process_join(msg_process_t process, double timeout)
1366 - msg_bar_t MSG_barrier_init(unsigned int count)
1367 - int MSG_barrier_wait(msg_bar_t barrier)
1368 - void MSG_barrier_destroy(msg_bar_t barrier)
1369 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1370 * New option "msg/debug_multiple_use" to help debugging when a task is used
1373 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1375 - Increase code coverage in test suites
1379 * Protect context stacks against stack overflow. The number of protected memory
1380 pages allocated on the top of each stack (1 by default) can be configured
1381 with the new command line option --cfg=contexts/guard_size:<value>.
1382 * Simcalls are now generated by a python script that
1383 - generates files included by SimGrid
1384 - checks that all the functions exist, and proposes prototypes
1386 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1387 host_set_data, host_get_data
1389 - simcall_process_join(smx_process_t process, double timeout)
1390 * Fix bug where sleeping processing could not be suspended.
1393 * Translate surf models from C to C++
1394 - Generic classes for all models: Model, Resource, Action
1395 - A generic interface for each kind of model (CPU, Network, Storage
1396 Workstation, WorkstationVM)
1398 * Translate surf routings from C to C++
1399 * Add callbacks using sigc++ or boost::signals2
1400 - Add callback functions for resource creation/destruction
1401 - Add callback functions for action state change
1402 - Handle Energy as a plugin
1403 * Replace swag by boost::intrusive
1404 * Add new routing models for clusters. For documentation, see
1405 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1406 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1407 parameters for cluster tag
1408 - Fat trees, with topology="FAT_TREE" and
1409 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1410 - see examples/platforms/torus_cluster.xml and
1411 examples/platforms/fat_tree_cluster.xml
1412 * More documentation
1415 * Hostfiles support host:nb_processes construct to deploy several processes on
1417 * Collective communication algorithms should not crash if used with
1418 improper number of nodes and report the error.
1419 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1420 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1421 MPI_Dims_create, MPI_Cart_sub are supported.
1422 * New interface to use SMPI programmatically (still depends on MSG for
1423 some parts, see examples/smpi/smpi_msg_masterslave):
1424 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1428 * Global variables privatization in MPI executables is now performed at runtime
1429 with the option smpi/privatize_global_variables (default:no).
1430 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1431 dynamic libraries loaded by the application are not privatized (static
1432 linking with these libraries is advised in this case)
1435 * Options defined in XML work correctly now.
1438 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1439 copied into simgrid.jar or not (ON by default). Use this option if you want
1440 to reduce the size of the installed simgrid.jar, *and* the native libraries
1441 are kept installed elsewhere.
1442 * Surf binding with SWIG (code generated in maintainer mode only):
1443 - plugin to handle callbacks
1444 - CPU model only for the moment
1447 * Supernovae build mode is definitively removed. It was used to improve
1448 inlining and inter-module optimizations. It is nowadays superseded by
1449 link-time optimizations commonly available in compilers.
1450 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1451 * Add boost dependency for surf++
1452 * Add new macro for tests
1453 - ADD_TESH(name <tesh_args>)
1454 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1458 - xbt_bar_t XBT_barrier_init(unsigned int count)
1459 - int XBT_barrier_wait(xbt_bar_t barrier)
1460 - void XBT_barrier_destroy(xbt_bar_t barrier)
1461 * Make the xbt_os_time module public
1463 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1465 SimGrid (3.10) stable; urgency=low
1467 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1470 * Reintegrate Java to the main archive as desynchronizing these
1471 package is not acceptable anymore (Java is now considered stable)
1472 * Add explicit synchronization facilities through semaphores
1473 * Bug fix: Task.setDataSize() only changed the C world, not the value
1474 cached in the Java world
1477 * Dramatically change the way files are handled. API and internals changed, but
1478 this part of MSG was not considered as production grade either.
1479 * Add explicit synchronization facilities through semaphores
1480 * Add a new function MSG_host_get_process_list()
1481 * Preliminary DVFS support (see examples/msg/energy/ for details)
1484 * SMPI is now included directly in the libsimgrid as the windows
1485 linker doesn't force us on splitting it anymore.
1486 * Improvements of the SMPI replay tool:
1487 - Most of the collective communications are now rooted in the same process as
1488 in the original application.
1489 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1490 used until now). Multiple data types can now be used in a trace.
1491 - The replay tool now supports traces produce either by TAU or a modified
1493 - Bug Fix: the compute part of the reduce action is now taken into account.
1494 - Gatherv collective is now supported
1495 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1496 of smpirun outputs time independent traces for the current run. One file
1497 is created per process. If too many processes are simulated, this behavior
1498 can be changed to one file for all processes by using the
1499 tracing/smpi/format/ti_one_file flag
1500 * smpirun generates the host file if needed (with given host count and platform)
1501 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1502 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1503 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1504 --cfg=smpi/coll_selector:(mpich/ompi)
1505 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1507 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1508 happening inside a collective SMPI call.
1509 * Fix the behavior of complex data types handling.
1510 * Make MPI_Wtime another synchronization point to take computations into
1512 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1513 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1514 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1515 (no privatization of global variables yet)
1516 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1519 * Verification of liveness properties is now available for SMPI applications
1520 (in addition to MSG applications)
1521 * Bugged examples using SMPI in examples/smpi/mc/
1522 * Add --cfg=model-check/visited option. Allows the verification of infinite
1523 programs. Detection of loops in the execution thanks to the system state
1524 comparison and reduction of the state space to explore. Can be combined with
1525 DPOR for safety properties.
1528 * Allow to change SimGrid configuration (see --help) within the code
1529 thanks to SD_config() as it can be done in MSG.
1530 * Add a new function SD_task_set_amount() upon user request.
1533 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1534 * Remove rule based routing (no more PCRE dependency)
1535 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1536 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1537 * Add a loopback_bw and loopback_lat options to cluster tag.
1538 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1539 routing. Now peers are encapsulated in an AS and have their own private
1540 router but this is transparent.
1543 * Our own implementation of getline is renamed xbt_getline, and gets
1544 used even if the OS provide a getline(). This should reduce the
1545 configuration complexity by using the same code on all platforms.
1546 * New type: xbt_cfg_elm_boolean.
1547 * Allow to use yes/no for boolean configuration options in the command line.
1548 * Allow to disable SimGrid cleanups at exit from command line option.
1549 There are situations where one may want a simulation to end with an exit.
1550 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1551 annoying when scripting around the simulator. Adding a
1552 --cfg=clean_atexit:no allows to circumvent this issue.
1555 * Lots of memory leaks were corrected in this release.
1556 There is no known memory leaks anymore, in all of our 600+ tests.
1557 * New command line option --version, to get SimGrid version information.
1558 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1560 * Supernovae builds are deprecated, and expected to be removed in the next
1563 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1565 SimGrid (3.9) stable; urgency=low
1567 The Grasgory release: GRAS is really dead now.
1569 * Complete overhaul of the internal host structures scheme.
1572 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1573 was considered as experimental and badly maintained since then.
1574 * Keeping it was thus a trap to our potential users, that could take
1575 it instead of MSG or SMPI by mistake despite is pity state.
1576 * GRAS seems to have very few users (if any), and no one volunteered
1577 to maintain it further. It also induces a lot of XBT code (for
1578 portability sake), that must be maintained too.
1579 * For all these reasons, we killed GRAS. If someone wants to revive it
1580 in the future, don't cry, our git history still remembers of GRAS.
1583 * Major overhaul. Merge our documentation again as time proved that
1584 splitting it was really not helping our users.
1585 * Further improve the developer documentation to help newcomers
1586 hacking on SimGrid itself. The user documentation (and in
1587 particular, the beginner documentation) is still in a sorry state.
1590 * Now works on Windows too!
1591 * Much more extensive test suite, from MPICH
1594 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1595 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1596 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1597 * Bug fix: let task be scheduled when the last dependency to be solved is
1598 a control dependency.
1599 * Remove SD_load_environment_script function.
1600 Use the C sg_platf function if you want to declare a platform
1604 * New function: MSG_process_get_number()
1605 * Old function documented: MSG_config()
1606 * Remove MSG_load_platform_script function
1607 Use the C sg_platf function if you want to declare a platform
1611 * Change the default value of the TCP_gamma constant (maximal size of TCP
1612 congestion window) to a more realistic 4MiB value. If you notice changes in
1613 your simulation results, you can fall back to the previous 20k tiny window
1614 by adding --cfg=network/TCP_gamma:20000 on command line.
1615 * (Hopefully) fix a bug wrt periodic availability and state traces
1616 * Bug fix: use default values at start when first event in availability/state
1617 trace is not at time 0.
1620 * remove the "new_" part of function name sg_platf_new_trace_connect
1621 (resulting in sg_platf_trace_connect), since it does not create
1625 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1626 use case where it's really mandatory, and maintaining it was a pain in
1628 * New: xbt_fifo_search(), search an item with a user-provided
1629 comparison function instead of dumb pointer comparison.
1632 * Fix the lua deployment:
1633 Use `simgrid.init_application()` before deployment instead of
1634 `simgrid.msg_register_application()` after.
1637 * Transfer the tracing files into the corresponding modules.
1639 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1641 SimGrid (3.8.1) stable; urgency=low
1643 The "we are told that some people want to also *install* the simgrid
1646 * Add missing file "tesh.1" to the archive.
1648 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1650 SimGrid (3.8) stable; urgency=low
1652 The Psssshiiiit release: SimGrid jumps into the Cloud.
1655 * Add an experimental interface to manipulate VMs. They are mainly
1656 process groups with very few intrinsic semantic, but they should
1657 allow you to build the semantic you want easily.
1658 * New function: MSG_host_set_property_value()
1659 * New function: MSG_process_on_exit(). To clean memory in all cases.
1660 * Bug fixes that made the host (and link) failures unusable.
1661 * Add a way to auto-restart process when the host in which they are
1662 executing comes back (ON_FAILURE="RESTART" on deployment file,
1663 MSG_process_auto_restart_set).
1664 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1665 please stop using the old ones, they are DEPRECATED.
1667 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1668 Please use MSG_init() instead. (reducing the amount of entry
1669 points in the library helps us).
1670 * Make it impossible to link against the wrong version of the lib
1671 * Deprecate MSG_clean(). No need to call it anymore.
1672 * Function MSG_get_host_number() is not deprecated anymore.
1675 * Split the doc into a user guide and a reference guide.
1676 * Start a developper guide to help people hacking on SimGrid.
1679 * Enable tracing by default. This modules rocks you should use it.
1680 * Remove option custom_flags. Now use environment variables CFLAGS
1682 * Use default cmake things to detect lua instead of home grown ones.
1683 * New option "enable_mallocators" to disable mallocators, for debugging
1684 purpose ("on" by default).
1687 * Bug fixes around the resource failures: don't let the processes
1688 survive the host they are running onto.
1689 * Add an interface to auto-restart processes when the host in which they are
1690 executing comes back.
1691 * Ensures that SIMIX_clean is called automatically. It's not part of
1692 the public interface anymore (bindings should be updated).
1695 * Bug fix for when SD_Simulate is called with a positive value: be careful
1696 when comparing doubles. Sometimes they are different for non significant
1698 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1699 parallel task whose initial work is distributed among host according
1700 to the Amdahl's law. Such tasks are created with a parameter alpha
1701 that corresponds to the non-parallelizable part of the computation.
1702 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1703 between two sets of workstations assuming a 1D block distribution (each
1704 workstation owns a similar share of data) on both sides.
1706 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1707 Data redistribution will be automatically scheduled once parent and child
1708 are both scheduled. The filling of computation_amount and
1709 communication_amount structures is now done seamlessly thanks to the chosen
1711 * New function SD_workstation_dump to display various information
1712 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1713 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1714 bandwidth on the route onto which the task is scheduled and the amount of
1716 To divide the nominal bandwidth by 2, the rate then has to be:
1717 rate = bandwidth/(2*amount)
1718 * Compute tasks that have failed can now be rescheduled and executed again
1719 (from their beginning)
1720 * Increasing source code coverage (src/simdag is now covered at 95.8%
1724 * Re-implement time-independent trace replay using SMPI (at the
1725 smpi_smp_* level) instead of MSG. This should replace
1726 examples/msg/actions/actions.c
1727 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1728 hindexed and structs)
1729 * Implement the exchange of non-contiguous data.
1730 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1731 * Correct behavior of smpi/sender_gap and set its default value to 0
1732 * Add option to asynchronously send small messages to allow better
1733 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1734 specifies the size in bytes under which messages will be asynchronously sent.
1735 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1736 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1739 * New C interface to define a platform: XML is now optional.
1740 For more info, please check include/simgrid/platf.h.
1741 * New interface to define random platforms from the C:
1742 For more info, please check include/simgrid/platf_generator.h and
1743 examples/msg/masterslave/masterslave_platfgen.c
1744 * Export a sg_cmdline dynar containing all the arguments we got from
1748 * Two new tracing options for adding comments to trace file so you
1749 can track your experiments (see --help-tracing for details).
1750 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1751 * Adding the SimGrid version that generated the trace file as a comment.
1752 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1753 * Fix to avoid key clashes on Paje links
1754 * Other minor fixes related to the Paje specification
1757 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1758 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1759 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1760 * Greatly improve the robustness of mmalloc to user errors (such as
1761 using an area after freeing it, or freeing it twice)
1763 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1765 SimGrid-java (3.8.1) stable; urgency=low
1767 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1768 Warning: all methods are visible, but only some of them are
1769 implemented so far. Check the source (src/jtrace.c)
1770 for further information.
1771 * New module: org.simgrid.msg.File (SimGrid File management functions)
1772 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1773 * Change the meaning of Process.restart: now restart the process from
1774 the beginning, like MSG_process_restart in C.
1775 * Add Process.setAutoRestart: handling of process restart when failed
1777 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1778 you to retrieve the properties of the processes/hosts
1779 * Deprecate Msg.clean(): you can just forget about it now.
1780 * New function Process.getCount(), that only works when compiling
1781 with the not yet released version 3.9 of the C library.
1783 * New context factory based on Coroutines. It mandates a modified JVM
1784 but then, the simulations run about five times faster, and there is
1785 no limit to the amount of processes (beside of the available memory).
1787 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1789 SimGrid (3.7.1) stable; urgency=low
1792 * Restore the prototype of MSG_process_create_with_environment() to
1793 the pre-3.7 situation by removing the kill_time argument.
1794 * Add a MSG_process_set_kill_time() function instead.
1797 * Fix weird behaviors when dealing with parallel tasks.
1800 * Simgrid is now built as a dll.
1801 * Simgrid-java now works on Windows.
1802 * Simgrid-Java is now included into Windows package.
1805 * First pre-build package for MacOSX.
1808 * Fix compilation when using MSG_USE_DEPRECATED.
1809 * Fix some compilation issues on Macs and Windows.
1810 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1811 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1813 We discovered that the Lua console is broken, but we are missing the
1814 manpower to fix it right now. The problem existed in 3.7 too, so we
1815 are not blocking the release for that. Sorry if you depended on this
1816 feature, any help would be really welcome.
1818 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1821 SimGrid-java (3.7.1) stable; urgency=low
1823 The "Java aint got to be bloated and slow" release
1826 * Various internal cleanups and performance improvement
1827 Simulations are expected to run up to twice faster or so
1828 * Make Process.kill(process) an instance method, not a static one
1829 * User processes are not java.lang.Thread subclasses.
1830 This breaks the compatibility (sorry), but previous API was
1831 brain-dead, making it impossible to have non-trivial
1832 initializations in the process constructor.
1833 * Require a full constructor per Process sub-class.
1834 Kinda breaks the compatibility (sorry), but this allows a much more
1835 efficient way to launch the processes at simulation startup.
1836 * Do not embeed our version of semaphores, java 1.5 can be considered
1837 as sufficiently prevalent for us to not dupplicate its features.
1841 * Add examples for almost every part of the API
1842 We spotted and fixed a lot of bugs in the process
1844 * New module: asynchronous communication API
1845 * New function: Process.sleep()
1846 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1847 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1848 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1850 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1852 SimGrid (3.7) stable; urgency=low
1854 The "spring cleanups (before the next Big Project kicks in)" release.
1857 * We can specify the SMPI latency/bandwidth factor with command line
1858 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1859 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1860 You can also use the "config tag" from platform file by adding this line
1861 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1862 (see "example/platforms/tag_config.xml" to use "config tag").
1863 Note that the command line supersedes the platform file configuration.
1864 * Change the correction factors used in LMM model, according to
1865 the latest experiments described in INRIA RR-7821.
1866 Accuracy should be improved this way.
1867 * Use the partial invalidation optimization by default for the
1868 network too. Should produce the exact same results, only faster.
1869 * Major cleanup in surf to merge models and split some optimization
1870 mechanisms from the core of the models. As a result you can now
1871 specify which model to use (e.g., --cfg=network/model:LV08
1872 --cfg=cpu/model:Cas01) and which optimization mode to use
1873 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1874 Incompatible combinations should err at initialization. See
1875 --help-models for the list of all models and optimization modes.
1876 * The CLM03 workstation model was dropped for simplicity because it
1877 used the deprecated CM02 network model. Use default instead.
1878 * Rename the TCP_gamma configuration option to network/TCP_gamma
1879 * Rename the coordinates configuration option to
1880 network/coordinates, and document it
1881 * Use now crosstraffic keyword instead of the terribly misleading
1882 fullduplex keyword. It is activated by default now in the current
1883 default model, use --cfg=network/crosstraffic:0 to turn it off.
1884 * Ongoing refactoring the model parsing to make XML files optional
1885 See include/simgrid/platf.h for details (still to be completed)
1888 * Major overhaul of the documentation. Almost instructive now :/
1889 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1890 functions and friends. This interface was considered as
1891 deprecated since over 2 years, it's time to inform our users that it is.
1892 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1893 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1894 use these (crappy) functions in your code.
1895 These functions will be removed soon. Stop using them now.
1896 * Deprecate MSG_get_host_{table,number}
1897 Implement MSG_hosts_as_dynar() instead.
1898 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1899 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1902 * Stabilize the parallel execution mode of user contexts
1903 * Introduce configuration variables to control parallel execution:
1904 - contexts/synchro: Synchronization mode to use when running
1905 contexts in parallel (either futex, posix or busy_wait)
1906 - contexts/parallel_threshold: Minimal number of user contexts
1907 that must be part of a scheduling round to switch to parallel
1908 execution mode (raw contexts only)
1909 * Fix bugs that prevented to use suspend/resume along with
1910 synchronization structures.
1911 * Fix bugs in process termination that lead to invalid memory access
1912 in very specific conditions.
1915 * Introduce a parallel mode for the models (controlled by surf/nthreads
1916 configuration item). In our tests, running the models in parallel
1917 never lead to any speedups because they are so fast that the gain
1918 of computing each model in parallel does not amortizes the
1919 synchronization costs, even when ultra fast futexes are used.
1920 This is released anyway because YMMV.
1923 * Performance boost by using a swag internally to compute the set of
1924 tasks that are finished and should constitute the return value of
1928 * Enable it by default now that it is considered rather stable.
1931 * Documentation of the tracing functions.
1932 * Performance gains when tracing categorized/uncategorized resource
1933 utilization by avoiding calls to get route when updating resource
1934 variables. LMM constraints are being used instead.
1935 * API changed to set task categories. Use MSG_task_set_category instead
1936 of TRACE_msg_set_task_category, and SD_task_set_category instead
1937 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1938 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1939 * Examples for link user variables added (see at examples/msg/tracing/)
1940 * Deprecated function TRACE_msg_set_process_category completely removed
1941 * Trace header updated according to the latest Paje file format
1942 * Tracing network lazy updates, no longer obligate users to use full updates
1943 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1944 * Experimental: let user code declare/set/push/pop application states for hosts
1945 * API changed to allow the manual creation of graph configuration files
1946 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1949 * Improve the API of Lua MSG bindings, using the Lua spirit.
1950 * Each simulated process now lives in its own Lua world (globals are
1951 automatically duplicated). It helps writing simulators. Will allow
1952 to run Splay programs within SimGrid in the future.
1953 * Add a Chord example in Lua, equivalent to the MSG one.
1956 * Start the implementation of a solution to express temporal
1957 properties, not only local assertions. This is still an
1958 experimental work in progress, stay clear from it to be safe.
1962 - Add new runtime parameters --help-logs and --help-log-categories
1963 to display information about supported logging parameters and
1965 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1966 don't exists anymore.
1967 * Mallocators: allow value NULL for the reset function.
1969 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1970 create homogeneous dictionaries, where all the elements share the
1971 same free function. Non homogeneous dictionaries will be
1972 deprecated in the next release.
1973 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1974 - Multi-level dictionaries are deprecated.
1976 - new function xbt_dynar_search_or_negative() that is useful when
1977 you have less than 2 million elements in your dynar and don't
1978 want of the extra complexity of catching exceptions when the
1979 element is not found.
1981 - Make xbt_os_thread module (for thread portability) public.
1982 Documentation is still to come, sorry.
1984 - Cleanups and simplifications to make it maintainable again.
1985 - Exotic features (such as memalign and valloc) were removed.
1986 - The metadata were extended and improved so that the
1987 model-checker becomes able to explore and inspect the heaps.
1988 - This may induce a performance drop when enable_model-checking is
1989 ON in cmake (even if it's not used in the simulation), but it is
1990 necessary at this point to get MC working.
1992 Turn model-checking OFF if simulation performance matters to you.
1993 Not enabling it at runtime is not enough, disable it in cmake.
1995 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1998 SimGrid (3.6.2) stable; urgency=low
2000 The "Not coding new stuff allows to polish old things" release.
2003 * New bindings to the NS3 packet level simulator (experimental)
2004 * Use the raw (efficient) execution contexts instead of the sysv
2005 (portable) ones when possible.
2006 * libpcre is now mandatory in any cases since not using it led to
2007 severe performance loss and possibly other issues
2008 * Update the XML platforms:
2009 - G5K: include the latest machine in Nancy
2010 - GridPP and LCG: new platforms
2011 * Documentation was partially updated, at least (more to come)
2013 Bug fixes, cosmetics and small improvements
2014 * Free terminated processes before the end of the simulation to avoid
2015 exhausting the memory of users having very dynamic amount of
2017 * Bug fix and cosmetics about canceling non-running tasks
2018 * Bug fix about the dot loader's issues when using libcgraph
2021 * Create an installer for windows with nsis (amd64 and win32)
2022 - Add an hello world project to illustrate simgrid project creation.
2023 - Embed libpcre into the Simgrid installer to avoid
2024 its compilation burden
2025 * The raw execution contexts should work on Apple now
2026 * Port to Windows 64 bits
2027 - Sysv contexts now have an implementation for this arch
2028 - GRAS communication features now support this arch
2029 * Drop support for borland compiler on windows
2030 - this code was not maintained, and we kinda depend on gcc nowadays
2031 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2032 * Fix portability issue on unstable ubuntu: linker became picky on
2035 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2037 SimGrid (3.6.1) stable; urgency=low
2039 The "Oops, we broke Macs too" release
2042 * Fixed contexts detection so that raw ones are used when possible
2043 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2044 leads to a strange error, with user code segfaulting sometimes when
2045 the generated code is not perfectly aligned (which is not
2046 controllable from the user side, depends on the amount of code)
2049 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2051 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2053 SimGrid-java (3.6) unstable; urgency=low
2056 * Split of every thing from simgrid v3.5 into a separate package.
2058 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2060 SimGrid (3.6) stable; urgency=medium
2062 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2065 * Bindings now constitute their own package, separated from the main one.
2066 Rationale: reduce our maintenance nightmare by reducing the module coupling
2067 They will soon be released on their own on gforge.
2069 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2070 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2072 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2073 * It's not quite deprecated for now because we have no replacement,
2074 but it may soon become the case.
2077 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2078 * Fortran: New user-level cache variable to store the rank of the running
2079 process. This improves performance by an order of magnitude.
2080 * C: New coccinelle script to automatically locate and modify global and
2081 local static variables.
2082 * Improved SMPI network model with a sender-side gap to account for multiple
2086 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2087 only say if a communication is finished, no matter whether it succeeded or
2088 failed. You can call MSG_comm_get_status() to know the status of a finished
2090 * New function MSG_task_dsend() to send a task and detach it. When a
2091 communication is detached, you are never notified of its success or failure
2092 and the memory is released automatically once it is finished. This function
2093 is useful when you don't care about the end nor the success of a
2095 * Change the prototypes of action replay. Sorry for inconvenience,
2096 but this is really more efficient this way (and to adapt your code,
2097 you just have to fix the initialization, that shouldn't be too long)
2098 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2099 ever managed to do anything useful with it.
2100 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2101 takes a dynar of communications. It returns immediately and gives the
2102 index of a finished communication (if any).
2103 * New example: a basic implementation of the Chord P2P algorithm.
2106 * New model for multi-core CPUs. You can now use the core attribute to
2107 precise the number of cores of a host. This is a basic model. Every
2108 process running on the host receives at most the power provided in
2109 the DTD (throughput<=power). Total throughput of process cannot exceed
2111 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2112 router linked by an up-link and a down-link (possibly asymmetrical).
2113 This kind of pattern allows to easily build last-mile model style platforms.
2114 Aggregating such patterns in a rule-based AS is thus the technique of
2115 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2116 * New model for Vivaldi routing. We transformed the Vivaldi network model
2117 into a Vivaldi routing model (based on the rule-based model). This allows to
2118 combine Vivaldi based latencies with last-mile platforms.
2121 * Added a check for NaN of IEEE754 infinite in the double entries of
2123 * Introduce a new context factory "raw", highly inspirated from the
2124 ucontext factory, but using manually crafted functions in assembly to
2125 do the work in an efficient manner.
2126 * Allow to change the used context factory at run time, not only at
2127 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2128 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2129 contexts at runtime (only with raw contexts or ucontexts).
2130 * Completely rewrote this module to allow parallel execution of user
2131 processes. Use --cfg=contexts/nthreads:N to execute user processes
2132 with N parallel threads (the default is 1, meaning no parallelism).
2133 * Allow to decide dynamically between sequential and parallel modes.
2134 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2135 processes in parallel only when their number is greater than or equal to P
2137 * Added a check for NaN of IEEE754 infinite in the double entries of
2141 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2142 won't output the state of processes when interrupted with Ctrl-C
2143 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2144 NULL-terminated array. This may solve backward compatibility issues
2145 due to the change to return type of SD_simulate. See also:
2146 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2147 * Add new macros with variable number of arguments.
2148 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2149 - in xbt/asserts.h: xbt_assert
2150 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2151 - in xbt/ex.h: THROWF and RETHROWF.
2152 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2154 * Change xbt_die() to accept a format string with arguments, just like printf.
2155 * New data structure: xbt_lib_t, like a dict but more general and with better
2159 * New configuration options
2160 Options triva/categorized and triva/uncategorized can be used to generate
2161 graph configuration files for Triva visualization tool.
2162 * Configuration option tracing/platform is renamed to tracing/categorized
2163 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2164 * New TRACE_user_link_variable interface:
2165 User provides the name of the link and the tracing variable to attach to it
2166 * the declaration of tracing categories must be done after the environment
2168 * simpler tracing interface, just one way to declare categories
2169 TRACE_category or TRACE_category_with_color, it is up to you
2170 * links in the trace file are again identified by their names
2171 * trace contains the full platform hierarchy exactly as declared using the ASes
2172 * Options tracing/msg/[task|process]:1 groups the process by hosts
2173 for both cases, tasks and processes must have names that are unique during the simulation
2174 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2175 * The experimental option tracing/msg/volume is deprecated
2176 its functionality may be reincorporated if needed
2178 The tracing generates a trace file with unordered timestamped events,
2179 because of the way the core simulator (surf) works. A script available
2180 at the tools directory (fix-paje-trace.sh) can be used to put the events
2181 in order. We have changed the tracing so it can generate ordered timestamped
2182 events in the final trace, but depending on the simulator (and how much time
2183 is simulated) that can lead to a huge memory utilization. It is deactivated
2184 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2186 Build Infrastructure
2187 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2188 - We are trying hard to keep the API stable, but it may happen that
2189 some things change (we're a research project after all, not a
2190 nuclear plant operating system). If such things should happen, you
2191 could rely on that macro to adapt.
2192 - current value: 30600 for 3.06.00, aka 3.6
2193 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2194 a number that can be compared with SIMGRID_VERSION.
2195 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2196 assertions and verbose/debug logging events are disabled at compile time.
2198 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2200 SimGrid (3.5) stable; urgency=medium
2203 * New feature to any SimGrid-based simulator: Model-Checking
2204 Check SIN#1 for more details.
2207 * New Model SMPI (three-interval linear regression for correction factors)
2208 See RR-7426, available at http://hal.inria.fr/inria-00527150
2209 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2210 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2211 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2212 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2213 * See also src/smpi/README
2218 - Tracing API changes: TRACE_start and TRACE_end should not be called
2219 by user-code. They are automatically called by simulators created
2220 with SimDAG, MSG and SMPI if the toolkit is compiled with
2221 tracing_enabled=ON. Categories declaration and utilization remain the
2222 same for MSG and SimDag.
2223 - A function was added to the tracing API to declare categories with
2225 - TRACE_category_with_color (char *category, char *color)
2226 where color must be in the following format
2227 "%f %f %f", red, green, blue
2228 and red, green, blue are float values in the interval [0, 1]
2229 - User can specify NULL as color parameter, or continue calling
2230 TRACE_category (cat)
2231 On that case, the tracing system will define random colors
2232 - The following command-line options are supported:
2233 --cfg=tracing/filename:msg.trace
2234 --cfg=tracing:1 (activate tracing, needed to use others)
2235 --cfg=tracing/platform:1 (categorized resource use)
2236 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2237 --cfg=tracing/msg/task:1 (task creation)
2238 --cfg=tracing/msg/process:1 (process creation, migration)
2239 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2240 --cfg=tracing/smpi:1 (SMPI interface tracing)
2241 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2242 - examples of examples/msg/tracing updated
2244 - DAXLoader and DOTLoader functions can generate tasks with categories
2245 - A new function to attribute a category to SD tasks:
2246 TRACE_sd_set_task_category (SD_task_t task, char *category)
2247 * Tracing the MPI interface implemented by SMPI
2248 - Collective operations are traced with states
2249 - Point-to-Point operations are traced with states/links
2250 - Tracing activated by a parameter "-trace filename" passed
2251 to smpirun during execution (considering that simgrid
2252 is compiled with tracing enabled)
2253 - To run the simulation with gdb, the simulator
2254 accepts --cfg=tracing/smpi:1 to trace SMPI
2255 - tesh files to check if smpi tracing is ok
2256 - See examples/smpi/NAS/DT-trace
2257 * GTNetS tracing re-worked
2258 - adaptation to the tracing system of GTNets to cope
2259 with modifications regarding the fullduplex mode
2260 - new tesh files to check if gtnets tracing is ok
2263 * Asynchronous communications through the functions:
2264 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2265 * New function: MSG_load_platform_script()
2266 to make possible using a lua script instead of XML files to set up platforms
2267 * New function: MSG_set_function
2268 to associate functions to processes, used when bypassing the parser
2269 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2271 Platforms: Add some more examples in examples/platforms
2272 * Grid'5000: see www.grid5000.fr
2273 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2276 * Change the XML format. This is a very important modification. SimGrid
2277 3.5 introduces a new hierarchical format based on the notion of
2278 Autonomous Systems. Compatibility with old format is ensured
2279 through the perl script provided in the install bin directory
2280 bin/simgrid_update_xml.
2281 It is now possible to build platforms with specific routing
2282 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2283 connect several platforms together. We will try to provide soon
2284 set of realistic platforms exploiting these properties (have a
2285 look at examples/platforms/ for the moment).
2286 * Take the opportunity of the XML format change to be a good XML citizen:
2287 rename link:ctn to link_ctn and similar changes (also dealt with by
2289 * Add a new routing scheme (rule-based) using regular expressions. It
2290 enables to have an extremely low memory footprint when the
2291 underlying routing is simple and can be compactly described. You
2292 need to have libpcre4-dev (perl regular expressions) installed if
2293 you want to use this routing scheme.
2294 * Revive the cluster TAG and allow to easily and efficiently (both in
2295 term of memory and speed) connect clusters together. Have a look
2296 at teshsuite/simdag/platforms/ to see how this can be done. With
2297 this tag, you can create clusters with thousands of tasks at no
2298 cost (have a look at examples/platforms/).
2299 Note: clusters are implemented as ASes, so there is no need for an
2300 enclosing AS tag if you have only one cluster in your platform.
2301 * Add new generic functions in the public interface that allows the user
2302 to call SURF 'create_resource' methods from your code (same
2303 functionality as the XML bypass mechanism but with a much lighter
2305 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2306 that uses a piecewise linear approximation to produce better
2307 results when exchanging small messages.
2308 * Add a new parameter to handle correctly full duplex link and account
2309 for interferences between uplink and downlink communications
2310 (activate with --cfg=fullduplex:1).
2313 * Rename the SD_READY (all dependencies are satisfied and task is
2314 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2315 dependencies are satisfied) state.
2316 This prevents a confusion between the notion of "ready to schedule"
2317 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2318 simulated" (SD_RUNNABLE) used by the simulation kernel.
2319 * Change the way a task is considered as ready. Instead of removing
2320 dependencies when a task is done, a counter is decreased. This way,
2321 it is always possible to reach ancestors thanks to the
2322 SD_taks_get_parents function (even after the end of the simulation.)
2323 * Change the return type of SD_Simulate from (SD_task_t*) into
2324 xbt_dynar_t. This function was in handling a dynar internally and
2325 converted it into a NULL terminated array for historical reasons.
2326 * New function SD_dotload(char*) to load a DAG described in dot
2327 format. This loader and the corresponding examples require the
2328 installation of the graphviz library.
2329 * Fix a bug in the management of tasks of size 0 in the surf network
2330 models. This problem was only visible with SIMDAG and you should
2331 thus disregard results produced with earlier versions if you
2332 relied on this feature (some tasks were blocked because of this).
2333 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2334 even though your applications were doing only point-to-point
2335 communications and sequential computations. Now you can really use any
2336 model you want (of course, if you create real parallel tasks, which are
2337 not implemented in most models beside ptaskL07, this will abort).
2338 * Add an example that schedules a DAX on an heterogeneous platform
2339 using a Min-Min strategy.
2340 * New function SD_workstation_get_current_task() that returns the kind
2341 of task currently running on a workstation in the sequential access
2343 * Raise some warnings when unexecuted tasks remains at the end of the
2344 simulation. This is usually caused by cycles in the DAG.
2347 * New function: SIMIX_process_set_function() called by MSG_set_function
2348 * Change the underlying waiting queue in semaphores so that a process
2349 can wait on several of them simultaneously (as in waitany).
2350 * Fix the way to handle tokens in semaphores so that all access patterns
2351 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2352 * kill the dirty pimple SIMIX_message_sizes_output()
2353 Please use (proper) visualization instead
2356 * New data container: setset (set of sets of elements)
2357 * New module: mmalloc (mapped malloc, allowing to have several
2358 independent segments of malloc)
2359 * New function: xbt_dict_cursor_set_data()
2360 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2361 * New function: xbt_dynar_is_empty()
2362 * New function: xbt_fifo_get_last_item()
2363 * Fix xbt_dynar_shrink(): use the right element size.
2364 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2365 uninitialized areas during expand.
2366 * Fix semaphores: previous implementation was severely broken.
2367 * Use library init/fini functions for our initialization.
2368 - you can use logs and other feature as soon as you want in your
2369 code (even before the xbt_init / MSG_init)
2370 - xbt_exit is now a no-op and produce a warning when used.
2373 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2374 benefit from the latest improvement to the simulation kernel.
2375 * Kill measurement sockets for now. If you rely on them, sorry. This
2376 release is not for you. This feature will be reintroduced in the
2377 future, but we cannot delay the release any further.
2378 * New function: gras_msgtype_get_name().
2379 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2380 * Fix (at last) the pmm example: it should not randomly fail anymore.
2382 Build chain: bug fixes and overall polishing
2383 * Cmake is now stable enough. Hence, we killed the autotools.
2384 * Port to windows ( TM :)
2385 * Fix the 'make install' target.
2386 No need to use 'make install-simgrid' anymore
2387 * Introduce a 'make dist' target compiling a *source* archive
2388 'make package' compiles a binary archive
2389 * Compile java files only on need
2390 * Add --cd and --setenv command line options to tesh
2391 * Out of source builds are not fully supported yet, but we are close
2392 * Enable supernovae and optimization flags by default for our users
2395 * Add layer to set up environment directly from lua, without XML.
2396 * The effect of gras_stub_generator can be achieved through
2397 lua too (check examples/gras/console/ping_generator.lua)
2399 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2401 SimGrid (3.4.1) stable; urgency=low
2403 The "Polishing easter eggs is probably a good idea" release.
2404 This is a bug fixes release only.
2408 * Fix a bug preventing the tasks from begin garbage collected.
2411 * Fix a bug occurring when a host involved in a communication fails.
2412 This was not detected properly by the other peer involved in the
2413 communication. Now, it's reported as a network error.
2416 * Warn the user about loop dependencies in data flow of DAX files
2417 * Obey the control-flow dependencies of DAX files
2420 * Add option "enable_smpi" allowing to not compile SMPI.
2421 Probably useful for the (Mac) users experiencing a build error here
2422 * Improve the detection of lua5.1 and ruby1.8
2424 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2426 SimGrid (3.4) stable; urgency=low
2428 The "Easter in Cargese" release. Also known as (major changes):
2430 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2431 ~> bindings were greatly improved
2432 ~> new tracing infrastructure for better visualization introduced
2434 * the "Welcome to configury modernity" release.
2435 ~> we switched from autotools to cmake, and improved our cdash
2438 A more detailled list of changes follow (full detail in svn log).
2440 Java Bindings: Various Cleanups
2441 * (install java-gcj-compat-dev on debian-like to use them)
2442 * Remove put/get: no need to export deprecated interface in Java
2443 Use send/receive instead.
2444 * Cleanup the examples and add a README per directory
2445 * Remove example autoDestination (that's the only way to go now)
2446 * Remove example explicitDestination (was a plain copy of basic)
2447 * Make JniException a runtime exception, so that there is no need to
2448 declare the fact that you may encounter such a beast. I guess that
2449 nobody will ever want to survive such error.
2450 * Create specific errors for each MSG case of failure:
2451 host failure, transfer failure, timeout, task cancelled
2452 * Cleanup the exceptions that may get thrown by each function
2453 * Other internal cleanups in Java bindings. Performance still bad :/
2454 Ruby and Lua Bindings: create them
2455 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2456 * That's new and great, you should try them out.
2457 Same functionalities than Java bindings, only even less polished
2459 * Kill the useless "rate" argument of SD_task_get_execution_time()
2460 Everyone used to provide -1 as a value, it was not used, and the
2461 semantic of a possible use wasn't even clear.
2462 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2463 either as comm costs or compute costs to mean that there is no
2464 such thing for that specific task.
2465 * Add a SD_task_set_name() function
2466 * Fix SD_task_unschedule() on typed tasks
2467 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2468 * In DAX loader, accept useless 'level' attributes to <job> since
2469 LIGO DAGs have them (seem to be to ease graphical representation).
2471 * Add an example masterslave_mailbox.c using send/receive and not
2472 the deprecated put/get interface.
2473 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2474 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2476 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2477 (declare MSG_USE_DEPRECATED to still have the old name)
2478 * Add a MSG_task_set_data() function
2479 * About trace replay (see examples/msg/actions):
2481 - Allow to work with splitted trace files for each process
2482 Give the specific trace file as argument of each process,
2483 and call MSG_action_trace_run(NULL)
2484 You can still have one merged file for all processes.
2485 - Fix implementation of collective operations
2486 * Allow task_execute() on 0-sized tasks (closes #10063)
2488 * This is the first release of SimGrid where SMPI is not considered
2489 beta anymore (even if some corners should still be improved)
2490 * Port over the new SIMIX_network submodule (internal refactoring)
2491 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2492 * Implement more missing elements of the standard:
2494 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2496 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2497 MPI_Scan() MPI_Get_processor_name()
2498 - Added implementation of missing case for Alltoall (warning: it's
2499 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2500 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2501 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2502 SMPI_MPI_Allgatherv()
2503 * Bug fixes include:
2504 - MPI_Waitsome() was broken
2505 - Allow relative includes in smpicc
2506 - Command line cfg argument 'reference_speed' was ignored...
2507 - Some functions did not properly lead to auto-benching of user code
2508 - smpicc passes -O2 by default (just like openmpi one)
2510 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2511 * Bug fixes about timeouts during communications
2512 * add SIMIX_message_sizes_output() as a pimple to write to file the
2513 amount of messages per size. Use gnuplot to get histogram.
2514 Pimple because that's the only user-visible function of simix,
2515 defined directly in xbt.h (irk, sorry)
2517 - Add a SIMIX_sem_get_capacity() function
2518 - Fix interactions with processe resume/suspende
2519 - release_forever() was stupidly broken
2520 - Fix SIMIX_display_process_status() for processes in a semaphore
2521 - Make SIMIX_sem_block_onto() user-visible
2522 * Refactoring context stuff:
2523 - Use pseudo-OOP for better modularity
2524 - reimplement SIMIX_process_kill() without process_schedule() so
2525 that the latter can take as invariant that it is called from
2527 - Merge context_start into context_new for sake of simplicity
2529 * Add a Vivaldi network model, coded live during SUD'10 ;)
2530 * Rename configuration variables to start a hierarchy:
2531 o cpu_model -> cpu/model
2532 o network_model -> network/model
2533 o workstation_model -> workstation/model
2534 * New configuration variables:
2535 o network/bandwidth_factor: correction to bandwidth
2536 o network/latency_factor: correction to latency
2537 o netwotk/weight_S: correction to the weight of competing streams
2538 * Add a long description to the models, that users can see with such
2539 argument on the command line: --cfg=cpu/model:help
2540 * --help-models display the long description of all known models
2542 * config: add the ability to set a default value after registration
2543 Does not override any previously set value (e.g. from cmd line)
2544 * dict: allow to have integer key and data.
2545 When so, you need to use the following functions
2546 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2547 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2548 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2549 In contrary to regular dicts, the key is not malloced before copy.
2550 Mixing scalar and regular elements in the same dict is not tested
2552 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2553 Tracing for Visualization:
2554 * SimGrid is now instrumented in order to generate a trace file for
2555 visualization analysis: to use it, need to compile SimGrid with the
2556 "tracing" option enabled, and instrument the program using SimGrid with
2557 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2558 (among other functions).
2559 * The instrumentation only traces the platform utilization for now
2560 * Documentation to use the tracing functions and how to analyze the
2561 traces with the Triva tool is written.
2562 * More information about: SimGrid FAQ (in the section Tracing Simulations
2565 * We moved to cmake as default build system. Autotools support will
2566 be dropped soon. Check the FAQ for more info about how to use it.
2567 * Greatly improved our cdash/ctest interactions
2568 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2569 * Added memory checking tests with valgrind; lot of memleak fixing.
2570 This may be the first release of simgrid with so few memory issues
2571 * Added code coverage tests.
2572 Our coverage is still improvable, but at least we see it on cdash.
2574 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2576 SimGrid (3.3.4) stable; urgency=low
2578 The "Desktop Grid needs love too" release (also called Xmas release).
2580 Models improvements:
2581 * Major speedup in the maxmin system solving by using lazy evaluation
2582 Instead of solving completely the maxmin system at each iteration,
2583 only invalidate (and recompute) the modified parts.
2584 This new feature is enabled in default models but you can try to
2585 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2586 * Cas01 IMproved as default CPU model
2587 This CPU model is the same Cas01 model, but it uses the
2588 maxmin-selective-update flag and a heap structure to manage
2589 actions on SURF kernel.
2590 It reduces the complexity to find the next action to finish and,
2591 consequently, it's faster than the old Cas01.
2592 This is the new default CPU model (Cas01).
2593 * Rename the old Cas01 model to Cas01_fullupdate
2594 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2595 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2596 * CpuTI (CPU Trace Integration)
2597 A new CPU model whose objective is simulate faster when using
2598 availability trace files.
2599 Instead of using a full featured, over engineered maxmin system for
2600 CPU modeling, this model does the pre-integration of traces files
2601 to calculate the amount of CPU power available, and so, executes
2602 faster than the old CPU models.
2603 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2604 * Use LV08 as default network model since it gives better accuracy
2605 for small messages and shouldn't change things for big ones.
2606 Use --cfg=network_model:CM02 to get the previous behavior.
2609 ******************************************
2610 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2611 ******************************************
2612 * The new CPU model may changes simulations!
2613 The point is that events occurring at the exact same timestamp
2614 are not scheduled in the same order with the old and new
2615 version. This may be enough to completely change the execution
2616 of simulations in some cases.
2617 * The new network model will change simulations!
2618 This new model is more realistic than the previous one, so you
2619 should consider redoing your old experiments with this model.
2620 Sorry for the inconvenience.
2623 * Introduce the supernovae compilation mode
2624 When compiled that way, the whole SimGrid (or almost) is put in a
2625 single compilation unit and compiled in one shoot.
2626 This is to help gcc which has difficulties to inline stuff from one
2628 The speedup seem to be above 15%, althrough more tests are needed on
2629 amd64 to confirm that gain.
2632 * Port of MSG's mailbox on top of SIMIX network
2633 The put/get mechanism was greatly simplified on the way.
2636 * New SIMIX network module. Provides:
2637 - Mailbox: rendez-vous mechanism to find with who you want to speak
2638 - Synchronous send/recv: easier and hopefully faster since the
2639 logic is handled in the maestro process directly now
2640 - Asynchronous send/recv: you dreamt of it? It's here now
2641 Too bad that nobody cared enough to propagate the change to MSG.
2642 * Add semaphores as SIMIX synchronization mechanism.
2645 * new function SD_daxload(char*) to load a DAX file
2646 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2647 * Introduce typed tasks. Specify its kind and cost at creation.
2648 At scheduling, just give where it should be placed, and the cost
2649 for each involved resource is automatically computed.
2650 Existing constructors so far (more to come of course):
2651 - SD_task_create_comm_e2e() for end-to-end communication
2652 - SD_task_create_comp_seq() for sequential computation
2653 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2654 * new function SD_task_dump() for debugging display
2655 * new function SD_task_dotty(task,FILE*) writing to file the info
2656 about the task in dotty format
2657 * SD_task_dependency_exists() can now cope with having one of its
2658 arguments NULL. If so, it tests whether the other argument has any
2660 * Add getters on list of preceding/following tasks:
2661 SD_task_get_parents(task) and SD_task_get_children(task)
2662 * Add getters on amount of workstations and list:
2663 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2664 * Add getter on task kind: SD_task_get_kind(task)
2665 * Update the start_time and finish_time of tasks on completion/failure
2666 * Bugfix: Remove task from state swags when destroyed
2669 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2672 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2673 dynar with one separate thread per value of the dynar.
2674 * Change the prototype of xbt_thread_create(), sorry.
2675 Added a boolean parameter indicating whether we want to join this
2676 thread (used in SG only for now)
2677 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2680 * GTNetS wrappers should now be usable again (and betterly tested too)
2681 * Fix a major regression from 3.2 where the timeout provided to
2682 MSG_task_put_with_timeout() was used as absolute time before which
2683 the comm should be done.
2684 * Start to fix the <cluster> tag.
2685 - Internal links should be good now (beside of the loopback, which
2686 use the private link instead)
2687 - paths to the external world is still rather broken
2688 - the <route:multi> tag is just broken. Actually that's brain-dead.
2689 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2690 to make it less stupid
2691 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2692 * Fix a source-level compatibility glitch from 3.2: after defining
2693 MSG_USE_DEPRECATED, you can use the old name
2694 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2695 * Allow to compile from the SVN with automake 1.11
2696 * Fix some problems when using the "start_time" tag in deployment XMLs.
2697 * Fix #8569: XBT/synchro.h has redundant declarations
2698 * Fix #8563: MSG return values and exceptions
2699 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2700 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2701 Thanks to Robson Peixoto.
2702 * Fix a vicious bug in dictionaries inducing that some elements were
2703 not freed on xbt_dict_free()
2705 Portability report of this version:
2706 * Main portability targets:
2707 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2708 on (amd64/i386/ia64)
2709 - mac leopard on i386
2710 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2711 but nothing critical.
2712 * Other platforms: windows, AIX and others were not tested for this release
2714 Timing report of this version:
2715 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2716 scenario parameters). From 8h to a few seconds in desktop grid settings.
2717 * Supernovae brings about 25% speedup on i386.
2719 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2721 SimGrid (3.3.3) stable; urgency=low
2723 The "Need for Speed" release.
2725 The timings done to validate the 3.3.2 were faulty.
2726 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2728 The problem was a conversion from a manually handled vector to
2729 xbt_dynar_t on the critical path.
2730 xbt_dynar_foreach calls functions, inducing stack management crap.
2732 We inlined these functions and xbt_dynar_foreach is now breath taking.
2733 We also inlined xbt_swag_belong on the way.
2735 Here are some approximate speedup measurements (on master/slaves
2736 simulations lasting between 10s and 20s each):
2737 3.3.1 -> 3.3.2: about same performance
2738 3.3.2 -> 3.3.3: 40% speedup
2739 3.3.1 -> 3.3.3: 40% speedup
2740 3.3.1 with inline patch -> 3.3.3: 30% speedup
2742 Our reading is that the refactoring which occurred in 3.3.2 made us
2743 suffer much more from the xbt_dynar_foreach low performance, but
2744 once we solved this, this refactoring proved to be very performance
2745 effective. From the 40% speedup, somehow, 10% are due to the
2746 inlining and 30% to the refactoring.
2748 That's a pitty that gcc cannot inline functions placed in other files
2749 alone. We have to choose between:
2750 - break the encapsulation (by putting private data structures and
2751 accessors in headers files to help gcc)
2752 - live with low performance
2753 - switch to a decent compiler such as icc (not quite possible).
2755 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2757 SimGrid (3.3.2) stable; urgency=low
2759 The "Simplicity does not preceed complexity, but follows it" release.
2761 The main contributors of this release were (lexical order):
2762 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2765 * Extract the routing logic into its own object.
2766 (was dupplicated in network.c and workstation_LV07.c;
2767 Allows to implement other ways of storing that info)
2768 => kill now useless network_card concept
2769 - Use dynar to represent routes (instead of void** + int*)
2770 - kill link_set (use surf_network_model->resource_set instead)
2771 - Add a command-line option to choose the routing schema to use
2772 - Add three new models:
2773 * Floyd (shortest path computed at initialization)
2774 * Dijikstra (shortest path recomputed all the time)
2775 * Cached Dijikstra (shortest path computed on need)
2776 All these models where contributed by Silas De Munck, and are
2777 described in his ICCS09 paper.
2779 * Simplify model declaration
2780 (less redirections, less function to write when defining a model)
2781 - Factorize stuff between models:
2784 surf_model_resource_set(model)
2785 surf_model_resource_by_name(model, name)
2786 - Unify the types of models in s_surf_model_t (using an union)
2787 - Embeed fields of common_public directly into s_surf_model_t
2788 - Rename model methods:
2789 action_free ~> action_unref
2790 action_change_state ~> action_state_set
2791 action_get_state ~> action_state_get
2792 - Change model methods into functions:
2793 (model)->common_public->action_use ~> surf_action_ref
2795 * Implement a generic resource; use it as ancestor to specific ones
2796 (allows to kill duplicated code in models)
2797 Drawback: timer command don't need no name nor properties;
2798 workstation_CLM03 don't need no properties
2799 (but I guess we can live with those few bytes wasted)
2801 * Improve the action object model
2802 - implement a constructor avoiding dupplicated code about field
2803 initialization in generic_action part.
2805 * Kill the SDP model: it has an external dependency, is deprecated
2806 in flavor of modern lmm models, and didn't compile since a while
2809 * Relocation of the context module from XBT to SIMIX.
2810 (the context were decoupled from the simix processes, duplicating a lot of code)
2811 => a lot of code was factorized
2812 - less overhead is introduced during scheduling
2813 - simpler API for the context factory
2814 - the logic for process creation,destruction and manipulation was simplified
2815 * Simplification of the s_smx_process_t data structure.
2816 => accesing the simix level data associated to a process is faster now,
2817 and the code is a lot more readable.
2820 * Implement some more MPI primitives:
2821 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2822 -implementation: Bcast: flat or 2-ary tree (default),
2823 Barrier: 4-ary tree,
2825 Allreduce: Reduce then Bcast
2826 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2827 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2828 Alltoallv: flat tree, like ompi
2830 * Add support for optimized collectives (Bcast is now binomial by default)
2831 * Port smpirun and smpicc to OS X
2834 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2835 Shout out if you used it.
2838 * Display the list of still queued messages in SG mode when existing
2842 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2843 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2844 * Add xbt_str_from_file(FILE*)
2845 * Add xbt_dict_get_key achieving a linear reverse search
2846 * Remove the context module
2848 Portability report of this version:
2849 * Main portability targets:
2850 - Linux(debian)/x86/context
2851 - Linux(debian)/x86/pthreads
2852 - Linux(debian)/amd64/context
2853 - Linux(debian)/amd64/pthreads
2854 On these, we still have the eratic breakages of gras/pmm and
2855 amok/saturate_sg reported in previous version. We still think
2856 that the tests are the cause of the fault, not the tested code.
2858 - Mac OSX Leopard/x86/context
2859 Still false negative in tesh autotesting.
2860 Smpi still fails, but this time because readlink does not accept -f
2861 Everything seems to work properly beside of that.
2864 - AIX version 5.3 (only tested contexts this time)
2865 Smpi still fails there because mktemp is not installed.
2866 Everything seems to work properly beside of that.
2868 I managed to compile it for the first time, but several breakages.
2869 Won't delay the release for this exotic platform.
2871 * Windows: it's still lagging behind. If you want to help, please
2874 Timing report of this version:
2875 This version seem to be more than 5% faster than 3.3.1 (on linux
2876 64bits with contextes). The gain is less than expected, we are
2877 investigating this for next release.
2879 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2881 SimGrid (3.3.1) stable; urgency=low
2884 * Implement a --cfg-help to show existing configuration variables
2885 * Build chain do not require doxygen in maintainer mode
2888 * fix a bug on struct sizeof computation, which prevented the
2889 exchange of arrays of structs in some conditions
2890 - added a regression test about this in datadesc_usage
2891 * Allow the exchange of 0-long dynamic vectors.
2892 - for that, use -1 as indicator of dynamic size instead of 0
2893 - This implied to change any size from unsigned long to long,
2894 reducing a bit communication abilities, but I guess that with
2895 64bits being quite common, this is more than enough.
2896 - This also induce a protocol change, thus bumping network protocol
2897 version from 0 to 1 (if we have external users, we have to get
2898 clean on that point too ;)
2899 - added two regression tests about this in datadesc_usage
2900 * Be more verbose when propagating local exceptions
2901 This helps debugging.
2902 * Display the status of simulated processes when receiving SIGINT in
2906 * Allow to control the simulation from a trace file.
2907 New functions MSG_action_register() and MSG_action_trace_run()
2908 The first one allows to associate a function execution to each
2909 kind of action while the second one parses a trace file and
2910 triggers the corresponding actions within the system.
2911 For now, only a toy example is provided in examples/msg/actions
2912 * Add an example of process migration in examples/msg/migration
2913 * Fix a bug in task exchange which broke MSG_task_get_sender()
2914 Add a teshsuite regression test for that.
2915 [Bug: if MSG_task_get_sender() is called after sender exit,
2917 * Fix a bug which prevented suspend/resume to work properly
2918 * Display the status of simulated processes when receiving SIGINT
2919 This fixes a regression of v3.3. due to the introduction of SIMIX
2920 * Bug fixing in failure management:
2921 - trace could not start by a failure at time 0
2922 - failure during communications were not working
2925 * Add SIMIX_process_set_name() to change the name of the current
2926 process in the log messages.
2927 * Store smx_hosts in a dict since we only retrieve them by name
2928 * Move the configuration infrastructure to surf
2931 * Move the configuration infrastructure to surf
2934 * Massive internal cleanups:
2935 - Store internal structures on processes instead of hosts (allows
2936 to have more than one process per host, in addition of being more
2938 - Cleanup the initialization/finalization process
2939 - Kill a whole bunch of unneeded synchronization:
2940 processes run in exclusive manner within the simulator
2941 - Move queues from global tables to process data fields
2943 - now accept -platform and -hostfile arguments
2944 - Pass the right rank value to processes according to the hostfile
2945 * Compile the examples by default, and use them as regression tests
2946 * Implement MPI_Wtime()
2947 * Change the reference speed to a command line option
2950 * TCP_gamma can now be specified as command line option using
2951 --cfg=TCP_gamma:10000000.0
2952 * Change the --surf-path cmd line option into --cfg=path:
2955 * Also include strbuff from xbt.h public header
2956 * xbt_ex_display(): do not free the exception after displaying
2957 This allows to do more with the given exception afterward.
2958 Users should call xbt_ex_free() themselves.
2962 Portability report of this version:
2963 * Main portability targets:
2964 - Linux(debian)/x86/context
2965 - Linux(debian)/x86/pthreads
2966 - Linux(debian)/amd64/context
2967 - Linux(debian)/amd64/pthreads
2968 These targets fail about 1/10 of times on gras/pmm, but we believe
2969 that this is because of the test, not because of simgrid.
2970 amok/saturate_sg fails even more rarely, and the test may not be
2973 - Mac OSX Leopard/x86/context
2974 The test suite still spits tons of errors because some obscure
2975 force prevents us from removing the temporary directories
2976 arguing that they still contain some metadata I've never heard of.
2977 Smpi fails because seq is not installed.
2978 Everything seems to work properly beside of that.
2981 - AIX version 5.3 (both contexts and pthread)
2982 Smpi still fails there because mktemp is not installed.
2983 XML inclusions seems rosty on AIX.
2985 * Windows: it's still lagging behind. If you want to help, please
2988 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2990 SimGrid (3.3) stable; urgency=high
2994 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2995 [Malek Cherier & Mt]
2997 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3000 Until now, GRAS were using MSG as an interface to SURF. It was
3001 quite difficult because both interface have several differences
3002 (MSG channels vs GRAS sockets were the most notable point).
3004 This also opens the gate to SMPI (which should occur soon) and speed
3005 up simulations by to 40% (even if it were not the main goal).
3007 **************************************
3008 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3009 **************************************
3010 The point is that events occurring at the exact same timestamp are
3011 not scheduled in the same order with the old and new version. This
3012 may be enough to completely change the execution of simulations in
3013 some cases. Sorry for the inconvenience.
3015 * Cleanup and upgrade the XML format to push further scalability
3016 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3018 * Improve the testing infrastructure with tesh. Now a very large part of
3019 the code is tested not only by being run but also by checking that the
3020 output match an expected output [Mt].
3022 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3023 is really less memory-demanding, which should allow you to use
3024 larger files in SimGrid [AL].
3026 * Inform valgrind about our contextes, so that it becomes usable
3027 with the default (and more effecient) version of SimGrid
3028 [contributed by Sékou Diakite, many thanks]
3031 * Introduce a listener thread in charge of receiving incoming
3032 messages from the network. It allows to overlap communication and
3033 computation but most notably, it removes some stupid deadlocks due
3034 to the fact that so far, a process could not send and receive at
3035 the same time. This made most non trivial communication schema
3037 * Convert the PIDs from long int to int to match the MSG ones (and
3038 linux ones too) [Mt]
3039 * New function: gras_agent_spawn() to launch a new process on
3040 current host. Only working in simulation for now. [Mt]
3041 * New function: gras_os_hostport() returning a constant form (ie,
3042 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3045 * Make the backtrace of exceptions more human readable [Mt]
3046 * New module: xbt/str [Mt]
3047 a ton of string utility functions (split, join, printf to a newly
3048 allocated buffer, trim, etc)
3049 * New module: xbt/hash [Mt]
3050 SHA1 hashing algorithm (more to come if needed)
3051 * New module: xbt/synchro [Mt]
3052 synchronization tools (mutex and conditions) working the same way
3053 in simulation and in real life (mainly useful for GRAS, but not
3055 * New module: xbt/queue [Mt]
3056 classical producer/consumer synchronization scheme
3057 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3058 (using the classical functions will get serialized) [Mt]
3059 * Make dictionary internal table dynamic. No need to specify its size
3060 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3062 * Make sure the log channels are organized as a tree under windows
3063 (because of ANSI C compatibility issue, any channel were child of
3064 root directly) [Mt].
3067 * Cleaned many thing in surf and fixed a few bugs [AL].
3068 * Add a nice command line configuration mechanism to compose models [AL].
3069 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3070 the previous one (KCCFLN05). It relies on something that looks like
3071 a max-min sharing mechanism but cannot be written as such. A new solver
3072 was thus designed [AL].
3073 * Add a new solver to lmm. Based on Lagrange optimization and
3074 gradient-based descent, it enables to efficiently maximise systems s.a
3076 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3078 This solver enables to propose two new network models for TCP Reno and
3079 TCP Vegas based on Low's work. These models still need to be fully
3080 tested though [Pedro Velho].
3083 * Bug fix in SD_simulate. Now the time bound given as argument is
3085 * Use the new parallel task model (ptask_L07) as default.
3086 * Use the SURF command line configuration mechanism.
3087 * 0-size tasks (for synchronization) should now work.
3089 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3091 SimGrid (3.2) stable; urgency=high
3095 We still experience issues on this platform, but we believe that at
3096 least MSG is usable.
3098 GRAS API BREAKAGE (for simplification purpose, sorry):
3099 * the gras_msgtype_by_name is not used anymore. Instead of
3100 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3101 you can write (and must)
3102 gras_msg_send(toserver, "request", &request);
3103 - If you still want to pass a gras_msgtype_t to the function (to cache
3104 the type and avoid the lookup time), use the gras_msg_send_() variant.
3105 - Impacted functions:
3106 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3107 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3108 * The callbacks are now expected to return 0 when everything went well
3109 (just like the main() function)
3111 GRAS new features and improvements:
3112 * New module mechanism where user code can use per process globals [Mt]
3113 This is similar to gras_userdata_*() functions, but for libraries. It
3114 factorize some code developped over and over in the examples and AMOK.
3115 It has still to be documented and used (only amok/peermanagement is
3117 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3118 loss when they were small enough to fit into the buffer and sent quickly
3119 enough so that they can all get received in one shoot.
3120 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3121 if not found. Use the *_or_null() variant for the old semantic.
3122 * In gras_msg_handle, do not discard messages without callback.
3123 They are probably messages to be explicitly awaited later (ie, proofs of
3124 mis-synchronization in userland since they are sent before being awaited)
3125 No big deal usually.
3126 * gras_socket_meas_send/recv: semantic changed!
3127 The numerical arguments used to be (1) the total amount of data to send
3128 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3129 messages. This was need for the fool willing to send more than MAXINT
3130 bytes on quite fat pipes.
3133 * Do really rename the hostmanagement module to peermanagement. [Mt]
3134 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3135 but this is rather new and this was documented in the module
3136 documentation (poor excuses, I admit)
3137 * Bandwidth measurement semantic changed! This follows the changes to
3138 gras_socket_meas_send/recv explained above.
3141 * A sequential mode has been added to the workstations. When a workstation
3142 is in sequential mode, it can execute only one task, and the other tasks
3143 are waiting in a FIFO. [Christophe Thiery]
3146 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3147 model for SIMDAG. [Christophe Thiery]
3148 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3149 equal to 0, which caused some bad behaviors in
3150 saturated_constraint_set_update. I now use a threshold mechanism like in
3154 * When running manually src/testall, you select specific units [Mt]
3155 testall is the result of our cunit mechanism, and should replace all
3156 the scripty thingy around since bash don't run easily on billware.
3158 * A mallocator system has been added. [Christophe Thiery]
3159 Mallocators allow you to recycle your unused objects instead of freeing them
3160 and allocating new ones.
3162 Documentation update:
3163 * FAQ reworking + New FAQs:
3164 - "Valgrind spits tons of errors!" [Mt]
3165 - "How to repport bugs" [Mt]
3166 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3167 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3168 - Communication time measurement within MSG [AL]
3169 - I experience weird communication times when I change the latency [AL]
3170 * GRAS tutorial [Mt]
3172 - an introduction to the framework and to the used communication model
3173 - an initiatic tour introducing the most proheminent features:
3175 . Lesson 0: Installing GRAS
3176 . Lesson 1: Setting up your own project
3177 o Part 2: Message passing
3178 . Lesson 2: Exchanging simple messages
3179 . Lesson 3: Passing arguments to the processes (in SG)
3180 . Lesson 4: Attaching callbacks to messages
3181 . Lesson 5: Using globals in processes
3182 . Lesson 6: Logging information properly
3183 . Lesson 7: Using internal timers
3184 . Lesson 8: Handling errors through exceptions
3185 . Lesson 9: Exchanging simple data
3186 . Lesson 10: Remote Procedure Calling (RPC)
3187 . Lesson 11: Explicitly waiting for messages
3188 . Recapping of message passing features in GRAS
3189 - A HOWTO section containing:
3190 o HOWTO design a GRAS application
3191 More are due, of course. They will come latter. In the meanwhile, you can
3192 check the examples which are still here.
3194 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3196 SimGrid (3.1) stable; urgency=high
3200 There was a stack corruption somewhere, visible only when optimizing
3201 with these versions. [Vince]
3204 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3205 module that enabled to play with Directed Acyclic Graphs. It is built
3206 directly on top of SURF and provides an API rather close to the old
3207 SG. Some old codes using SG are currently under rewrite to check that
3208 all needful functions are provided. [Christophe Thiery]
3211 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3212 extension of the classical CLM03 model that gracefully handles
3213 failures. This is now the default model for MSG and GRAS. It doesn't
3214 handle parallel tasks yet though. [AL]
3215 * Bug fix: Weights were not correctly set in the network part.
3216 WARNING: This may have resulted in incorrect results with simulations
3217 where there are more than one flow on a given link. [AL]
3220 * After a (long ?) discussion on simgrid-devel, we have decided that the
3221 convention we had on units was stupid. That is why it has been decided
3222 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3223 WARNING: This means that all previous platform files will not work as
3224 such with this version! A warning is issued to ask users to update
3226 A conversion script can be found in the contrib module of the CVS, under
3227 the name contrib/platform_generation/surfxml_update.pl [MQ]
3230 * Bug fix: Processes were started in reverse order, wrt deployment file.
3231 WARNING: if your code relies on this bug, please fix it. [AL]
3232 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3233 being executed on two different locations. [AL]
3234 * Bug fix: Failures are now better supported thanks to Derrick's tests
3235 (there was many failure situations I hadn't thought of and that weren't
3236 correctly handled). [AL]
3237 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3241 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3242 Exception occurring on server-side are propagated back to client (!).
3244 API CHANGE: the callback changed their prototype. Change:
3245 int my_handler(gras_socket_t expeditor, void *payload_data) {
3247 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3248 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3250 * New! function: gras_msg_handleall to deal with all messages arriving
3251 within a given period.
3252 * New! function: gras_socket_server_range to get a server socket in a
3253 range of port numbers (ease to avoid port number conflicts) [MQ]
3254 * New! gras processes display their backtrace when they get a SIGUSR1
3255 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3256 Sweet to debug RL processes [MQ]
3260 - Do not force experiment sizes to be expressed in kb, or it becomes
3261 impossible to measure the latency this way (needs one byte-long tests)
3262 WARNING: this changes the amok_bw_* function semantic. [MQ]
3263 - Implements the link saturation stuff. [MQ]
3264 * Peer management module:
3265 New! module factorizing code that we wrote over and over [MQ].
3268 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3269 - Test units are placed directly into the library code, they get extracted
3270 automatically and placed into the src/testall binary.
3271 - Convert most of the XBT tests to this system.
3272 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3273 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3274 live in the same C code. This required to change a little bit the API
3275 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3276 * New module: sparse graph structure with basic algorithms (this is work
3277 in progress and the API is not considered to be frozen yet). [AL]
3278 * Display more information on backtraces: source line & function names are
3279 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3280 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3281 * Reworked a little bit some #include statements to load only required
3282 headers. Some user code that relied on SimGrid to include stdlib or
3283 stdio may need to include it by themselves. [AL]
3284 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3286 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3287 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3288 but I guess that almost nobody used those parts. [MQ]
3290 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3292 SimGrid (3.0.1) stable; urgency=low
3295 * Unfortunately, I had missed 5 misnamed functions:
3296 xbt_fifo_item_t xbt_fifo_newitem(void);
3297 void xbt_fifo_freeitem(xbt_fifo_item_t);
3298 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3299 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3300 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3301 They're now deprecated. Please use their new versions:
3302 xbt_fifo_item_t xbt_fifo_new_item(void);
3303 void xbt_fifo_free_item(xbt_fifo_item_t);
3304 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3305 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3306 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3308 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3309 * Documentation: xbt_log module unmercifully reworked [MQ]
3310 * Bugfix: there was a problem with the ending of contexts with
3311 the pthread backend. It caused some weird deadlock or behavior
3312 depending on the pthread implementation. [AL]
3313 * Bugfix: get the exceptions raised in the simulator repport where
3314 and why they come from when they are not caught in time [AL, MQ]
3317 * Bugfix: Do repport the error when two non-connected hosts try to
3318 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3321 * Add additionnal checkings on communications. Assert that two
3322 communicating hosts are connected by a set of links... [AL]
3325 * Add additionnal checkings on channel values in communication [AL]
3326 * New: MSG_task_get_source to see on which host a task was generated [HC]
3327 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3328 the number of tasks waiting to be received on channel and sent
3330 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3331 waits for the first task coming from a given host.. [AL]
3333 GRAS new functionnalities: [MQ]
3334 * Enhance the parsing macro to allow the size of multidimentional objects
3335 to be given thru annotations.
3336 * New example (and documentation): Matrix Multiplication a la RPC
3337 (as when I was young!) and fix a bunch of bugs found on the way.
3339 GRAS performance improvements: [MQ]
3341 * Reduce the amount of cbps creation/destruction by making it static to
3342 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3345 * Change libdata to a set so that we can search for stuff by ID (and thus
3346 reduce the insane amount of dict lookups)
3349 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3350 RL communication on top of SG since it's so uneffective.
3351 It may also be used for inter-thread communication in RL, one day.
3352 * Use gras_datadesc_copy() to exchange messages on top of SG
3354 - improve message exchange performance on top of SG
3355 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3356 * Don't exchange on the network the size of the used part of buffer,
3357 instead, specify the possible buffer size to read().
3359 - reduces the amount of read/write calls (one pair per exchange)
3360 - reduces the amount of exchanged data (the size)
3361 - allows to retrieve all arrived data on receiver side, if we don't need
3362 it right now (subsequent read will peek the buffer)
3363 - allows the receiver to proceed with the beginning of the stream before
3364 everything is arrived
3365 - make it possible to build an iov transport (using readv/writev)
3367 - take care of the data with non-stable storage (like stacked data),
3369 * If possible, TCP send uses vector I/O (when writev() is here)
3370 - Don't use it for receive since we send data sizes and data on the
3371 same stream, so we wouldn't be able to chain large amount of chunks
3372 before having to flush the stuff to read the size.
3373 * Rework the transport plugin mechanism to simplify it and reduce the
3374 amount of pointer dereferencement when searching for the right function
3377 * I guess that now, we do almost as few system calls as possible while
3378 doing as few data copy as possible.
3380 To improve it further, we could try to send all the sizes first and then
3381 all the data (to use iov on receiving size), but it's only a partial
3382 solution: when you have 2 dimensional data, the sizes of the second
3383 dimension is data of the first dimension, so you need 3 streams.
3385 I'm not sure the potential performance gains justify the coding burden.
3387 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3389 SimGrid (3.00) stable; urgency=high
3392 * New! Give the possibility to hijack the surf parser and thus bypass
3393 MSG_create_environment and MSG_launch_application. Have a look at
3394 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3396 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3398 SimGrid (2.96) unstable; urgency=low
3403 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3404 This deprecates the xbt_error_t mechanisms.
3405 It modifies (simplifies) all XBT and GRAS API.
3406 MSG API keeps unchanged (exceptions raised by XBT are caught from
3407 within MSG and masked with existing error handling facilities)
3410 * New! Add a FATPIPE model. [AL]
3411 * New! Add a parallel task model. [AL]
3412 * New! Add automatically a loopback interface (in the default
3413 network model) if none was precised.
3416 * Bugfix: MSG_process_resume now works with the current running process.
3418 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3419 * Modification of MSG_task_get_compute_duration. Once a task has been
3420 processed, the value returned by this function is now equal to 0. [AL]
3421 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3422 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3423 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3424 of a m_task. For now, MSG_task_cancel only works with computation
3426 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3427 of the processor (in Mflop/s) regardless of the current load on the
3429 * API Change: use proper naming convention for MSG_getClock and
3430 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3432 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3433 This function changes the priority of a computation task. This priority
3434 doesn't affect the transfer rate. A priority of 2 will make a task
3435 receive two times more cpu power than the other ones. This function
3436 has been added to suit the needs of Nguyen The Loc and hasn't been that
3437 much tested yet. So if it fails, please report it and send me your code.
3439 * API Change: removed all functions and types that were marked "deprecated"
3440 since many months. Renamed MSG_global_init_args to MSG_global_init.
3442 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3444 SimGrid (2.95) unstable; urgency=low
3447 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3448 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3449 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3452 * Bugfix: Allow absolute paths for platform description files [MQ]
3453 * Bugfix: do free the variables after use. Leads to drastic performance
3455 * Implement max_duration (ie, timeouts) on resources [AL]
3458 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3460 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3461 the needed support (timeouts) [AL]
3464 * Implement measurement sockets. You can now get the bandwidth between two
3465 hosts thanks to AMOK (see below). [MQ]
3466 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3467 dynar over the network (yeah) [MQ]
3468 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3469 * simple/static token-ring example. [Alexandre Colucci and MQ]
3470 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3471 in place before [MQ]
3474 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3475 type parsing. "short" and "long" modifiers were also ignored; other
3476 modifier (such as reference level) are still ignored. [MQ]
3477 * Embeed the buffer size within the buffer itself on SG. [MQ]
3478 That way, send() are atomic and cannot get intermixed anymore (at least
3479 the ones which are less than 100k; bigger messages still have the issue)
3480 * Array size pushed by the field, not by the field type (or each
3481 and every long int will push stuff to the cbps) [MQ]
3482 * use select() to sleep since it allows to portably sleep less than one
3485 GRAS (minor cleanups)
3486 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3489 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3490 * type category 'ignored' killed as it was never used and were difficult
3492 * whether a type can cycle or not is now a flag, leaving room for more
3493 flags (as "ignored", if we feel the need one day ;)
3494 * Rename raw sockets to measurement sockets since "raw" has another
3495 meaning in networking community.
3498 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3499 and offers features not belonging to GRAS but that most applications
3500 need. One day, it may be a set of plugins loadable at runtime.
3501 * New module: bandwidth
3502 bandwidth measurement between arbitrary nodes running this module. [MQ]
3504 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3506 SimGrid (2.94) unstable; urgency=low
3508 The first beta release of SimGrid 3 !
3512 * Update the main page and the FAQ. Adding references to gforge.
3515 * Add a gras_os_getpid function.
3518 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3519 * Extend the logs so that they also print PID, hostname, date, ... if
3521 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3522 and kill the old version which were in testsuite/
3523 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3527 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3528 for users. Instead, each programming environment (GRAS, MSG,...) use it
3529 when needed to provide such a feature to users.
3530 Don't shortcut the mechanism or you will also shortcut the virtualization
3531 you need on the simulator.
3535 * Cleanups in configury with regard to compile optimization/warning flags.
3536 Also add -fno-loop-optimize to any powerpc since it's the optimization
3537 killing gcc (< 3.4.0).
3538 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3539 needed by MSG examples complications
3540 * Borrow configury beautifications from PHP
3543 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3544 warning (thanks loris for stumbling into this one).
3545 * Bugfix: stop loading private headers (gras_config.h) from the public
3549 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3550 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3551 RL would imply a DNS resolver, which is *hard* to do in a portable way
3552 (and therefore delayed).
3553 * Implement a real timer mechanism and use it in timing macros. This allows
3554 to avoid rounding errors and get a 0.000005 sec precision in timing
3555 macros. While I was at it, various cleanups:
3556 - allow to declare more than one timed section per file (fix a stupid bug)
3557 - move some private declaration to the right place
3558 - merge conditional execution and timing macros into emulation module
3559 - document the module
3560 - make sure the module cleanups its mess on gras_exit
3561 * Documentation improvements:
3562 - (new) how to compile applications using GRAS
3563 - (new) emulation support (timing macros)
3565 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3567 SimGrid (2.93) unstable; urgency=low
3569 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3572 - Use Paje properly where used. Still to be sanitized properly.
3573 - Portability fix: Add an implementation of the contexts using pthread
3577 - Add xbt_procname(): returns the name of the current process.
3578 Use it to show the current process's name in all logging.
3580 - fix detection of older flex version and the reaction, since we do
3581 depend on modern ones (we use lex_destroy)
3582 - Better separation of SG and RL in the libs: remove all simulation code
3583 from libgras. As a result, this lib is now only 200k when stripped.
3584 Some of the xbt modules may also be duplicated (two sets and such) and
3585 should be cleaned/killed before SG3.
3586 - Insist on using xlC on AIX because of weird problems involving gcc there.
3587 - Cleanup the make remote stuff. This is now done by scripts
3588 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3589 mainly for our private use, but we're working on changing them to user
3592 - Bugfix: flush the socket on close only if there is some *output*.
3593 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3595 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3598 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3600 SimGrid (2.92) unstable; urgency=low
3602 Alpha 3 on the path to SimGrid 3
3606 - New! First try of benchmarking macros.
3607 - New! First try so that gras_stub_generator generate deployment and
3608 remote compilation helpers.
3610 - Bugfix: Initialization fix in msg_test.
3614 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3616 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3618 - API Change: gras_os_sleep to take the amount of seconds as a double.
3619 Accepting an int was error prone since it was the only location where
3620 seconds were coded as such. It leaded to damn rounding errors.
3621 - Bugfix: Hard to belive that timers ever worked before this.
3623 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3625 SimGrid (2.91) unstable; urgency=low
3627 Alpha 2 on the path to SimGrid 3
3631 - Bug fix in the lmm_solver.
3633 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3634 through the function MSG_paje_output.
3635 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3636 - It is possible to bound the rate of a communication in MSG with
3637 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3638 to write it in the changelog).
3639 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3644 - Various cleanups to the autotools stuff
3645 - Begin to move Gras examples to examples/gras/
3646 - Let make distcheck work again (yeah!)
3648 - documentation overhauled using doxygen.
3649 gtk-doc-tools is dead in SimGrid now.
3650 - Automatically extract all existing logging categories, and add the list
3651 to the documentation (long standing one, to say the less)
3653 - Cleanup the known architecture table. Reorder the entries to group what
3654 should be, and use a more consistent naming scheme.
3655 (some of the test dataset are still to be regenerated)
3656 - New! Allow library to register globals on each process just as userdata
3658 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3659 lookup time (for now).
3660 Use it in msg and trp.
3661 This cleans a lot the internals and helps enforcing privacy of the
3662 headers between the gras components.
3663 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3664 - Bugfix: gras_os_time was delirious in RL.
3665 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3667 - Reenable GRAS now that it works.
3669 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3671 SimGrid (2.90) unstable; urgency=low
3673 Alpha 1 on the path to SimGrid 3
3675 * It is a long time since the last release of SimGrid. I'm sorry about
3676 that but as I had told you, I was rewriting a lot of things. I apologize
3677 to those who had been reporting bugs to me and that I had not answered.
3678 If your bug is still in the new version, please tell me. Here is a
3679 summary of the main changes.
3681 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3682 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3683 portability, speed, and a lot more but you'll figure it out later.
3684 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3685 I wanted to release the new MSG as soon as possible and I have
3686 broken GRAS, which is the reason why, for now, only MSG is fully
3687 functional. A laconic description of these projects is available
3688 in the documentation.
3690 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3691 kernel optimized for our needs (called SURF but only the developers
3692 should use it). Hence, MSG is now roughly 30 times faster and I think
3693 that by rewriting a little bit MSG, I could event speed it up a little
3694 bit more. Beside the gain in speed, it is also much easier to encode a
3695 new platform model with SURF than it was with SG. More to come...
3697 * REVOLUTION 3: I have tried to change a little as possible the API of
3698 MSG but a few things really had to disappear. The main differences
3699 with the previous version are:
3700 1) no more m_links_t and the corresponding functions. Platforms are
3701 directly read from a XML description and cannot be hard-coded
3702 anymore. The same format is used for application deployment
3703 description. The new format is described in the documentation.
3704 Have a look in tools/platform_generation. There is a tiny script
3705 that converts from the old platform format to the new one. Concerning
3706 the application deployment format, parsing the old one is tricky.
3707 I think most of you should however be able to convert your files. If
3708 it is really an issue, I can write a C code that does the conversion.
3710 2) the toolbox tbx does not exist anymore. We now have a library
3711 with much more data-structures but without the hash-tables (we have
3712 dictionaries that are much faster).
3714 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3716 *****************************************************************************
3717 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3718 * don't feel like converting the oldies. (Mt) *
3719 *****************************************************************************
3722 Version 2.90: "the long awaited one"
3723 - Finished rewriting and debugging MSG. Rewrote the documentation.
3724 - disable GRAS for now since it needs to be ported to the newest SG
3727 - Finish the port to windows (using mingw32 for cross-compile)
3730 - Main loop and datastructures of SURF. A cpu resource object is
3731 functional. Surf can thus be used to create cpu's with variable
3732 performance on which you can execute some actions.
3734 2004-11-15 Martin Quinson
3735 - Port to ARM. Simply added the alignment and size descriptions. Should
3736 work, but the ARM machines are so slow that I didn't had the opportunity
3737 to 'make check' over there yet.
3739 2004-11-15 Arnaud Legrand
3740 - Trace manager now written. It uses a heap structure and is therefore
3741 expected to be efficient. It may however be speeded up (particularly
3742 when many events occur at the same date) by using red and black
3743 trees. One day maybe...
3744 - Max-min linear system solver written. It uses a sparse matrix
3745 structure taking advantage of its expected use. Most operations are
3746 O(1) and free/calloc are called as few as possible. The computation of
3747 the minimum could however be improved by using a red and black tree
3750 2004-11-03 Arnaud Legrand
3751 - Rename every gras_* function that was in xbt/ to its xbt_
3753 - Add a heap and a doubly-linked list to xbt
3754 - Added a dichotomy to the dictionaries. make check works as well before
3755 so I assume that the patch is correct. I do not know however if things
3756 run effectively faster than before now. :)
3758 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3759 SimGrid, and the version number is bumped to 2.x
3761 2004-10-29 Martin Quinson
3762 - Introduction of the remote errors.
3763 They are the result of a RMI/RPC on the remote machine.
3764 ErrCodes being scalar values, you can't get the host on which those
3765 errors did happen. Extending the error mechanism as in Gnome is possible.
3766 No idea yet whether it is a good idea.
3768 2004-10-28 Martin Quinson
3769 - Interface revolution: the Starred Structure Eradication.
3770 I used to do typedef struct {} toto_t; and then handle *toto_t.
3771 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3772 - ???_t is a valid type (builded with typedef)
3773 - s_toto_t is a structure (access to fields with .)
3774 - s_toto is a structure needing 'struct' keyword to be used
3775 - e_toto_t is an enum
3776 - toto_t is an 'object' (struct*)
3778 typedef struct s_toto {} s_toto_t, *toto_t;
3779 typedef enum {} e_toto_t;
3780 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3781 s_toto_t) is private.
3783 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3784 so that it can set the variable to NULL. It was so for dicts and sets,
3785 it changed for dynars.
3787 - Fix a bunch of memleaks in dict_remove
3788 - Fix a bug in sg/server_socket opening: it failed all the time.
3790 2004-10-07 Martin Quinson
3791 - Speed up dynar lookup operation a bit.
3793 gras_dynar_get is dead.
3795 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3796 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3797 (which gives you the address of the stored data).
3799 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3800 copy of the data using an affectation to do the job and not a memcpy.
3802 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3803 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3805 It does not really speedup the dynar test because they are
3806 setting elements all the time (and look them seldom). But the dict does
3807 far more lookup than setting.
3809 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3811 2004-10-05 Martin Quinson
3812 - Allow to (en/dis)able the cycle detection at run time.
3814 Whether we should check for cycle or not is now a property of each
3815 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3816 datadesc_cycle_unset allow to remove this property when previously set.
3818 Note that the cycle detection is off by default since it impacts the
3819 performance. Watch the data you feed GRAS with ;)
3821 This property is hereditary. Any element embedded in a structure having it
3822 set have it set for the time of this data exchange.
3824 You should set it both on sender and receiver side. If you don't set it on
3825 sender side, it will enter an endless loop. If you forget on receiver
3826 side, the cycles won't be recreated after communication.
3828 - Header reorganization.
3829 Kill gras_private.h, each submodule must load the headers it needs.
3831 2004-10-04 Martin Quinson
3832 - Interface revolution: do not try to survive to malloc failure.
3834 Now, gras_malloc and friends call gras_abort() on failure.
3835 As a conclusion, malloc_error is not a valid error anymore, and all
3836 functions for which it was the only gras_error_t return value are
3837 changed. They now return void, or there result directly.
3838 This simplify the API a lot.
3840 2004-09-29 Martin Quinson
3841 - Re-enable raw sockets.
3842 Created by gras_socket_{client,server}_ext;
3843 Used with gras_raw_{send,recv}
3846 It should allow to kill the last bits of gras first version soon.
3848 This is not completely satisfactory yet (duplicate code with
3849 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3852 - Simplify transport plugin (internal) interface by not passing any
3853 argument to _server and _client, but embedding them in the socket
3856 2004-09-28 Martin Quinson
3857 - Finish the port to AIX.
3858 autoconf was my problem (segfault within the malloc replacement
3859 function. No idea why)
3861 2004-09-16 Martin Quinson
3862 - Fix some size_t madness on 64bit architectures.
3864 2004-09-08 Martin Quinson
3865 - Reduce the number of system headers loaded, overload some more system
3866 calls (such as malloc to cast the result of the system one, and work
3868 - Fix and reintroduce the config support
3870 2004-09-07 Martin Quinson
3871 - Source code reorganization to allow Arnaud to surf all over there.
3872 - Allow to document the logging categories.
3873 - Remove all uppercase from logging categories and useless cleanup in names.
3875 2004-08-18 Martin Quinson
3876 Version 0.6.2 (protocol not changed; API changed)
3877 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3878 gras_error_t), and NULL when not found. Functions expecting a msgtype
3879 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3880 hopefully usefull message.
3881 - Portability to prehistoric sparcs again
3883 2004-08-17 Martin Quinson
3884 Version 0.6.1 (protocol not changed; ABI not changed)
3885 - prealloc some buffers to speed things up
3887 2004-08-11 Martin Quinson
3888 Version 0.6 (protocol not changed; ABI expended)
3889 - The parsing macro can deal with the references, provided that you add
3890 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3892 2004-08-09 Martin Quinson
3893 Version 0.5 (protocol not changed; ABI changed)
3894 - Allow to off turn the cycle detection code in data exchange at
3895 compilation time. It should be at run time, but I'm short of time (and
3896 the config stuff is still broken). That way, we keep dict out of the
3897 critical path, which is good because the performance is poor:
3898 - search not dichotomial yet
3899 - dynar give no way to access their content and memcpy everytime
3900 - In composed data description (struct, ref and so on), stop foolness of
3901 keeping the subtype's ID, but store the type itself. This keeps sets out
3902 of the critical path, which is good since they rely on dynar and
3903 dictionaries. The only loose of that is that we cannot detect the
3904 redeclaration of a structure/union with another content (but I'm not sure
3905 the code detected well this error before anyway). We still can detect
3906 the redefinition discrepancy for the other types.
3907 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3908 it breaks the code because of type-punning used all over the place).
3909 This breaks on all non-gcc architectures (for now).
3911 All those changes (plus the buffer of last time) allow me to gain 2 order
3912 of magnitude on cruel tests consisting of 800000 array of integers on two
3913 level of a hierarchical structure (200 secondes -> 4 secondes)
3916 - the selector of reference must now return the type it points to, not
3917 the ID of this type.
3919 2004-08-06 Martin Quinson
3920 Version 0.4 (protocol changed; ABI not changed)
3921 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3922 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3923 - Implement a buffer transport, and use it by default (it relies on tcp in
3924 real life and on sg in simulation).
3925 That's a bit hackish since I had a new field to the structure to store
3926 its data without interfering with the subtype ones. Inheritance
3927 is tricky in C. And that's a kind of reverse inheritance with one class
3928 derivating two classes. Or maybe a game with java interfaces. Anyway,
3929 that's damn hard in C (at least).
3930 Moreover, I got tired while trying to ensure plugin separation and
3931 genericity in SG mode. MSG wants me to do weird things, so let's go for
3932 cruel hacks (temporarily of course ;).
3933 See comment in transport_private.h:71
3934 - do not include all the _interface headers in private but in the files
3935 which really need them (to cut the compilation time when they are
3938 2004-07-26 Martin Quinson
3939 Version 0.3 (protocol not changed; ABI changed)
3940 - Major overhault of the datadesc interface to simplify it:
3941 - shorted the function names:
3942 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3943 - add a trivial way to push/pop integers into the cbps without malloc.
3944 This allows to make really generic sub_type description, which simply
3945 pop their size of the stack.
3946 - add a function gras_datadesc_ref_pop_arr() which does what users want
3947 most of the time: Declare a dynamic array (which pops its size of the
3948 stack) and declare a reference to it. Poor name, but anyway.
3949 - kill the post-send callback, add a post-receive one
3951 2004-07-23 Martin Quinson
3952 Version 0.2 (protocol changed; ABI changed)
3953 - add some testing for cpbs in the test cases, and fix some more bugs.
3954 This invalidate again the little64 data file, since I cannot regenerate
3956 - remove an awfull optimization in the logging stuff, allowing me to:
3957 - understand it again
3958 - learn gcc how to check that the argument match the provided format
3959 - fix all errors revealed by gcc after that
3960 - internal keys of dict are not \0 terminated. Deal with it properly in
3961 loggings instead of segfaulting when the user want to see the logs :-/
3963 2004-07-22 Martin Quinson
3964 - Fix some stupid bug preventing cbps (callback postit) from working
3966 2004-07-21 Martin Quinson
3967 - Some documentation cleanups
3968 - remove the useless last argument of msgtype_declare
3969 - rename the Virtu functions to fit into the 'os' namespace
3970 - move headers src/include -> src/include/gras/ and stop fooling with
3971 gras -> . symbolic link
3972 - make distcheck is now successful
3974 2004-07-19 Martin Quinson
3976 - Build shared library also
3977 - Install html doc to the right location
3978 - stop removing maintainer files in make clean
3979 - build tests only on make check
3981 2004-07-13 Martin Quinson
3983 - No major issue in previous version => change versioning schema
3984 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3985 corresponding dataset.
3987 2004-07-11 Martin Quinson
3989 - Get it working with any kind of structure (we can compute the padding
3990 bytes remotely for all the architectures I have access to)
3991 - Implement the structure parsing macro (still not quite robust/complete)
3992 - Improvement to the remote testing toysuite
3994 2004-07-10 Martin Quinson
3995 [autoconf mechanism]
3996 - get ride of a bunch of deprecated macros
3997 - actually run the test for two-compliment, not only compile it :-/
3998 - test whether the structures get packed (and bail out if yes. Damn.
3999 Alignment is a serious matter)
4000 - test whether the structures get compacted (but respecting the alignment
4001 constraints of each types)
4002 - test whether the array fields of structures can straddle alignment boundaries
4004 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4005 (took me 2 hours to find that bug, looking at the wrong place)
4006 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4007 before sending/receiving and are used to compute the offsets of fields
4008 - Given that padding size depend even on compiler options, keep track of
4009 alignment and aligned_size only for the current architecture. Not a big
4010 deal since we send structure fields one after the other (seems
4012 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4013 program, let it work on linux/gcc/little32. portability todo.
4015 2004-07-08 Martin Quinson
4016 - import and improve remote compilation support from FAST
4017 - make sure make check works on half a dozen of machines out there
4019 2004-07-07 Martin Quinson
4020 Let's say it's version 0.0.3 ;)
4021 - Implement conversions (yuhu!)
4022 - Let it work on solaris (beside conversion, of course)
4023 - Stupid me, using rand() to generate the conversion datatests in not wise.
4025 2004-07-06 Martin Quinson
4026 - Let make dist work, since I'm gonna need it to compile on remote hosts
4027 - Let Tests/datadesc_usage write the architecture on which the file was
4028 generated as first byte.
4029 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4030 architecture descriptions.
4031 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4032 on those architectures.
4033 - Optimization: send/recv array of scalar in one shoot
4035 2004-07-05 Martin Quinson
4036 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4038 - Plug a whole bunch of memleaks
4039 - each process now have to call gras_{init,exit}. One day, their log
4040 settings will be separated
4041 - Continue the code factorisation between SG, RL and common in Transport.
4043 2004-07-04 Martin Quinson
4045 - Redistribution between SG and RL.
4046 We wanna have to accept in SG, so move accepted related parts of RL in
4047 the common part. (more precisely, the dynar of all known sockets is no
4048 more a static in transport.c, but part of the process_data)
4050 [gras_stub_generator]
4051 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4053 2004-07-03 Martin Quinson
4054 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4055 them all up in one shot)
4057 - Fix the ugly case of reference to dynamic array.
4058 - New (semi-public) function gras_datadesc_size to allow the messaging
4059 layer to malloc the needed space for the buffer.
4061 - gras_socket_close now expect the socket to close (and not its address to
4062 put NULL in it after it). This is because the socket passed to handlers
4063 is one of their argument (=> not writable).
4065 - propagate the interface cleanup from last week in datadesc, ie remove a
4066 superfluous level of indirection. User pass adress of variable
4067 containing data (both when sending and receiving), and not of a variable
4068 being a pointer to the data. Let's say that I like it better ;)
4069 The price for that is constructs like "int msg=*(int*)payload" in
4070 handlers, but it's a fine price, IMHO.
4072 - Let it work in RL (yuhu)
4074 2004-06-21 Martin Quinson
4076 - porting SG plugin and SG select to new standards (works almost).
4077 - plug memleaks and fix bugs around.
4080 - cleanup the prototype of data recv and force users to specify when they
4081 want to handle references to objects. Test case working even for cycles.
4082 - plug memleaks. Valgrind is perfectly ok with this.
4084 2004-06-12 Martin Quinson
4086 - cleanup the separation between plugin and main code in plugin creation
4088 2004-06-11 Martin Quinson
4090 - Reput hook for raw sockets, needed for BW experiments
4091 - kill a few lines of dead code
4092 [Data description] Interface cleanup
4093 - gras_datadesc_by_name returns the searched type or NULL.
4094 That way, no variable is needed to use a type desc once, which makes
4096 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4097 two parameters were strings (field name, type name), leading to
4099 [Dicos] Interface cleanup
4100 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4101 This is consistant with the dynar API.
4103 2004-04-21 Martin Quinson
4105 - Porting to new standards.
4107 - interface cleanup.
4108 There is no bag anymore, no need to take extra provision to mask the
4109 pointers behind "ID".
4110 Better splitup of functions between files create/exchange/convert.
4111 This is still a bit artificial since convert and receive are so
4112 interleaved, but anyway.
4114 - add a queued message list to procdata (the ones not matching criteria
4116 - factorize some more code between SG and RL wrt procdata
4118 - use gras_exit in example to track memleaks
4119 - get rid of gs_example now that GS is properly integrated into gras
4120 - update run_test to integrate the latest tests (datadesc)
4122 - rename WARNINGn macros to WARNn since it prooved error-prone
4124 2004-04-19 Martin Quinson
4126 - register init/exit functions within gras module mechanism
4127 - send/receive function.
4128 Convertion is not implemented, but short-cutted if not needed.
4129 struct/array elements are sent one by one (instead of block-wise), but
4130 nobody really cares (yet). Get a prototype before optimizing.
4131 - tests (using a file socket) for DD send/receive on:
4132 - base types: int, float
4133 - array: fixed size, string (ie ref to dynamic string)
4134 - structure: homogeneous, heterogeneous
4135 - chained list, graph with cycle
4136 Believe it or not, valgrind is not too unhappy with the results. The
4137 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4141 - Bugfix when using a filename explicitly (instead of '-')
4143 2004-04-09 Martin Quinson
4145 - factorize more code between RL and SG in socket creation
4146 - Complete the implementation and tests of:
4148 o file (only in RL, and mainly for debugging)
4150 I lost 3 days to design a portable address resolver, and then decided
4151 that the prototype mainly have to run on my box.
4152 Addressing portability too early may be like optimizing too early :-/
4154 - use gras_init in the Tests instead of the crappy parse_log_opt
4155 (the latter function is removed)
4156 [Conditional execution]
4157 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4158 [Code reorganisation]
4159 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4160 Build examples against the RL library, since there is no way to disable
4161 its creation for now.
4163 For information, the beginning of coding on GRAS was back in june
4164 2003. I guess that every line has been rewritten at least twice since