1 ----------------------------------------------------------------------------
3 SimGrid (3.25.1) UNRELEASED (target: March 20 2020 or v3.26 on 3/26/20)
5 Important user-visible changes:
6 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
7 several users had difficult to understand issues when using high bandwidth or
8 small latency events. The new value was already the default for SMPI and
9 should not cause too much performance hit. It can change some simulation
10 timings if simulation was skipping a lot of small events (which was actually
11 an issue of said simulation). The value can still be changed back using
12 --cfg=surf/precision:1e-5
15 - LTO is now enabled for Intel/clang compilers.
16 - LTO behavior on GCC can be parametrized using LTO_EXTRA_FLAG in cmake.
17 Setting it to "auto" will use all available cores, while setting it to n will
18 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
21 - Many MSG tests were converted to the new S4U's interface in C, that
22 was extended for that.
24 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
25 - FG#43: xbt::random needs some care
26 - FG#48: The Impossible Did Happen (yet again)
27 - FG!24: Documentation and fix for xbt/random
28 - GH#139: Allow pthread creation in SMPI
30 ----------------------------------------------------------------------------
32 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
34 The Palindrom Day Release.
36 Important user-visible changes:
37 - Improve the Python usability (stability and documentation).
38 - A nasty synchronization bug was ironed out, see also below.
39 - Python's doc was organized as a separate tree, now integrated with C++.
40 - C bindings of S4U were not part of the doc.
41 - The C++ doc was also improved as methods are now split by theme.
42 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
43 - This is mandatory to use the Java bindings.
44 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
45 - SimDAG++: Automatic dependencies on S4U activities (experimental)
46 - Some features are already implemented but not all of them
47 - Cannot block an activity until it's scheduled on a resource
48 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
49 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
54 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
55 - Drop support for Viva/Triva (old visualization tools), hence removed
56 TRACE_get_node_types() and TRACE_get_edge_types() functions.
59 - Require pybind11 version 2.4.
60 - Greatly improve locking during context switches with Python.
61 - Declaring Python callbacks to this_actor::on_exit is now working.
62 - No more tests are failing currently, even on win32.
65 - Actor: Merge signals on_migration_start/end into on_host_change
66 - Actor: Rename migrate() into set_host()
67 - Disk: Allow users to get the read and write nominal bandwidth values
68 - Exec: Implement wait_for(timeout)
69 - Io: Implement wait_for(timeout)
72 - Parse errors now raise a simgrid::ParseError that you may want to catch.
75 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
76 been renamed to fix a spelling error. As usual, the previous definitions are
77 kept with a deprecation warning. However, to avoid runtime errors, it is now
78 forbidden to override these deprecated methods in a derived class. Please use
79 the new names immediately if you need to override them.
82 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
83 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
84 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
85 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
86 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
87 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
88 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
89 - New module: random, providing classical random numbers generators.
92 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
93 allocations' size above which they are considered "shared" by default (as if
94 they were performed through SMPI_SHARED_MALLOC macros).
95 Default = 0 = disabled feature.
96 Note : malloc, calloc and free are now overriden by smpicc/cxx by default.
97 This can cause some troubles if codes are already overriding these. If this
98 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
99 help, but will make this feature unusable.
100 - Input check is now performed much more consistently and outputs precise
101 warnings for each error found
102 - New supported calls : MPI_File_set_view, MPI_File_get_view
103 - MPI I/O now only supports disks, as storage is discontinued.
104 - MPI I/O allows opening files without specifying absolute path
105 - Note : SMPI configuration options are now initialized by smpi_init_options()
106 call, which should be called if SMPI is not being used through smpirun.
108 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
109 - FG#9: Python bindings crashing
110 - FG#39: Missing s4u::Comm::wait_any_for example
111 - FG#42: Add support for ThreadSanitizer (TSan)
112 - FG!19: Removing RngStream
113 - FG!20: A module for RNG calls
114 - FG!21: Choice between ad-hoc and standard distributions implementations
116 - FG!25: Fix link in Mutex doc
117 - FG!26: Fix links in SMPI interface doc
118 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
119 - GH#305: Unscheduled tasks are still excuted
120 - GH#313: smpirun: manual outdated w.r.t. --help
121 - GH#321: [S4U] Get task remaining work ratio
122 - GH#323: Crash when an actor turn off his physical host
123 - GH#335: Missing links on dragonfly example svg
124 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
125 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
127 ----------------------------------------------------------------------------
129 SimGrid (3.24) October 9. 2019
131 The Clean Disk Release.
134 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
135 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
136 thus becomes the default resource to execute I/O operations. Examples
137 and the file system plugin have been modified to use disks instead of
138 storage elements. Storage can still be used but is doomed to disappear.
139 Users are thus adviced to update their simulators accordingly.
140 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
141 for consistency with pthread_barrier_wait()
142 - Host::get_englobing_zone() returns the englobing netzone
143 - Actor::on_destruction is now called in the destructor
144 Actor::on_termination new signal called when the actor terminates
146 - Global signals are now part of the Engine:
147 - on_platform_creation: after config settings, before the XML parsing
148 - on_platform_created: right after the XML parsing
149 - on_time_advance: each time the clock advances
150 - on_simulation_end: after simulation, before cleanups
151 - on_deadlock: as the name implies.
153 - sg_{actor,host,link}_{data,data_set}() now all exist.
154 Use them to attach user data to the object and retrieve it.
157 - Introduce an experimental Wifi model. It sounds reasonable
158 according to the state of the art, but it still has to be properly
159 validated, at least against ns-3 if not against reality.
160 - Improved the usability of ns-3. Several bugs were ironed out.
161 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
162 It is now expressed as 'idle:epsilon:allCores' to properly model the
163 consumption of non-whole tasks on mono-core hosts. More info in the doc.
166 - convert a new set of functions to the S4U C interface and move the old MSG
167 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
170 - Fortran bindings for DVFS have been removed.
171 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
172 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
173 - SMPI can now be selected by cmake's find_module(MPI) with
174 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
175 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
176 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
177 - trace-call-location can be used with TI traces, and replayed, and is compatible
178 with smpi/comp-adjustment-file.
179 - sleep events are now correctly traced/replayed.
180 - Default for trace-call-location is now to use file names and not full paths.
181 To revert to previous behavior (in case of collision of filenames), option
182 "smpi/trace-call-use-absolute-path" can be set to yes.
185 - Use the included xxHash as an hash implem when C++14 is usable.
186 - Option model-checker/hash was removed. This is always activated now.
187 - New option smpi/buffering controls the MPI buffering in MC mode.
188 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
189 This is useful to check for MPI compliance.
192 - New section on plugins: how to define a new one, and existing ones.
195 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
196 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
199 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
200 and <mount> tags. The rationale is finish to decouple what should be handle
201 by SimGrid kernel (i.e., resources and activities) from plugins built in
202 user space (e.g., the file system plugin). This new tag currently accepts
203 only three attributes (id, read_bw, and write_bw). All the other information
204 that was declared with the storage related tags now has to be expressed as
205 properties. An example of platform using this new tag is available at
206 examples/platforms/hosts_with_disks.xml
209 - 'expect signal' can now accept more than one potential signal.
211 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
212 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
213 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
214 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
215 - FG#31: per-actor data
216 - FG#34: SG_BARRIER_SERIAL_THREAD?
217 - FG#35: model-checker does not like buster-produced binaries
218 - FG!13: MC: complete workaround in the error msg seen on modern systems
219 - FG!15: execute_flops now logs compute
220 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
221 - FG!17: ns-3: unblock the right number of communications + others issues
222 - FG!18: Improving the performance of the ns-3 bindings
223 - GH#189: Energy consumption of parallel tasks
224 - GH#219: Error in the throughput of TCP transfer
225 - GH!330: Minor polishing of English: Configuring_Simgrid
226 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
227 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
228 - GH!334: Rework energy plugin, again
230 ----------------------------------------------------------------------------
232 SimGrid (3.23.2) July 8. 2019
235 - Nicer introduction page.
236 - Migrate the "Deploy your application" page to the new doc.
237 - Move Java as a subtree of MSG.
240 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
243 - Drop xbt_dynar_sort_strings().
246 - Really fix FG#26: Turning off a link should raise NetworkFailureException
247 - FG#27: Wrong exception thrown to wait_any when link is turned off
248 - GH#328: Java: Canceling multiple tasks in a single vm/host
250 ----------------------------------------------------------------------------
252 SimGrid (3.23) June 25. 2019
254 The Exotic Solstice Release.
257 - SunOS and Haiku OS support. Because exotic platforms are fun.
258 - Stop setting random seed with srand48() at initialization.
259 - Use addr2line as a fallback for stacktraces when backtrace is not available.
260 - Build option -Denable_documentation is now OFF by default.
261 - Network model 'NS3' was renamed into 'ns-3'.
264 - Simgrid can now hopefully be installed with pip.
267 - wait_any can now be used for asynchronous executions too.
270 - New log appenders: stdout and stderr. Use stdout for xbt_help.
271 - Drop xbt_dict_dump.
274 - 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
275 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
276 - MPI interface is now const correct for input parameters
277 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
278 - Fortran bindings for async collectives.
279 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
282 - Remove option 'model-check/record': Paths are recorded in any cases now.
283 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
284 always sparse. That's an immense gain in memory, and can even be
285 faster because of cache effects. So there is no need to clutter the
286 code to allow the user to go for the unefficient mode.
289 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
290 models proved to be more accurate than these old experiments.
292 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
293 - FG#1: Broken link in error messages
294 - FG#2: missing installation documentation
295 - FG#3: missing documentation in smpirun
296 - FG#6: Python bindings not available on PyPI
297 - FG#7: simple cmake call requires doxygen
298 - FG#8: make python bindings an optional dependency
299 - FG#10: Can not use MSG_process_set_data from SMPI any more
300 - FG#11: Auto-restart actors forget their on_exit behavior
301 - FG#12: -Denable_lto=OFF doesn't disable LTO
302 - FG#13: Installs unstripped file 'bin/graphicator'
303 - FG#14: Installs the empty directory 'doc/simgrid/html'
304 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
305 - FG#17: Dead link in doc (pls_ns3)
306 - FG#20: 'tesh --help' should return 0
307 - FG#21: Documentation link on http://simgrid.org/ broken
308 - FG#22: Debian installation instruction are broken
309 - FG#26: Turning off a link should raise NetworkFailureException exceptions
310 - GH#133: Java: a process can run on a VM even if its host is off
311 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
312 - GH#326: Valgrind-detected error for join() when energy plugin is activated
314 ----------------------------------------------------------------------------
316 SimGrid (3.22) April 1. 2019
318 The Easter Christmas Release (this one is somewhat late).
321 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
322 - Not complete yet: asynchronous activities (amongst others) are still missing
323 - Still ongoing: the interface may change in the future. We need more
324 testers! Please report any glitches.
325 - No new project using Java should start now. Please switch to Python.
328 - Some of the internal cleanups may lead to speed improvements:
329 - The hard limitation on the amount of simulated actors with
330 Java+Mac was removed. Now, the available memory is the only limit.
331 - Our refcounting was tidyied, leading to 10% speedups in some cases.
332 - We are still working on making our code robust to the actor kills
333 and hosts' churn. Things are improving, but it's not perfect yet.
334 - Replaced our own code to display a backtrace (that was forking addr2line)
335 with the Boost.Stacktrace library.
336 You won't see your backtraces without this optional dependency.
337 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
338 - Stop setting random seed with srand() at initialization.
341 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
342 XML file version remains 4.2 since old files are still compatible.
345 - Process termination which was broken at version 3.21 has been repaired.
346 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
347 - Hide the examples into examples/deprecated. New users should use Python.
350 - Drop MSG_process_create_from_stdfunc() from the API.
351 This C++ function was a pimple in the C API, made necessary at some
352 point by the Java bindings. This is fixed now.
353 - Hide the examples into examples/deprecated. New users should use S4U.
354 - MSG_process_create and MSG_process_attach now crash if the host used
355 to run this process is off.
356 - Fix the protype of MSG_process_on_exit()
357 Now use (int,void*) callbacks instead of (void*,void*) ones.
358 The implementation was ways too messy to actually work, I guess.
361 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
362 too often and provide more accurate timings.
363 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
364 - MPI_Alltoallw support
365 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
366 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
367 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
368 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
370 - Basic implementation of generalized requests (SMPI doesn't
371 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
374 - Drop sg_cmdline. Please use xbt_cmdline instead.
375 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
376 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
377 OsSemaphore is implemented in a portable way with C++11 threads.
378 This should allow much more threads to be created at the same time,
379 allowing Mac and Java users to simulate many more actors.
380 - Implement the 'thread' factory with std::thread instead of xbt ones.
381 It is not possible to set the stack size with threads anymore, but
382 -fsplit-stack is the way to go nowadays when using threads.
383 - Drop the xbt_os_thread_t module (now unused)
384 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
385 - Drop xbt_str_join_array().
386 - Drop cunit, use Catch2 instead.
389 - Many cleanups in the kernel::activity namespace. This was long
390 overdue, and shall open the path to many future endeavors.
393 - #132: Java: a process can not shut down its own host
394 - #220: S4U: detached send causes exception if sender terminates soon after sending
395 - #261: Document the parameters of parallel execution's constructor
396 - #300: [s4u] BarrierPtr is missing
397 - #314: SMPI args internal cleanup
398 - #316: Fix a bug related to the CPU utilization of multi-core VM
399 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
400 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
401 - #325: Turning off a host has different behavior on sleeping actors and computing actors
403 ----------------------------------------------------------------------------
405 SimGrid (3.21) October 3. 2018
407 The Restarting Documentation (TRD) Release.
410 - Convert the user manual to Sphinx for improved usability.
411 Unstable version now hosted on https://simgrid.frama.io/simgrid
412 Many glitches remain; stay tuned.
413 - Restrict the Doxygen reference API to the installed header files.
414 (documenting all internals this way is overkill + doxygen scales badly)
415 - New tutorial on S4U (highly inspired from the old MSG one, but with
416 a git repository to fork as a starting point, and a docker image)
417 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
418 - The Developper manual is still to be converted (not compiled until then)
419 - Some parts are still missing in this conversion (such as the
420 platform and deployment sections) while others were blindly converted
421 and would need more love (such as the configuration flags).
422 Things will certainly further improve in the future releases.
425 - Fully reimplement auto_restart mechanism that was utterly broken
426 (fix #22, #131 and #281 that were all very related)
427 - Implement semaphores (fix #303 and #291)
428 - When creating an actor from a function and its parameters,
429 move-only parameters are not allowed anymore, as it would prevent
430 the actor to be restartable if its parameters are consumed on run.
431 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
432 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
433 IO operations that can be started, waited for, or canceled as a regular
435 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
439 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
440 - New host variable: 'core_count'
443 - Remove xbt_os_thread_specific features
444 - Remove portability wrapper to condition variables
445 - Remove xbt_os_thread_yield()
448 - MPICH collective selector now mimics MPICH 3.3b
449 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
450 - MPI_Init(NULL,NULL) is now allowed
451 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
452 - Command-line arguments should now be handled properly in Fortran simulations
453 - MPI Attributes and related callbacks should work in Fortran
454 - Apps using C + Fortran should now work
455 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
456 - Support MPI_Error_String
459 - Due to an internal bug, Msg.run() must now be your last line.
460 We hope to fix it in a future release, and we are sorry for the inconvenience.
463 - #22: Process autorestart seem to only work with CAS01 cpus
464 - #56: Feature request: dynamic SMPI replay
465 - #93: simgrid should not eat --help
466 - #111: How to change the loopback link characteristics?
467 - #116: Platform cluster radical
468 - #117: Platform cluster/host order
469 - #129: Replay_multiple on steroid
470 - #131: Java : autostart does not work
471 - #143: Setting a breakpoint at a given time
472 - #235: Network energy models should be integrated
473 - #264: Add ptask L07 resource tracing
474 - #271: Dynamic and manual replay of SMPI traces
475 - #279: Breakpoints option uses time deltas instead of absolute time
476 - #281: Daemonized actor no longer daemonized after an auto-restart
477 - #282: TIT convertor erase traces when using absolute path in the trace list file
478 - #285: segfault when a s4u actor kills itself with msg process tracing activated
479 - #286: Pajé traces are not exposing the number of cores
480 - #287: Command-line parsing should stop when encountering '--'
481 - #288: MPI_Init(NULL, NULL)
482 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
483 - #290: Method to check if a mailbox has a message ready to be consumed
484 - #291: [s4u] semaphores do not seem available yet
485 - #294: Alltoall collective implementation of mpich
486 - #295: Starting a Host as OFF is not supported
487 - #296: DTD too permissive
488 - #299: Add a s4u barrier example
490 ----------------------------------------------------------------------------
492 SimGrid (3.20) Released June 24. 2018
494 The proxy snake_case() release.
497 - The whole API is now using snake_case() instead of mixing it with camelCase.
498 Compatibility wrappers in place for one year.
499 - We took the opportunity to increase the consistency of the API.
500 Please report any remaining glitches.
501 - New function simgrid::s4u::Host::get_actor_count:
502 Returns the number of actors running on a specific host.
505 - SMPI is now tested with ~45 proxy apps from various sources, with none or
506 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
507 - Replay: The replay file has been re-written in C++.
508 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
509 supported. They are stored in the trace and used when replayed.
510 - Basic support of MPI_Cancel. Robustness not guaranteed.
511 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
512 (only for MPI_COMM_TYPE_SHARED).
513 - New option: smpi/privatize-libs, to add external shared libs to be privatized
514 by SMPI. They will be copied locally and loaded separately by each process.
515 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
516 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
517 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
518 with internal index optind with multiple processes. Only works if MPI_Init has
520 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
521 syntax without preprocessing tricks.
524 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
525 behavior tracing is no longer limited to MSG
526 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
527 this means that traces now need to be updated or re-obtained; both Irecv
528 and Isend lines in your traces have now in total 5 mandatory fields:
529 <rankid> <command> <to/from rankid> <tag> <size>
530 To update your traces, it suffices to add a 0 for the tag here.
531 - TIT now also supports waiting for a distinct request via MPI_Wait.
532 Wait/Test now wait for a specific request, not just the last one that was
533 issued. This unfortunately means another update, because we need to
534 identify which request you want to wait for. We do this via the
535 triplet (sender, receiver, tag), which needs to be added:
536 <rankid> <command> <sender> <receiver> <tag>
537 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
538 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
541 - The deprecation of MSG is ongoing (but this should not impact you).
542 Many MSG functions are now simple wrappers on the C API of S4U. If
543 you wish to convert your code to S4U, find the S4U counterparts of
544 your MSG calls in src/msg/msg_legacy.cpp.
545 - MSG can still be used, but won't evolve anymore.
548 - Allow to run the Link energy plugin from the command line with
549 --cfg=plugin:link_energy
550 - Rename Energy plugin into host_energy
551 - Rename Load plugin into host_load
554 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
555 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
556 - Improve performance when handling timeouts of simix synchros.
559 - Config: the C API is now deprecated (will be removed in 3.23), and
560 the C++ API has been slightly improved.
563 - Fix several build issues on OSX.
564 - Move simgrid_config.h to simgrid/config.h (old header still working)
567 - #143: Setting a breakpoint at a given time
568 - #150: Inconsistent event names in SMPI replay
569 - #258: daemonized actors hang after all non-daemonized actors have completed
570 - #267: Linker error on unit_tmgr
571 - #269: SMPI: tracing of MPI_Wait/all/any broken
572 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
574 ----------------------------------------------------------------------------
576 SimGrid (3.19.1) Released March 22 2018
578 The Fixed ABI Release.
580 As you may know, we are currently refactoring SimGrid in deep.
581 Upcoming SimGrid4 will be really different from SimGrid3: modular,
582 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
584 Our goal is to smooth this transition, with backward compatibility
585 and automatic update paths, while still progressing toward SimGrid4.
587 SimGrid remains open during works: The last pure SimGrid3 release was
588 v3.12 while all subsequent versions are usable alpha versions of
589 SimGrid4: Existing interfaces remain unchanged, but the new S4U
590 interface is budding and the internals are deeply reorganized.
592 Since 2015, we work hard to reduce the changes to public APIs. When
593 we need to rename a public library symbol in S4U, we let your
594 compiler issue an explicative warning when you use the deprecated
595 function. These messages remain for four releases, i.e. for one full
596 year, before turning into an error. Starting with v3.15, your can
597 also adapt to API changes with the SIMGRID_VERSION macro, that is
598 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
600 Starting with this v3.19.1, our commitment to reduce the changes to
601 the public interfaces is extended from the API to the ABI: a program
602 using only MSG or SimDag and compiled against a given version of
603 simgrid can probably be used with a later version of SimGrid without
604 recompilation. We will do our best... but don't expect too much of
605 it, that's a really difficult goal during such profund refactoring.
607 The difference between v3.19 and v3.19.1 is that the former was
608 accidentally breaking the ABI of MSG, while the later is restoring
611 S4U and kernel APIs will still evolve until SimGrid4, with one-year
612 deprecation warnings as currently. In fact, cleaning up these
613 interfaces and converting them to snake_case() is one release goal of
614 v3.20. But don't worry, we are working to smooth this upgrade path.
616 In summary, new projects should start with S4U to benefit of the
617 future, but old MSG projects should still be usable with no change.
619 ----------------------------------------------------------------------------
621 SimGrid (3.19) Released March 20 2018
623 The Moscovitly-cold Spring release.
625 SMPI user-visible improvements:
626 - Support of MPI_Type_create_subarray
627 - Unimplemented MPI calls now abort (instead of issuing a warning)
628 Some harmless calls around errhandlers still only issue a warning.
629 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
630 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
631 is now possible (mmap privatization was refactored to this end).
632 Adding new SMPI instances should also be technically doable now.
634 SMPI internal cleanups:
635 - Removed index notion from SMPI -- no more "getPid() - 1"!
636 In tracing files, actors are called rank-<PROCESS-ID> but these are
637 NOT the real ranks (tracing several communicators is not supported yet).
638 - Deprecate smpi_process_index() (will be removed in 3.22)
639 - Better testing on classical proxy apps, through an external project:
640 https://github.com/simgrid/SMPI-proxy-apps/
643 - Use a graphical TOC to make it easier to find the documentation you need
644 - Revamp the MSG tutorial
647 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
648 - Remove parameter of MSG_process_killall().
649 Resetting the PID was bogus anyway (several actors could have the same PID).
652 - Execution->setHost() can be called after start(), to migrate it.
653 - Comm::test_any() is now implemented.
654 - s4u::Actor now has onCreation() and onDestruction() signals.
655 - Install some kernel header files for the users' plugins and more.
656 Warning, their API is really not stable yet.
657 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
658 include/simgrid/kernel/routing/*Zone.hpp
661 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
663 Build System and other cleanups:
664 - Remove unused run-time parameter "tracing/onelink-only".
665 - SimGrid now works with intel compilers (MC still buggy)
666 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
667 - tesh kills subprocesses on timeout (but not on windows)
670 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
673 - #194: Feature request: simgrid::s4u::Comm::test_any()
674 - #245: migrating an actor does not migrate its execution
675 - #253: Feature Request: expose clusters as objects
676 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
677 - #255: Tesh broken on Windows
678 - #256: Modernize FindSimGrid.cmake
679 - #257: Fix (ab)use of CMake install
681 ----------------------------------------------------------------------------
683 SimGrid (3.18) Released December 24 2017
685 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
687 MSG IS NOW DEPRECATED. Please use S4U for new projects.
688 - Support for MSG should not be removed from SimGrid before 2020,
689 but future evolutions will be limited to the integration of
690 user-provided patches.
693 - s4u::onDeadlock() signal, run before stopping the simulation.
694 - s4u::Exec: asynchronous executions (abstraction of background
696 - s4u::Host->getLoad() returns the achieved speed in flops/s
699 - Link::name() is deprecated and will be removed in v3.21, use
700 Link::getCname() instead.
701 - Mailbox::getName() changed to return a std::string, use
702 Mailbox::getCname() instead to get a char*.
703 - Storage::getName() changed to return a std::string, use
704 Storage::getCname() instead to get a char*.
705 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
706 to have both allocation and deallocation of the map in user space.
709 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
710 The amount of flops remaining to do is a vector, not a scalar.
711 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
712 name implies on both sequential and parallel tasks.
713 - Both changes fix GitHub's #223 using PR #237 as a basis.
714 Thanks Michael Mercier.
715 - Most examples were converted to S4U and hidden elsewhere at tests.
716 Rationale: we still want MSG to work; we want newcomers to use S4U.
719 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
720 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
723 - Switch to the faster dlopen privatization mechanism by default
724 - Documentation improvments
727 - Live migration is getting moved to a plugin. Dirty page tracking is
728 the first part of this plugin. This imply that VM migration is now
729 only possible if one this function is called:
730 - C/MSG: MSG_vm_live_migration_plugin_init()
731 - C/C++: sg_vm_live_migration_plugin_init()
732 - Java: Msg.liveMigrationInit()
733 For C and C++, "simgrid/plugins/live_migration.h" also has to be
737 - Define class simgrid::xbt::Path to manage file names.
738 - Removed unused functions:
739 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
740 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
741 xbt_graph_export_graphviz()
742 - xbt/str.h: xbt_str_join()
743 - Remove unused datatypes:
744 - xbt/heap.h: use std::priority_queue or boost::heap instead
745 - xbt/swag.h: use boost::intrusive::list instead
748 - New link_energy plugin for the consumption of the links.
749 - All of the operations on files and storage contents have been
750 packaged into a plugin (src/plugins/file_system). The current
751 public interface can be found in
752 include/simgrid/plugins/file_system.h
753 To use these functions you now have to initialize the plugin by
754 calling MSG_storage_file_system_init() just after calling
755 MSG_init() or sg_storage_file_system_init() just after creating
759 - Remove the undocumented/untested tag <include>
762 - Remove viva specific tracing as the tool is no longer maintained
765 - #248: Kill of finished processes leads to segfault
766 - #240: xbt_cond_wait_timeout should gracefully return for C apps
767 - #239: please implement signal s4u::onDeadlock()
768 - #230: segfaults when exit() without run()
769 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
770 - #223: MSG_task_get_flops_amount() not working with parallel tasks
771 - #222: Actor::kill() doesn't really kill and segfaults
772 - #221: odd LMM warning when killing an actor
773 - #120: Memory leak when the processes are forcefully killed
775 ----------------------------------------------------------------------------
777 SimGrid (3.17) Released October 8 2017
779 The Drained Leaks release: (almost) no known leaks despite the tests.
781 Even very long simulations will preserve your memory: our testsuite
782 stresses SimGrid for over 45mn on fast machines for a coverage of
783 over 80%, and there is only one single known leak, of about 4kb.
787 - Comm.detach(): start and forget about asynchronous emission. A cleanup
788 handler may be given to free resources if the comm cannot be completed.
789 - this_actor::send(mailbox) is now mailbox->put()
790 - New: simgrid::s4u::Comm::wait_all()
791 - New: Host.execute() for remote executions.
794 - Unused option network/sender-gap is removed.
797 - Deprecate MSG_task_isend_with_matching(): This unused feature
798 really complicates our internals. Will be removed in v3.20.
801 - Improved context termination. It is now possible to free resources, even
802 when a process is forcibly killed.
803 - ContextBoost: add support for Boost versions above 1.61.
806 - Bring back run-time option --cfg=exception/cutpath to remove exception
808 - Removed unused functions:
809 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
810 xbt_str_rtrim(), xbt_str_trim().
811 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
814 - Removed header files obsolete since SimGrid 3.12:
815 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
817 - #3: SD_exit should be made optional
818 - #120: Memory leak when the processes are forcefully killed
819 - #159: Threading test regression in Actor refcounting
820 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
821 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
822 - #186: Actor::killAll() segfaults if some process is blocked on wait()
823 - #191: VM migration and pstate
824 - #192: Updating the energy consumptions of all hosts crashes with VMs
825 - #195: All actors have PID=0 in the logs
826 - #204: Sometimes segfault with thread contexts and mmap privatization
827 - #222: Actor::kill() doesn't really kill and segfaults
828 - #225: Actor::kill() doesn't really kill when victims are doing a join()
830 ----------------------------------------------------------------------------
832 SimGrid (3.16) Released June 22. 2017.
834 The Blooming Spring Release: developments are budding.
837 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
838 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
839 - Java JAR file should be finally fully working on Mac OSX too.
841 XML platforms: Switch to platform v4.1 format.
842 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
843 - <zone> can be used as a synonym for the now deprecated <as>
844 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
845 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
846 - <actor> can be used as a synonym for the now deprecated <process>
847 - state_file and avail_file periodicity is now easier to express
848 (check the documentation) the old behavior should still work.
849 * Storage: not backward compatible, but it seems that nobody ever used it
850 - Remove attribute 'content_type' of <storage_type>: was never used
851 - Make attribute 'model' of <storage_type> optional: for future usage
852 - Remove Bconnection model property: was never (in)validated. Replaced by
853 the maximum of the read and write bandwidth as a resource constraint.
856 - New and Backwards Compatibility break:
857 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
858 When one wants to get the list of tasks whose states have changed during a
859 simulation round, s/he has to allocate and free a dynar and use it as
860 argument to this function. The former SD_simulate (double how_long)
864 - Allow multicore VMs, with the correct sharing (unless you overcommit)
865 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
868 - New function to update the consumption of all hosts at once.
869 - Fix the model for multi-core hosts, linear in the amount of busy
870 cores with an abnormality for pIdle (see paper).
871 BUG: Still not satisfactory for ptask on mono-cores.
874 - The netzone are now available from the MSG API.
875 The old names still work, but are now deprecated.
878 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
879 - New option: smpi/keep-temps to not cleanup temp files
880 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
881 mallocs mode is used, allows to change the size of the fake file used
882 (default 1MB), to potentially limit the number of mappings for large runs.
883 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
884 - Fortran ifort and flang compilers support
885 - New RMA calls supported (experimental):
886 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
887 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
888 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
889 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
890 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
891 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
894 - New: this_actor::isMaestro() returns whether we are in kernel mode.
895 - Behind the scene, ActivityImpl are now automatically refcounted.
896 Removing this nasty bottleneck will greatly simplify our internals.
899 - Replay: New function xbt_replay_action_get():
900 Retrieve the function previously associated to an event type.
901 - DROPPED FUNCTION: xbt_str_varsubst()
902 - DROPPED MODULE: strbuff. We don't need it anymore.
903 - DROPPED MODULE: matrix. We don't need it anymore.
904 - DROPPED MODULE: lib. We don't need it anymore.
906 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
908 SimGrid (3.15) Released March 22. 2017
910 The Spring Release: continuous integration servers become green
911 We fixed even the transient bugs on all target architectures:
912 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
913 Windows, FreeBSD, NetBSD.
915 - Rename NetCards to NetPoints.
916 This was intended to help NS3 users, but that's not a netcard.
917 That's a point in the routing algorithm, let's avoid wrong simplifications.
920 - New: MSG_process_yield(). Stop and yield to other processes.
921 - New: MSG_process_daemon(). Daemon processes are automatically killed
922 when the last non-daemon process terminates
923 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
924 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
925 to make room for the upcoming network energy plugin.
926 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
929 - Ensure that an actor can kill itself with Process::exit()
930 - Kill the obscure NativeException. Nobody want to survive the issues
931 it denotes, so use JniException that is a RuntimeException (not to
932 be caught explicitly).
933 - Partial bug fix in initialization. SimGrid flags on command line were
934 consumed at C level but stayed in the original Java String[] args.
935 This could mess users' args[i] if SG flags were not put at the end of
937 The SimGrid flags are now removed from the Java arguments. However,
938 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
939 if args.length is greater than the number of YOUR OWN ARGUMENTS.
940 It might be if you have --log or --cfg flags in the command line.
941 - Fix numerous memleaks all around the place. In particular, around VMs.
945 - s4u::onPlatformCreated: right before the simulation starts
946 - s4u::onSimulationEnd: right after the main simulation loop
947 - s4u::onTimeAdvance: right after a clock change (time is discrete)
948 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
949 event from the availability_file changes the avail speed.
950 - Links are now usable from s4u
951 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
952 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
953 - New examples: The conversion of MSG examples to S4U has begun
954 - Actors: create, kill, migration, and suspend
955 - Applications: master-worker and token-ring
956 - Action replay: communications and storage
957 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
960 - Backwards Compatibility breaks
961 - The SD_route_get_list and SD_route_get_size functions have been removed.
962 They are replaced by the unique
963 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
964 The route from src to dst is built in the links dynar whose size can
965 be obtained with xbt_dynar_length.
966 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
967 sg_host_route_bandwidth and sg_host_route_latency.
968 Macros ensure the backwards compatibility, but you should fix your code
971 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
972 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
973 - Bump our claim of support from MPI 1.1 to MPI 2.2.
974 We don't support 100% of it, but it should be enough. Contact us if not.
975 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
976 naive and probably inaccurate)
977 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
978 mapped onto a unique small file using some system magic.
979 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
980 - Fortran: cleanups, fixes, support of user-added operations
981 - RMA: MPI_Accumulate are applied in correct order.
982 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
983 - Better support for MPI_IN_PLACE
984 - Support for MPI_Win attrs and keyvals.
985 - Support MPI_Comm_group_create, MPI_Type_size_x
986 - MPI ops should be properly applied to all allowed types
987 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
988 algorithm by a non failing one from mpich.
991 - Kill the fifo data container: we don't use it anymore.
994 - A <cluster> can now be created with different speed values to represent pstates.
995 see examples/platforms/energy_cluster.xml for an example.
997 SimGrid (3.14.159) Released December 28. 2016
999 The previous release was only a crude approximation of a Pi release;
1000 this one should be a bit better.
1002 - Revert a last minute change that broke on Mac OSX.
1003 - Fix the changelog and NEWS files.
1004 - Improve the documentation, in particular of the routing module.
1005 - Integrate some patches flying around in Debian.
1007 SimGrid (3.14) Released December 24. 2016
1010 * The whole documentation was reworked and reorganized. There is
1011 still much room for improvement here, but we're on it.
1014 * We now need python3 for our testing infrastructure.
1015 * Model checking needs libevent
1018 * Now works on FreeBSD
1021 * Privatization now works on FreeBSD
1023 * Privatization is now activated by default in smpirun
1024 You can turn it off with -no-privatize if something goes wrong.
1026 * Call-location tracing for SMPI.
1027 You can add the exact location (filename / linenumber) of an MPI call to
1028 your trace files and slow down or speed up the simulation between two
1029 consecutive calls by using an adjustment file (see the documentation).
1031 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1032 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1033 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1034 values were ignored in some cases. The timings of these functions can now
1035 be significantly different.
1037 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1038 smpi/running-power is renamed to smpi/host-speed
1040 * smpi/grow-injected-times option to enable or disable multiplication of the
1041 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1042 make simulation less precise (but also much faster).
1044 * smpirun script should be (much) faster for large deployments.
1046 * SMPI tracing: fixed issue with poor matching of send/receives.
1048 * Replay: Fix broken waitall
1050 New functions and features
1051 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1053 Dropped / renamed functions and features
1054 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1055 - MSG_mailbox_is_empty() -> Mailbox::empty()
1056 - MSG_mailbox_front() -> Mailbox::front()
1057 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1058 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1059 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1060 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1061 * Don't pass the free_f parameter to property related functions:
1062 - MSG_host_set_property_value()
1063 - MSG_as_router_set_property_value()
1064 - MSG_storage_set_property_value()
1065 * VM properties. Since msg_vm_t are msg_host_t, just use
1066 MSG_host_get_property_value() and friends
1067 * VM I/O related things:
1068 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1069 - Unimplemented save/restore methods
1070 * MSG_as_router_get_property_value() was redundant with
1071 MSG_environment_as_get_property_value().
1072 - Removed MSG_as_router_*propert*() functions
1073 - Added MSG_environment_as_set_property_value() for consistency
1074 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1075 Well, they are still there for now, but deprecated with a warning.
1076 Please switch to xbt_dict_new_homogeneous() before this is removed
1078 * Task affinity. Its intended behavior (that was very badly tested
1079 and probably not really working) was deceiving what most users
1080 would have hoped here.
1081 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1082 * Option network/coordinates is now useless and should be dropped.
1085 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1088 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1089 * Added Dragonfly topology support
1092 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1093 EiBps, EBps, Eibps, Ebps for bandwidth.
1094 They may become useful to some lucky ones.
1097 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1098 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1099 * ex/async_waitAll: new example, on asynchronous communications
1102 * Memory usage should be decreased for simulations with a large number
1103 of processes. This also helps for SMPI.
1105 SimGrid (3.13) stable; urgency=low
1107 The Half Release, a.k.a. the Zealous Easter Trim.
1109 - We removed half of the lines, that were mostly experimental cruft.
1110 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1111 (not counting blanks and comments -- according to openhub.net)
1112 - The internals are now compiled in C++ (and will soon be clean C++)
1113 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1114 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1116 * Backwards Compatibility breaks
1117 - Removed Lua simulation bindings (switch to C or Java for that).
1118 Lua can still be used to describe platforms
1119 - Removed Java kernel plug-ins.
1120 Will be reintroduced after the ongoing major internals reorg.
1122 - the following functions were removed.
1123 They were too specific and should be reimplemented in a generic
1124 way, with filter function.
1125 - MSG_task_listen_from_host
1126 - MSG_mailbox_get_count_host_waiting_tasks
1127 - MSG_mailbox_put_with_timeout was removed.
1128 Please use MSG_task_send_with_timeout instead.
1130 - the SD_application_reinit function was removed. It has been a noop for a while.
1131 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1132 replaced by a more flexible mechanism.
1133 - The following functions thus do not exist anymore
1134 - SD_workstation_get_access_mode
1135 - SD_workstation_set_access_mode
1136 - SD_workstation_get_current_task
1137 - Basic estimation functions have been removed but can easily be replaced
1138 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1139 - SD_workstation_get_computation_time => amount / sg_host_speed()
1141 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1142 Use VM.getSpeed()*load/100 for the legacy behavior.
1144 - option enable_tracing was removed. It was not doing anything for a while.
1145 - In the ModelChecker:
1146 - the model-checker now ptraces the model-checked process which means
1147 you cannot use a debugger on the latter anymore (we might make this
1148 optional in the feature);
1149 - removed soft-dirty page tracking;
1150 - remove model-checked side snapshot management,
1151 MC_snapshot() and MC_compare_snapshot();
1152 - keep the MC_cut() function as a stub (it was not really working
1153 in the previous release).
1156 * All options are consistently in kebab-case. Old names are kept as alias.
1159 * Switch to platform v4 format.
1160 - Rename from 'power' to 'speed' the attributes describing the amount of
1161 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1162 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1163 - In <host> and <link>, attributes availability and state are gone.
1164 It was redundant with state and availability traces, and with peak values.
1165 - In <cluster>, attributes availability_file and state_file are gone.
1166 It was too complex and unused.
1167 - Kill <gpu>. Was not doing anything.
1168 - The DOCTYPE points to the right URL:
1169 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1170 (the file at this address now documents the changelog since its v1)
1171 - A warning is emitted for unit-less values (they are still accepted).
1172 - speed. Default: 'f' or 'flops'. Also defined:
1173 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1174 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1175 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1176 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1177 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1178 - latency. Default: 's' second. Also defined:
1179 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1181 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1182 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1185 * s4u::Host is now the preferred public interface to the Host features.
1186 sg_host_* functions are C bindings to the exact same behavior
1187 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1190 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1191 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1192 They were deprecated since a few years, and probably did not even compile anymore.
1195 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1196 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1197 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1198 encouraged to update their codes. A list of the modified functions can be found at the end of
1199 include/simgrid/simdag.h
1202 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1203 the lambda or closure passed as a parameter will run in kernel mode.
1204 All the callback functions should be rewritten to that interface at some point.
1207 * Reorganizing and cleaning the internals all around the place.
1210 * Remove old default barrier/bcast buggy algorithms (see #18407)
1211 * Various bug fixes to handle more codes
1212 * Remove the need for the --foreground option of smpirun (it is still
1213 accepted for backward compatibility).
1216 * Kill the setset data container: MC don't use it anymore.
1217 * Kill the queue data container: it made more sense with GRAS.
1218 * Kill the xbt_peer_t data type: it's useless without GRAS.
1219 * Kill rm_cb feature of config sets: it was never useful.
1220 * Kill graphxml parsing feature. It was not used.
1221 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1223 - xbt_dynar_sort_strings(), when the content is char*
1224 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1225 They throw exceptions on invalid input;
1226 * C++ support for declaring CLI flags (simgrid::config::Flag);
1227 * class for abstracting different signal backends (simgrid::xbt::signal).
1228 with no external dependencies (we need very simple signals).
1231 * refactoring and cleanup of the code;
1232 * ongoing process to cleanly separate the model-checking algorithms
1233 from the code model-checking support.
1235 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1237 SimGrid (3.12) stable; urgency=low
1239 The Facelift Release.
1242 * Require g++ v4.7 at least to not speak prehistorical C++.
1243 * Require Boost 1.48 (for signal2 component).
1244 * Java must be version 7 at least when activated.
1245 * Builds on Windows again (including Java bindings).
1246 * Tracing is now always enabled (no way to turn it out)
1247 * Remove GTNetS. It was not working anyway.
1248 * Various cleanups in the cmake scripts.
1249 * Move headers around to sort them out on installed systems:
1250 - instr/instr.h -> simgrid/instr.h
1251 - instr/jedule/* -> simgrid/jedule
1252 - simdag/datatypes.h was removed
1253 - simdag/simdag.h -> simgrid/simdag.h
1254 - msg/datatypes.h was removed
1255 - msg/msg.h -> simgrid/msg.h
1258 * Interface improvement:
1259 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1260 - Sanitize the interface in MSG_task_ module:
1261 - Merge two functions that were close enough but misleading:
1262 set_compute_duration(t) -> set_flops_amount(t)
1263 get_remaining_computation(t) -> get_flops_amount(t)
1264 - set_data_size(t) -> set_bytes_amount(t)
1265 get_data_size(t) -> get_bytes_amount(t)
1266 - Massive cleanups in the functions related to the energy
1267 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1268 - New: MSG_host_get_pstate()
1270 - msg/energy/onoff: switching hosts on and off
1272 * Interface improvement:
1273 - Rename Host.isAvail() to Host.isOn()
1274 - Rename Process.currentProcess() to Process.getCurrentProcess()
1275 - Rename Task.setDataSize() to Task.setBytesAmount()
1276 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1278 - #18874: Actually allows the GC to reclaim tasks
1282 - SIMIX_process_throw: raises an exception in a remote process
1283 * Refactoring: Separate sub-modules
1284 - libsmx: the public interface, as libc in a real system
1285 - popping: the strange dance that converts a user request into a kernel handling
1286 - smx_context_*: the virtualization mechanisms that embed the user code
1287 - smx_*: the handling of each simcalls
1288 * Interface cleanup:
1289 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1290 * Rename smx_action_t into smx_synchro_t, making explicit that these
1291 things are used to synchronize processes with their environment.
1292 For example, a communication is a sort of synchronization involving
1293 the communicating processes (that may block until the exchange) and
1294 the platform. The same can be said from computations, etc.
1296 - #18888: segfault when a process ends before its kill_time
1300 - 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.
1301 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1302 - MPI_Type_set_name, MPI_Type_get_name
1303 - MPI_*_c2f and MPI_*_f2c functions
1304 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1305 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1306 - Activate a lot of new tests from the mpich 3 testsuite
1308 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1309 - InfiniBand network model added: Based on the works of Jerome Vienne
1310 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1311 - When smpi/display_timing is set, also display global simulation time and application times
1312 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1313 * Collective communications
1314 - 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
1315 - 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/).
1316 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1318 - Replay now uses algorithms from wanted collective selector
1319 - Replay can be used with SMP-aware algorithms
1320 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1321 - 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.
1323 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1324 - Correctly use loopback on fat-tree clusters
1325 - Asynchronous small messages shouldn't trigger deadlocks anymore
1326 * Energy/DVFS cleanup and improvement
1327 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1328 - new: smpi_get_host_pstate()
1331 - "Full" network optimization flag was broken since Surf++
1332 - Better handling of precision flags in maxmin
1333 - Fix bug causing sometimes "Impossible" errors
1334 - Properly pass cluster properties to included hosts
1335 * Improvement of the Energy plugin.
1336 - Always update the consumption before returning that value
1337 - New property: watt_off to denote the dissipation when the host is off
1338 - New functions getWattMinAt and getWattMaxAt to retrieve the
1339 dissipation of pstates that we are not currently at.
1340 * Java: class NetworkLink renamed to Link
1341 * New function: simcall_process_get_kill_time()
1342 * Massive rename s/workstation/host/
1343 - That's intrusive, but that's good for the project consistency. Sorry.
1344 - Change config option "workstation/model" into "host/model"
1347 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1348 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1349 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1350 - xbt_dirname and xbt_basename for non-POSIX systems
1352 * The model checker now runs as a separate process.
1353 * The model checker runs must now be launched with the new simgrid-mc program.
1354 * Record/Replay: the MC can display a textual representation of a path in the
1355 execution graph. It can then be replayed outside of the model checker.
1357 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1359 SimGrid (3.11) stable; urgency=low
1364 * Normalizing pointers addresses tool for better diff between logs
1367 * Add cloud examples using new VMs
1368 - examples/msg/cloud/two_tasks_vm.tesh
1369 - examples/msg/cloud/simple_vm.tesh
1370 - examples/java/cloud/cloud.tesh
1371 - examples/java/cloud/migration/migration.tesh
1372 * Add java surf examples:
1373 - examples/java/surfPlugin/surf_plugin.tesh
1374 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1375 - examples/java/surfCpuModel/surf_cpu_model.tesh
1376 * Add SMPI+MSG example:
1377 - examples/smpi/smpi_msg_masterslave/
1382 - msg task destroy cancel test
1383 - msg_host on/off test
1384 * Move all tests in testsuite to teshsuite (adding tesh files)
1385 * Restructure teshsuites
1386 - one folder for each kind of test
1387 * Restructure AddTests.cmake
1389 - structure the order of tests (with sections)
1392 * Add virtual machine
1393 - creation of a VM on a PM
1394 - migration of a VM from a PM to another PM
1396 - MSG_process_join(msg_process_t process, double timeout)
1397 - msg_bar_t MSG_barrier_init(unsigned int count)
1398 - int MSG_barrier_wait(msg_bar_t barrier)
1399 - void MSG_barrier_destroy(msg_bar_t barrier)
1400 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1401 * New option "msg/debug_multiple_use" to help debugging when a task is used
1404 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1406 - Increase code coverage in test suites
1410 * Protect context stacks against stack overflow. The number of protected memory
1411 pages allocated on the top of each stack (1 by default) can be configured
1412 with the new command line option --cfg=contexts/guard_size:<value>.
1413 * Simcalls are now generated by a python script that
1414 - generates files included by SimGrid
1415 - checks that all the functions exist, and proposes prototypes
1417 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1418 host_set_data, host_get_data
1420 - simcall_process_join(smx_process_t process, double timeout)
1421 * Fix bug where sleeping processing could not be suspended.
1424 * Translate surf models from C to C++
1425 - Generic classes for all models: Model, Resource, Action
1426 - A generic interface for each kind of model (CPU, Network, Storage
1427 Workstation, WorkstationVM)
1429 * Translate surf routings from C to C++
1430 * Add callbacks using sigc++ or boost::signals2
1431 - Add callback functions for resource creation/destruction
1432 - Add callback functions for action state change
1433 - Handle Energy as a plugin
1434 * Replace swag by boost::intrusive
1435 * Add new routing models for clusters. For documentation, see
1436 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1437 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1438 parameters for cluster tag
1439 - Fat trees, with topology="FAT_TREE" and
1440 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1441 - see examples/platforms/torus_cluster.xml and
1442 examples/platforms/fat_tree_cluster.xml
1443 * More documentation
1446 * Hostfiles support host:nb_processes construct to deploy several processes on
1448 * Collective communication algorithms should not crash if used with
1449 improper number of nodes and report the error.
1450 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1451 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1452 MPI_Dims_create, MPI_Cart_sub are supported.
1453 * New interface to use SMPI programmatically (still depends on MSG for
1454 some parts, see examples/smpi/smpi_msg_masterslave):
1455 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1459 * Global variables privatization in MPI executables is now performed at runtime
1460 with the option smpi/privatize_global_variables (default:no).
1461 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1462 dynamic libraries loaded by the application are not privatized (static
1463 linking with these libraries is advised in this case)
1466 * Options defined in XML work correctly now.
1469 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1470 copied into simgrid.jar or not (ON by default). Use this option if you want
1471 to reduce the size of the installed simgrid.jar, *and* the native libraries
1472 are kept installed elsewhere.
1473 * Surf binding with SWIG (code generated in maintainer mode only):
1474 - plugin to handle callbacks
1475 - CPU model only for the moment
1478 * Supernovae build mode is definitively removed. It was used to improve
1479 inlining and inter-module optimizations. It is nowadays superseded by
1480 link-time optimizations commonly available in compilers.
1481 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1482 * Add boost dependency for surf++
1483 * Add new macro for tests
1484 - ADD_TESH(name <tesh_args>)
1485 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1489 - xbt_bar_t XBT_barrier_init(unsigned int count)
1490 - int XBT_barrier_wait(xbt_bar_t barrier)
1491 - void XBT_barrier_destroy(xbt_bar_t barrier)
1492 * Make the xbt_os_time module public
1494 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1496 SimGrid (3.10) stable; urgency=low
1498 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1501 * Reintegrate Java to the main archive as desynchronizing these
1502 package is not acceptable anymore (Java is now considered stable)
1503 * Add explicit synchronization facilities through semaphores
1504 * Bug fix: Task.setDataSize() only changed the C world, not the value
1505 cached in the Java world
1508 * Dramatically change the way files are handled. API and internals changed, but
1509 this part of MSG was not considered as production grade either.
1510 * Add explicit synchronization facilities through semaphores
1511 * Add a new function MSG_host_get_process_list()
1512 * Preliminary DVFS support (see examples/msg/energy/ for details)
1515 * SMPI is now included directly in the libsimgrid as the windows
1516 linker doesn't force us on splitting it anymore.
1517 * Improvements of the SMPI replay tool:
1518 - Most of the collective communications are now rooted in the same process as
1519 in the original application.
1520 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1521 used until now). Multiple data types can now be used in a trace.
1522 - The replay tool now supports traces produce either by TAU or a modified
1524 - Bug Fix: the compute part of the reduce action is now taken into account.
1525 - Gatherv collective is now supported
1526 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1527 of smpirun outputs time independent traces for the current run. One file
1528 is created per process. If too many processes are simulated, this behavior
1529 can be changed to one file for all processes by using the
1530 tracing/smpi/format/ti_one_file flag
1531 * smpirun generates the host file if needed (with given host count and platform)
1532 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1533 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1534 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1535 --cfg=smpi/coll_selector:(mpich/ompi)
1536 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1538 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1539 happening inside a collective SMPI call.
1540 * Fix the behavior of complex data types handling.
1541 * Make MPI_Wtime another synchronization point to take computations into
1543 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1544 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1545 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1546 (no privatization of global variables yet)
1547 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1550 * Verification of liveness properties is now available for SMPI applications
1551 (in addition to MSG applications)
1552 * Bugged examples using SMPI in examples/smpi/mc/
1553 * Add --cfg=model-check/visited option. Allows the verification of infinite
1554 programs. Detection of loops in the execution thanks to the system state
1555 comparison and reduction of the state space to explore. Can be combined with
1556 DPOR for safety properties.
1559 * Allow to change SimGrid configuration (see --help) within the code
1560 thanks to SD_config() as it can be done in MSG.
1561 * Add a new function SD_task_set_amount() upon user request.
1564 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1565 * Remove rule based routing (no more PCRE dependency)
1566 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1567 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1568 * Add a loopback_bw and loopback_lat options to cluster tag.
1569 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1570 routing. Now peers are encapsulated in an AS and have their own private
1571 router but this is transparent.
1574 * Our own implementation of getline is renamed xbt_getline, and gets
1575 used even if the OS provide a getline(). This should reduce the
1576 configuration complexity by using the same code on all platforms.
1577 * New type: xbt_cfg_elm_boolean.
1578 * Allow to use yes/no for boolean configuration options in the command line.
1579 * Allow to disable SimGrid cleanups at exit from command line option.
1580 There are situations where one may want a simulation to end with an exit.
1581 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1582 annoying when scripting around the simulator. Adding a
1583 --cfg=clean_atexit:no allows to circumvent this issue.
1586 * Lots of memory leaks were corrected in this release.
1587 There is no known memory leaks anymore, in all of our 600+ tests.
1588 * New command line option --version, to get SimGrid version information.
1589 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1591 * Supernovae builds are deprecated, and expected to be removed in the next
1594 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1596 SimGrid (3.9) stable; urgency=low
1598 The Grasgory release: GRAS is really dead now.
1600 * Complete overhaul of the internal host structures scheme.
1603 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1604 was considered as experimental and badly maintained since then.
1605 * Keeping it was thus a trap to our potential users, that could take
1606 it instead of MSG or SMPI by mistake despite is pity state.
1607 * GRAS seems to have very few users (if any), and no one volunteered
1608 to maintain it further. It also induces a lot of XBT code (for
1609 portability sake), that must be maintained too.
1610 * For all these reasons, we killed GRAS. If someone wants to revive it
1611 in the future, don't cry, our git history still remembers of GRAS.
1614 * Major overhaul. Merge our documentation again as time proved that
1615 splitting it was really not helping our users.
1616 * Further improve the developer documentation to help newcomers
1617 hacking on SimGrid itself. The user documentation (and in
1618 particular, the beginner documentation) is still in a sorry state.
1621 * Now works on Windows too!
1622 * Much more extensive test suite, from MPICH
1625 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1626 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1627 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1628 * Bug fix: let task be scheduled when the last dependency to be solved is
1629 a control dependency.
1630 * Remove SD_load_environment_script function.
1631 Use the C sg_platf function if you want to declare a platform
1635 * New function: MSG_process_get_number()
1636 * Old function documented: MSG_config()
1637 * Remove MSG_load_platform_script function
1638 Use the C sg_platf function if you want to declare a platform
1642 * Change the default value of the TCP_gamma constant (maximal size of TCP
1643 congestion window) to a more realistic 4MiB value. If you notice changes in
1644 your simulation results, you can fall back to the previous 20k tiny window
1645 by adding --cfg=network/TCP_gamma:20000 on command line.
1646 * (Hopefully) fix a bug wrt periodic availability and state traces
1647 * Bug fix: use default values at start when first event in availability/state
1648 trace is not at time 0.
1651 * remove the "new_" part of function name sg_platf_new_trace_connect
1652 (resulting in sg_platf_trace_connect), since it does not create
1656 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1657 use case where it's really mandatory, and maintaining it was a pain in
1659 * New: xbt_fifo_search(), search an item with a user-provided
1660 comparison function instead of dumb pointer comparison.
1663 * Fix the lua deployment:
1664 Use `simgrid.init_application()` before deployment instead of
1665 `simgrid.msg_register_application()` after.
1668 * Transfer the tracing files into the corresponding modules.
1670 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1672 SimGrid (3.8.1) stable; urgency=low
1674 The "we are told that some people want to also *install* the simgrid
1677 * Add missing file "tesh.1" to the archive.
1679 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1681 SimGrid (3.8) stable; urgency=low
1683 The Psssshiiiit release: SimGrid jumps into the Cloud.
1686 * Add an experimental interface to manipulate VMs. They are mainly
1687 process groups with very few intrinsic semantic, but they should
1688 allow you to build the semantic you want easily.
1689 * New function: MSG_host_set_property_value()
1690 * New function: MSG_process_on_exit(). To clean memory in all cases.
1691 * Bug fixes that made the host (and link) failures unusable.
1692 * Add a way to auto-restart process when the host in which they are
1693 executing comes back (ON_FAILURE="RESTART" on deployment file,
1694 MSG_process_auto_restart_set).
1695 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1696 please stop using the old ones, they are DEPRECATED.
1698 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1699 Please use MSG_init() instead. (reducing the amount of entry
1700 points in the library helps us).
1701 * Make it impossible to link against the wrong version of the lib
1702 * Deprecate MSG_clean(). No need to call it anymore.
1703 * Function MSG_get_host_number() is not deprecated anymore.
1706 * Split the doc into a user guide and a reference guide.
1707 * Start a developper guide to help people hacking on SimGrid.
1710 * Enable tracing by default. This modules rocks you should use it.
1711 * Remove option custom_flags. Now use environment variables CFLAGS
1713 * Use default cmake things to detect lua instead of home grown ones.
1714 * New option "enable_mallocators" to disable mallocators, for debugging
1715 purpose ("on" by default).
1718 * Bug fixes around the resource failures: don't let the processes
1719 survive the host they are running onto.
1720 * Add an interface to auto-restart processes when the host in which they are
1721 executing comes back.
1722 * Ensures that SIMIX_clean is called automatically. It's not part of
1723 the public interface anymore (bindings should be updated).
1726 * Bug fix for when SD_Simulate is called with a positive value: be careful
1727 when comparing doubles. Sometimes they are different for non significant
1729 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1730 parallel task whose initial work is distributed among host according
1731 to the Amdahl's law. Such tasks are created with a parameter alpha
1732 that corresponds to the non-parallelizable part of the computation.
1733 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1734 between two sets of workstations assuming a 1D block distribution (each
1735 workstation owns a similar share of data) on both sides.
1737 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1738 Data redistribution will be automatically scheduled once parent and child
1739 are both scheduled. The filling of computation_amount and
1740 communication_amount structures is now done seamlessly thanks to the chosen
1742 * New function SD_workstation_dump to display various information
1743 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1744 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1745 bandwidth on the route onto which the task is scheduled and the amount of
1747 To divide the nominal bandwidth by 2, the rate then has to be:
1748 rate = bandwidth/(2*amount)
1749 * Compute tasks that have failed can now be rescheduled and executed again
1750 (from their beginning)
1751 * Increasing source code coverage (src/simdag is now covered at 95.8%
1755 * Re-implement time-independent trace replay using SMPI (at the
1756 smpi_smp_* level) instead of MSG. This should replace
1757 examples/msg/actions/actions.c
1758 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1759 hindexed and structs)
1760 * Implement the exchange of non-contiguous data.
1761 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1762 * Correct behavior of smpi/sender_gap and set its default value to 0
1763 * Add option to asynchronously send small messages to allow better
1764 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1765 specifies the size in bytes under which messages will be asynchronously sent.
1766 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1767 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1770 * New C interface to define a platform: XML is now optional.
1771 For more info, please check include/simgrid/platf.h.
1772 * New interface to define random platforms from the C:
1773 For more info, please check include/simgrid/platf_generator.h and
1774 examples/msg/masterslave/masterslave_platfgen.c
1775 * Export a sg_cmdline dynar containing all the arguments we got from
1779 * Two new tracing options for adding comments to trace file so you
1780 can track your experiments (see --help-tracing for details).
1781 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1782 * Adding the SimGrid version that generated the trace file as a comment.
1783 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1784 * Fix to avoid key clashes on Paje links
1785 * Other minor fixes related to the Paje specification
1788 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1789 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1790 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1791 * Greatly improve the robustness of mmalloc to user errors (such as
1792 using an area after freeing it, or freeing it twice)
1794 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1796 SimGrid-java (3.8.1) stable; urgency=low
1798 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1799 Warning: all methods are visible, but only some of them are
1800 implemented so far. Check the source (src/jtrace.c)
1801 for further information.
1802 * New module: org.simgrid.msg.File (SimGrid File management functions)
1803 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1804 * Change the meaning of Process.restart: now restart the process from
1805 the beginning, like MSG_process_restart in C.
1806 * Add Process.setAutoRestart: handling of process restart when failed
1808 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1809 you to retrieve the properties of the processes/hosts
1810 * Deprecate Msg.clean(): you can just forget about it now.
1811 * New function Process.getCount(), that only works when compiling
1812 with the not yet released version 3.9 of the C library.
1814 * New context factory based on Coroutines. It mandates a modified JVM
1815 but then, the simulations run about five times faster, and there is
1816 no limit to the amount of processes (beside of the available memory).
1818 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1820 SimGrid (3.7.1) stable; urgency=low
1823 * Restore the prototype of MSG_process_create_with_environment() to
1824 the pre-3.7 situation by removing the kill_time argument.
1825 * Add a MSG_process_set_kill_time() function instead.
1828 * Fix weird behaviors when dealing with parallel tasks.
1831 * Simgrid is now built as a dll.
1832 * Simgrid-java now works on Windows.
1833 * Simgrid-Java is now included into Windows package.
1836 * First pre-build package for MacOSX.
1839 * Fix compilation when using MSG_USE_DEPRECATED.
1840 * Fix some compilation issues on Macs and Windows.
1841 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1842 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1844 We discovered that the Lua console is broken, but we are missing the
1845 manpower to fix it right now. The problem existed in 3.7 too, so we
1846 are not blocking the release for that. Sorry if you depended on this
1847 feature, any help would be really welcome.
1849 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1852 SimGrid-java (3.7.1) stable; urgency=low
1854 The "Java aint got to be bloated and slow" release
1857 * Various internal cleanups and performance improvement
1858 Simulations are expected to run up to twice faster or so
1859 * Make Process.kill(process) an instance method, not a static one
1860 * User processes are not java.lang.Thread subclasses.
1861 This breaks the compatibility (sorry), but previous API was
1862 brain-dead, making it impossible to have non-trivial
1863 initializations in the process constructor.
1864 * Require a full constructor per Process sub-class.
1865 Kinda breaks the compatibility (sorry), but this allows a much more
1866 efficient way to launch the processes at simulation startup.
1867 * Do not embeed our version of semaphores, java 1.5 can be considered
1868 as sufficiently prevalent for us to not dupplicate its features.
1872 * Add examples for almost every part of the API
1873 We spotted and fixed a lot of bugs in the process
1875 * New module: asynchronous communication API
1876 * New function: Process.sleep()
1877 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1878 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1879 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1881 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1883 SimGrid (3.7) stable; urgency=low
1885 The "spring cleanups (before the next Big Project kicks in)" release.
1888 * We can specify the SMPI latency/bandwidth factor with command line
1889 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1890 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1891 You can also use the "config tag" from platform file by adding this line
1892 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1893 (see "example/platforms/tag_config.xml" to use "config tag").
1894 Note that the command line supersedes the platform file configuration.
1895 * Change the correction factors used in LMM model, according to
1896 the latest experiments described in INRIA RR-7821.
1897 Accuracy should be improved this way.
1898 * Use the partial invalidation optimization by default for the
1899 network too. Should produce the exact same results, only faster.
1900 * Major cleanup in surf to merge models and split some optimization
1901 mechanisms from the core of the models. As a result you can now
1902 specify which model to use (e.g., --cfg=network/model:LV08
1903 --cfg=cpu/model:Cas01) and which optimization mode to use
1904 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1905 Incompatible combinations should err at initialization. See
1906 --help-models for the list of all models and optimization modes.
1907 * The CLM03 workstation model was dropped for simplicity because it
1908 used the deprecated CM02 network model. Use default instead.
1909 * Rename the TCP_gamma configuration option to network/TCP_gamma
1910 * Rename the coordinates configuration option to
1911 network/coordinates, and document it
1912 * Use now crosstraffic keyword instead of the terribly misleading
1913 fullduplex keyword. It is activated by default now in the current
1914 default model, use --cfg=network/crosstraffic:0 to turn it off.
1915 * Ongoing refactoring the model parsing to make XML files optional
1916 See include/simgrid/platf.h for details (still to be completed)
1919 * Major overhaul of the documentation. Almost instructive now :/
1920 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1921 functions and friends. This interface was considered as
1922 deprecated since over 2 years, it's time to inform our users that it is.
1923 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1924 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1925 use these (crappy) functions in your code.
1926 These functions will be removed soon. Stop using them now.
1927 * Deprecate MSG_get_host_{table,number}
1928 Implement MSG_hosts_as_dynar() instead.
1929 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1930 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1933 * Stabilize the parallel execution mode of user contexts
1934 * Introduce configuration variables to control parallel execution:
1935 - contexts/synchro: Synchronization mode to use when running
1936 contexts in parallel (either futex, posix or busy_wait)
1937 - contexts/parallel_threshold: Minimal number of user contexts
1938 that must be part of a scheduling round to switch to parallel
1939 execution mode (raw contexts only)
1940 * Fix bugs that prevented to use suspend/resume along with
1941 synchronization structures.
1942 * Fix bugs in process termination that lead to invalid memory access
1943 in very specific conditions.
1946 * Introduce a parallel mode for the models (controlled by surf/nthreads
1947 configuration item). In our tests, running the models in parallel
1948 never lead to any speedups because they are so fast that the gain
1949 of computing each model in parallel does not amortizes the
1950 synchronization costs, even when ultra fast futexes are used.
1951 This is released anyway because YMMV.
1954 * Performance boost by using a swag internally to compute the set of
1955 tasks that are finished and should constitute the return value of
1959 * Enable it by default now that it is considered rather stable.
1962 * Documentation of the tracing functions.
1963 * Performance gains when tracing categorized/uncategorized resource
1964 utilization by avoiding calls to get route when updating resource
1965 variables. LMM constraints are being used instead.
1966 * API changed to set task categories. Use MSG_task_set_category instead
1967 of TRACE_msg_set_task_category, and SD_task_set_category instead
1968 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1969 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1970 * Examples for link user variables added (see at examples/msg/tracing/)
1971 * Deprecated function TRACE_msg_set_process_category completely removed
1972 * Trace header updated according to the latest Paje file format
1973 * Tracing network lazy updates, no longer obligate users to use full updates
1974 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1975 * Experimental: let user code declare/set/push/pop application states for hosts
1976 * API changed to allow the manual creation of graph configuration files
1977 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1980 * Improve the API of Lua MSG bindings, using the Lua spirit.
1981 * Each simulated process now lives in its own Lua world (globals are
1982 automatically duplicated). It helps writing simulators. Will allow
1983 to run Splay programs within SimGrid in the future.
1984 * Add a Chord example in Lua, equivalent to the MSG one.
1987 * Start the implementation of a solution to express temporal
1988 properties, not only local assertions. This is still an
1989 experimental work in progress, stay clear from it to be safe.
1993 - Add new runtime parameters --help-logs and --help-log-categories
1994 to display information about supported logging parameters and
1996 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1997 don't exists anymore.
1998 * Mallocators: allow value NULL for the reset function.
2000 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2001 create homogeneous dictionaries, where all the elements share the
2002 same free function. Non homogeneous dictionaries will be
2003 deprecated in the next release.
2004 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2005 - Multi-level dictionaries are deprecated.
2007 - new function xbt_dynar_search_or_negative() that is useful when
2008 you have less than 2 million elements in your dynar and don't
2009 want of the extra complexity of catching exceptions when the
2010 element is not found.
2012 - Make xbt_os_thread module (for thread portability) public.
2013 Documentation is still to come, sorry.
2015 - Cleanups and simplifications to make it maintainable again.
2016 - Exotic features (such as memalign and valloc) were removed.
2017 - The metadata were extended and improved so that the
2018 model-checker becomes able to explore and inspect the heaps.
2019 - This may induce a performance drop when enable_model-checking is
2020 ON in cmake (even if it's not used in the simulation), but it is
2021 necessary at this point to get MC working.
2023 Turn model-checking OFF if simulation performance matters to you.
2024 Not enabling it at runtime is not enough, disable it in cmake.
2026 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2029 SimGrid (3.6.2) stable; urgency=low
2031 The "Not coding new stuff allows to polish old things" release.
2034 * New bindings to the NS3 packet level simulator (experimental)
2035 * Use the raw (efficient) execution contexts instead of the sysv
2036 (portable) ones when possible.
2037 * libpcre is now mandatory in any cases since not using it led to
2038 severe performance loss and possibly other issues
2039 * Update the XML platforms:
2040 - G5K: include the latest machine in Nancy
2041 - GridPP and LCG: new platforms
2042 * Documentation was partially updated, at least (more to come)
2044 Bug fixes, cosmetics and small improvements
2045 * Free terminated processes before the end of the simulation to avoid
2046 exhausting the memory of users having very dynamic amount of
2048 * Bug fix and cosmetics about canceling non-running tasks
2049 * Bug fix about the dot loader's issues when using libcgraph
2052 * Create an installer for windows with nsis (amd64 and win32)
2053 - Add an hello world project to illustrate simgrid project creation.
2054 - Embed libpcre into the Simgrid installer to avoid
2055 its compilation burden
2056 * The raw execution contexts should work on Apple now
2057 * Port to Windows 64 bits
2058 - Sysv contexts now have an implementation for this arch
2059 - GRAS communication features now support this arch
2060 * Drop support for borland compiler on windows
2061 - this code was not maintained, and we kinda depend on gcc nowadays
2062 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2063 * Fix portability issue on unstable ubuntu: linker became picky on
2066 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2068 SimGrid (3.6.1) stable; urgency=low
2070 The "Oops, we broke Macs too" release
2073 * Fixed contexts detection so that raw ones are used when possible
2074 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2075 leads to a strange error, with user code segfaulting sometimes when
2076 the generated code is not perfectly aligned (which is not
2077 controllable from the user side, depends on the amount of code)
2080 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2082 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2084 SimGrid-java (3.6) unstable; urgency=low
2087 * Split of every thing from simgrid v3.5 into a separate package.
2089 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2091 SimGrid (3.6) stable; urgency=medium
2093 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2096 * Bindings now constitute their own package, separated from the main one.
2097 Rationale: reduce our maintenance nightmare by reducing the module coupling
2098 They will soon be released on their own on gforge.
2100 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2101 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2103 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2104 * It's not quite deprecated for now because we have no replacement,
2105 but it may soon become the case.
2108 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2109 * Fortran: New user-level cache variable to store the rank of the running
2110 process. This improves performance by an order of magnitude.
2111 * C: New coccinelle script to automatically locate and modify global and
2112 local static variables.
2113 * Improved SMPI network model with a sender-side gap to account for multiple
2117 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2118 only say if a communication is finished, no matter whether it succeeded or
2119 failed. You can call MSG_comm_get_status() to know the status of a finished
2121 * New function MSG_task_dsend() to send a task and detach it. When a
2122 communication is detached, you are never notified of its success or failure
2123 and the memory is released automatically once it is finished. This function
2124 is useful when you don't care about the end nor the success of a
2126 * Change the prototypes of action replay. Sorry for inconvenience,
2127 but this is really more efficient this way (and to adapt your code,
2128 you just have to fix the initialization, that shouldn't be too long)
2129 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2130 ever managed to do anything useful with it.
2131 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2132 takes a dynar of communications. It returns immediately and gives the
2133 index of a finished communication (if any).
2134 * New example: a basic implementation of the Chord P2P algorithm.
2137 * New model for multi-core CPUs. You can now use the core attribute to
2138 precise the number of cores of a host. This is a basic model. Every
2139 process running on the host receives at most the power provided in
2140 the DTD (throughput<=power). Total throughput of process cannot exceed
2142 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2143 router linked by an up-link and a down-link (possibly asymmetrical).
2144 This kind of pattern allows to easily build last-mile model style platforms.
2145 Aggregating such patterns in a rule-based AS is thus the technique of
2146 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2147 * New model for Vivaldi routing. We transformed the Vivaldi network model
2148 into a Vivaldi routing model (based on the rule-based model). This allows to
2149 combine Vivaldi based latencies with last-mile platforms.
2152 * Added a check for NaN of IEEE754 infinite in the double entries of
2154 * Introduce a new context factory "raw", highly inspirated from the
2155 ucontext factory, but using manually crafted functions in assembly to
2156 do the work in an efficient manner.
2157 * Allow to change the used context factory at run time, not only at
2158 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2159 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2160 contexts at runtime (only with raw contexts or ucontexts).
2161 * Completely rewrote this module to allow parallel execution of user
2162 processes. Use --cfg=contexts/nthreads:N to execute user processes
2163 with N parallel threads (the default is 1, meaning no parallelism).
2164 * Allow to decide dynamically between sequential and parallel modes.
2165 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2166 processes in parallel only when their number is greater than or equal to P
2168 * Added a check for NaN of IEEE754 infinite in the double entries of
2172 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2173 won't output the state of processes when interrupted with Ctrl-C
2174 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2175 NULL-terminated array. This may solve backward compatibility issues
2176 due to the change to return type of SD_simulate. See also:
2177 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2178 * Add new macros with variable number of arguments.
2179 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2180 - in xbt/asserts.h: xbt_assert
2181 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2182 - in xbt/ex.h: THROWF and RETHROWF.
2183 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2185 * Change xbt_die() to accept a format string with arguments, just like printf.
2186 * New data structure: xbt_lib_t, like a dict but more general and with better
2190 * New configuration options
2191 Options triva/categorized and triva/uncategorized can be used to generate
2192 graph configuration files for Triva visualization tool.
2193 * Configuration option tracing/platform is renamed to tracing/categorized
2194 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2195 * New TRACE_user_link_variable interface:
2196 User provides the name of the link and the tracing variable to attach to it
2197 * the declaration of tracing categories must be done after the environment
2199 * simpler tracing interface, just one way to declare categories
2200 TRACE_category or TRACE_category_with_color, it is up to you
2201 * links in the trace file are again identified by their names
2202 * trace contains the full platform hierarchy exactly as declared using the ASes
2203 * Options tracing/msg/[task|process]:1 groups the process by hosts
2204 for both cases, tasks and processes must have names that are unique during the simulation
2205 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2206 * The experimental option tracing/msg/volume is deprecated
2207 its functionality may be reincorporated if needed
2209 The tracing generates a trace file with unordered timestamped events,
2210 because of the way the core simulator (surf) works. A script available
2211 at the tools directory (fix-paje-trace.sh) can be used to put the events
2212 in order. We have changed the tracing so it can generate ordered timestamped
2213 events in the final trace, but depending on the simulator (and how much time
2214 is simulated) that can lead to a huge memory utilization. It is deactivated
2215 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2217 Build Infrastructure
2218 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2219 - We are trying hard to keep the API stable, but it may happen that
2220 some things change (we're a research project after all, not a
2221 nuclear plant operating system). If such things should happen, you
2222 could rely on that macro to adapt.
2223 - current value: 30600 for 3.06.00, aka 3.6
2224 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2225 a number that can be compared with SIMGRID_VERSION.
2226 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2227 assertions and verbose/debug logging events are disabled at compile time.
2229 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2231 SimGrid (3.5) stable; urgency=medium
2234 * New feature to any SimGrid-based simulator: Model-Checking
2235 Check SIN#1 for more details.
2238 * New Model SMPI (three-interval linear regression for correction factors)
2239 See RR-7426, available at http://hal.inria.fr/inria-00527150
2240 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2241 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2242 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2243 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2244 * See also src/smpi/README
2249 - Tracing API changes: TRACE_start and TRACE_end should not be called
2250 by user-code. They are automatically called by simulators created
2251 with SimDAG, MSG and SMPI if the toolkit is compiled with
2252 tracing_enabled=ON. Categories declaration and utilization remain the
2253 same for MSG and SimDag.
2254 - A function was added to the tracing API to declare categories with
2256 - TRACE_category_with_color (char *category, char *color)
2257 where color must be in the following format
2258 "%f %f %f", red, green, blue
2259 and red, green, blue are float values in the interval [0, 1]
2260 - User can specify NULL as color parameter, or continue calling
2261 TRACE_category (cat)
2262 On that case, the tracing system will define random colors
2263 - The following command-line options are supported:
2264 --cfg=tracing/filename:msg.trace
2265 --cfg=tracing:1 (activate tracing, needed to use others)
2266 --cfg=tracing/platform:1 (categorized resource use)
2267 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2268 --cfg=tracing/msg/task:1 (task creation)
2269 --cfg=tracing/msg/process:1 (process creation, migration)
2270 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2271 --cfg=tracing/smpi:1 (SMPI interface tracing)
2272 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2273 - examples of examples/msg/tracing updated
2275 - DAXLoader and DOTLoader functions can generate tasks with categories
2276 - A new function to attribute a category to SD tasks:
2277 TRACE_sd_set_task_category (SD_task_t task, char *category)
2278 * Tracing the MPI interface implemented by SMPI
2279 - Collective operations are traced with states
2280 - Point-to-Point operations are traced with states/links
2281 - Tracing activated by a parameter "-trace filename" passed
2282 to smpirun during execution (considering that simgrid
2283 is compiled with tracing enabled)
2284 - To run the simulation with gdb, the simulator
2285 accepts --cfg=tracing/smpi:1 to trace SMPI
2286 - tesh files to check if smpi tracing is ok
2287 - See examples/smpi/NAS/DT-trace
2288 * GTNetS tracing re-worked
2289 - adaptation to the tracing system of GTNets to cope
2290 with modifications regarding the fullduplex mode
2291 - new tesh files to check if gtnets tracing is ok
2294 * Asynchronous communications through the functions:
2295 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2296 * New function: MSG_load_platform_script()
2297 to make possible using a lua script instead of XML files to set up platforms
2298 * New function: MSG_set_function
2299 to associate functions to processes, used when bypassing the parser
2300 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2302 Platforms: Add some more examples in examples/platforms
2303 * Grid'5000: see www.grid5000.fr
2304 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2307 * Change the XML format. This is a very important modification. SimGrid
2308 3.5 introduces a new hierarchical format based on the notion of
2309 Autonomous Systems. Compatibility with old format is ensured
2310 through the perl script provided in the install bin directory
2311 bin/simgrid_update_xml.
2312 It is now possible to build platforms with specific routing
2313 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2314 connect several platforms together. We will try to provide soon
2315 set of realistic platforms exploiting these properties (have a
2316 look at examples/platforms/ for the moment).
2317 * Take the opportunity of the XML format change to be a good XML citizen:
2318 rename link:ctn to link_ctn and similar changes (also dealt with by
2320 * Add a new routing scheme (rule-based) using regular expressions. It
2321 enables to have an extremely low memory footprint when the
2322 underlying routing is simple and can be compactly described. You
2323 need to have libpcre4-dev (perl regular expressions) installed if
2324 you want to use this routing scheme.
2325 * Revive the cluster TAG and allow to easily and efficiently (both in
2326 term of memory and speed) connect clusters together. Have a look
2327 at teshsuite/simdag/platforms/ to see how this can be done. With
2328 this tag, you can create clusters with thousands of tasks at no
2329 cost (have a look at examples/platforms/).
2330 Note: clusters are implemented as ASes, so there is no need for an
2331 enclosing AS tag if you have only one cluster in your platform.
2332 * Add new generic functions in the public interface that allows the user
2333 to call SURF 'create_resource' methods from your code (same
2334 functionality as the XML bypass mechanism but with a much lighter
2336 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2337 that uses a piecewise linear approximation to produce better
2338 results when exchanging small messages.
2339 * Add a new parameter to handle correctly full duplex link and account
2340 for interferences between uplink and downlink communications
2341 (activate with --cfg=fullduplex:1).
2344 * Rename the SD_READY (all dependencies are satisfied and task is
2345 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2346 dependencies are satisfied) state.
2347 This prevents a confusion between the notion of "ready to schedule"
2348 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2349 simulated" (SD_RUNNABLE) used by the simulation kernel.
2350 * Change the way a task is considered as ready. Instead of removing
2351 dependencies when a task is done, a counter is decreased. This way,
2352 it is always possible to reach ancestors thanks to the
2353 SD_taks_get_parents function (even after the end of the simulation.)
2354 * Change the return type of SD_Simulate from (SD_task_t*) into
2355 xbt_dynar_t. This function was in handling a dynar internally and
2356 converted it into a NULL terminated array for historical reasons.
2357 * New function SD_dotload(char*) to load a DAG described in dot
2358 format. This loader and the corresponding examples require the
2359 installation of the graphviz library.
2360 * Fix a bug in the management of tasks of size 0 in the surf network
2361 models. This problem was only visible with SIMDAG and you should
2362 thus disregard results produced with earlier versions if you
2363 relied on this feature (some tasks were blocked because of this).
2364 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2365 even though your applications were doing only point-to-point
2366 communications and sequential computations. Now you can really use any
2367 model you want (of course, if you create real parallel tasks, which are
2368 not implemented in most models beside ptaskL07, this will abort).
2369 * Add an example that schedules a DAX on an heterogeneous platform
2370 using a Min-Min strategy.
2371 * New function SD_workstation_get_current_task() that returns the kind
2372 of task currently running on a workstation in the sequential access
2374 * Raise some warnings when unexecuted tasks remains at the end of the
2375 simulation. This is usually caused by cycles in the DAG.
2378 * New function: SIMIX_process_set_function() called by MSG_set_function
2379 * Change the underlying waiting queue in semaphores so that a process
2380 can wait on several of them simultaneously (as in waitany).
2381 * Fix the way to handle tokens in semaphores so that all access patterns
2382 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2383 * kill the dirty pimple SIMIX_message_sizes_output()
2384 Please use (proper) visualization instead
2387 * New data container: setset (set of sets of elements)
2388 * New module: mmalloc (mapped malloc, allowing to have several
2389 independent segments of malloc)
2390 * New function: xbt_dict_cursor_set_data()
2391 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2392 * New function: xbt_dynar_is_empty()
2393 * New function: xbt_fifo_get_last_item()
2394 * Fix xbt_dynar_shrink(): use the right element size.
2395 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2396 uninitialized areas during expand.
2397 * Fix semaphores: previous implementation was severely broken.
2398 * Use library init/fini functions for our initialization.
2399 - you can use logs and other feature as soon as you want in your
2400 code (even before the xbt_init / MSG_init)
2401 - xbt_exit is now a no-op and produce a warning when used.
2404 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2405 benefit from the latest improvement to the simulation kernel.
2406 * Kill measurement sockets for now. If you rely on them, sorry. This
2407 release is not for you. This feature will be reintroduced in the
2408 future, but we cannot delay the release any further.
2409 * New function: gras_msgtype_get_name().
2410 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2411 * Fix (at last) the pmm example: it should not randomly fail anymore.
2413 Build chain: bug fixes and overall polishing
2414 * Cmake is now stable enough. Hence, we killed the autotools.
2415 * Port to windows ( TM :)
2416 * Fix the 'make install' target.
2417 No need to use 'make install-simgrid' anymore
2418 * Introduce a 'make dist' target compiling a *source* archive
2419 'make package' compiles a binary archive
2420 * Compile java files only on need
2421 * Add --cd and --setenv command line options to tesh
2422 * Out of source builds are not fully supported yet, but we are close
2423 * Enable supernovae and optimization flags by default for our users
2426 * Add layer to set up environment directly from lua, without XML.
2427 * The effect of gras_stub_generator can be achieved through
2428 lua too (check examples/gras/console/ping_generator.lua)
2430 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2432 SimGrid (3.4.1) stable; urgency=low
2434 The "Polishing easter eggs is probably a good idea" release.
2435 This is a bug fixes release only.
2439 * Fix a bug preventing the tasks from begin garbage collected.
2442 * Fix a bug occurring when a host involved in a communication fails.
2443 This was not detected properly by the other peer involved in the
2444 communication. Now, it's reported as a network error.
2447 * Warn the user about loop dependencies in data flow of DAX files
2448 * Obey the control-flow dependencies of DAX files
2451 * Add option "enable_smpi" allowing to not compile SMPI.
2452 Probably useful for the (Mac) users experiencing a build error here
2453 * Improve the detection of lua5.1 and ruby1.8
2455 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2457 SimGrid (3.4) stable; urgency=low
2459 The "Easter in Cargese" release. Also known as (major changes):
2461 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2462 ~> bindings were greatly improved
2463 ~> new tracing infrastructure for better visualization introduced
2465 * the "Welcome to configury modernity" release.
2466 ~> we switched from autotools to cmake, and improved our cdash
2469 A more detailled list of changes follow (full detail in svn log).
2471 Java Bindings: Various Cleanups
2472 * (install java-gcj-compat-dev on debian-like to use them)
2473 * Remove put/get: no need to export deprecated interface in Java
2474 Use send/receive instead.
2475 * Cleanup the examples and add a README per directory
2476 * Remove example autoDestination (that's the only way to go now)
2477 * Remove example explicitDestination (was a plain copy of basic)
2478 * Make JniException a runtime exception, so that there is no need to
2479 declare the fact that you may encounter such a beast. I guess that
2480 nobody will ever want to survive such error.
2481 * Create specific errors for each MSG case of failure:
2482 host failure, transfer failure, timeout, task cancelled
2483 * Cleanup the exceptions that may get thrown by each function
2484 * Other internal cleanups in Java bindings. Performance still bad :/
2485 Ruby and Lua Bindings: create them
2486 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2487 * That's new and great, you should try them out.
2488 Same functionalities than Java bindings, only even less polished
2490 * Kill the useless "rate" argument of SD_task_get_execution_time()
2491 Everyone used to provide -1 as a value, it was not used, and the
2492 semantic of a possible use wasn't even clear.
2493 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2494 either as comm costs or compute costs to mean that there is no
2495 such thing for that specific task.
2496 * Add a SD_task_set_name() function
2497 * Fix SD_task_unschedule() on typed tasks
2498 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2499 * In DAX loader, accept useless 'level' attributes to <job> since
2500 LIGO DAGs have them (seem to be to ease graphical representation).
2502 * Add an example masterslave_mailbox.c using send/receive and not
2503 the deprecated put/get interface.
2504 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2505 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2507 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2508 (declare MSG_USE_DEPRECATED to still have the old name)
2509 * Add a MSG_task_set_data() function
2510 * About trace replay (see examples/msg/actions):
2512 - Allow to work with splitted trace files for each process
2513 Give the specific trace file as argument of each process,
2514 and call MSG_action_trace_run(NULL)
2515 You can still have one merged file for all processes.
2516 - Fix implementation of collective operations
2517 * Allow task_execute() on 0-sized tasks (closes #10063)
2519 * This is the first release of SimGrid where SMPI is not considered
2520 beta anymore (even if some corners should still be improved)
2521 * Port over the new SIMIX_network submodule (internal refactoring)
2522 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2523 * Implement more missing elements of the standard:
2525 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2527 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2528 MPI_Scan() MPI_Get_processor_name()
2529 - Added implementation of missing case for Alltoall (warning: it's
2530 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2531 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2532 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2533 SMPI_MPI_Allgatherv()
2534 * Bug fixes include:
2535 - MPI_Waitsome() was broken
2536 - Allow relative includes in smpicc
2537 - Command line cfg argument 'reference_speed' was ignored...
2538 - Some functions did not properly lead to auto-benching of user code
2539 - smpicc passes -O2 by default (just like openmpi one)
2541 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2542 * Bug fixes about timeouts during communications
2543 * add SIMIX_message_sizes_output() as a pimple to write to file the
2544 amount of messages per size. Use gnuplot to get histogram.
2545 Pimple because that's the only user-visible function of simix,
2546 defined directly in xbt.h (irk, sorry)
2548 - Add a SIMIX_sem_get_capacity() function
2549 - Fix interactions with processe resume/suspende
2550 - release_forever() was stupidly broken
2551 - Fix SIMIX_display_process_status() for processes in a semaphore
2552 - Make SIMIX_sem_block_onto() user-visible
2553 * Refactoring context stuff:
2554 - Use pseudo-OOP for better modularity
2555 - reimplement SIMIX_process_kill() without process_schedule() so
2556 that the latter can take as invariant that it is called from
2558 - Merge context_start into context_new for sake of simplicity
2560 * Add a Vivaldi network model, coded live during SUD'10 ;)
2561 * Rename configuration variables to start a hierarchy:
2562 o cpu_model -> cpu/model
2563 o network_model -> network/model
2564 o workstation_model -> workstation/model
2565 * New configuration variables:
2566 o network/bandwidth_factor: correction to bandwidth
2567 o network/latency_factor: correction to latency
2568 o netwotk/weight_S: correction to the weight of competing streams
2569 * Add a long description to the models, that users can see with such
2570 argument on the command line: --cfg=cpu/model:help
2571 * --help-models display the long description of all known models
2573 * config: add the ability to set a default value after registration
2574 Does not override any previously set value (e.g. from cmd line)
2575 * dict: allow to have integer key and data.
2576 When so, you need to use the following functions
2577 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2578 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2579 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2580 In contrary to regular dicts, the key is not malloced before copy.
2581 Mixing scalar and regular elements in the same dict is not tested
2583 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2584 Tracing for Visualization:
2585 * SimGrid is now instrumented in order to generate a trace file for
2586 visualization analysis: to use it, need to compile SimGrid with the
2587 "tracing" option enabled, and instrument the program using SimGrid with
2588 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2589 (among other functions).
2590 * The instrumentation only traces the platform utilization for now
2591 * Documentation to use the tracing functions and how to analyze the
2592 traces with the Triva tool is written.
2593 * More information about: SimGrid FAQ (in the section Tracing Simulations
2596 * We moved to cmake as default build system. Autotools support will
2597 be dropped soon. Check the FAQ for more info about how to use it.
2598 * Greatly improved our cdash/ctest interactions
2599 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2600 * Added memory checking tests with valgrind; lot of memleak fixing.
2601 This may be the first release of simgrid with so few memory issues
2602 * Added code coverage tests.
2603 Our coverage is still improvable, but at least we see it on cdash.
2605 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2607 SimGrid (3.3.4) stable; urgency=low
2609 The "Desktop Grid needs love too" release (also called Xmas release).
2611 Models improvements:
2612 * Major speedup in the maxmin system solving by using lazy evaluation
2613 Instead of solving completely the maxmin system at each iteration,
2614 only invalidate (and recompute) the modified parts.
2615 This new feature is enabled in default models but you can try to
2616 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2617 * Cas01 IMproved as default CPU model
2618 This CPU model is the same Cas01 model, but it uses the
2619 maxmin-selective-update flag and a heap structure to manage
2620 actions on SURF kernel.
2621 It reduces the complexity to find the next action to finish and,
2622 consequently, it's faster than the old Cas01.
2623 This is the new default CPU model (Cas01).
2624 * Rename the old Cas01 model to Cas01_fullupdate
2625 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2626 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2627 * CpuTI (CPU Trace Integration)
2628 A new CPU model whose objective is simulate faster when using
2629 availability trace files.
2630 Instead of using a full featured, over engineered maxmin system for
2631 CPU modeling, this model does the pre-integration of traces files
2632 to calculate the amount of CPU power available, and so, executes
2633 faster than the old CPU models.
2634 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2635 * Use LV08 as default network model since it gives better accuracy
2636 for small messages and shouldn't change things for big ones.
2637 Use --cfg=network_model:CM02 to get the previous behavior.
2640 ******************************************
2641 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2642 ******************************************
2643 * The new CPU model may changes simulations!
2644 The point is that events occurring at the exact same timestamp
2645 are not scheduled in the same order with the old and new
2646 version. This may be enough to completely change the execution
2647 of simulations in some cases.
2648 * The new network model will change simulations!
2649 This new model is more realistic than the previous one, so you
2650 should consider redoing your old experiments with this model.
2651 Sorry for the inconvenience.
2654 * Introduce the supernovae compilation mode
2655 When compiled that way, the whole SimGrid (or almost) is put in a
2656 single compilation unit and compiled in one shoot.
2657 This is to help gcc which has difficulties to inline stuff from one
2659 The speedup seem to be above 15%, althrough more tests are needed on
2660 amd64 to confirm that gain.
2663 * Port of MSG's mailbox on top of SIMIX network
2664 The put/get mechanism was greatly simplified on the way.
2667 * New SIMIX network module. Provides:
2668 - Mailbox: rendez-vous mechanism to find with who you want to speak
2669 - Synchronous send/recv: easier and hopefully faster since the
2670 logic is handled in the maestro process directly now
2671 - Asynchronous send/recv: you dreamt of it? It's here now
2672 Too bad that nobody cared enough to propagate the change to MSG.
2673 * Add semaphores as SIMIX synchronization mechanism.
2676 * new function SD_daxload(char*) to load a DAX file
2677 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2678 * Introduce typed tasks. Specify its kind and cost at creation.
2679 At scheduling, just give where it should be placed, and the cost
2680 for each involved resource is automatically computed.
2681 Existing constructors so far (more to come of course):
2682 - SD_task_create_comm_e2e() for end-to-end communication
2683 - SD_task_create_comp_seq() for sequential computation
2684 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2685 * new function SD_task_dump() for debugging display
2686 * new function SD_task_dotty(task,FILE*) writing to file the info
2687 about the task in dotty format
2688 * SD_task_dependency_exists() can now cope with having one of its
2689 arguments NULL. If so, it tests whether the other argument has any
2691 * Add getters on list of preceding/following tasks:
2692 SD_task_get_parents(task) and SD_task_get_children(task)
2693 * Add getters on amount of workstations and list:
2694 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2695 * Add getter on task kind: SD_task_get_kind(task)
2696 * Update the start_time and finish_time of tasks on completion/failure
2697 * Bugfix: Remove task from state swags when destroyed
2700 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2703 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2704 dynar with one separate thread per value of the dynar.
2705 * Change the prototype of xbt_thread_create(), sorry.
2706 Added a boolean parameter indicating whether we want to join this
2707 thread (used in SG only for now)
2708 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2711 * GTNetS wrappers should now be usable again (and betterly tested too)
2712 * Fix a major regression from 3.2 where the timeout provided to
2713 MSG_task_put_with_timeout() was used as absolute time before which
2714 the comm should be done.
2715 * Start to fix the <cluster> tag.
2716 - Internal links should be good now (beside of the loopback, which
2717 use the private link instead)
2718 - paths to the external world is still rather broken
2719 - the <route:multi> tag is just broken. Actually that's brain-dead.
2720 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2721 to make it less stupid
2722 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2723 * Fix a source-level compatibility glitch from 3.2: after defining
2724 MSG_USE_DEPRECATED, you can use the old name
2725 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2726 * Allow to compile from the SVN with automake 1.11
2727 * Fix some problems when using the "start_time" tag in deployment XMLs.
2728 * Fix #8569: XBT/synchro.h has redundant declarations
2729 * Fix #8563: MSG return values and exceptions
2730 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2731 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2732 Thanks to Robson Peixoto.
2733 * Fix a vicious bug in dictionaries inducing that some elements were
2734 not freed on xbt_dict_free()
2736 Portability report of this version:
2737 * Main portability targets:
2738 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2739 on (amd64/i386/ia64)
2740 - mac leopard on i386
2741 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2742 but nothing critical.
2743 * Other platforms: windows, AIX and others were not tested for this release
2745 Timing report of this version:
2746 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2747 scenario parameters). From 8h to a few seconds in desktop grid settings.
2748 * Supernovae brings about 25% speedup on i386.
2750 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2752 SimGrid (3.3.3) stable; urgency=low
2754 The "Need for Speed" release.
2756 The timings done to validate the 3.3.2 were faulty.
2757 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2759 The problem was a conversion from a manually handled vector to
2760 xbt_dynar_t on the critical path.
2761 xbt_dynar_foreach calls functions, inducing stack management crap.
2763 We inlined these functions and xbt_dynar_foreach is now breath taking.
2764 We also inlined xbt_swag_belong on the way.
2766 Here are some approximate speedup measurements (on master/slaves
2767 simulations lasting between 10s and 20s each):
2768 3.3.1 -> 3.3.2: about same performance
2769 3.3.2 -> 3.3.3: 40% speedup
2770 3.3.1 -> 3.3.3: 40% speedup
2771 3.3.1 with inline patch -> 3.3.3: 30% speedup
2773 Our reading is that the refactoring which occurred in 3.3.2 made us
2774 suffer much more from the xbt_dynar_foreach low performance, but
2775 once we solved this, this refactoring proved to be very performance
2776 effective. From the 40% speedup, somehow, 10% are due to the
2777 inlining and 30% to the refactoring.
2779 That's a pitty that gcc cannot inline functions placed in other files
2780 alone. We have to choose between:
2781 - break the encapsulation (by putting private data structures and
2782 accessors in headers files to help gcc)
2783 - live with low performance
2784 - switch to a decent compiler such as icc (not quite possible).
2786 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2788 SimGrid (3.3.2) stable; urgency=low
2790 The "Simplicity does not preceed complexity, but follows it" release.
2792 The main contributors of this release were (lexical order):
2793 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2796 * Extract the routing logic into its own object.
2797 (was dupplicated in network.c and workstation_LV07.c;
2798 Allows to implement other ways of storing that info)
2799 => kill now useless network_card concept
2800 - Use dynar to represent routes (instead of void** + int*)
2801 - kill link_set (use surf_network_model->resource_set instead)
2802 - Add a command-line option to choose the routing schema to use
2803 - Add three new models:
2804 * Floyd (shortest path computed at initialization)
2805 * Dijikstra (shortest path recomputed all the time)
2806 * Cached Dijikstra (shortest path computed on need)
2807 All these models where contributed by Silas De Munck, and are
2808 described in his ICCS09 paper.
2810 * Simplify model declaration
2811 (less redirections, less function to write when defining a model)
2812 - Factorize stuff between models:
2815 surf_model_resource_set(model)
2816 surf_model_resource_by_name(model, name)
2817 - Unify the types of models in s_surf_model_t (using an union)
2818 - Embeed fields of common_public directly into s_surf_model_t
2819 - Rename model methods:
2820 action_free ~> action_unref
2821 action_change_state ~> action_state_set
2822 action_get_state ~> action_state_get
2823 - Change model methods into functions:
2824 (model)->common_public->action_use ~> surf_action_ref
2826 * Implement a generic resource; use it as ancestor to specific ones
2827 (allows to kill duplicated code in models)
2828 Drawback: timer command don't need no name nor properties;
2829 workstation_CLM03 don't need no properties
2830 (but I guess we can live with those few bytes wasted)
2832 * Improve the action object model
2833 - implement a constructor avoiding dupplicated code about field
2834 initialization in generic_action part.
2836 * Kill the SDP model: it has an external dependency, is deprecated
2837 in flavor of modern lmm models, and didn't compile since a while
2840 * Relocation of the context module from XBT to SIMIX.
2841 (the context were decoupled from the simix processes, duplicating a lot of code)
2842 => a lot of code was factorized
2843 - less overhead is introduced during scheduling
2844 - simpler API for the context factory
2845 - the logic for process creation,destruction and manipulation was simplified
2846 * Simplification of the s_smx_process_t data structure.
2847 => accesing the simix level data associated to a process is faster now,
2848 and the code is a lot more readable.
2851 * Implement some more MPI primitives:
2852 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2853 -implementation: Bcast: flat or 2-ary tree (default),
2854 Barrier: 4-ary tree,
2856 Allreduce: Reduce then Bcast
2857 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2858 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2859 Alltoallv: flat tree, like ompi
2861 * Add support for optimized collectives (Bcast is now binomial by default)
2862 * Port smpirun and smpicc to OS X
2865 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2866 Shout out if you used it.
2869 * Display the list of still queued messages in SG mode when existing
2873 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2874 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2875 * Add xbt_str_from_file(FILE*)
2876 * Add xbt_dict_get_key achieving a linear reverse search
2877 * Remove the context module
2879 Portability report of this version:
2880 * Main portability targets:
2881 - Linux(debian)/x86/context
2882 - Linux(debian)/x86/pthreads
2883 - Linux(debian)/amd64/context
2884 - Linux(debian)/amd64/pthreads
2885 On these, we still have the eratic breakages of gras/pmm and
2886 amok/saturate_sg reported in previous version. We still think
2887 that the tests are the cause of the fault, not the tested code.
2889 - Mac OSX Leopard/x86/context
2890 Still false negative in tesh autotesting.
2891 Smpi still fails, but this time because readlink does not accept -f
2892 Everything seems to work properly beside of that.
2895 - AIX version 5.3 (only tested contexts this time)
2896 Smpi still fails there because mktemp is not installed.
2897 Everything seems to work properly beside of that.
2899 I managed to compile it for the first time, but several breakages.
2900 Won't delay the release for this exotic platform.
2902 * Windows: it's still lagging behind. If you want to help, please
2905 Timing report of this version:
2906 This version seem to be more than 5% faster than 3.3.1 (on linux
2907 64bits with contextes). The gain is less than expected, we are
2908 investigating this for next release.
2910 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2912 SimGrid (3.3.1) stable; urgency=low
2915 * Implement a --cfg-help to show existing configuration variables
2916 * Build chain do not require doxygen in maintainer mode
2919 * fix a bug on struct sizeof computation, which prevented the
2920 exchange of arrays of structs in some conditions
2921 - added a regression test about this in datadesc_usage
2922 * Allow the exchange of 0-long dynamic vectors.
2923 - for that, use -1 as indicator of dynamic size instead of 0
2924 - This implied to change any size from unsigned long to long,
2925 reducing a bit communication abilities, but I guess that with
2926 64bits being quite common, this is more than enough.
2927 - This also induce a protocol change, thus bumping network protocol
2928 version from 0 to 1 (if we have external users, we have to get
2929 clean on that point too ;)
2930 - added two regression tests about this in datadesc_usage
2931 * Be more verbose when propagating local exceptions
2932 This helps debugging.
2933 * Display the status of simulated processes when receiving SIGINT in
2937 * Allow to control the simulation from a trace file.
2938 New functions MSG_action_register() and MSG_action_trace_run()
2939 The first one allows to associate a function execution to each
2940 kind of action while the second one parses a trace file and
2941 triggers the corresponding actions within the system.
2942 For now, only a toy example is provided in examples/msg/actions
2943 * Add an example of process migration in examples/msg/migration
2944 * Fix a bug in task exchange which broke MSG_task_get_sender()
2945 Add a teshsuite regression test for that.
2946 [Bug: if MSG_task_get_sender() is called after sender exit,
2948 * Fix a bug which prevented suspend/resume to work properly
2949 * Display the status of simulated processes when receiving SIGINT
2950 This fixes a regression of v3.3. due to the introduction of SIMIX
2951 * Bug fixing in failure management:
2952 - trace could not start by a failure at time 0
2953 - failure during communications were not working
2956 * Add SIMIX_process_set_name() to change the name of the current
2957 process in the log messages.
2958 * Store smx_hosts in a dict since we only retrieve them by name
2959 * Move the configuration infrastructure to surf
2962 * Move the configuration infrastructure to surf
2965 * Massive internal cleanups:
2966 - Store internal structures on processes instead of hosts (allows
2967 to have more than one process per host, in addition of being more
2969 - Cleanup the initialization/finalization process
2970 - Kill a whole bunch of unneeded synchronization:
2971 processes run in exclusive manner within the simulator
2972 - Move queues from global tables to process data fields
2974 - now accept -platform and -hostfile arguments
2975 - Pass the right rank value to processes according to the hostfile
2976 * Compile the examples by default, and use them as regression tests
2977 * Implement MPI_Wtime()
2978 * Change the reference speed to a command line option
2981 * TCP_gamma can now be specified as command line option using
2982 --cfg=TCP_gamma:10000000.0
2983 * Change the --surf-path cmd line option into --cfg=path:
2986 * Also include strbuff from xbt.h public header
2987 * xbt_ex_display(): do not free the exception after displaying
2988 This allows to do more with the given exception afterward.
2989 Users should call xbt_ex_free() themselves.
2993 Portability report of this version:
2994 * Main portability targets:
2995 - Linux(debian)/x86/context
2996 - Linux(debian)/x86/pthreads
2997 - Linux(debian)/amd64/context
2998 - Linux(debian)/amd64/pthreads
2999 These targets fail about 1/10 of times on gras/pmm, but we believe
3000 that this is because of the test, not because of simgrid.
3001 amok/saturate_sg fails even more rarely, and the test may not be
3004 - Mac OSX Leopard/x86/context
3005 The test suite still spits tons of errors because some obscure
3006 force prevents us from removing the temporary directories
3007 arguing that they still contain some metadata I've never heard of.
3008 Smpi fails because seq is not installed.
3009 Everything seems to work properly beside of that.
3012 - AIX version 5.3 (both contexts and pthread)
3013 Smpi still fails there because mktemp is not installed.
3014 XML inclusions seems rosty on AIX.
3016 * Windows: it's still lagging behind. If you want to help, please
3019 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3021 SimGrid (3.3) stable; urgency=high
3025 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3026 [Malek Cherier & Mt]
3028 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3031 Until now, GRAS were using MSG as an interface to SURF. It was
3032 quite difficult because both interface have several differences
3033 (MSG channels vs GRAS sockets were the most notable point).
3035 This also opens the gate to SMPI (which should occur soon) and speed
3036 up simulations by to 40% (even if it were not the main goal).
3038 **************************************
3039 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3040 **************************************
3041 The point is that events occurring at the exact same timestamp are
3042 not scheduled in the same order with the old and new version. This
3043 may be enough to completely change the execution of simulations in
3044 some cases. Sorry for the inconvenience.
3046 * Cleanup and upgrade the XML format to push further scalability
3047 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3049 * Improve the testing infrastructure with tesh. Now a very large part of
3050 the code is tested not only by being run but also by checking that the
3051 output match an expected output [Mt].
3053 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3054 is really less memory-demanding, which should allow you to use
3055 larger files in SimGrid [AL].
3057 * Inform valgrind about our contextes, so that it becomes usable
3058 with the default (and more effecient) version of SimGrid
3059 [contributed by Sékou Diakite, many thanks]
3062 * Introduce a listener thread in charge of receiving incoming
3063 messages from the network. It allows to overlap communication and
3064 computation but most notably, it removes some stupid deadlocks due
3065 to the fact that so far, a process could not send and receive at
3066 the same time. This made most non trivial communication schema
3068 * Convert the PIDs from long int to int to match the MSG ones (and
3069 linux ones too) [Mt]
3070 * New function: gras_agent_spawn() to launch a new process on
3071 current host. Only working in simulation for now. [Mt]
3072 * New function: gras_os_hostport() returning a constant form (ie,
3073 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3076 * Make the backtrace of exceptions more human readable [Mt]
3077 * New module: xbt/str [Mt]
3078 a ton of string utility functions (split, join, printf to a newly
3079 allocated buffer, trim, etc)
3080 * New module: xbt/hash [Mt]
3081 SHA1 hashing algorithm (more to come if needed)
3082 * New module: xbt/synchro [Mt]
3083 synchronization tools (mutex and conditions) working the same way
3084 in simulation and in real life (mainly useful for GRAS, but not
3086 * New module: xbt/queue [Mt]
3087 classical producer/consumer synchronization scheme
3088 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3089 (using the classical functions will get serialized) [Mt]
3090 * Make dictionary internal table dynamic. No need to specify its size
3091 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3093 * Make sure the log channels are organized as a tree under windows
3094 (because of ANSI C compatibility issue, any channel were child of
3095 root directly) [Mt].
3098 * Cleaned many thing in surf and fixed a few bugs [AL].
3099 * Add a nice command line configuration mechanism to compose models [AL].
3100 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3101 the previous one (KCCFLN05). It relies on something that looks like
3102 a max-min sharing mechanism but cannot be written as such. A new solver
3103 was thus designed [AL].
3104 * Add a new solver to lmm. Based on Lagrange optimization and
3105 gradient-based descent, it enables to efficiently maximise systems s.a
3107 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3109 This solver enables to propose two new network models for TCP Reno and
3110 TCP Vegas based on Low's work. These models still need to be fully
3111 tested though [Pedro Velho].
3114 * Bug fix in SD_simulate. Now the time bound given as argument is
3116 * Use the new parallel task model (ptask_L07) as default.
3117 * Use the SURF command line configuration mechanism.
3118 * 0-size tasks (for synchronization) should now work.
3120 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3122 SimGrid (3.2) stable; urgency=high
3126 We still experience issues on this platform, but we believe that at
3127 least MSG is usable.
3129 GRAS API BREAKAGE (for simplification purpose, sorry):
3130 * the gras_msgtype_by_name is not used anymore. Instead of
3131 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3132 you can write (and must)
3133 gras_msg_send(toserver, "request", &request);
3134 - If you still want to pass a gras_msgtype_t to the function (to cache
3135 the type and avoid the lookup time), use the gras_msg_send_() variant.
3136 - Impacted functions:
3137 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3138 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3139 * The callbacks are now expected to return 0 when everything went well
3140 (just like the main() function)
3142 GRAS new features and improvements:
3143 * New module mechanism where user code can use per process globals [Mt]
3144 This is similar to gras_userdata_*() functions, but for libraries. It
3145 factorize some code developped over and over in the examples and AMOK.
3146 It has still to be documented and used (only amok/peermanagement is
3148 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3149 loss when they were small enough to fit into the buffer and sent quickly
3150 enough so that they can all get received in one shoot.
3151 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3152 if not found. Use the *_or_null() variant for the old semantic.
3153 * In gras_msg_handle, do not discard messages without callback.
3154 They are probably messages to be explicitly awaited later (ie, proofs of
3155 mis-synchronization in userland since they are sent before being awaited)
3156 No big deal usually.
3157 * gras_socket_meas_send/recv: semantic changed!
3158 The numerical arguments used to be (1) the total amount of data to send
3159 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3160 messages. This was need for the fool willing to send more than MAXINT
3161 bytes on quite fat pipes.
3164 * Do really rename the hostmanagement module to peermanagement. [Mt]
3165 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3166 but this is rather new and this was documented in the module
3167 documentation (poor excuses, I admit)
3168 * Bandwidth measurement semantic changed! This follows the changes to
3169 gras_socket_meas_send/recv explained above.
3172 * A sequential mode has been added to the workstations. When a workstation
3173 is in sequential mode, it can execute only one task, and the other tasks
3174 are waiting in a FIFO. [Christophe Thiery]
3177 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3178 model for SIMDAG. [Christophe Thiery]
3179 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3180 equal to 0, which caused some bad behaviors in
3181 saturated_constraint_set_update. I now use a threshold mechanism like in
3185 * When running manually src/testall, you select specific units [Mt]
3186 testall is the result of our cunit mechanism, and should replace all
3187 the scripty thingy around since bash don't run easily on billware.
3189 * A mallocator system has been added. [Christophe Thiery]
3190 Mallocators allow you to recycle your unused objects instead of freeing them
3191 and allocating new ones.
3193 Documentation update:
3194 * FAQ reworking + New FAQs:
3195 - "Valgrind spits tons of errors!" [Mt]
3196 - "How to repport bugs" [Mt]
3197 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3198 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3199 - Communication time measurement within MSG [AL]
3200 - I experience weird communication times when I change the latency [AL]
3201 * GRAS tutorial [Mt]
3203 - an introduction to the framework and to the used communication model
3204 - an initiatic tour introducing the most proheminent features:
3206 . Lesson 0: Installing GRAS
3207 . Lesson 1: Setting up your own project
3208 o Part 2: Message passing
3209 . Lesson 2: Exchanging simple messages
3210 . Lesson 3: Passing arguments to the processes (in SG)
3211 . Lesson 4: Attaching callbacks to messages
3212 . Lesson 5: Using globals in processes
3213 . Lesson 6: Logging information properly
3214 . Lesson 7: Using internal timers
3215 . Lesson 8: Handling errors through exceptions
3216 . Lesson 9: Exchanging simple data
3217 . Lesson 10: Remote Procedure Calling (RPC)
3218 . Lesson 11: Explicitly waiting for messages
3219 . Recapping of message passing features in GRAS
3220 - A HOWTO section containing:
3221 o HOWTO design a GRAS application
3222 More are due, of course. They will come latter. In the meanwhile, you can
3223 check the examples which are still here.
3225 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3227 SimGrid (3.1) stable; urgency=high
3231 There was a stack corruption somewhere, visible only when optimizing
3232 with these versions. [Vince]
3235 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3236 module that enabled to play with Directed Acyclic Graphs. It is built
3237 directly on top of SURF and provides an API rather close to the old
3238 SG. Some old codes using SG are currently under rewrite to check that
3239 all needful functions are provided. [Christophe Thiery]
3242 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3243 extension of the classical CLM03 model that gracefully handles
3244 failures. This is now the default model for MSG and GRAS. It doesn't
3245 handle parallel tasks yet though. [AL]
3246 * Bug fix: Weights were not correctly set in the network part.
3247 WARNING: This may have resulted in incorrect results with simulations
3248 where there are more than one flow on a given link. [AL]
3251 * After a (long ?) discussion on simgrid-devel, we have decided that the
3252 convention we had on units was stupid. That is why it has been decided
3253 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3254 WARNING: This means that all previous platform files will not work as
3255 such with this version! A warning is issued to ask users to update
3257 A conversion script can be found in the contrib module of the CVS, under
3258 the name contrib/platform_generation/surfxml_update.pl [MQ]
3261 * Bug fix: Processes were started in reverse order, wrt deployment file.
3262 WARNING: if your code relies on this bug, please fix it. [AL]
3263 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3264 being executed on two different locations. [AL]
3265 * Bug fix: Failures are now better supported thanks to Derrick's tests
3266 (there was many failure situations I hadn't thought of and that weren't
3267 correctly handled). [AL]
3268 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3272 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3273 Exception occurring on server-side are propagated back to client (!).
3275 API CHANGE: the callback changed their prototype. Change:
3276 int my_handler(gras_socket_t expeditor, void *payload_data) {
3278 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3279 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3281 * New! function: gras_msg_handleall to deal with all messages arriving
3282 within a given period.
3283 * New! function: gras_socket_server_range to get a server socket in a
3284 range of port numbers (ease to avoid port number conflicts) [MQ]
3285 * New! gras processes display their backtrace when they get a SIGUSR1
3286 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3287 Sweet to debug RL processes [MQ]
3291 - Do not force experiment sizes to be expressed in kb, or it becomes
3292 impossible to measure the latency this way (needs one byte-long tests)
3293 WARNING: this changes the amok_bw_* function semantic. [MQ]
3294 - Implements the link saturation stuff. [MQ]
3295 * Peer management module:
3296 New! module factorizing code that we wrote over and over [MQ].
3299 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3300 - Test units are placed directly into the library code, they get extracted
3301 automatically and placed into the src/testall binary.
3302 - Convert most of the XBT tests to this system.
3303 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3304 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3305 live in the same C code. This required to change a little bit the API
3306 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3307 * New module: sparse graph structure with basic algorithms (this is work
3308 in progress and the API is not considered to be frozen yet). [AL]
3309 * Display more information on backtraces: source line & function names are
3310 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3311 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3312 * Reworked a little bit some #include statements to load only required
3313 headers. Some user code that relied on SimGrid to include stdlib or
3314 stdio may need to include it by themselves. [AL]
3315 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3317 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3318 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3319 but I guess that almost nobody used those parts. [MQ]
3321 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3323 SimGrid (3.0.1) stable; urgency=low
3326 * Unfortunately, I had missed 5 misnamed functions:
3327 xbt_fifo_item_t xbt_fifo_newitem(void);
3328 void xbt_fifo_freeitem(xbt_fifo_item_t);
3329 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3330 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3331 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3332 They're now deprecated. Please use their new versions:
3333 xbt_fifo_item_t xbt_fifo_new_item(void);
3334 void xbt_fifo_free_item(xbt_fifo_item_t);
3335 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3336 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3337 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3339 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3340 * Documentation: xbt_log module unmercifully reworked [MQ]
3341 * Bugfix: there was a problem with the ending of contexts with
3342 the pthread backend. It caused some weird deadlock or behavior
3343 depending on the pthread implementation. [AL]
3344 * Bugfix: get the exceptions raised in the simulator repport where
3345 and why they come from when they are not caught in time [AL, MQ]
3348 * Bugfix: Do repport the error when two non-connected hosts try to
3349 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3352 * Add additionnal checkings on communications. Assert that two
3353 communicating hosts are connected by a set of links... [AL]
3356 * Add additionnal checkings on channel values in communication [AL]
3357 * New: MSG_task_get_source to see on which host a task was generated [HC]
3358 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3359 the number of tasks waiting to be received on channel and sent
3361 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3362 waits for the first task coming from a given host.. [AL]
3364 GRAS new functionnalities: [MQ]
3365 * Enhance the parsing macro to allow the size of multidimentional objects
3366 to be given thru annotations.
3367 * New example (and documentation): Matrix Multiplication a la RPC
3368 (as when I was young!) and fix a bunch of bugs found on the way.
3370 GRAS performance improvements: [MQ]
3372 * Reduce the amount of cbps creation/destruction by making it static to
3373 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3376 * Change libdata to a set so that we can search for stuff by ID (and thus
3377 reduce the insane amount of dict lookups)
3380 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3381 RL communication on top of SG since it's so uneffective.
3382 It may also be used for inter-thread communication in RL, one day.
3383 * Use gras_datadesc_copy() to exchange messages on top of SG
3385 - improve message exchange performance on top of SG
3386 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3387 * Don't exchange on the network the size of the used part of buffer,
3388 instead, specify the possible buffer size to read().
3390 - reduces the amount of read/write calls (one pair per exchange)
3391 - reduces the amount of exchanged data (the size)
3392 - allows to retrieve all arrived data on receiver side, if we don't need
3393 it right now (subsequent read will peek the buffer)
3394 - allows the receiver to proceed with the beginning of the stream before
3395 everything is arrived
3396 - make it possible to build an iov transport (using readv/writev)
3398 - take care of the data with non-stable storage (like stacked data),
3400 * If possible, TCP send uses vector I/O (when writev() is here)
3401 - Don't use it for receive since we send data sizes and data on the
3402 same stream, so we wouldn't be able to chain large amount of chunks
3403 before having to flush the stuff to read the size.
3404 * Rework the transport plugin mechanism to simplify it and reduce the
3405 amount of pointer dereferencement when searching for the right function
3408 * I guess that now, we do almost as few system calls as possible while
3409 doing as few data copy as possible.
3411 To improve it further, we could try to send all the sizes first and then
3412 all the data (to use iov on receiving size), but it's only a partial
3413 solution: when you have 2 dimensional data, the sizes of the second
3414 dimension is data of the first dimension, so you need 3 streams.
3416 I'm not sure the potential performance gains justify the coding burden.
3418 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3420 SimGrid (3.00) stable; urgency=high
3423 * New! Give the possibility to hijack the surf parser and thus bypass
3424 MSG_create_environment and MSG_launch_application. Have a look at
3425 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3427 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3429 SimGrid (2.96) unstable; urgency=low
3434 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3435 This deprecates the xbt_error_t mechanisms.
3436 It modifies (simplifies) all XBT and GRAS API.
3437 MSG API keeps unchanged (exceptions raised by XBT are caught from
3438 within MSG and masked with existing error handling facilities)
3441 * New! Add a FATPIPE model. [AL]
3442 * New! Add a parallel task model. [AL]
3443 * New! Add automatically a loopback interface (in the default
3444 network model) if none was precised.
3447 * Bugfix: MSG_process_resume now works with the current running process.
3449 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3450 * Modification of MSG_task_get_compute_duration. Once a task has been
3451 processed, the value returned by this function is now equal to 0. [AL]
3452 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3453 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3454 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3455 of a m_task. For now, MSG_task_cancel only works with computation
3457 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3458 of the processor (in Mflop/s) regardless of the current load on the
3460 * API Change: use proper naming convention for MSG_getClock and
3461 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3463 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3464 This function changes the priority of a computation task. This priority
3465 doesn't affect the transfer rate. A priority of 2 will make a task
3466 receive two times more cpu power than the other ones. This function
3467 has been added to suit the needs of Nguyen The Loc and hasn't been that
3468 much tested yet. So if it fails, please report it and send me your code.
3470 * API Change: removed all functions and types that were marked "deprecated"
3471 since many months. Renamed MSG_global_init_args to MSG_global_init.
3473 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3475 SimGrid (2.95) unstable; urgency=low
3478 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3479 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3480 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3483 * Bugfix: Allow absolute paths for platform description files [MQ]
3484 * Bugfix: do free the variables after use. Leads to drastic performance
3486 * Implement max_duration (ie, timeouts) on resources [AL]
3489 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3491 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3492 the needed support (timeouts) [AL]
3495 * Implement measurement sockets. You can now get the bandwidth between two
3496 hosts thanks to AMOK (see below). [MQ]
3497 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3498 dynar over the network (yeah) [MQ]
3499 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3500 * simple/static token-ring example. [Alexandre Colucci and MQ]
3501 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3502 in place before [MQ]
3505 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3506 type parsing. "short" and "long" modifiers were also ignored; other
3507 modifier (such as reference level) are still ignored. [MQ]
3508 * Embeed the buffer size within the buffer itself on SG. [MQ]
3509 That way, send() are atomic and cannot get intermixed anymore (at least
3510 the ones which are less than 100k; bigger messages still have the issue)
3511 * Array size pushed by the field, not by the field type (or each
3512 and every long int will push stuff to the cbps) [MQ]
3513 * use select() to sleep since it allows to portably sleep less than one
3516 GRAS (minor cleanups)
3517 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3520 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3521 * type category 'ignored' killed as it was never used and were difficult
3523 * whether a type can cycle or not is now a flag, leaving room for more
3524 flags (as "ignored", if we feel the need one day ;)
3525 * Rename raw sockets to measurement sockets since "raw" has another
3526 meaning in networking community.
3529 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3530 and offers features not belonging to GRAS but that most applications
3531 need. One day, it may be a set of plugins loadable at runtime.
3532 * New module: bandwidth
3533 bandwidth measurement between arbitrary nodes running this module. [MQ]
3535 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3537 SimGrid (2.94) unstable; urgency=low
3539 The first beta release of SimGrid 3 !
3543 * Update the main page and the FAQ. Adding references to gforge.
3546 * Add a gras_os_getpid function.
3549 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3550 * Extend the logs so that they also print PID, hostname, date, ... if
3552 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3553 and kill the old version which were in testsuite/
3554 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3558 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3559 for users. Instead, each programming environment (GRAS, MSG,...) use it
3560 when needed to provide such a feature to users.
3561 Don't shortcut the mechanism or you will also shortcut the virtualization
3562 you need on the simulator.
3566 * Cleanups in configury with regard to compile optimization/warning flags.
3567 Also add -fno-loop-optimize to any powerpc since it's the optimization
3568 killing gcc (< 3.4.0).
3569 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3570 needed by MSG examples complications
3571 * Borrow configury beautifications from PHP
3574 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3575 warning (thanks loris for stumbling into this one).
3576 * Bugfix: stop loading private headers (gras_config.h) from the public
3580 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3581 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3582 RL would imply a DNS resolver, which is *hard* to do in a portable way
3583 (and therefore delayed).
3584 * Implement a real timer mechanism and use it in timing macros. This allows
3585 to avoid rounding errors and get a 0.000005 sec precision in timing
3586 macros. While I was at it, various cleanups:
3587 - allow to declare more than one timed section per file (fix a stupid bug)
3588 - move some private declaration to the right place
3589 - merge conditional execution and timing macros into emulation module
3590 - document the module
3591 - make sure the module cleanups its mess on gras_exit
3592 * Documentation improvements:
3593 - (new) how to compile applications using GRAS
3594 - (new) emulation support (timing macros)
3596 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3598 SimGrid (2.93) unstable; urgency=low
3600 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3603 - Use Paje properly where used. Still to be sanitized properly.
3604 - Portability fix: Add an implementation of the contexts using pthread
3608 - Add xbt_procname(): returns the name of the current process.
3609 Use it to show the current process's name in all logging.
3611 - fix detection of older flex version and the reaction, since we do
3612 depend on modern ones (we use lex_destroy)
3613 - Better separation of SG and RL in the libs: remove all simulation code
3614 from libgras. As a result, this lib is now only 200k when stripped.
3615 Some of the xbt modules may also be duplicated (two sets and such) and
3616 should be cleaned/killed before SG3.
3617 - Insist on using xlC on AIX because of weird problems involving gcc there.
3618 - Cleanup the make remote stuff. This is now done by scripts
3619 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3620 mainly for our private use, but we're working on changing them to user
3623 - Bugfix: flush the socket on close only if there is some *output*.
3624 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3626 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3629 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3631 SimGrid (2.92) unstable; urgency=low
3633 Alpha 3 on the path to SimGrid 3
3637 - New! First try of benchmarking macros.
3638 - New! First try so that gras_stub_generator generate deployment and
3639 remote compilation helpers.
3641 - Bugfix: Initialization fix in msg_test.
3645 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3647 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3649 - API Change: gras_os_sleep to take the amount of seconds as a double.
3650 Accepting an int was error prone since it was the only location where
3651 seconds were coded as such. It leaded to damn rounding errors.
3652 - Bugfix: Hard to belive that timers ever worked before this.
3654 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3656 SimGrid (2.91) unstable; urgency=low
3658 Alpha 2 on the path to SimGrid 3
3662 - Bug fix in the lmm_solver.
3664 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3665 through the function MSG_paje_output.
3666 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3667 - It is possible to bound the rate of a communication in MSG with
3668 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3669 to write it in the changelog).
3670 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3675 - Various cleanups to the autotools stuff
3676 - Begin to move Gras examples to examples/gras/
3677 - Let make distcheck work again (yeah!)
3679 - documentation overhauled using doxygen.
3680 gtk-doc-tools is dead in SimGrid now.
3681 - Automatically extract all existing logging categories, and add the list
3682 to the documentation (long standing one, to say the less)
3684 - Cleanup the known architecture table. Reorder the entries to group what
3685 should be, and use a more consistent naming scheme.
3686 (some of the test dataset are still to be regenerated)
3687 - New! Allow library to register globals on each process just as userdata
3689 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3690 lookup time (for now).
3691 Use it in msg and trp.
3692 This cleans a lot the internals and helps enforcing privacy of the
3693 headers between the gras components.
3694 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3695 - Bugfix: gras_os_time was delirious in RL.
3696 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3698 - Reenable GRAS now that it works.
3700 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3702 SimGrid (2.90) unstable; urgency=low
3704 Alpha 1 on the path to SimGrid 3
3706 * It is a long time since the last release of SimGrid. I'm sorry about
3707 that but as I had told you, I was rewriting a lot of things. I apologize
3708 to those who had been reporting bugs to me and that I had not answered.
3709 If your bug is still in the new version, please tell me. Here is a
3710 summary of the main changes.
3712 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3713 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3714 portability, speed, and a lot more but you'll figure it out later.
3715 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3716 I wanted to release the new MSG as soon as possible and I have
3717 broken GRAS, which is the reason why, for now, only MSG is fully
3718 functional. A laconic description of these projects is available
3719 in the documentation.
3721 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3722 kernel optimized for our needs (called SURF but only the developers
3723 should use it). Hence, MSG is now roughly 30 times faster and I think
3724 that by rewriting a little bit MSG, I could event speed it up a little
3725 bit more. Beside the gain in speed, it is also much easier to encode a
3726 new platform model with SURF than it was with SG. More to come...
3728 * REVOLUTION 3: I have tried to change a little as possible the API of
3729 MSG but a few things really had to disappear. The main differences
3730 with the previous version are:
3731 1) no more m_links_t and the corresponding functions. Platforms are
3732 directly read from a XML description and cannot be hard-coded
3733 anymore. The same format is used for application deployment
3734 description. The new format is described in the documentation.
3735 Have a look in tools/platform_generation. There is a tiny script
3736 that converts from the old platform format to the new one. Concerning
3737 the application deployment format, parsing the old one is tricky.
3738 I think most of you should however be able to convert your files. If
3739 it is really an issue, I can write a C code that does the conversion.
3741 2) the toolbox tbx does not exist anymore. We now have a library
3742 with much more data-structures but without the hash-tables (we have
3743 dictionaries that are much faster).
3745 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3747 *****************************************************************************
3748 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3749 * don't feel like converting the oldies. (Mt) *
3750 *****************************************************************************
3753 Version 2.90: "the long awaited one"
3754 - Finished rewriting and debugging MSG. Rewrote the documentation.
3755 - disable GRAS for now since it needs to be ported to the newest SG
3758 - Finish the port to windows (using mingw32 for cross-compile)
3761 - Main loop and datastructures of SURF. A cpu resource object is
3762 functional. Surf can thus be used to create cpu's with variable
3763 performance on which you can execute some actions.
3765 2004-11-15 Martin Quinson
3766 - Port to ARM. Simply added the alignment and size descriptions. Should
3767 work, but the ARM machines are so slow that I didn't had the opportunity
3768 to 'make check' over there yet.
3770 2004-11-15 Arnaud Legrand
3771 - Trace manager now written. It uses a heap structure and is therefore
3772 expected to be efficient. It may however be speeded up (particularly
3773 when many events occur at the same date) by using red and black
3774 trees. One day maybe...
3775 - Max-min linear system solver written. It uses a sparse matrix
3776 structure taking advantage of its expected use. Most operations are
3777 O(1) and free/calloc are called as few as possible. The computation of
3778 the minimum could however be improved by using a red and black tree
3781 2004-11-03 Arnaud Legrand
3782 - Rename every gras_* function that was in xbt/ to its xbt_
3784 - Add a heap and a doubly-linked list to xbt
3785 - Added a dichotomy to the dictionaries. make check works as well before
3786 so I assume that the patch is correct. I do not know however if things
3787 run effectively faster than before now. :)
3789 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3790 SimGrid, and the version number is bumped to 2.x
3792 2004-10-29 Martin Quinson
3793 - Introduction of the remote errors.
3794 They are the result of a RMI/RPC on the remote machine.
3795 ErrCodes being scalar values, you can't get the host on which those
3796 errors did happen. Extending the error mechanism as in Gnome is possible.
3797 No idea yet whether it is a good idea.
3799 2004-10-28 Martin Quinson
3800 - Interface revolution: the Starred Structure Eradication.
3801 I used to do typedef struct {} toto_t; and then handle *toto_t.
3802 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3803 - ???_t is a valid type (builded with typedef)
3804 - s_toto_t is a structure (access to fields with .)
3805 - s_toto is a structure needing 'struct' keyword to be used
3806 - e_toto_t is an enum
3807 - toto_t is an 'object' (struct*)
3809 typedef struct s_toto {} s_toto_t, *toto_t;
3810 typedef enum {} e_toto_t;
3811 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3812 s_toto_t) is private.
3814 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3815 so that it can set the variable to NULL. It was so for dicts and sets,
3816 it changed for dynars.
3818 - Fix a bunch of memleaks in dict_remove
3819 - Fix a bug in sg/server_socket opening: it failed all the time.
3821 2004-10-07 Martin Quinson
3822 - Speed up dynar lookup operation a bit.
3824 gras_dynar_get is dead.
3826 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3827 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3828 (which gives you the address of the stored data).
3830 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3831 copy of the data using an affectation to do the job and not a memcpy.
3833 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3834 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3836 It does not really speedup the dynar test because they are
3837 setting elements all the time (and look them seldom). But the dict does
3838 far more lookup than setting.
3840 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3842 2004-10-05 Martin Quinson
3843 - Allow to (en/dis)able the cycle detection at run time.
3845 Whether we should check for cycle or not is now a property of each
3846 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3847 datadesc_cycle_unset allow to remove this property when previously set.
3849 Note that the cycle detection is off by default since it impacts the
3850 performance. Watch the data you feed GRAS with ;)
3852 This property is hereditary. Any element embedded in a structure having it
3853 set have it set for the time of this data exchange.
3855 You should set it both on sender and receiver side. If you don't set it on
3856 sender side, it will enter an endless loop. If you forget on receiver
3857 side, the cycles won't be recreated after communication.
3859 - Header reorganization.
3860 Kill gras_private.h, each submodule must load the headers it needs.
3862 2004-10-04 Martin Quinson
3863 - Interface revolution: do not try to survive to malloc failure.
3865 Now, gras_malloc and friends call gras_abort() on failure.
3866 As a conclusion, malloc_error is not a valid error anymore, and all
3867 functions for which it was the only gras_error_t return value are
3868 changed. They now return void, or there result directly.
3869 This simplify the API a lot.
3871 2004-09-29 Martin Quinson
3872 - Re-enable raw sockets.
3873 Created by gras_socket_{client,server}_ext;
3874 Used with gras_raw_{send,recv}
3877 It should allow to kill the last bits of gras first version soon.
3879 This is not completely satisfactory yet (duplicate code with
3880 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3883 - Simplify transport plugin (internal) interface by not passing any
3884 argument to _server and _client, but embedding them in the socket
3887 2004-09-28 Martin Quinson
3888 - Finish the port to AIX.
3889 autoconf was my problem (segfault within the malloc replacement
3890 function. No idea why)
3892 2004-09-16 Martin Quinson
3893 - Fix some size_t madness on 64bit architectures.
3895 2004-09-08 Martin Quinson
3896 - Reduce the number of system headers loaded, overload some more system
3897 calls (such as malloc to cast the result of the system one, and work
3899 - Fix and reintroduce the config support
3901 2004-09-07 Martin Quinson
3902 - Source code reorganization to allow Arnaud to surf all over there.
3903 - Allow to document the logging categories.
3904 - Remove all uppercase from logging categories and useless cleanup in names.
3906 2004-08-18 Martin Quinson
3907 Version 0.6.2 (protocol not changed; API changed)
3908 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3909 gras_error_t), and NULL when not found. Functions expecting a msgtype
3910 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3911 hopefully usefull message.
3912 - Portability to prehistoric sparcs again
3914 2004-08-17 Martin Quinson
3915 Version 0.6.1 (protocol not changed; ABI not changed)
3916 - prealloc some buffers to speed things up
3918 2004-08-11 Martin Quinson
3919 Version 0.6 (protocol not changed; ABI expended)
3920 - The parsing macro can deal with the references, provided that you add
3921 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3923 2004-08-09 Martin Quinson
3924 Version 0.5 (protocol not changed; ABI changed)
3925 - Allow to off turn the cycle detection code in data exchange at
3926 compilation time. It should be at run time, but I'm short of time (and
3927 the config stuff is still broken). That way, we keep dict out of the
3928 critical path, which is good because the performance is poor:
3929 - search not dichotomial yet
3930 - dynar give no way to access their content and memcpy everytime
3931 - In composed data description (struct, ref and so on), stop foolness of
3932 keeping the subtype's ID, but store the type itself. This keeps sets out
3933 of the critical path, which is good since they rely on dynar and
3934 dictionaries. The only loose of that is that we cannot detect the
3935 redeclaration of a structure/union with another content (but I'm not sure
3936 the code detected well this error before anyway). We still can detect
3937 the redefinition discrepancy for the other types.
3938 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3939 it breaks the code because of type-punning used all over the place).
3940 This breaks on all non-gcc architectures (for now).
3942 All those changes (plus the buffer of last time) allow me to gain 2 order
3943 of magnitude on cruel tests consisting of 800000 array of integers on two
3944 level of a hierarchical structure (200 secondes -> 4 secondes)
3947 - the selector of reference must now return the type it points to, not
3948 the ID of this type.
3950 2004-08-06 Martin Quinson
3951 Version 0.4 (protocol changed; ABI not changed)
3952 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3953 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3954 - Implement a buffer transport, and use it by default (it relies on tcp in
3955 real life and on sg in simulation).
3956 That's a bit hackish since I had a new field to the structure to store
3957 its data without interfering with the subtype ones. Inheritance
3958 is tricky in C. And that's a kind of reverse inheritance with one class
3959 derivating two classes. Or maybe a game with java interfaces. Anyway,
3960 that's damn hard in C (at least).
3961 Moreover, I got tired while trying to ensure plugin separation and
3962 genericity in SG mode. MSG wants me to do weird things, so let's go for
3963 cruel hacks (temporarily of course ;).
3964 See comment in transport_private.h:71
3965 - do not include all the _interface headers in private but in the files
3966 which really need them (to cut the compilation time when they are
3969 2004-07-26 Martin Quinson
3970 Version 0.3 (protocol not changed; ABI changed)
3971 - Major overhault of the datadesc interface to simplify it:
3972 - shorted the function names:
3973 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3974 - add a trivial way to push/pop integers into the cbps without malloc.
3975 This allows to make really generic sub_type description, which simply
3976 pop their size of the stack.
3977 - add a function gras_datadesc_ref_pop_arr() which does what users want
3978 most of the time: Declare a dynamic array (which pops its size of the
3979 stack) and declare a reference to it. Poor name, but anyway.
3980 - kill the post-send callback, add a post-receive one
3982 2004-07-23 Martin Quinson
3983 Version 0.2 (protocol changed; ABI changed)
3984 - add some testing for cpbs in the test cases, and fix some more bugs.
3985 This invalidate again the little64 data file, since I cannot regenerate
3987 - remove an awfull optimization in the logging stuff, allowing me to:
3988 - understand it again
3989 - learn gcc how to check that the argument match the provided format
3990 - fix all errors revealed by gcc after that
3991 - internal keys of dict are not \0 terminated. Deal with it properly in
3992 loggings instead of segfaulting when the user want to see the logs :-/
3994 2004-07-22 Martin Quinson
3995 - Fix some stupid bug preventing cbps (callback postit) from working
3997 2004-07-21 Martin Quinson
3998 - Some documentation cleanups
3999 - remove the useless last argument of msgtype_declare
4000 - rename the Virtu functions to fit into the 'os' namespace
4001 - move headers src/include -> src/include/gras/ and stop fooling with
4002 gras -> . symbolic link
4003 - make distcheck is now successful
4005 2004-07-19 Martin Quinson
4007 - Build shared library also
4008 - Install html doc to the right location
4009 - stop removing maintainer files in make clean
4010 - build tests only on make check
4012 2004-07-13 Martin Quinson
4014 - No major issue in previous version => change versioning schema
4015 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4016 corresponding dataset.
4018 2004-07-11 Martin Quinson
4020 - Get it working with any kind of structure (we can compute the padding
4021 bytes remotely for all the architectures I have access to)
4022 - Implement the structure parsing macro (still not quite robust/complete)
4023 - Improvement to the remote testing toysuite
4025 2004-07-10 Martin Quinson
4026 [autoconf mechanism]
4027 - get ride of a bunch of deprecated macros
4028 - actually run the test for two-compliment, not only compile it :-/
4029 - test whether the structures get packed (and bail out if yes. Damn.
4030 Alignment is a serious matter)
4031 - test whether the structures get compacted (but respecting the alignment
4032 constraints of each types)
4033 - test whether the array fields of structures can straddle alignment boundaries
4035 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4036 (took me 2 hours to find that bug, looking at the wrong place)
4037 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4038 before sending/receiving and are used to compute the offsets of fields
4039 - Given that padding size depend even on compiler options, keep track of
4040 alignment and aligned_size only for the current architecture. Not a big
4041 deal since we send structure fields one after the other (seems
4043 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4044 program, let it work on linux/gcc/little32. portability todo.
4046 2004-07-08 Martin Quinson
4047 - import and improve remote compilation support from FAST
4048 - make sure make check works on half a dozen of machines out there
4050 2004-07-07 Martin Quinson
4051 Let's say it's version 0.0.3 ;)
4052 - Implement conversions (yuhu!)
4053 - Let it work on solaris (beside conversion, of course)
4054 - Stupid me, using rand() to generate the conversion datatests in not wise.
4056 2004-07-06 Martin Quinson
4057 - Let make dist work, since I'm gonna need it to compile on remote hosts
4058 - Let Tests/datadesc_usage write the architecture on which the file was
4059 generated as first byte.
4060 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4061 architecture descriptions.
4062 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4063 on those architectures.
4064 - Optimization: send/recv array of scalar in one shoot
4066 2004-07-05 Martin Quinson
4067 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4069 - Plug a whole bunch of memleaks
4070 - each process now have to call gras_{init,exit}. One day, their log
4071 settings will be separated
4072 - Continue the code factorisation between SG, RL and common in Transport.
4074 2004-07-04 Martin Quinson
4076 - Redistribution between SG and RL.
4077 We wanna have to accept in SG, so move accepted related parts of RL in
4078 the common part. (more precisely, the dynar of all known sockets is no
4079 more a static in transport.c, but part of the process_data)
4081 [gras_stub_generator]
4082 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4084 2004-07-03 Martin Quinson
4085 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4086 them all up in one shot)
4088 - Fix the ugly case of reference to dynamic array.
4089 - New (semi-public) function gras_datadesc_size to allow the messaging
4090 layer to malloc the needed space for the buffer.
4092 - gras_socket_close now expect the socket to close (and not its address to
4093 put NULL in it after it). This is because the socket passed to handlers
4094 is one of their argument (=> not writable).
4096 - propagate the interface cleanup from last week in datadesc, ie remove a
4097 superfluous level of indirection. User pass adress of variable
4098 containing data (both when sending and receiving), and not of a variable
4099 being a pointer to the data. Let's say that I like it better ;)
4100 The price for that is constructs like "int msg=*(int*)payload" in
4101 handlers, but it's a fine price, IMHO.
4103 - Let it work in RL (yuhu)
4105 2004-06-21 Martin Quinson
4107 - porting SG plugin and SG select to new standards (works almost).
4108 - plug memleaks and fix bugs around.
4111 - cleanup the prototype of data recv and force users to specify when they
4112 want to handle references to objects. Test case working even for cycles.
4113 - plug memleaks. Valgrind is perfectly ok with this.
4115 2004-06-12 Martin Quinson
4117 - cleanup the separation between plugin and main code in plugin creation
4119 2004-06-11 Martin Quinson
4121 - Reput hook for raw sockets, needed for BW experiments
4122 - kill a few lines of dead code
4123 [Data description] Interface cleanup
4124 - gras_datadesc_by_name returns the searched type or NULL.
4125 That way, no variable is needed to use a type desc once, which makes
4127 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4128 two parameters were strings (field name, type name), leading to
4130 [Dicos] Interface cleanup
4131 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4132 This is consistant with the dynar API.
4134 2004-04-21 Martin Quinson
4136 - Porting to new standards.
4138 - interface cleanup.
4139 There is no bag anymore, no need to take extra provision to mask the
4140 pointers behind "ID".
4141 Better splitup of functions between files create/exchange/convert.
4142 This is still a bit artificial since convert and receive are so
4143 interleaved, but anyway.
4145 - add a queued message list to procdata (the ones not matching criteria
4147 - factorize some more code between SG and RL wrt procdata
4149 - use gras_exit in example to track memleaks
4150 - get rid of gs_example now that GS is properly integrated into gras
4151 - update run_test to integrate the latest tests (datadesc)
4153 - rename WARNINGn macros to WARNn since it prooved error-prone
4155 2004-04-19 Martin Quinson
4157 - register init/exit functions within gras module mechanism
4158 - send/receive function.
4159 Convertion is not implemented, but short-cutted if not needed.
4160 struct/array elements are sent one by one (instead of block-wise), but
4161 nobody really cares (yet). Get a prototype before optimizing.
4162 - tests (using a file socket) for DD send/receive on:
4163 - base types: int, float
4164 - array: fixed size, string (ie ref to dynamic string)
4165 - structure: homogeneous, heterogeneous
4166 - chained list, graph with cycle
4167 Believe it or not, valgrind is not too unhappy with the results. The
4168 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4172 - Bugfix when using a filename explicitly (instead of '-')
4174 2004-04-09 Martin Quinson
4176 - factorize more code between RL and SG in socket creation
4177 - Complete the implementation and tests of:
4179 o file (only in RL, and mainly for debugging)
4181 I lost 3 days to design a portable address resolver, and then decided
4182 that the prototype mainly have to run on my box.
4183 Addressing portability too early may be like optimizing too early :-/
4185 - use gras_init in the Tests instead of the crappy parse_log_opt
4186 (the latter function is removed)
4187 [Conditional execution]
4188 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4189 [Code reorganisation]
4190 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4191 Build examples against the RL library, since there is no way to disable
4192 its creation for now.
4194 For information, the beginning of coding on GRAS was back in june
4195 2003. I guess that every line has been rewritten at least twice since