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#41: Add sg_actor_create C interface
26 - FG#43: xbt::random needs some care
27 - FG#48: The Impossible Did Happen (yet again)
28 - FG!24: Documentation and fix for xbt/random
29 - GH#139: Allow pthread creation in SMPI
31 ----------------------------------------------------------------------------
33 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
35 The Palindrom Day Release.
37 Important user-visible changes:
38 - Improve the Python usability (stability and documentation).
39 - A nasty synchronization bug was ironed out, see also below.
40 - Python's doc was organized as a separate tree, now integrated with C++.
41 - C bindings of S4U were not part of the doc.
42 - The C++ doc was also improved as methods are now split by theme.
43 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
44 - This is mandatory to use the Java bindings.
45 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
46 - SimDAG++: Automatic dependencies on S4U activities (experimental)
47 - Some features are already implemented but not all of them
48 - Cannot block an activity until it's scheduled on a resource
49 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
50 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
55 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
56 - Drop support for Viva/Triva (old visualization tools), hence removed
57 TRACE_get_node_types() and TRACE_get_edge_types() functions.
60 - Require pybind11 version 2.4.
61 - Greatly improve locking during context switches with Python.
62 - Declaring Python callbacks to this_actor::on_exit is now working.
63 - No more tests are failing currently, even on win32.
66 - Actor: Merge signals on_migration_start/end into on_host_change
67 - Actor: Rename migrate() into set_host()
68 - Disk: Allow users to get the read and write nominal bandwidth values
69 - Exec: Implement wait_for(timeout)
70 - Io: Implement wait_for(timeout)
73 - Parse errors now raise a simgrid::ParseError that you may want to catch.
76 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
77 been renamed to fix a spelling error. As usual, the previous definitions are
78 kept with a deprecation warning. However, to avoid runtime errors, it is now
79 forbidden to override these deprecated methods in a derived class. Please use
80 the new names immediately if you need to override them.
83 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
84 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
85 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
86 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
87 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
88 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
89 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
90 - New module: random, providing classical random numbers generators.
93 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
94 allocations' size above which they are considered "shared" by default (as if
95 they were performed through SMPI_SHARED_MALLOC macros).
96 Default = 0 = disabled feature.
97 Note : malloc, calloc and free are now overriden by smpicc/cxx by default.
98 This can cause some troubles if codes are already overriding these. If this
99 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
100 help, but will make this feature unusable.
101 - Input check is now performed much more consistently and outputs precise
102 warnings for each error found
103 - New supported calls : MPI_File_set_view, MPI_File_get_view
104 - MPI I/O now only supports disks, as storage is discontinued.
105 - MPI I/O allows opening files without specifying absolute path
106 - Note : SMPI configuration options are now initialized by smpi_init_options()
107 call, which should be called if SMPI is not being used through smpirun.
109 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
110 - FG#9: Python bindings crashing
111 - FG#39: Missing s4u::Comm::wait_any_for example
112 - FG#42: Add support for ThreadSanitizer (TSan)
113 - FG!19: Removing RngStream
114 - FG!20: A module for RNG calls
115 - FG!21: Choice between ad-hoc and standard distributions implementations
117 - FG!25: Fix link in Mutex doc
118 - FG!26: Fix links in SMPI interface doc
119 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
120 - GH#305: Unscheduled tasks are still excuted
121 - GH#313: smpirun: manual outdated w.r.t. --help
122 - GH#321: [S4U] Get task remaining work ratio
123 - GH#323: Crash when an actor turn off his physical host
124 - GH#335: Missing links on dragonfly example svg
125 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
126 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
128 ----------------------------------------------------------------------------
130 SimGrid (3.24) October 9. 2019
132 The Clean Disk Release.
135 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
136 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
137 thus becomes the default resource to execute I/O operations. Examples
138 and the file system plugin have been modified to use disks instead of
139 storage elements. Storage can still be used but is doomed to disappear.
140 Users are thus adviced to update their simulators accordingly.
141 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
142 for consistency with pthread_barrier_wait()
143 - Host::get_englobing_zone() returns the englobing netzone
144 - Actor::on_destruction is now called in the destructor
145 Actor::on_termination new signal called when the actor terminates
147 - Global signals are now part of the Engine:
148 - on_platform_creation: after config settings, before the XML parsing
149 - on_platform_created: right after the XML parsing
150 - on_time_advance: each time the clock advances
151 - on_simulation_end: after simulation, before cleanups
152 - on_deadlock: as the name implies.
154 - sg_{actor,host,link}_{data,data_set}() now all exist.
155 Use them to attach user data to the object and retrieve it.
158 - Introduce an experimental Wifi model. It sounds reasonable
159 according to the state of the art, but it still has to be properly
160 validated, at least against ns-3 if not against reality.
161 - Improved the usability of ns-3. Several bugs were ironed out.
162 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
163 It is now expressed as 'idle:epsilon:allCores' to properly model the
164 consumption of non-whole tasks on mono-core hosts. More info in the doc.
167 - convert a new set of functions to the S4U C interface and move the old MSG
168 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
171 - Fortran bindings for DVFS have been removed.
172 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
173 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
174 - SMPI can now be selected by cmake's find_module(MPI) with
175 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
176 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
177 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
178 - trace-call-location can be used with TI traces, and replayed, and is compatible
179 with smpi/comp-adjustment-file.
180 - sleep events are now correctly traced/replayed.
181 - Default for trace-call-location is now to use file names and not full paths.
182 To revert to previous behavior (in case of collision of filenames), option
183 "smpi/trace-call-use-absolute-path" can be set to yes.
186 - Use the included xxHash as an hash implem when C++14 is usable.
187 - Option model-checker/hash was removed. This is always activated now.
188 - New option smpi/buffering controls the MPI buffering in MC mode.
189 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
190 This is useful to check for MPI compliance.
193 - New section on plugins: how to define a new one, and existing ones.
196 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
197 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
200 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
201 and <mount> tags. The rationale is finish to decouple what should be handle
202 by SimGrid kernel (i.e., resources and activities) from plugins built in
203 user space (e.g., the file system plugin). This new tag currently accepts
204 only three attributes (id, read_bw, and write_bw). All the other information
205 that was declared with the storage related tags now has to be expressed as
206 properties. An example of platform using this new tag is available at
207 examples/platforms/hosts_with_disks.xml
210 - 'expect signal' can now accept more than one potential signal.
212 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
213 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
214 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
215 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
216 - FG#31: per-actor data
217 - FG#34: SG_BARRIER_SERIAL_THREAD?
218 - FG#35: model-checker does not like buster-produced binaries
219 - FG!13: MC: complete workaround in the error msg seen on modern systems
220 - FG!15: execute_flops now logs compute
221 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
222 - FG!17: ns-3: unblock the right number of communications + others issues
223 - FG!18: Improving the performance of the ns-3 bindings
224 - GH#189: Energy consumption of parallel tasks
225 - GH#219: Error in the throughput of TCP transfer
226 - GH!330: Minor polishing of English: Configuring_Simgrid
227 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
228 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
229 - GH!334: Rework energy plugin, again
231 ----------------------------------------------------------------------------
233 SimGrid (3.23.2) July 8. 2019
236 - Nicer introduction page.
237 - Migrate the "Deploy your application" page to the new doc.
238 - Move Java as a subtree of MSG.
241 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
244 - Drop xbt_dynar_sort_strings().
247 - Really fix FG#26: Turning off a link should raise NetworkFailureException
248 - FG#27: Wrong exception thrown to wait_any when link is turned off
249 - GH#328: Java: Canceling multiple tasks in a single vm/host
251 ----------------------------------------------------------------------------
253 SimGrid (3.23) June 25. 2019
255 The Exotic Solstice Release.
258 - SunOS and Haiku OS support. Because exotic platforms are fun.
259 - Stop setting random seed with srand48() at initialization.
260 - Use addr2line as a fallback for stacktraces when backtrace is not available.
261 - Build option -Denable_documentation is now OFF by default.
262 - Network model 'NS3' was renamed into 'ns-3'.
265 - Simgrid can now hopefully be installed with pip.
268 - wait_any can now be used for asynchronous executions too.
271 - New log appenders: stdout and stderr. Use stdout for xbt_help.
272 - Drop xbt_dict_dump.
275 - 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
276 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
277 - MPI interface is now const correct for input parameters
278 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
279 - Fortran bindings for async collectives.
280 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
283 - Remove option 'model-check/record': Paths are recorded in any cases now.
284 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
285 always sparse. That's an immense gain in memory, and can even be
286 faster because of cache effects. So there is no need to clutter the
287 code to allow the user to go for the unefficient mode.
290 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
291 models proved to be more accurate than these old experiments.
293 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
294 - FG#1: Broken link in error messages
295 - FG#2: missing installation documentation
296 - FG#3: missing documentation in smpirun
297 - FG#6: Python bindings not available on PyPI
298 - FG#7: simple cmake call requires doxygen
299 - FG#8: make python bindings an optional dependency
300 - FG#10: Can not use MSG_process_set_data from SMPI any more
301 - FG#11: Auto-restart actors forget their on_exit behavior
302 - FG#12: -Denable_lto=OFF doesn't disable LTO
303 - FG#13: Installs unstripped file 'bin/graphicator'
304 - FG#14: Installs the empty directory 'doc/simgrid/html'
305 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
306 - FG#17: Dead link in doc (pls_ns3)
307 - FG#20: 'tesh --help' should return 0
308 - FG#21: Documentation link on http://simgrid.org/ broken
309 - FG#22: Debian installation instruction are broken
310 - FG#26: Turning off a link should raise NetworkFailureException exceptions
311 - GH#133: Java: a process can run on a VM even if its host is off
312 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
313 - GH#326: Valgrind-detected error for join() when energy plugin is activated
315 ----------------------------------------------------------------------------
317 SimGrid (3.22) April 1. 2019
319 The Easter Christmas Release (this one is somewhat late).
322 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
323 - Not complete yet: asynchronous activities (amongst others) are still missing
324 - Still ongoing: the interface may change in the future. We need more
325 testers! Please report any glitches.
326 - No new project using Java should start now. Please switch to Python.
329 - Some of the internal cleanups may lead to speed improvements:
330 - The hard limitation on the amount of simulated actors with
331 Java+Mac was removed. Now, the available memory is the only limit.
332 - Our refcounting was tidyied, leading to 10% speedups in some cases.
333 - We are still working on making our code robust to the actor kills
334 and hosts' churn. Things are improving, but it's not perfect yet.
335 - Replaced our own code to display a backtrace (that was forking addr2line)
336 with the Boost.Stacktrace library.
337 You won't see your backtraces without this optional dependency.
338 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
339 - Stop setting random seed with srand() at initialization.
342 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
343 XML file version remains 4.2 since old files are still compatible.
346 - Process termination which was broken at version 3.21 has been repaired.
347 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
348 - Hide the examples into examples/deprecated. New users should use Python.
351 - Drop MSG_process_create_from_stdfunc() from the API.
352 This C++ function was a pimple in the C API, made necessary at some
353 point by the Java bindings. This is fixed now.
354 - Hide the examples into examples/deprecated. New users should use S4U.
355 - MSG_process_create and MSG_process_attach now crash if the host used
356 to run this process is off.
357 - Fix the protype of MSG_process_on_exit()
358 Now use (int,void*) callbacks instead of (void*,void*) ones.
359 The implementation was ways too messy to actually work, I guess.
362 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
363 too often and provide more accurate timings.
364 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
365 - MPI_Alltoallw support
366 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
367 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
368 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
369 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
371 - Basic implementation of generalized requests (SMPI doesn't
372 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
375 - Drop sg_cmdline. Please use xbt_cmdline instead.
376 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
377 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
378 OsSemaphore is implemented in a portable way with C++11 threads.
379 This should allow much more threads to be created at the same time,
380 allowing Mac and Java users to simulate many more actors.
381 - Implement the 'thread' factory with std::thread instead of xbt ones.
382 It is not possible to set the stack size with threads anymore, but
383 -fsplit-stack is the way to go nowadays when using threads.
384 - Drop the xbt_os_thread_t module (now unused)
385 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
386 - Drop xbt_str_join_array().
387 - Drop cunit, use Catch2 instead.
390 - Many cleanups in the kernel::activity namespace. This was long
391 overdue, and shall open the path to many future endeavors.
394 - #132: Java: a process can not shut down its own host
395 - #220: S4U: detached send causes exception if sender terminates soon after sending
396 - #261: Document the parameters of parallel execution's constructor
397 - #300: [s4u] BarrierPtr is missing
398 - #314: SMPI args internal cleanup
399 - #316: Fix a bug related to the CPU utilization of multi-core VM
400 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
401 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
402 - #325: Turning off a host has different behavior on sleeping actors and computing actors
404 ----------------------------------------------------------------------------
406 SimGrid (3.21) October 3. 2018
408 The Restarting Documentation (TRD) Release.
411 - Convert the user manual to Sphinx for improved usability.
412 Unstable version now hosted on https://simgrid.frama.io/simgrid
413 Many glitches remain; stay tuned.
414 - Restrict the Doxygen reference API to the installed header files.
415 (documenting all internals this way is overkill + doxygen scales badly)
416 - New tutorial on S4U (highly inspired from the old MSG one, but with
417 a git repository to fork as a starting point, and a docker image)
418 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
419 - The Developper manual is still to be converted (not compiled until then)
420 - Some parts are still missing in this conversion (such as the
421 platform and deployment sections) while others were blindly converted
422 and would need more love (such as the configuration flags).
423 Things will certainly further improve in the future releases.
426 - Fully reimplement auto_restart mechanism that was utterly broken
427 (fix #22, #131 and #281 that were all very related)
428 - Implement semaphores (fix #303 and #291)
429 - When creating an actor from a function and its parameters,
430 move-only parameters are not allowed anymore, as it would prevent
431 the actor to be restartable if its parameters are consumed on run.
432 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
433 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
434 IO operations that can be started, waited for, or canceled as a regular
436 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
440 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
441 - New host variable: 'core_count'
444 - Remove xbt_os_thread_specific features
445 - Remove portability wrapper to condition variables
446 - Remove xbt_os_thread_yield()
449 - MPICH collective selector now mimics MPICH 3.3b
450 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
451 - MPI_Init(NULL,NULL) is now allowed
452 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
453 - Command-line arguments should now be handled properly in Fortran simulations
454 - MPI Attributes and related callbacks should work in Fortran
455 - Apps using C + Fortran should now work
456 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
457 - Support MPI_Error_String
460 - Due to an internal bug, Msg.run() must now be your last line.
461 We hope to fix it in a future release, and we are sorry for the inconvenience.
464 - #22: Process autorestart seem to only work with CAS01 cpus
465 - #56: Feature request: dynamic SMPI replay
466 - #93: simgrid should not eat --help
467 - #111: How to change the loopback link characteristics?
468 - #116: Platform cluster radical
469 - #117: Platform cluster/host order
470 - #129: Replay_multiple on steroid
471 - #131: Java : autostart does not work
472 - #143: Setting a breakpoint at a given time
473 - #235: Network energy models should be integrated
474 - #264: Add ptask L07 resource tracing
475 - #271: Dynamic and manual replay of SMPI traces
476 - #279: Breakpoints option uses time deltas instead of absolute time
477 - #281: Daemonized actor no longer daemonized after an auto-restart
478 - #282: TIT convertor erase traces when using absolute path in the trace list file
479 - #285: segfault when a s4u actor kills itself with msg process tracing activated
480 - #286: Pajé traces are not exposing the number of cores
481 - #287: Command-line parsing should stop when encountering '--'
482 - #288: MPI_Init(NULL, NULL)
483 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
484 - #290: Method to check if a mailbox has a message ready to be consumed
485 - #291: [s4u] semaphores do not seem available yet
486 - #294: Alltoall collective implementation of mpich
487 - #295: Starting a Host as OFF is not supported
488 - #296: DTD too permissive
489 - #299: Add a s4u barrier example
491 ----------------------------------------------------------------------------
493 SimGrid (3.20) Released June 24. 2018
495 The proxy snake_case() release.
498 - The whole API is now using snake_case() instead of mixing it with camelCase.
499 Compatibility wrappers in place for one year.
500 - We took the opportunity to increase the consistency of the API.
501 Please report any remaining glitches.
502 - New function simgrid::s4u::Host::get_actor_count:
503 Returns the number of actors running on a specific host.
506 - SMPI is now tested with ~45 proxy apps from various sources, with none or
507 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
508 - Replay: The replay file has been re-written in C++.
509 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
510 supported. They are stored in the trace and used when replayed.
511 - Basic support of MPI_Cancel. Robustness not guaranteed.
512 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
513 (only for MPI_COMM_TYPE_SHARED).
514 - New option: smpi/privatize-libs, to add external shared libs to be privatized
515 by SMPI. They will be copied locally and loaded separately by each process.
516 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
517 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
518 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
519 with internal index optind with multiple processes. Only works if MPI_Init has
521 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
522 syntax without preprocessing tricks.
525 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
526 behavior tracing is no longer limited to MSG
527 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
528 this means that traces now need to be updated or re-obtained; both Irecv
529 and Isend lines in your traces have now in total 5 mandatory fields:
530 <rankid> <command> <to/from rankid> <tag> <size>
531 To update your traces, it suffices to add a 0 for the tag here.
532 - TIT now also supports waiting for a distinct request via MPI_Wait.
533 Wait/Test now wait for a specific request, not just the last one that was
534 issued. This unfortunately means another update, because we need to
535 identify which request you want to wait for. We do this via the
536 triplet (sender, receiver, tag), which needs to be added:
537 <rankid> <command> <sender> <receiver> <tag>
538 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
539 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
542 - The deprecation of MSG is ongoing (but this should not impact you).
543 Many MSG functions are now simple wrappers on the C API of S4U. If
544 you wish to convert your code to S4U, find the S4U counterparts of
545 your MSG calls in src/msg/msg_legacy.cpp.
546 - MSG can still be used, but won't evolve anymore.
549 - Allow to run the Link energy plugin from the command line with
550 --cfg=plugin:link_energy
551 - Rename Energy plugin into host_energy
552 - Rename Load plugin into host_load
555 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
556 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
557 - Improve performance when handling timeouts of simix synchros.
560 - Config: the C API is now deprecated (will be removed in 3.23), and
561 the C++ API has been slightly improved.
564 - Fix several build issues on OSX.
565 - Move simgrid_config.h to simgrid/config.h (old header still working)
568 - #143: Setting a breakpoint at a given time
569 - #150: Inconsistent event names in SMPI replay
570 - #258: daemonized actors hang after all non-daemonized actors have completed
571 - #267: Linker error on unit_tmgr
572 - #269: SMPI: tracing of MPI_Wait/all/any broken
573 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
575 ----------------------------------------------------------------------------
577 SimGrid (3.19.1) Released March 22 2018
579 The Fixed ABI Release.
581 As you may know, we are currently refactoring SimGrid in deep.
582 Upcoming SimGrid4 will be really different from SimGrid3: modular,
583 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
585 Our goal is to smooth this transition, with backward compatibility
586 and automatic update paths, while still progressing toward SimGrid4.
588 SimGrid remains open during works: The last pure SimGrid3 release was
589 v3.12 while all subsequent versions are usable alpha versions of
590 SimGrid4: Existing interfaces remain unchanged, but the new S4U
591 interface is budding and the internals are deeply reorganized.
593 Since 2015, we work hard to reduce the changes to public APIs. When
594 we need to rename a public library symbol in S4U, we let your
595 compiler issue an explicative warning when you use the deprecated
596 function. These messages remain for four releases, i.e. for one full
597 year, before turning into an error. Starting with v3.15, your can
598 also adapt to API changes with the SIMGRID_VERSION macro, that is
599 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
601 Starting with this v3.19.1, our commitment to reduce the changes to
602 the public interfaces is extended from the API to the ABI: a program
603 using only MSG or SimDag and compiled against a given version of
604 simgrid can probably be used with a later version of SimGrid without
605 recompilation. We will do our best... but don't expect too much of
606 it, that's a really difficult goal during such profund refactoring.
608 The difference between v3.19 and v3.19.1 is that the former was
609 accidentally breaking the ABI of MSG, while the later is restoring
612 S4U and kernel APIs will still evolve until SimGrid4, with one-year
613 deprecation warnings as currently. In fact, cleaning up these
614 interfaces and converting them to snake_case() is one release goal of
615 v3.20. But don't worry, we are working to smooth this upgrade path.
617 In summary, new projects should start with S4U to benefit of the
618 future, but old MSG projects should still be usable with no change.
620 ----------------------------------------------------------------------------
622 SimGrid (3.19) Released March 20 2018
624 The Moscovitly-cold Spring release.
626 SMPI user-visible improvements:
627 - Support of MPI_Type_create_subarray
628 - Unimplemented MPI calls now abort (instead of issuing a warning)
629 Some harmless calls around errhandlers still only issue a warning.
630 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
631 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
632 is now possible (mmap privatization was refactored to this end).
633 Adding new SMPI instances should also be technically doable now.
635 SMPI internal cleanups:
636 - Removed index notion from SMPI -- no more "getPid() - 1"!
637 In tracing files, actors are called rank-<PROCESS-ID> but these are
638 NOT the real ranks (tracing several communicators is not supported yet).
639 - Deprecate smpi_process_index() (will be removed in 3.22)
640 - Better testing on classical proxy apps, through an external project:
641 https://github.com/simgrid/SMPI-proxy-apps/
644 - Use a graphical TOC to make it easier to find the documentation you need
645 - Revamp the MSG tutorial
648 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
649 - Remove parameter of MSG_process_killall().
650 Resetting the PID was bogus anyway (several actors could have the same PID).
653 - Execution->setHost() can be called after start(), to migrate it.
654 - Comm::test_any() is now implemented.
655 - s4u::Actor now has onCreation() and onDestruction() signals.
656 - Install some kernel header files for the users' plugins and more.
657 Warning, their API is really not stable yet.
658 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
659 include/simgrid/kernel/routing/*Zone.hpp
662 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
664 Build System and other cleanups:
665 - Remove unused run-time parameter "tracing/onelink-only".
666 - SimGrid now works with intel compilers (MC still buggy)
667 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
668 - tesh kills subprocesses on timeout (but not on windows)
671 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
674 - #194: Feature request: simgrid::s4u::Comm::test_any()
675 - #245: migrating an actor does not migrate its execution
676 - #253: Feature Request: expose clusters as objects
677 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
678 - #255: Tesh broken on Windows
679 - #256: Modernize FindSimGrid.cmake
680 - #257: Fix (ab)use of CMake install
682 ----------------------------------------------------------------------------
684 SimGrid (3.18) Released December 24 2017
686 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
688 MSG IS NOW DEPRECATED. Please use S4U for new projects.
689 - Support for MSG should not be removed from SimGrid before 2020,
690 but future evolutions will be limited to the integration of
691 user-provided patches.
694 - s4u::onDeadlock() signal, run before stopping the simulation.
695 - s4u::Exec: asynchronous executions (abstraction of background
697 - s4u::Host->getLoad() returns the achieved speed in flops/s
700 - Link::name() is deprecated and will be removed in v3.21, use
701 Link::getCname() instead.
702 - Mailbox::getName() changed to return a std::string, use
703 Mailbox::getCname() instead to get a char*.
704 - Storage::getName() changed to return a std::string, use
705 Storage::getCname() instead to get a char*.
706 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
707 to have both allocation and deallocation of the map in user space.
710 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
711 The amount of flops remaining to do is a vector, not a scalar.
712 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
713 name implies on both sequential and parallel tasks.
714 - Both changes fix GitHub's #223 using PR #237 as a basis.
715 Thanks Michael Mercier.
716 - Most examples were converted to S4U and hidden elsewhere at tests.
717 Rationale: we still want MSG to work; we want newcomers to use S4U.
720 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
721 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
724 - Switch to the faster dlopen privatization mechanism by default
725 - Documentation improvments
728 - Live migration is getting moved to a plugin. Dirty page tracking is
729 the first part of this plugin. This imply that VM migration is now
730 only possible if one this function is called:
731 - C/MSG: MSG_vm_live_migration_plugin_init()
732 - C/C++: sg_vm_live_migration_plugin_init()
733 - Java: Msg.liveMigrationInit()
734 For C and C++, "simgrid/plugins/live_migration.h" also has to be
738 - Define class simgrid::xbt::Path to manage file names.
739 - Removed unused functions:
740 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
741 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
742 xbt_graph_export_graphviz()
743 - xbt/str.h: xbt_str_join()
744 - Remove unused datatypes:
745 - xbt/heap.h: use std::priority_queue or boost::heap instead
746 - xbt/swag.h: use boost::intrusive::list instead
749 - New link_energy plugin for the consumption of the links.
750 - All of the operations on files and storage contents have been
751 packaged into a plugin (src/plugins/file_system). The current
752 public interface can be found in
753 include/simgrid/plugins/file_system.h
754 To use these functions you now have to initialize the plugin by
755 calling MSG_storage_file_system_init() just after calling
756 MSG_init() or sg_storage_file_system_init() just after creating
760 - Remove the undocumented/untested tag <include>
763 - Remove viva specific tracing as the tool is no longer maintained
766 - #248: Kill of finished processes leads to segfault
767 - #240: xbt_cond_wait_timeout should gracefully return for C apps
768 - #239: please implement signal s4u::onDeadlock()
769 - #230: segfaults when exit() without run()
770 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
771 - #223: MSG_task_get_flops_amount() not working with parallel tasks
772 - #222: Actor::kill() doesn't really kill and segfaults
773 - #221: odd LMM warning when killing an actor
774 - #120: Memory leak when the processes are forcefully killed
776 ----------------------------------------------------------------------------
778 SimGrid (3.17) Released October 8 2017
780 The Drained Leaks release: (almost) no known leaks despite the tests.
782 Even very long simulations will preserve your memory: our testsuite
783 stresses SimGrid for over 45mn on fast machines for a coverage of
784 over 80%, and there is only one single known leak, of about 4kb.
788 - Comm.detach(): start and forget about asynchronous emission. A cleanup
789 handler may be given to free resources if the comm cannot be completed.
790 - this_actor::send(mailbox) is now mailbox->put()
791 - New: simgrid::s4u::Comm::wait_all()
792 - New: Host.execute() for remote executions.
795 - Unused option network/sender-gap is removed.
798 - Deprecate MSG_task_isend_with_matching(): This unused feature
799 really complicates our internals. Will be removed in v3.20.
802 - Improved context termination. It is now possible to free resources, even
803 when a process is forcibly killed.
804 - ContextBoost: add support for Boost versions above 1.61.
807 - Bring back run-time option --cfg=exception/cutpath to remove exception
809 - Removed unused functions:
810 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
811 xbt_str_rtrim(), xbt_str_trim().
812 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
815 - Removed header files obsolete since SimGrid 3.12:
816 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
818 - #3: SD_exit should be made optional
819 - #120: Memory leak when the processes are forcefully killed
820 - #159: Threading test regression in Actor refcounting
821 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
822 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
823 - #186: Actor::killAll() segfaults if some process is blocked on wait()
824 - #191: VM migration and pstate
825 - #192: Updating the energy consumptions of all hosts crashes with VMs
826 - #195: All actors have PID=0 in the logs
827 - #204: Sometimes segfault with thread contexts and mmap privatization
828 - #222: Actor::kill() doesn't really kill and segfaults
829 - #225: Actor::kill() doesn't really kill when victims are doing a join()
831 ----------------------------------------------------------------------------
833 SimGrid (3.16) Released June 22. 2017.
835 The Blooming Spring Release: developments are budding.
838 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
839 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
840 - Java JAR file should be finally fully working on Mac OSX too.
842 XML platforms: Switch to platform v4.1 format.
843 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
844 - <zone> can be used as a synonym for the now deprecated <as>
845 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
846 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
847 - <actor> can be used as a synonym for the now deprecated <process>
848 - state_file and avail_file periodicity is now easier to express
849 (check the documentation) the old behavior should still work.
850 * Storage: not backward compatible, but it seems that nobody ever used it
851 - Remove attribute 'content_type' of <storage_type>: was never used
852 - Make attribute 'model' of <storage_type> optional: for future usage
853 - Remove Bconnection model property: was never (in)validated. Replaced by
854 the maximum of the read and write bandwidth as a resource constraint.
857 - New and Backwards Compatibility break:
858 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
859 When one wants to get the list of tasks whose states have changed during a
860 simulation round, s/he has to allocate and free a dynar and use it as
861 argument to this function. The former SD_simulate (double how_long)
865 - Allow multicore VMs, with the correct sharing (unless you overcommit)
866 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
869 - New function to update the consumption of all hosts at once.
870 - Fix the model for multi-core hosts, linear in the amount of busy
871 cores with an abnormality for pIdle (see paper).
872 BUG: Still not satisfactory for ptask on mono-cores.
875 - The netzone are now available from the MSG API.
876 The old names still work, but are now deprecated.
879 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
880 - New option: smpi/keep-temps to not cleanup temp files
881 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
882 mallocs mode is used, allows to change the size of the fake file used
883 (default 1MB), to potentially limit the number of mappings for large runs.
884 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
885 - Fortran ifort and flang compilers support
886 - New RMA calls supported (experimental):
887 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
888 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
889 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
890 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
891 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
892 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
895 - New: this_actor::isMaestro() returns whether we are in kernel mode.
896 - Behind the scene, ActivityImpl are now automatically refcounted.
897 Removing this nasty bottleneck will greatly simplify our internals.
900 - Replay: New function xbt_replay_action_get():
901 Retrieve the function previously associated to an event type.
902 - DROPPED FUNCTION: xbt_str_varsubst()
903 - DROPPED MODULE: strbuff. We don't need it anymore.
904 - DROPPED MODULE: matrix. We don't need it anymore.
905 - DROPPED MODULE: lib. We don't need it anymore.
907 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
909 SimGrid (3.15) Released March 22. 2017
911 The Spring Release: continuous integration servers become green
912 We fixed even the transient bugs on all target architectures:
913 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
914 Windows, FreeBSD, NetBSD.
916 - Rename NetCards to NetPoints.
917 This was intended to help NS3 users, but that's not a netcard.
918 That's a point in the routing algorithm, let's avoid wrong simplifications.
921 - New: MSG_process_yield(). Stop and yield to other processes.
922 - New: MSG_process_daemon(). Daemon processes are automatically killed
923 when the last non-daemon process terminates
924 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
925 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
926 to make room for the upcoming network energy plugin.
927 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
930 - Ensure that an actor can kill itself with Process::exit()
931 - Kill the obscure NativeException. Nobody want to survive the issues
932 it denotes, so use JniException that is a RuntimeException (not to
933 be caught explicitly).
934 - Partial bug fix in initialization. SimGrid flags on command line were
935 consumed at C level but stayed in the original Java String[] args.
936 This could mess users' args[i] if SG flags were not put at the end of
938 The SimGrid flags are now removed from the Java arguments. However,
939 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
940 if args.length is greater than the number of YOUR OWN ARGUMENTS.
941 It might be if you have --log or --cfg flags in the command line.
942 - Fix numerous memleaks all around the place. In particular, around VMs.
946 - s4u::onPlatformCreated: right before the simulation starts
947 - s4u::onSimulationEnd: right after the main simulation loop
948 - s4u::onTimeAdvance: right after a clock change (time is discrete)
949 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
950 event from the availability_file changes the avail speed.
951 - Links are now usable from s4u
952 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
953 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
954 - New examples: The conversion of MSG examples to S4U has begun
955 - Actors: create, kill, migration, and suspend
956 - Applications: master-worker and token-ring
957 - Action replay: communications and storage
958 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
961 - Backwards Compatibility breaks
962 - The SD_route_get_list and SD_route_get_size functions have been removed.
963 They are replaced by the unique
964 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
965 The route from src to dst is built in the links dynar whose size can
966 be obtained with xbt_dynar_length.
967 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
968 sg_host_route_bandwidth and sg_host_route_latency.
969 Macros ensure the backwards compatibility, but you should fix your code
972 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
973 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
974 - Bump our claim of support from MPI 1.1 to MPI 2.2.
975 We don't support 100% of it, but it should be enough. Contact us if not.
976 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
977 naive and probably inaccurate)
978 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
979 mapped onto a unique small file using some system magic.
980 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
981 - Fortran: cleanups, fixes, support of user-added operations
982 - RMA: MPI_Accumulate are applied in correct order.
983 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
984 - Better support for MPI_IN_PLACE
985 - Support for MPI_Win attrs and keyvals.
986 - Support MPI_Comm_group_create, MPI_Type_size_x
987 - MPI ops should be properly applied to all allowed types
988 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
989 algorithm by a non failing one from mpich.
992 - Kill the fifo data container: we don't use it anymore.
995 - A <cluster> can now be created with different speed values to represent pstates.
996 see examples/platforms/energy_cluster.xml for an example.
998 SimGrid (3.14.159) Released December 28. 2016
1000 The previous release was only a crude approximation of a Pi release;
1001 this one should be a bit better.
1003 - Revert a last minute change that broke on Mac OSX.
1004 - Fix the changelog and NEWS files.
1005 - Improve the documentation, in particular of the routing module.
1006 - Integrate some patches flying around in Debian.
1008 SimGrid (3.14) Released December 24. 2016
1011 * The whole documentation was reworked and reorganized. There is
1012 still much room for improvement here, but we're on it.
1015 * We now need python3 for our testing infrastructure.
1016 * Model checking needs libevent
1019 * Now works on FreeBSD
1022 * Privatization now works on FreeBSD
1024 * Privatization is now activated by default in smpirun
1025 You can turn it off with -no-privatize if something goes wrong.
1027 * Call-location tracing for SMPI.
1028 You can add the exact location (filename / linenumber) of an MPI call to
1029 your trace files and slow down or speed up the simulation between two
1030 consecutive calls by using an adjustment file (see the documentation).
1032 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1033 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1034 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1035 values were ignored in some cases. The timings of these functions can now
1036 be significantly different.
1038 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1039 smpi/running-power is renamed to smpi/host-speed
1041 * smpi/grow-injected-times option to enable or disable multiplication of the
1042 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1043 make simulation less precise (but also much faster).
1045 * smpirun script should be (much) faster for large deployments.
1047 * SMPI tracing: fixed issue with poor matching of send/receives.
1049 * Replay: Fix broken waitall
1051 New functions and features
1052 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1054 Dropped / renamed functions and features
1055 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1056 - MSG_mailbox_is_empty() -> Mailbox::empty()
1057 - MSG_mailbox_front() -> Mailbox::front()
1058 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1059 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1060 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1061 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1062 * Don't pass the free_f parameter to property related functions:
1063 - MSG_host_set_property_value()
1064 - MSG_as_router_set_property_value()
1065 - MSG_storage_set_property_value()
1066 * VM properties. Since msg_vm_t are msg_host_t, just use
1067 MSG_host_get_property_value() and friends
1068 * VM I/O related things:
1069 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1070 - Unimplemented save/restore methods
1071 * MSG_as_router_get_property_value() was redundant with
1072 MSG_environment_as_get_property_value().
1073 - Removed MSG_as_router_*propert*() functions
1074 - Added MSG_environment_as_set_property_value() for consistency
1075 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1076 Well, they are still there for now, but deprecated with a warning.
1077 Please switch to xbt_dict_new_homogeneous() before this is removed
1079 * Task affinity. Its intended behavior (that was very badly tested
1080 and probably not really working) was deceiving what most users
1081 would have hoped here.
1082 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1083 * Option network/coordinates is now useless and should be dropped.
1086 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1089 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1090 * Added Dragonfly topology support
1093 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1094 EiBps, EBps, Eibps, Ebps for bandwidth.
1095 They may become useful to some lucky ones.
1098 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1099 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1100 * ex/async_waitAll: new example, on asynchronous communications
1103 * Memory usage should be decreased for simulations with a large number
1104 of processes. This also helps for SMPI.
1106 SimGrid (3.13) stable; urgency=low
1108 The Half Release, a.k.a. the Zealous Easter Trim.
1110 - We removed half of the lines, that were mostly experimental cruft.
1111 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1112 (not counting blanks and comments -- according to openhub.net)
1113 - The internals are now compiled in C++ (and will soon be clean C++)
1114 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1115 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1117 * Backwards Compatibility breaks
1118 - Removed Lua simulation bindings (switch to C or Java for that).
1119 Lua can still be used to describe platforms
1120 - Removed Java kernel plug-ins.
1121 Will be reintroduced after the ongoing major internals reorg.
1123 - the following functions were removed.
1124 They were too specific and should be reimplemented in a generic
1125 way, with filter function.
1126 - MSG_task_listen_from_host
1127 - MSG_mailbox_get_count_host_waiting_tasks
1128 - MSG_mailbox_put_with_timeout was removed.
1129 Please use MSG_task_send_with_timeout instead.
1131 - the SD_application_reinit function was removed. It has been a noop for a while.
1132 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1133 replaced by a more flexible mechanism.
1134 - The following functions thus do not exist anymore
1135 - SD_workstation_get_access_mode
1136 - SD_workstation_set_access_mode
1137 - SD_workstation_get_current_task
1138 - Basic estimation functions have been removed but can easily be replaced
1139 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1140 - SD_workstation_get_computation_time => amount / sg_host_speed()
1142 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1143 Use VM.getSpeed()*load/100 for the legacy behavior.
1145 - option enable_tracing was removed. It was not doing anything for a while.
1146 - In the ModelChecker:
1147 - the model-checker now ptraces the model-checked process which means
1148 you cannot use a debugger on the latter anymore (we might make this
1149 optional in the feature);
1150 - removed soft-dirty page tracking;
1151 - remove model-checked side snapshot management,
1152 MC_snapshot() and MC_compare_snapshot();
1153 - keep the MC_cut() function as a stub (it was not really working
1154 in the previous release).
1157 * All options are consistently in kebab-case. Old names are kept as alias.
1160 * Switch to platform v4 format.
1161 - Rename from 'power' to 'speed' the attributes describing the amount of
1162 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1163 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1164 - In <host> and <link>, attributes availability and state are gone.
1165 It was redundant with state and availability traces, and with peak values.
1166 - In <cluster>, attributes availability_file and state_file are gone.
1167 It was too complex and unused.
1168 - Kill <gpu>. Was not doing anything.
1169 - The DOCTYPE points to the right URL:
1170 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1171 (the file at this address now documents the changelog since its v1)
1172 - A warning is emitted for unit-less values (they are still accepted).
1173 - speed. Default: 'f' or 'flops'. Also defined:
1174 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1175 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1176 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1177 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1178 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1179 - latency. Default: 's' second. Also defined:
1180 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1182 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1183 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1186 * s4u::Host is now the preferred public interface to the Host features.
1187 sg_host_* functions are C bindings to the exact same behavior
1188 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1191 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1192 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1193 They were deprecated since a few years, and probably did not even compile anymore.
1196 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1197 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1198 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1199 encouraged to update their codes. A list of the modified functions can be found at the end of
1200 include/simgrid/simdag.h
1203 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1204 the lambda or closure passed as a parameter will run in kernel mode.
1205 All the callback functions should be rewritten to that interface at some point.
1208 * Reorganizing and cleaning the internals all around the place.
1211 * Remove old default barrier/bcast buggy algorithms (see #18407)
1212 * Various bug fixes to handle more codes
1213 * Remove the need for the --foreground option of smpirun (it is still
1214 accepted for backward compatibility).
1217 * Kill the setset data container: MC don't use it anymore.
1218 * Kill the queue data container: it made more sense with GRAS.
1219 * Kill the xbt_peer_t data type: it's useless without GRAS.
1220 * Kill rm_cb feature of config sets: it was never useful.
1221 * Kill graphxml parsing feature. It was not used.
1222 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1224 - xbt_dynar_sort_strings(), when the content is char*
1225 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1226 They throw exceptions on invalid input;
1227 * C++ support for declaring CLI flags (simgrid::config::Flag);
1228 * class for abstracting different signal backends (simgrid::xbt::signal).
1229 with no external dependencies (we need very simple signals).
1232 * refactoring and cleanup of the code;
1233 * ongoing process to cleanly separate the model-checking algorithms
1234 from the code model-checking support.
1236 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1238 SimGrid (3.12) stable; urgency=low
1240 The Facelift Release.
1243 * Require g++ v4.7 at least to not speak prehistorical C++.
1244 * Require Boost 1.48 (for signal2 component).
1245 * Java must be version 7 at least when activated.
1246 * Builds on Windows again (including Java bindings).
1247 * Tracing is now always enabled (no way to turn it out)
1248 * Remove GTNetS. It was not working anyway.
1249 * Various cleanups in the cmake scripts.
1250 * Move headers around to sort them out on installed systems:
1251 - instr/instr.h -> simgrid/instr.h
1252 - instr/jedule/* -> simgrid/jedule
1253 - simdag/datatypes.h was removed
1254 - simdag/simdag.h -> simgrid/simdag.h
1255 - msg/datatypes.h was removed
1256 - msg/msg.h -> simgrid/msg.h
1259 * Interface improvement:
1260 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1261 - Sanitize the interface in MSG_task_ module:
1262 - Merge two functions that were close enough but misleading:
1263 set_compute_duration(t) -> set_flops_amount(t)
1264 get_remaining_computation(t) -> get_flops_amount(t)
1265 - set_data_size(t) -> set_bytes_amount(t)
1266 get_data_size(t) -> get_bytes_amount(t)
1267 - Massive cleanups in the functions related to the energy
1268 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1269 - New: MSG_host_get_pstate()
1271 - msg/energy/onoff: switching hosts on and off
1273 * Interface improvement:
1274 - Rename Host.isAvail() to Host.isOn()
1275 - Rename Process.currentProcess() to Process.getCurrentProcess()
1276 - Rename Task.setDataSize() to Task.setBytesAmount()
1277 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1279 - #18874: Actually allows the GC to reclaim tasks
1283 - SIMIX_process_throw: raises an exception in a remote process
1284 * Refactoring: Separate sub-modules
1285 - libsmx: the public interface, as libc in a real system
1286 - popping: the strange dance that converts a user request into a kernel handling
1287 - smx_context_*: the virtualization mechanisms that embed the user code
1288 - smx_*: the handling of each simcalls
1289 * Interface cleanup:
1290 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1291 * Rename smx_action_t into smx_synchro_t, making explicit that these
1292 things are used to synchronize processes with their environment.
1293 For example, a communication is a sort of synchronization involving
1294 the communicating processes (that may block until the exchange) and
1295 the platform. The same can be said from computations, etc.
1297 - #18888: segfault when a process ends before its kill_time
1301 - 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.
1302 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1303 - MPI_Type_set_name, MPI_Type_get_name
1304 - MPI_*_c2f and MPI_*_f2c functions
1305 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1306 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1307 - Activate a lot of new tests from the mpich 3 testsuite
1309 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1310 - InfiniBand network model added: Based on the works of Jerome Vienne
1311 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1312 - When smpi/display_timing is set, also display global simulation time and application times
1313 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1314 * Collective communications
1315 - 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
1316 - 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/).
1317 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1319 - Replay now uses algorithms from wanted collective selector
1320 - Replay can be used with SMP-aware algorithms
1321 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1322 - 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.
1324 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1325 - Correctly use loopback on fat-tree clusters
1326 - Asynchronous small messages shouldn't trigger deadlocks anymore
1327 * Energy/DVFS cleanup and improvement
1328 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1329 - new: smpi_get_host_pstate()
1332 - "Full" network optimization flag was broken since Surf++
1333 - Better handling of precision flags in maxmin
1334 - Fix bug causing sometimes "Impossible" errors
1335 - Properly pass cluster properties to included hosts
1336 * Improvement of the Energy plugin.
1337 - Always update the consumption before returning that value
1338 - New property: watt_off to denote the dissipation when the host is off
1339 - New functions getWattMinAt and getWattMaxAt to retrieve the
1340 dissipation of pstates that we are not currently at.
1341 * Java: class NetworkLink renamed to Link
1342 * New function: simcall_process_get_kill_time()
1343 * Massive rename s/workstation/host/
1344 - That's intrusive, but that's good for the project consistency. Sorry.
1345 - Change config option "workstation/model" into "host/model"
1348 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1349 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1350 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1351 - xbt_dirname and xbt_basename for non-POSIX systems
1353 * The model checker now runs as a separate process.
1354 * The model checker runs must now be launched with the new simgrid-mc program.
1355 * Record/Replay: the MC can display a textual representation of a path in the
1356 execution graph. It can then be replayed outside of the model checker.
1358 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1360 SimGrid (3.11) stable; urgency=low
1365 * Normalizing pointers addresses tool for better diff between logs
1368 * Add cloud examples using new VMs
1369 - examples/msg/cloud/two_tasks_vm.tesh
1370 - examples/msg/cloud/simple_vm.tesh
1371 - examples/java/cloud/cloud.tesh
1372 - examples/java/cloud/migration/migration.tesh
1373 * Add java surf examples:
1374 - examples/java/surfPlugin/surf_plugin.tesh
1375 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1376 - examples/java/surfCpuModel/surf_cpu_model.tesh
1377 * Add SMPI+MSG example:
1378 - examples/smpi/smpi_msg_masterslave/
1383 - msg task destroy cancel test
1384 - msg_host on/off test
1385 * Move all tests in testsuite to teshsuite (adding tesh files)
1386 * Restructure teshsuites
1387 - one folder for each kind of test
1388 * Restructure AddTests.cmake
1390 - structure the order of tests (with sections)
1393 * Add virtual machine
1394 - creation of a VM on a PM
1395 - migration of a VM from a PM to another PM
1397 - MSG_process_join(msg_process_t process, double timeout)
1398 - msg_bar_t MSG_barrier_init(unsigned int count)
1399 - int MSG_barrier_wait(msg_bar_t barrier)
1400 - void MSG_barrier_destroy(msg_bar_t barrier)
1401 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1402 * New option "msg/debug_multiple_use" to help debugging when a task is used
1405 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1407 - Increase code coverage in test suites
1411 * Protect context stacks against stack overflow. The number of protected memory
1412 pages allocated on the top of each stack (1 by default) can be configured
1413 with the new command line option --cfg=contexts/guard_size:<value>.
1414 * Simcalls are now generated by a python script that
1415 - generates files included by SimGrid
1416 - checks that all the functions exist, and proposes prototypes
1418 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1419 host_set_data, host_get_data
1421 - simcall_process_join(smx_process_t process, double timeout)
1422 * Fix bug where sleeping processing could not be suspended.
1425 * Translate surf models from C to C++
1426 - Generic classes for all models: Model, Resource, Action
1427 - A generic interface for each kind of model (CPU, Network, Storage
1428 Workstation, WorkstationVM)
1430 * Translate surf routings from C to C++
1431 * Add callbacks using sigc++ or boost::signals2
1432 - Add callback functions for resource creation/destruction
1433 - Add callback functions for action state change
1434 - Handle Energy as a plugin
1435 * Replace swag by boost::intrusive
1436 * Add new routing models for clusters. For documentation, see
1437 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1438 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1439 parameters for cluster tag
1440 - Fat trees, with topology="FAT_TREE" and
1441 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1442 - see examples/platforms/torus_cluster.xml and
1443 examples/platforms/fat_tree_cluster.xml
1444 * More documentation
1447 * Hostfiles support host:nb_processes construct to deploy several processes on
1449 * Collective communication algorithms should not crash if used with
1450 improper number of nodes and report the error.
1451 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1452 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1453 MPI_Dims_create, MPI_Cart_sub are supported.
1454 * New interface to use SMPI programmatically (still depends on MSG for
1455 some parts, see examples/smpi/smpi_msg_masterslave):
1456 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1460 * Global variables privatization in MPI executables is now performed at runtime
1461 with the option smpi/privatize_global_variables (default:no).
1462 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1463 dynamic libraries loaded by the application are not privatized (static
1464 linking with these libraries is advised in this case)
1467 * Options defined in XML work correctly now.
1470 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1471 copied into simgrid.jar or not (ON by default). Use this option if you want
1472 to reduce the size of the installed simgrid.jar, *and* the native libraries
1473 are kept installed elsewhere.
1474 * Surf binding with SWIG (code generated in maintainer mode only):
1475 - plugin to handle callbacks
1476 - CPU model only for the moment
1479 * Supernovae build mode is definitively removed. It was used to improve
1480 inlining and inter-module optimizations. It is nowadays superseded by
1481 link-time optimizations commonly available in compilers.
1482 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1483 * Add boost dependency for surf++
1484 * Add new macro for tests
1485 - ADD_TESH(name <tesh_args>)
1486 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1490 - xbt_bar_t XBT_barrier_init(unsigned int count)
1491 - int XBT_barrier_wait(xbt_bar_t barrier)
1492 - void XBT_barrier_destroy(xbt_bar_t barrier)
1493 * Make the xbt_os_time module public
1495 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1497 SimGrid (3.10) stable; urgency=low
1499 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1502 * Reintegrate Java to the main archive as desynchronizing these
1503 package is not acceptable anymore (Java is now considered stable)
1504 * Add explicit synchronization facilities through semaphores
1505 * Bug fix: Task.setDataSize() only changed the C world, not the value
1506 cached in the Java world
1509 * Dramatically change the way files are handled. API and internals changed, but
1510 this part of MSG was not considered as production grade either.
1511 * Add explicit synchronization facilities through semaphores
1512 * Add a new function MSG_host_get_process_list()
1513 * Preliminary DVFS support (see examples/msg/energy/ for details)
1516 * SMPI is now included directly in the libsimgrid as the windows
1517 linker doesn't force us on splitting it anymore.
1518 * Improvements of the SMPI replay tool:
1519 - Most of the collective communications are now rooted in the same process as
1520 in the original application.
1521 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1522 used until now). Multiple data types can now be used in a trace.
1523 - The replay tool now supports traces produce either by TAU or a modified
1525 - Bug Fix: the compute part of the reduce action is now taken into account.
1526 - Gatherv collective is now supported
1527 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1528 of smpirun outputs time independent traces for the current run. One file
1529 is created per process. If too many processes are simulated, this behavior
1530 can be changed to one file for all processes by using the
1531 tracing/smpi/format/ti_one_file flag
1532 * smpirun generates the host file if needed (with given host count and platform)
1533 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1534 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1535 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1536 --cfg=smpi/coll_selector:(mpich/ompi)
1537 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1539 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1540 happening inside a collective SMPI call.
1541 * Fix the behavior of complex data types handling.
1542 * Make MPI_Wtime another synchronization point to take computations into
1544 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1545 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1546 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1547 (no privatization of global variables yet)
1548 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1551 * Verification of liveness properties is now available for SMPI applications
1552 (in addition to MSG applications)
1553 * Bugged examples using SMPI in examples/smpi/mc/
1554 * Add --cfg=model-check/visited option. Allows the verification of infinite
1555 programs. Detection of loops in the execution thanks to the system state
1556 comparison and reduction of the state space to explore. Can be combined with
1557 DPOR for safety properties.
1560 * Allow to change SimGrid configuration (see --help) within the code
1561 thanks to SD_config() as it can be done in MSG.
1562 * Add a new function SD_task_set_amount() upon user request.
1565 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1566 * Remove rule based routing (no more PCRE dependency)
1567 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1568 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1569 * Add a loopback_bw and loopback_lat options to cluster tag.
1570 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1571 routing. Now peers are encapsulated in an AS and have their own private
1572 router but this is transparent.
1575 * Our own implementation of getline is renamed xbt_getline, and gets
1576 used even if the OS provide a getline(). This should reduce the
1577 configuration complexity by using the same code on all platforms.
1578 * New type: xbt_cfg_elm_boolean.
1579 * Allow to use yes/no for boolean configuration options in the command line.
1580 * Allow to disable SimGrid cleanups at exit from command line option.
1581 There are situations where one may want a simulation to end with an exit.
1582 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1583 annoying when scripting around the simulator. Adding a
1584 --cfg=clean_atexit:no allows to circumvent this issue.
1587 * Lots of memory leaks were corrected in this release.
1588 There is no known memory leaks anymore, in all of our 600+ tests.
1589 * New command line option --version, to get SimGrid version information.
1590 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1592 * Supernovae builds are deprecated, and expected to be removed in the next
1595 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1597 SimGrid (3.9) stable; urgency=low
1599 The Grasgory release: GRAS is really dead now.
1601 * Complete overhaul of the internal host structures scheme.
1604 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1605 was considered as experimental and badly maintained since then.
1606 * Keeping it was thus a trap to our potential users, that could take
1607 it instead of MSG or SMPI by mistake despite is pity state.
1608 * GRAS seems to have very few users (if any), and no one volunteered
1609 to maintain it further. It also induces a lot of XBT code (for
1610 portability sake), that must be maintained too.
1611 * For all these reasons, we killed GRAS. If someone wants to revive it
1612 in the future, don't cry, our git history still remembers of GRAS.
1615 * Major overhaul. Merge our documentation again as time proved that
1616 splitting it was really not helping our users.
1617 * Further improve the developer documentation to help newcomers
1618 hacking on SimGrid itself. The user documentation (and in
1619 particular, the beginner documentation) is still in a sorry state.
1622 * Now works on Windows too!
1623 * Much more extensive test suite, from MPICH
1626 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1627 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1628 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1629 * Bug fix: let task be scheduled when the last dependency to be solved is
1630 a control dependency.
1631 * Remove SD_load_environment_script function.
1632 Use the C sg_platf function if you want to declare a platform
1636 * New function: MSG_process_get_number()
1637 * Old function documented: MSG_config()
1638 * Remove MSG_load_platform_script function
1639 Use the C sg_platf function if you want to declare a platform
1643 * Change the default value of the TCP_gamma constant (maximal size of TCP
1644 congestion window) to a more realistic 4MiB value. If you notice changes in
1645 your simulation results, you can fall back to the previous 20k tiny window
1646 by adding --cfg=network/TCP_gamma:20000 on command line.
1647 * (Hopefully) fix a bug wrt periodic availability and state traces
1648 * Bug fix: use default values at start when first event in availability/state
1649 trace is not at time 0.
1652 * remove the "new_" part of function name sg_platf_new_trace_connect
1653 (resulting in sg_platf_trace_connect), since it does not create
1657 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1658 use case where it's really mandatory, and maintaining it was a pain in
1660 * New: xbt_fifo_search(), search an item with a user-provided
1661 comparison function instead of dumb pointer comparison.
1664 * Fix the lua deployment:
1665 Use `simgrid.init_application()` before deployment instead of
1666 `simgrid.msg_register_application()` after.
1669 * Transfer the tracing files into the corresponding modules.
1671 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1673 SimGrid (3.8.1) stable; urgency=low
1675 The "we are told that some people want to also *install* the simgrid
1678 * Add missing file "tesh.1" to the archive.
1680 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1682 SimGrid (3.8) stable; urgency=low
1684 The Psssshiiiit release: SimGrid jumps into the Cloud.
1687 * Add an experimental interface to manipulate VMs. They are mainly
1688 process groups with very few intrinsic semantic, but they should
1689 allow you to build the semantic you want easily.
1690 * New function: MSG_host_set_property_value()
1691 * New function: MSG_process_on_exit(). To clean memory in all cases.
1692 * Bug fixes that made the host (and link) failures unusable.
1693 * Add a way to auto-restart process when the host in which they are
1694 executing comes back (ON_FAILURE="RESTART" on deployment file,
1695 MSG_process_auto_restart_set).
1696 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1697 please stop using the old ones, they are DEPRECATED.
1699 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1700 Please use MSG_init() instead. (reducing the amount of entry
1701 points in the library helps us).
1702 * Make it impossible to link against the wrong version of the lib
1703 * Deprecate MSG_clean(). No need to call it anymore.
1704 * Function MSG_get_host_number() is not deprecated anymore.
1707 * Split the doc into a user guide and a reference guide.
1708 * Start a developper guide to help people hacking on SimGrid.
1711 * Enable tracing by default. This modules rocks you should use it.
1712 * Remove option custom_flags. Now use environment variables CFLAGS
1714 * Use default cmake things to detect lua instead of home grown ones.
1715 * New option "enable_mallocators" to disable mallocators, for debugging
1716 purpose ("on" by default).
1719 * Bug fixes around the resource failures: don't let the processes
1720 survive the host they are running onto.
1721 * Add an interface to auto-restart processes when the host in which they are
1722 executing comes back.
1723 * Ensures that SIMIX_clean is called automatically. It's not part of
1724 the public interface anymore (bindings should be updated).
1727 * Bug fix for when SD_Simulate is called with a positive value: be careful
1728 when comparing doubles. Sometimes they are different for non significant
1730 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1731 parallel task whose initial work is distributed among host according
1732 to the Amdahl's law. Such tasks are created with a parameter alpha
1733 that corresponds to the non-parallelizable part of the computation.
1734 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1735 between two sets of workstations assuming a 1D block distribution (each
1736 workstation owns a similar share of data) on both sides.
1738 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1739 Data redistribution will be automatically scheduled once parent and child
1740 are both scheduled. The filling of computation_amount and
1741 communication_amount structures is now done seamlessly thanks to the chosen
1743 * New function SD_workstation_dump to display various information
1744 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1745 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1746 bandwidth on the route onto which the task is scheduled and the amount of
1748 To divide the nominal bandwidth by 2, the rate then has to be:
1749 rate = bandwidth/(2*amount)
1750 * Compute tasks that have failed can now be rescheduled and executed again
1751 (from their beginning)
1752 * Increasing source code coverage (src/simdag is now covered at 95.8%
1756 * Re-implement time-independent trace replay using SMPI (at the
1757 smpi_smp_* level) instead of MSG. This should replace
1758 examples/msg/actions/actions.c
1759 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1760 hindexed and structs)
1761 * Implement the exchange of non-contiguous data.
1762 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1763 * Correct behavior of smpi/sender_gap and set its default value to 0
1764 * Add option to asynchronously send small messages to allow better
1765 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1766 specifies the size in bytes under which messages will be asynchronously sent.
1767 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1768 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1771 * New C interface to define a platform: XML is now optional.
1772 For more info, please check include/simgrid/platf.h.
1773 * New interface to define random platforms from the C:
1774 For more info, please check include/simgrid/platf_generator.h and
1775 examples/msg/masterslave/masterslave_platfgen.c
1776 * Export a sg_cmdline dynar containing all the arguments we got from
1780 * Two new tracing options for adding comments to trace file so you
1781 can track your experiments (see --help-tracing for details).
1782 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1783 * Adding the SimGrid version that generated the trace file as a comment.
1784 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1785 * Fix to avoid key clashes on Paje links
1786 * Other minor fixes related to the Paje specification
1789 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1790 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1791 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1792 * Greatly improve the robustness of mmalloc to user errors (such as
1793 using an area after freeing it, or freeing it twice)
1795 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1797 SimGrid-java (3.8.1) stable; urgency=low
1799 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1800 Warning: all methods are visible, but only some of them are
1801 implemented so far. Check the source (src/jtrace.c)
1802 for further information.
1803 * New module: org.simgrid.msg.File (SimGrid File management functions)
1804 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1805 * Change the meaning of Process.restart: now restart the process from
1806 the beginning, like MSG_process_restart in C.
1807 * Add Process.setAutoRestart: handling of process restart when failed
1809 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1810 you to retrieve the properties of the processes/hosts
1811 * Deprecate Msg.clean(): you can just forget about it now.
1812 * New function Process.getCount(), that only works when compiling
1813 with the not yet released version 3.9 of the C library.
1815 * New context factory based on Coroutines. It mandates a modified JVM
1816 but then, the simulations run about five times faster, and there is
1817 no limit to the amount of processes (beside of the available memory).
1819 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1821 SimGrid (3.7.1) stable; urgency=low
1824 * Restore the prototype of MSG_process_create_with_environment() to
1825 the pre-3.7 situation by removing the kill_time argument.
1826 * Add a MSG_process_set_kill_time() function instead.
1829 * Fix weird behaviors when dealing with parallel tasks.
1832 * Simgrid is now built as a dll.
1833 * Simgrid-java now works on Windows.
1834 * Simgrid-Java is now included into Windows package.
1837 * First pre-build package for MacOSX.
1840 * Fix compilation when using MSG_USE_DEPRECATED.
1841 * Fix some compilation issues on Macs and Windows.
1842 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1843 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1845 We discovered that the Lua console is broken, but we are missing the
1846 manpower to fix it right now. The problem existed in 3.7 too, so we
1847 are not blocking the release for that. Sorry if you depended on this
1848 feature, any help would be really welcome.
1850 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1853 SimGrid-java (3.7.1) stable; urgency=low
1855 The "Java aint got to be bloated and slow" release
1858 * Various internal cleanups and performance improvement
1859 Simulations are expected to run up to twice faster or so
1860 * Make Process.kill(process) an instance method, not a static one
1861 * User processes are not java.lang.Thread subclasses.
1862 This breaks the compatibility (sorry), but previous API was
1863 brain-dead, making it impossible to have non-trivial
1864 initializations in the process constructor.
1865 * Require a full constructor per Process sub-class.
1866 Kinda breaks the compatibility (sorry), but this allows a much more
1867 efficient way to launch the processes at simulation startup.
1868 * Do not embeed our version of semaphores, java 1.5 can be considered
1869 as sufficiently prevalent for us to not dupplicate its features.
1873 * Add examples for almost every part of the API
1874 We spotted and fixed a lot of bugs in the process
1876 * New module: asynchronous communication API
1877 * New function: Process.sleep()
1878 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1879 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1880 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1882 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1884 SimGrid (3.7) stable; urgency=low
1886 The "spring cleanups (before the next Big Project kicks in)" release.
1889 * We can specify the SMPI latency/bandwidth factor with command line
1890 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1891 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1892 You can also use the "config tag" from platform file by adding this line
1893 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1894 (see "example/platforms/tag_config.xml" to use "config tag").
1895 Note that the command line supersedes the platform file configuration.
1896 * Change the correction factors used in LMM model, according to
1897 the latest experiments described in INRIA RR-7821.
1898 Accuracy should be improved this way.
1899 * Use the partial invalidation optimization by default for the
1900 network too. Should produce the exact same results, only faster.
1901 * Major cleanup in surf to merge models and split some optimization
1902 mechanisms from the core of the models. As a result you can now
1903 specify which model to use (e.g., --cfg=network/model:LV08
1904 --cfg=cpu/model:Cas01) and which optimization mode to use
1905 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1906 Incompatible combinations should err at initialization. See
1907 --help-models for the list of all models and optimization modes.
1908 * The CLM03 workstation model was dropped for simplicity because it
1909 used the deprecated CM02 network model. Use default instead.
1910 * Rename the TCP_gamma configuration option to network/TCP_gamma
1911 * Rename the coordinates configuration option to
1912 network/coordinates, and document it
1913 * Use now crosstraffic keyword instead of the terribly misleading
1914 fullduplex keyword. It is activated by default now in the current
1915 default model, use --cfg=network/crosstraffic:0 to turn it off.
1916 * Ongoing refactoring the model parsing to make XML files optional
1917 See include/simgrid/platf.h for details (still to be completed)
1920 * Major overhaul of the documentation. Almost instructive now :/
1921 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1922 functions and friends. This interface was considered as
1923 deprecated since over 2 years, it's time to inform our users that it is.
1924 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1925 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1926 use these (crappy) functions in your code.
1927 These functions will be removed soon. Stop using them now.
1928 * Deprecate MSG_get_host_{table,number}
1929 Implement MSG_hosts_as_dynar() instead.
1930 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1931 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1934 * Stabilize the parallel execution mode of user contexts
1935 * Introduce configuration variables to control parallel execution:
1936 - contexts/synchro: Synchronization mode to use when running
1937 contexts in parallel (either futex, posix or busy_wait)
1938 - contexts/parallel_threshold: Minimal number of user contexts
1939 that must be part of a scheduling round to switch to parallel
1940 execution mode (raw contexts only)
1941 * Fix bugs that prevented to use suspend/resume along with
1942 synchronization structures.
1943 * Fix bugs in process termination that lead to invalid memory access
1944 in very specific conditions.
1947 * Introduce a parallel mode for the models (controlled by surf/nthreads
1948 configuration item). In our tests, running the models in parallel
1949 never lead to any speedups because they are so fast that the gain
1950 of computing each model in parallel does not amortizes the
1951 synchronization costs, even when ultra fast futexes are used.
1952 This is released anyway because YMMV.
1955 * Performance boost by using a swag internally to compute the set of
1956 tasks that are finished and should constitute the return value of
1960 * Enable it by default now that it is considered rather stable.
1963 * Documentation of the tracing functions.
1964 * Performance gains when tracing categorized/uncategorized resource
1965 utilization by avoiding calls to get route when updating resource
1966 variables. LMM constraints are being used instead.
1967 * API changed to set task categories. Use MSG_task_set_category instead
1968 of TRACE_msg_set_task_category, and SD_task_set_category instead
1969 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1970 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1971 * Examples for link user variables added (see at examples/msg/tracing/)
1972 * Deprecated function TRACE_msg_set_process_category completely removed
1973 * Trace header updated according to the latest Paje file format
1974 * Tracing network lazy updates, no longer obligate users to use full updates
1975 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1976 * Experimental: let user code declare/set/push/pop application states for hosts
1977 * API changed to allow the manual creation of graph configuration files
1978 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1981 * Improve the API of Lua MSG bindings, using the Lua spirit.
1982 * Each simulated process now lives in its own Lua world (globals are
1983 automatically duplicated). It helps writing simulators. Will allow
1984 to run Splay programs within SimGrid in the future.
1985 * Add a Chord example in Lua, equivalent to the MSG one.
1988 * Start the implementation of a solution to express temporal
1989 properties, not only local assertions. This is still an
1990 experimental work in progress, stay clear from it to be safe.
1994 - Add new runtime parameters --help-logs and --help-log-categories
1995 to display information about supported logging parameters and
1997 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1998 don't exists anymore.
1999 * Mallocators: allow value NULL for the reset function.
2001 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2002 create homogeneous dictionaries, where all the elements share the
2003 same free function. Non homogeneous dictionaries will be
2004 deprecated in the next release.
2005 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2006 - Multi-level dictionaries are deprecated.
2008 - new function xbt_dynar_search_or_negative() that is useful when
2009 you have less than 2 million elements in your dynar and don't
2010 want of the extra complexity of catching exceptions when the
2011 element is not found.
2013 - Make xbt_os_thread module (for thread portability) public.
2014 Documentation is still to come, sorry.
2016 - Cleanups and simplifications to make it maintainable again.
2017 - Exotic features (such as memalign and valloc) were removed.
2018 - The metadata were extended and improved so that the
2019 model-checker becomes able to explore and inspect the heaps.
2020 - This may induce a performance drop when enable_model-checking is
2021 ON in cmake (even if it's not used in the simulation), but it is
2022 necessary at this point to get MC working.
2024 Turn model-checking OFF if simulation performance matters to you.
2025 Not enabling it at runtime is not enough, disable it in cmake.
2027 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2030 SimGrid (3.6.2) stable; urgency=low
2032 The "Not coding new stuff allows to polish old things" release.
2035 * New bindings to the NS3 packet level simulator (experimental)
2036 * Use the raw (efficient) execution contexts instead of the sysv
2037 (portable) ones when possible.
2038 * libpcre is now mandatory in any cases since not using it led to
2039 severe performance loss and possibly other issues
2040 * Update the XML platforms:
2041 - G5K: include the latest machine in Nancy
2042 - GridPP and LCG: new platforms
2043 * Documentation was partially updated, at least (more to come)
2045 Bug fixes, cosmetics and small improvements
2046 * Free terminated processes before the end of the simulation to avoid
2047 exhausting the memory of users having very dynamic amount of
2049 * Bug fix and cosmetics about canceling non-running tasks
2050 * Bug fix about the dot loader's issues when using libcgraph
2053 * Create an installer for windows with nsis (amd64 and win32)
2054 - Add an hello world project to illustrate simgrid project creation.
2055 - Embed libpcre into the Simgrid installer to avoid
2056 its compilation burden
2057 * The raw execution contexts should work on Apple now
2058 * Port to Windows 64 bits
2059 - Sysv contexts now have an implementation for this arch
2060 - GRAS communication features now support this arch
2061 * Drop support for borland compiler on windows
2062 - this code was not maintained, and we kinda depend on gcc nowadays
2063 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2064 * Fix portability issue on unstable ubuntu: linker became picky on
2067 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2069 SimGrid (3.6.1) stable; urgency=low
2071 The "Oops, we broke Macs too" release
2074 * Fixed contexts detection so that raw ones are used when possible
2075 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2076 leads to a strange error, with user code segfaulting sometimes when
2077 the generated code is not perfectly aligned (which is not
2078 controllable from the user side, depends on the amount of code)
2081 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2083 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2085 SimGrid-java (3.6) unstable; urgency=low
2088 * Split of every thing from simgrid v3.5 into a separate package.
2090 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2092 SimGrid (3.6) stable; urgency=medium
2094 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2097 * Bindings now constitute their own package, separated from the main one.
2098 Rationale: reduce our maintenance nightmare by reducing the module coupling
2099 They will soon be released on their own on gforge.
2101 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2102 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2104 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2105 * It's not quite deprecated for now because we have no replacement,
2106 but it may soon become the case.
2109 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2110 * Fortran: New user-level cache variable to store the rank of the running
2111 process. This improves performance by an order of magnitude.
2112 * C: New coccinelle script to automatically locate and modify global and
2113 local static variables.
2114 * Improved SMPI network model with a sender-side gap to account for multiple
2118 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2119 only say if a communication is finished, no matter whether it succeeded or
2120 failed. You can call MSG_comm_get_status() to know the status of a finished
2122 * New function MSG_task_dsend() to send a task and detach it. When a
2123 communication is detached, you are never notified of its success or failure
2124 and the memory is released automatically once it is finished. This function
2125 is useful when you don't care about the end nor the success of a
2127 * Change the prototypes of action replay. Sorry for inconvenience,
2128 but this is really more efficient this way (and to adapt your code,
2129 you just have to fix the initialization, that shouldn't be too long)
2130 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2131 ever managed to do anything useful with it.
2132 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2133 takes a dynar of communications. It returns immediately and gives the
2134 index of a finished communication (if any).
2135 * New example: a basic implementation of the Chord P2P algorithm.
2138 * New model for multi-core CPUs. You can now use the core attribute to
2139 precise the number of cores of a host. This is a basic model. Every
2140 process running on the host receives at most the power provided in
2141 the DTD (throughput<=power). Total throughput of process cannot exceed
2143 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2144 router linked by an up-link and a down-link (possibly asymmetrical).
2145 This kind of pattern allows to easily build last-mile model style platforms.
2146 Aggregating such patterns in a rule-based AS is thus the technique of
2147 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2148 * New model for Vivaldi routing. We transformed the Vivaldi network model
2149 into a Vivaldi routing model (based on the rule-based model). This allows to
2150 combine Vivaldi based latencies with last-mile platforms.
2153 * Added a check for NaN of IEEE754 infinite in the double entries of
2155 * Introduce a new context factory "raw", highly inspirated from the
2156 ucontext factory, but using manually crafted functions in assembly to
2157 do the work in an efficient manner.
2158 * Allow to change the used context factory at run time, not only at
2159 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2160 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2161 contexts at runtime (only with raw contexts or ucontexts).
2162 * Completely rewrote this module to allow parallel execution of user
2163 processes. Use --cfg=contexts/nthreads:N to execute user processes
2164 with N parallel threads (the default is 1, meaning no parallelism).
2165 * Allow to decide dynamically between sequential and parallel modes.
2166 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2167 processes in parallel only when their number is greater than or equal to P
2169 * Added a check for NaN of IEEE754 infinite in the double entries of
2173 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2174 won't output the state of processes when interrupted with Ctrl-C
2175 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2176 NULL-terminated array. This may solve backward compatibility issues
2177 due to the change to return type of SD_simulate. See also:
2178 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2179 * Add new macros with variable number of arguments.
2180 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2181 - in xbt/asserts.h: xbt_assert
2182 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2183 - in xbt/ex.h: THROWF and RETHROWF.
2184 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2186 * Change xbt_die() to accept a format string with arguments, just like printf.
2187 * New data structure: xbt_lib_t, like a dict but more general and with better
2191 * New configuration options
2192 Options triva/categorized and triva/uncategorized can be used to generate
2193 graph configuration files for Triva visualization tool.
2194 * Configuration option tracing/platform is renamed to tracing/categorized
2195 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2196 * New TRACE_user_link_variable interface:
2197 User provides the name of the link and the tracing variable to attach to it
2198 * the declaration of tracing categories must be done after the environment
2200 * simpler tracing interface, just one way to declare categories
2201 TRACE_category or TRACE_category_with_color, it is up to you
2202 * links in the trace file are again identified by their names
2203 * trace contains the full platform hierarchy exactly as declared using the ASes
2204 * Options tracing/msg/[task|process]:1 groups the process by hosts
2205 for both cases, tasks and processes must have names that are unique during the simulation
2206 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2207 * The experimental option tracing/msg/volume is deprecated
2208 its functionality may be reincorporated if needed
2210 The tracing generates a trace file with unordered timestamped events,
2211 because of the way the core simulator (surf) works. A script available
2212 at the tools directory (fix-paje-trace.sh) can be used to put the events
2213 in order. We have changed the tracing so it can generate ordered timestamped
2214 events in the final trace, but depending on the simulator (and how much time
2215 is simulated) that can lead to a huge memory utilization. It is deactivated
2216 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2218 Build Infrastructure
2219 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2220 - We are trying hard to keep the API stable, but it may happen that
2221 some things change (we're a research project after all, not a
2222 nuclear plant operating system). If such things should happen, you
2223 could rely on that macro to adapt.
2224 - current value: 30600 for 3.06.00, aka 3.6
2225 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2226 a number that can be compared with SIMGRID_VERSION.
2227 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2228 assertions and verbose/debug logging events are disabled at compile time.
2230 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2232 SimGrid (3.5) stable; urgency=medium
2235 * New feature to any SimGrid-based simulator: Model-Checking
2236 Check SIN#1 for more details.
2239 * New Model SMPI (three-interval linear regression for correction factors)
2240 See RR-7426, available at http://hal.inria.fr/inria-00527150
2241 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2242 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2243 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2244 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2245 * See also src/smpi/README
2250 - Tracing API changes: TRACE_start and TRACE_end should not be called
2251 by user-code. They are automatically called by simulators created
2252 with SimDAG, MSG and SMPI if the toolkit is compiled with
2253 tracing_enabled=ON. Categories declaration and utilization remain the
2254 same for MSG and SimDag.
2255 - A function was added to the tracing API to declare categories with
2257 - TRACE_category_with_color (char *category, char *color)
2258 where color must be in the following format
2259 "%f %f %f", red, green, blue
2260 and red, green, blue are float values in the interval [0, 1]
2261 - User can specify NULL as color parameter, or continue calling
2262 TRACE_category (cat)
2263 On that case, the tracing system will define random colors
2264 - The following command-line options are supported:
2265 --cfg=tracing/filename:msg.trace
2266 --cfg=tracing:1 (activate tracing, needed to use others)
2267 --cfg=tracing/platform:1 (categorized resource use)
2268 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2269 --cfg=tracing/msg/task:1 (task creation)
2270 --cfg=tracing/msg/process:1 (process creation, migration)
2271 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2272 --cfg=tracing/smpi:1 (SMPI interface tracing)
2273 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2274 - examples of examples/msg/tracing updated
2276 - DAXLoader and DOTLoader functions can generate tasks with categories
2277 - A new function to attribute a category to SD tasks:
2278 TRACE_sd_set_task_category (SD_task_t task, char *category)
2279 * Tracing the MPI interface implemented by SMPI
2280 - Collective operations are traced with states
2281 - Point-to-Point operations are traced with states/links
2282 - Tracing activated by a parameter "-trace filename" passed
2283 to smpirun during execution (considering that simgrid
2284 is compiled with tracing enabled)
2285 - To run the simulation with gdb, the simulator
2286 accepts --cfg=tracing/smpi:1 to trace SMPI
2287 - tesh files to check if smpi tracing is ok
2288 - See examples/smpi/NAS/DT-trace
2289 * GTNetS tracing re-worked
2290 - adaptation to the tracing system of GTNets to cope
2291 with modifications regarding the fullduplex mode
2292 - new tesh files to check if gtnets tracing is ok
2295 * Asynchronous communications through the functions:
2296 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2297 * New function: MSG_load_platform_script()
2298 to make possible using a lua script instead of XML files to set up platforms
2299 * New function: MSG_set_function
2300 to associate functions to processes, used when bypassing the parser
2301 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2303 Platforms: Add some more examples in examples/platforms
2304 * Grid'5000: see www.grid5000.fr
2305 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2308 * Change the XML format. This is a very important modification. SimGrid
2309 3.5 introduces a new hierarchical format based on the notion of
2310 Autonomous Systems. Compatibility with old format is ensured
2311 through the perl script provided in the install bin directory
2312 bin/simgrid_update_xml.
2313 It is now possible to build platforms with specific routing
2314 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2315 connect several platforms together. We will try to provide soon
2316 set of realistic platforms exploiting these properties (have a
2317 look at examples/platforms/ for the moment).
2318 * Take the opportunity of the XML format change to be a good XML citizen:
2319 rename link:ctn to link_ctn and similar changes (also dealt with by
2321 * Add a new routing scheme (rule-based) using regular expressions. It
2322 enables to have an extremely low memory footprint when the
2323 underlying routing is simple and can be compactly described. You
2324 need to have libpcre4-dev (perl regular expressions) installed if
2325 you want to use this routing scheme.
2326 * Revive the cluster TAG and allow to easily and efficiently (both in
2327 term of memory and speed) connect clusters together. Have a look
2328 at teshsuite/simdag/platforms/ to see how this can be done. With
2329 this tag, you can create clusters with thousands of tasks at no
2330 cost (have a look at examples/platforms/).
2331 Note: clusters are implemented as ASes, so there is no need for an
2332 enclosing AS tag if you have only one cluster in your platform.
2333 * Add new generic functions in the public interface that allows the user
2334 to call SURF 'create_resource' methods from your code (same
2335 functionality as the XML bypass mechanism but with a much lighter
2337 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2338 that uses a piecewise linear approximation to produce better
2339 results when exchanging small messages.
2340 * Add a new parameter to handle correctly full duplex link and account
2341 for interferences between uplink and downlink communications
2342 (activate with --cfg=fullduplex:1).
2345 * Rename the SD_READY (all dependencies are satisfied and task is
2346 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2347 dependencies are satisfied) state.
2348 This prevents a confusion between the notion of "ready to schedule"
2349 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2350 simulated" (SD_RUNNABLE) used by the simulation kernel.
2351 * Change the way a task is considered as ready. Instead of removing
2352 dependencies when a task is done, a counter is decreased. This way,
2353 it is always possible to reach ancestors thanks to the
2354 SD_taks_get_parents function (even after the end of the simulation.)
2355 * Change the return type of SD_Simulate from (SD_task_t*) into
2356 xbt_dynar_t. This function was in handling a dynar internally and
2357 converted it into a NULL terminated array for historical reasons.
2358 * New function SD_dotload(char*) to load a DAG described in dot
2359 format. This loader and the corresponding examples require the
2360 installation of the graphviz library.
2361 * Fix a bug in the management of tasks of size 0 in the surf network
2362 models. This problem was only visible with SIMDAG and you should
2363 thus disregard results produced with earlier versions if you
2364 relied on this feature (some tasks were blocked because of this).
2365 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2366 even though your applications were doing only point-to-point
2367 communications and sequential computations. Now you can really use any
2368 model you want (of course, if you create real parallel tasks, which are
2369 not implemented in most models beside ptaskL07, this will abort).
2370 * Add an example that schedules a DAX on an heterogeneous platform
2371 using a Min-Min strategy.
2372 * New function SD_workstation_get_current_task() that returns the kind
2373 of task currently running on a workstation in the sequential access
2375 * Raise some warnings when unexecuted tasks remains at the end of the
2376 simulation. This is usually caused by cycles in the DAG.
2379 * New function: SIMIX_process_set_function() called by MSG_set_function
2380 * Change the underlying waiting queue in semaphores so that a process
2381 can wait on several of them simultaneously (as in waitany).
2382 * Fix the way to handle tokens in semaphores so that all access patterns
2383 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2384 * kill the dirty pimple SIMIX_message_sizes_output()
2385 Please use (proper) visualization instead
2388 * New data container: setset (set of sets of elements)
2389 * New module: mmalloc (mapped malloc, allowing to have several
2390 independent segments of malloc)
2391 * New function: xbt_dict_cursor_set_data()
2392 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2393 * New function: xbt_dynar_is_empty()
2394 * New function: xbt_fifo_get_last_item()
2395 * Fix xbt_dynar_shrink(): use the right element size.
2396 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2397 uninitialized areas during expand.
2398 * Fix semaphores: previous implementation was severely broken.
2399 * Use library init/fini functions for our initialization.
2400 - you can use logs and other feature as soon as you want in your
2401 code (even before the xbt_init / MSG_init)
2402 - xbt_exit is now a no-op and produce a warning when used.
2405 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2406 benefit from the latest improvement to the simulation kernel.
2407 * Kill measurement sockets for now. If you rely on them, sorry. This
2408 release is not for you. This feature will be reintroduced in the
2409 future, but we cannot delay the release any further.
2410 * New function: gras_msgtype_get_name().
2411 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2412 * Fix (at last) the pmm example: it should not randomly fail anymore.
2414 Build chain: bug fixes and overall polishing
2415 * Cmake is now stable enough. Hence, we killed the autotools.
2416 * Port to windows ( TM :)
2417 * Fix the 'make install' target.
2418 No need to use 'make install-simgrid' anymore
2419 * Introduce a 'make dist' target compiling a *source* archive
2420 'make package' compiles a binary archive
2421 * Compile java files only on need
2422 * Add --cd and --setenv command line options to tesh
2423 * Out of source builds are not fully supported yet, but we are close
2424 * Enable supernovae and optimization flags by default for our users
2427 * Add layer to set up environment directly from lua, without XML.
2428 * The effect of gras_stub_generator can be achieved through
2429 lua too (check examples/gras/console/ping_generator.lua)
2431 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2433 SimGrid (3.4.1) stable; urgency=low
2435 The "Polishing easter eggs is probably a good idea" release.
2436 This is a bug fixes release only.
2440 * Fix a bug preventing the tasks from begin garbage collected.
2443 * Fix a bug occurring when a host involved in a communication fails.
2444 This was not detected properly by the other peer involved in the
2445 communication. Now, it's reported as a network error.
2448 * Warn the user about loop dependencies in data flow of DAX files
2449 * Obey the control-flow dependencies of DAX files
2452 * Add option "enable_smpi" allowing to not compile SMPI.
2453 Probably useful for the (Mac) users experiencing a build error here
2454 * Improve the detection of lua5.1 and ruby1.8
2456 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2458 SimGrid (3.4) stable; urgency=low
2460 The "Easter in Cargese" release. Also known as (major changes):
2462 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2463 ~> bindings were greatly improved
2464 ~> new tracing infrastructure for better visualization introduced
2466 * the "Welcome to configury modernity" release.
2467 ~> we switched from autotools to cmake, and improved our cdash
2470 A more detailled list of changes follow (full detail in svn log).
2472 Java Bindings: Various Cleanups
2473 * (install java-gcj-compat-dev on debian-like to use them)
2474 * Remove put/get: no need to export deprecated interface in Java
2475 Use send/receive instead.
2476 * Cleanup the examples and add a README per directory
2477 * Remove example autoDestination (that's the only way to go now)
2478 * Remove example explicitDestination (was a plain copy of basic)
2479 * Make JniException a runtime exception, so that there is no need to
2480 declare the fact that you may encounter such a beast. I guess that
2481 nobody will ever want to survive such error.
2482 * Create specific errors for each MSG case of failure:
2483 host failure, transfer failure, timeout, task cancelled
2484 * Cleanup the exceptions that may get thrown by each function
2485 * Other internal cleanups in Java bindings. Performance still bad :/
2486 Ruby and Lua Bindings: create them
2487 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2488 * That's new and great, you should try them out.
2489 Same functionalities than Java bindings, only even less polished
2491 * Kill the useless "rate" argument of SD_task_get_execution_time()
2492 Everyone used to provide -1 as a value, it was not used, and the
2493 semantic of a possible use wasn't even clear.
2494 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2495 either as comm costs or compute costs to mean that there is no
2496 such thing for that specific task.
2497 * Add a SD_task_set_name() function
2498 * Fix SD_task_unschedule() on typed tasks
2499 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2500 * In DAX loader, accept useless 'level' attributes to <job> since
2501 LIGO DAGs have them (seem to be to ease graphical representation).
2503 * Add an example masterslave_mailbox.c using send/receive and not
2504 the deprecated put/get interface.
2505 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2506 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2508 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2509 (declare MSG_USE_DEPRECATED to still have the old name)
2510 * Add a MSG_task_set_data() function
2511 * About trace replay (see examples/msg/actions):
2513 - Allow to work with splitted trace files for each process
2514 Give the specific trace file as argument of each process,
2515 and call MSG_action_trace_run(NULL)
2516 You can still have one merged file for all processes.
2517 - Fix implementation of collective operations
2518 * Allow task_execute() on 0-sized tasks (closes #10063)
2520 * This is the first release of SimGrid where SMPI is not considered
2521 beta anymore (even if some corners should still be improved)
2522 * Port over the new SIMIX_network submodule (internal refactoring)
2523 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2524 * Implement more missing elements of the standard:
2526 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2528 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2529 MPI_Scan() MPI_Get_processor_name()
2530 - Added implementation of missing case for Alltoall (warning: it's
2531 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2532 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2533 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2534 SMPI_MPI_Allgatherv()
2535 * Bug fixes include:
2536 - MPI_Waitsome() was broken
2537 - Allow relative includes in smpicc
2538 - Command line cfg argument 'reference_speed' was ignored...
2539 - Some functions did not properly lead to auto-benching of user code
2540 - smpicc passes -O2 by default (just like openmpi one)
2542 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2543 * Bug fixes about timeouts during communications
2544 * add SIMIX_message_sizes_output() as a pimple to write to file the
2545 amount of messages per size. Use gnuplot to get histogram.
2546 Pimple because that's the only user-visible function of simix,
2547 defined directly in xbt.h (irk, sorry)
2549 - Add a SIMIX_sem_get_capacity() function
2550 - Fix interactions with processe resume/suspende
2551 - release_forever() was stupidly broken
2552 - Fix SIMIX_display_process_status() for processes in a semaphore
2553 - Make SIMIX_sem_block_onto() user-visible
2554 * Refactoring context stuff:
2555 - Use pseudo-OOP for better modularity
2556 - reimplement SIMIX_process_kill() without process_schedule() so
2557 that the latter can take as invariant that it is called from
2559 - Merge context_start into context_new for sake of simplicity
2561 * Add a Vivaldi network model, coded live during SUD'10 ;)
2562 * Rename configuration variables to start a hierarchy:
2563 o cpu_model -> cpu/model
2564 o network_model -> network/model
2565 o workstation_model -> workstation/model
2566 * New configuration variables:
2567 o network/bandwidth_factor: correction to bandwidth
2568 o network/latency_factor: correction to latency
2569 o netwotk/weight_S: correction to the weight of competing streams
2570 * Add a long description to the models, that users can see with such
2571 argument on the command line: --cfg=cpu/model:help
2572 * --help-models display the long description of all known models
2574 * config: add the ability to set a default value after registration
2575 Does not override any previously set value (e.g. from cmd line)
2576 * dict: allow to have integer key and data.
2577 When so, you need to use the following functions
2578 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2579 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2580 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2581 In contrary to regular dicts, the key is not malloced before copy.
2582 Mixing scalar and regular elements in the same dict is not tested
2584 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2585 Tracing for Visualization:
2586 * SimGrid is now instrumented in order to generate a trace file for
2587 visualization analysis: to use it, need to compile SimGrid with the
2588 "tracing" option enabled, and instrument the program using SimGrid with
2589 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2590 (among other functions).
2591 * The instrumentation only traces the platform utilization for now
2592 * Documentation to use the tracing functions and how to analyze the
2593 traces with the Triva tool is written.
2594 * More information about: SimGrid FAQ (in the section Tracing Simulations
2597 * We moved to cmake as default build system. Autotools support will
2598 be dropped soon. Check the FAQ for more info about how to use it.
2599 * Greatly improved our cdash/ctest interactions
2600 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2601 * Added memory checking tests with valgrind; lot of memleak fixing.
2602 This may be the first release of simgrid with so few memory issues
2603 * Added code coverage tests.
2604 Our coverage is still improvable, but at least we see it on cdash.
2606 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2608 SimGrid (3.3.4) stable; urgency=low
2610 The "Desktop Grid needs love too" release (also called Xmas release).
2612 Models improvements:
2613 * Major speedup in the maxmin system solving by using lazy evaluation
2614 Instead of solving completely the maxmin system at each iteration,
2615 only invalidate (and recompute) the modified parts.
2616 This new feature is enabled in default models but you can try to
2617 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2618 * Cas01 IMproved as default CPU model
2619 This CPU model is the same Cas01 model, but it uses the
2620 maxmin-selective-update flag and a heap structure to manage
2621 actions on SURF kernel.
2622 It reduces the complexity to find the next action to finish and,
2623 consequently, it's faster than the old Cas01.
2624 This is the new default CPU model (Cas01).
2625 * Rename the old Cas01 model to Cas01_fullupdate
2626 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2627 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2628 * CpuTI (CPU Trace Integration)
2629 A new CPU model whose objective is simulate faster when using
2630 availability trace files.
2631 Instead of using a full featured, over engineered maxmin system for
2632 CPU modeling, this model does the pre-integration of traces files
2633 to calculate the amount of CPU power available, and so, executes
2634 faster than the old CPU models.
2635 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2636 * Use LV08 as default network model since it gives better accuracy
2637 for small messages and shouldn't change things for big ones.
2638 Use --cfg=network_model:CM02 to get the previous behavior.
2641 ******************************************
2642 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2643 ******************************************
2644 * The new CPU model may changes simulations!
2645 The point is that events occurring at the exact same timestamp
2646 are not scheduled in the same order with the old and new
2647 version. This may be enough to completely change the execution
2648 of simulations in some cases.
2649 * The new network model will change simulations!
2650 This new model is more realistic than the previous one, so you
2651 should consider redoing your old experiments with this model.
2652 Sorry for the inconvenience.
2655 * Introduce the supernovae compilation mode
2656 When compiled that way, the whole SimGrid (or almost) is put in a
2657 single compilation unit and compiled in one shoot.
2658 This is to help gcc which has difficulties to inline stuff from one
2660 The speedup seem to be above 15%, althrough more tests are needed on
2661 amd64 to confirm that gain.
2664 * Port of MSG's mailbox on top of SIMIX network
2665 The put/get mechanism was greatly simplified on the way.
2668 * New SIMIX network module. Provides:
2669 - Mailbox: rendez-vous mechanism to find with who you want to speak
2670 - Synchronous send/recv: easier and hopefully faster since the
2671 logic is handled in the maestro process directly now
2672 - Asynchronous send/recv: you dreamt of it? It's here now
2673 Too bad that nobody cared enough to propagate the change to MSG.
2674 * Add semaphores as SIMIX synchronization mechanism.
2677 * new function SD_daxload(char*) to load a DAX file
2678 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2679 * Introduce typed tasks. Specify its kind and cost at creation.
2680 At scheduling, just give where it should be placed, and the cost
2681 for each involved resource is automatically computed.
2682 Existing constructors so far (more to come of course):
2683 - SD_task_create_comm_e2e() for end-to-end communication
2684 - SD_task_create_comp_seq() for sequential computation
2685 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2686 * new function SD_task_dump() for debugging display
2687 * new function SD_task_dotty(task,FILE*) writing to file the info
2688 about the task in dotty format
2689 * SD_task_dependency_exists() can now cope with having one of its
2690 arguments NULL. If so, it tests whether the other argument has any
2692 * Add getters on list of preceding/following tasks:
2693 SD_task_get_parents(task) and SD_task_get_children(task)
2694 * Add getters on amount of workstations and list:
2695 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2696 * Add getter on task kind: SD_task_get_kind(task)
2697 * Update the start_time and finish_time of tasks on completion/failure
2698 * Bugfix: Remove task from state swags when destroyed
2701 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2704 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2705 dynar with one separate thread per value of the dynar.
2706 * Change the prototype of xbt_thread_create(), sorry.
2707 Added a boolean parameter indicating whether we want to join this
2708 thread (used in SG only for now)
2709 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2712 * GTNetS wrappers should now be usable again (and betterly tested too)
2713 * Fix a major regression from 3.2 where the timeout provided to
2714 MSG_task_put_with_timeout() was used as absolute time before which
2715 the comm should be done.
2716 * Start to fix the <cluster> tag.
2717 - Internal links should be good now (beside of the loopback, which
2718 use the private link instead)
2719 - paths to the external world is still rather broken
2720 - the <route:multi> tag is just broken. Actually that's brain-dead.
2721 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2722 to make it less stupid
2723 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2724 * Fix a source-level compatibility glitch from 3.2: after defining
2725 MSG_USE_DEPRECATED, you can use the old name
2726 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2727 * Allow to compile from the SVN with automake 1.11
2728 * Fix some problems when using the "start_time" tag in deployment XMLs.
2729 * Fix #8569: XBT/synchro.h has redundant declarations
2730 * Fix #8563: MSG return values and exceptions
2731 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2732 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2733 Thanks to Robson Peixoto.
2734 * Fix a vicious bug in dictionaries inducing that some elements were
2735 not freed on xbt_dict_free()
2737 Portability report of this version:
2738 * Main portability targets:
2739 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2740 on (amd64/i386/ia64)
2741 - mac leopard on i386
2742 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2743 but nothing critical.
2744 * Other platforms: windows, AIX and others were not tested for this release
2746 Timing report of this version:
2747 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2748 scenario parameters). From 8h to a few seconds in desktop grid settings.
2749 * Supernovae brings about 25% speedup on i386.
2751 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2753 SimGrid (3.3.3) stable; urgency=low
2755 The "Need for Speed" release.
2757 The timings done to validate the 3.3.2 were faulty.
2758 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2760 The problem was a conversion from a manually handled vector to
2761 xbt_dynar_t on the critical path.
2762 xbt_dynar_foreach calls functions, inducing stack management crap.
2764 We inlined these functions and xbt_dynar_foreach is now breath taking.
2765 We also inlined xbt_swag_belong on the way.
2767 Here are some approximate speedup measurements (on master/slaves
2768 simulations lasting between 10s and 20s each):
2769 3.3.1 -> 3.3.2: about same performance
2770 3.3.2 -> 3.3.3: 40% speedup
2771 3.3.1 -> 3.3.3: 40% speedup
2772 3.3.1 with inline patch -> 3.3.3: 30% speedup
2774 Our reading is that the refactoring which occurred in 3.3.2 made us
2775 suffer much more from the xbt_dynar_foreach low performance, but
2776 once we solved this, this refactoring proved to be very performance
2777 effective. From the 40% speedup, somehow, 10% are due to the
2778 inlining and 30% to the refactoring.
2780 That's a pitty that gcc cannot inline functions placed in other files
2781 alone. We have to choose between:
2782 - break the encapsulation (by putting private data structures and
2783 accessors in headers files to help gcc)
2784 - live with low performance
2785 - switch to a decent compiler such as icc (not quite possible).
2787 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2789 SimGrid (3.3.2) stable; urgency=low
2791 The "Simplicity does not preceed complexity, but follows it" release.
2793 The main contributors of this release were (lexical order):
2794 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2797 * Extract the routing logic into its own object.
2798 (was dupplicated in network.c and workstation_LV07.c;
2799 Allows to implement other ways of storing that info)
2800 => kill now useless network_card concept
2801 - Use dynar to represent routes (instead of void** + int*)
2802 - kill link_set (use surf_network_model->resource_set instead)
2803 - Add a command-line option to choose the routing schema to use
2804 - Add three new models:
2805 * Floyd (shortest path computed at initialization)
2806 * Dijikstra (shortest path recomputed all the time)
2807 * Cached Dijikstra (shortest path computed on need)
2808 All these models where contributed by Silas De Munck, and are
2809 described in his ICCS09 paper.
2811 * Simplify model declaration
2812 (less redirections, less function to write when defining a model)
2813 - Factorize stuff between models:
2816 surf_model_resource_set(model)
2817 surf_model_resource_by_name(model, name)
2818 - Unify the types of models in s_surf_model_t (using an union)
2819 - Embeed fields of common_public directly into s_surf_model_t
2820 - Rename model methods:
2821 action_free ~> action_unref
2822 action_change_state ~> action_state_set
2823 action_get_state ~> action_state_get
2824 - Change model methods into functions:
2825 (model)->common_public->action_use ~> surf_action_ref
2827 * Implement a generic resource; use it as ancestor to specific ones
2828 (allows to kill duplicated code in models)
2829 Drawback: timer command don't need no name nor properties;
2830 workstation_CLM03 don't need no properties
2831 (but I guess we can live with those few bytes wasted)
2833 * Improve the action object model
2834 - implement a constructor avoiding dupplicated code about field
2835 initialization in generic_action part.
2837 * Kill the SDP model: it has an external dependency, is deprecated
2838 in flavor of modern lmm models, and didn't compile since a while
2841 * Relocation of the context module from XBT to SIMIX.
2842 (the context were decoupled from the simix processes, duplicating a lot of code)
2843 => a lot of code was factorized
2844 - less overhead is introduced during scheduling
2845 - simpler API for the context factory
2846 - the logic for process creation,destruction and manipulation was simplified
2847 * Simplification of the s_smx_process_t data structure.
2848 => accesing the simix level data associated to a process is faster now,
2849 and the code is a lot more readable.
2852 * Implement some more MPI primitives:
2853 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2854 -implementation: Bcast: flat or 2-ary tree (default),
2855 Barrier: 4-ary tree,
2857 Allreduce: Reduce then Bcast
2858 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2859 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2860 Alltoallv: flat tree, like ompi
2862 * Add support for optimized collectives (Bcast is now binomial by default)
2863 * Port smpirun and smpicc to OS X
2866 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2867 Shout out if you used it.
2870 * Display the list of still queued messages in SG mode when existing
2874 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2875 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2876 * Add xbt_str_from_file(FILE*)
2877 * Add xbt_dict_get_key achieving a linear reverse search
2878 * Remove the context module
2880 Portability report of this version:
2881 * Main portability targets:
2882 - Linux(debian)/x86/context
2883 - Linux(debian)/x86/pthreads
2884 - Linux(debian)/amd64/context
2885 - Linux(debian)/amd64/pthreads
2886 On these, we still have the eratic breakages of gras/pmm and
2887 amok/saturate_sg reported in previous version. We still think
2888 that the tests are the cause of the fault, not the tested code.
2890 - Mac OSX Leopard/x86/context
2891 Still false negative in tesh autotesting.
2892 Smpi still fails, but this time because readlink does not accept -f
2893 Everything seems to work properly beside of that.
2896 - AIX version 5.3 (only tested contexts this time)
2897 Smpi still fails there because mktemp is not installed.
2898 Everything seems to work properly beside of that.
2900 I managed to compile it for the first time, but several breakages.
2901 Won't delay the release for this exotic platform.
2903 * Windows: it's still lagging behind. If you want to help, please
2906 Timing report of this version:
2907 This version seem to be more than 5% faster than 3.3.1 (on linux
2908 64bits with contextes). The gain is less than expected, we are
2909 investigating this for next release.
2911 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2913 SimGrid (3.3.1) stable; urgency=low
2916 * Implement a --cfg-help to show existing configuration variables
2917 * Build chain do not require doxygen in maintainer mode
2920 * fix a bug on struct sizeof computation, which prevented the
2921 exchange of arrays of structs in some conditions
2922 - added a regression test about this in datadesc_usage
2923 * Allow the exchange of 0-long dynamic vectors.
2924 - for that, use -1 as indicator of dynamic size instead of 0
2925 - This implied to change any size from unsigned long to long,
2926 reducing a bit communication abilities, but I guess that with
2927 64bits being quite common, this is more than enough.
2928 - This also induce a protocol change, thus bumping network protocol
2929 version from 0 to 1 (if we have external users, we have to get
2930 clean on that point too ;)
2931 - added two regression tests about this in datadesc_usage
2932 * Be more verbose when propagating local exceptions
2933 This helps debugging.
2934 * Display the status of simulated processes when receiving SIGINT in
2938 * Allow to control the simulation from a trace file.
2939 New functions MSG_action_register() and MSG_action_trace_run()
2940 The first one allows to associate a function execution to each
2941 kind of action while the second one parses a trace file and
2942 triggers the corresponding actions within the system.
2943 For now, only a toy example is provided in examples/msg/actions
2944 * Add an example of process migration in examples/msg/migration
2945 * Fix a bug in task exchange which broke MSG_task_get_sender()
2946 Add a teshsuite regression test for that.
2947 [Bug: if MSG_task_get_sender() is called after sender exit,
2949 * Fix a bug which prevented suspend/resume to work properly
2950 * Display the status of simulated processes when receiving SIGINT
2951 This fixes a regression of v3.3. due to the introduction of SIMIX
2952 * Bug fixing in failure management:
2953 - trace could not start by a failure at time 0
2954 - failure during communications were not working
2957 * Add SIMIX_process_set_name() to change the name of the current
2958 process in the log messages.
2959 * Store smx_hosts in a dict since we only retrieve them by name
2960 * Move the configuration infrastructure to surf
2963 * Move the configuration infrastructure to surf
2966 * Massive internal cleanups:
2967 - Store internal structures on processes instead of hosts (allows
2968 to have more than one process per host, in addition of being more
2970 - Cleanup the initialization/finalization process
2971 - Kill a whole bunch of unneeded synchronization:
2972 processes run in exclusive manner within the simulator
2973 - Move queues from global tables to process data fields
2975 - now accept -platform and -hostfile arguments
2976 - Pass the right rank value to processes according to the hostfile
2977 * Compile the examples by default, and use them as regression tests
2978 * Implement MPI_Wtime()
2979 * Change the reference speed to a command line option
2982 * TCP_gamma can now be specified as command line option using
2983 --cfg=TCP_gamma:10000000.0
2984 * Change the --surf-path cmd line option into --cfg=path:
2987 * Also include strbuff from xbt.h public header
2988 * xbt_ex_display(): do not free the exception after displaying
2989 This allows to do more with the given exception afterward.
2990 Users should call xbt_ex_free() themselves.
2994 Portability report of this version:
2995 * Main portability targets:
2996 - Linux(debian)/x86/context
2997 - Linux(debian)/x86/pthreads
2998 - Linux(debian)/amd64/context
2999 - Linux(debian)/amd64/pthreads
3000 These targets fail about 1/10 of times on gras/pmm, but we believe
3001 that this is because of the test, not because of simgrid.
3002 amok/saturate_sg fails even more rarely, and the test may not be
3005 - Mac OSX Leopard/x86/context
3006 The test suite still spits tons of errors because some obscure
3007 force prevents us from removing the temporary directories
3008 arguing that they still contain some metadata I've never heard of.
3009 Smpi fails because seq is not installed.
3010 Everything seems to work properly beside of that.
3013 - AIX version 5.3 (both contexts and pthread)
3014 Smpi still fails there because mktemp is not installed.
3015 XML inclusions seems rosty on AIX.
3017 * Windows: it's still lagging behind. If you want to help, please
3020 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3022 SimGrid (3.3) stable; urgency=high
3026 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3027 [Malek Cherier & Mt]
3029 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3032 Until now, GRAS were using MSG as an interface to SURF. It was
3033 quite difficult because both interface have several differences
3034 (MSG channels vs GRAS sockets were the most notable point).
3036 This also opens the gate to SMPI (which should occur soon) and speed
3037 up simulations by to 40% (even if it were not the main goal).
3039 **************************************
3040 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3041 **************************************
3042 The point is that events occurring at the exact same timestamp are
3043 not scheduled in the same order with the old and new version. This
3044 may be enough to completely change the execution of simulations in
3045 some cases. Sorry for the inconvenience.
3047 * Cleanup and upgrade the XML format to push further scalability
3048 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3050 * Improve the testing infrastructure with tesh. Now a very large part of
3051 the code is tested not only by being run but also by checking that the
3052 output match an expected output [Mt].
3054 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3055 is really less memory-demanding, which should allow you to use
3056 larger files in SimGrid [AL].
3058 * Inform valgrind about our contextes, so that it becomes usable
3059 with the default (and more effecient) version of SimGrid
3060 [contributed by Sékou Diakite, many thanks]
3063 * Introduce a listener thread in charge of receiving incoming
3064 messages from the network. It allows to overlap communication and
3065 computation but most notably, it removes some stupid deadlocks due
3066 to the fact that so far, a process could not send and receive at
3067 the same time. This made most non trivial communication schema
3069 * Convert the PIDs from long int to int to match the MSG ones (and
3070 linux ones too) [Mt]
3071 * New function: gras_agent_spawn() to launch a new process on
3072 current host. Only working in simulation for now. [Mt]
3073 * New function: gras_os_hostport() returning a constant form (ie,
3074 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3077 * Make the backtrace of exceptions more human readable [Mt]
3078 * New module: xbt/str [Mt]
3079 a ton of string utility functions (split, join, printf to a newly
3080 allocated buffer, trim, etc)
3081 * New module: xbt/hash [Mt]
3082 SHA1 hashing algorithm (more to come if needed)
3083 * New module: xbt/synchro [Mt]
3084 synchronization tools (mutex and conditions) working the same way
3085 in simulation and in real life (mainly useful for GRAS, but not
3087 * New module: xbt/queue [Mt]
3088 classical producer/consumer synchronization scheme
3089 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3090 (using the classical functions will get serialized) [Mt]
3091 * Make dictionary internal table dynamic. No need to specify its size
3092 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3094 * Make sure the log channels are organized as a tree under windows
3095 (because of ANSI C compatibility issue, any channel were child of
3096 root directly) [Mt].
3099 * Cleaned many thing in surf and fixed a few bugs [AL].
3100 * Add a nice command line configuration mechanism to compose models [AL].
3101 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3102 the previous one (KCCFLN05). It relies on something that looks like
3103 a max-min sharing mechanism but cannot be written as such. A new solver
3104 was thus designed [AL].
3105 * Add a new solver to lmm. Based on Lagrange optimization and
3106 gradient-based descent, it enables to efficiently maximise systems s.a
3108 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3110 This solver enables to propose two new network models for TCP Reno and
3111 TCP Vegas based on Low's work. These models still need to be fully
3112 tested though [Pedro Velho].
3115 * Bug fix in SD_simulate. Now the time bound given as argument is
3117 * Use the new parallel task model (ptask_L07) as default.
3118 * Use the SURF command line configuration mechanism.
3119 * 0-size tasks (for synchronization) should now work.
3121 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3123 SimGrid (3.2) stable; urgency=high
3127 We still experience issues on this platform, but we believe that at
3128 least MSG is usable.
3130 GRAS API BREAKAGE (for simplification purpose, sorry):
3131 * the gras_msgtype_by_name is not used anymore. Instead of
3132 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3133 you can write (and must)
3134 gras_msg_send(toserver, "request", &request);
3135 - If you still want to pass a gras_msgtype_t to the function (to cache
3136 the type and avoid the lookup time), use the gras_msg_send_() variant.
3137 - Impacted functions:
3138 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3139 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3140 * The callbacks are now expected to return 0 when everything went well
3141 (just like the main() function)
3143 GRAS new features and improvements:
3144 * New module mechanism where user code can use per process globals [Mt]
3145 This is similar to gras_userdata_*() functions, but for libraries. It
3146 factorize some code developped over and over in the examples and AMOK.
3147 It has still to be documented and used (only amok/peermanagement is
3149 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3150 loss when they were small enough to fit into the buffer and sent quickly
3151 enough so that they can all get received in one shoot.
3152 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3153 if not found. Use the *_or_null() variant for the old semantic.
3154 * In gras_msg_handle, do not discard messages without callback.
3155 They are probably messages to be explicitly awaited later (ie, proofs of
3156 mis-synchronization in userland since they are sent before being awaited)
3157 No big deal usually.
3158 * gras_socket_meas_send/recv: semantic changed!
3159 The numerical arguments used to be (1) the total amount of data to send
3160 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3161 messages. This was need for the fool willing to send more than MAXINT
3162 bytes on quite fat pipes.
3165 * Do really rename the hostmanagement module to peermanagement. [Mt]
3166 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3167 but this is rather new and this was documented in the module
3168 documentation (poor excuses, I admit)
3169 * Bandwidth measurement semantic changed! This follows the changes to
3170 gras_socket_meas_send/recv explained above.
3173 * A sequential mode has been added to the workstations. When a workstation
3174 is in sequential mode, it can execute only one task, and the other tasks
3175 are waiting in a FIFO. [Christophe Thiery]
3178 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3179 model for SIMDAG. [Christophe Thiery]
3180 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3181 equal to 0, which caused some bad behaviors in
3182 saturated_constraint_set_update. I now use a threshold mechanism like in
3186 * When running manually src/testall, you select specific units [Mt]
3187 testall is the result of our cunit mechanism, and should replace all
3188 the scripty thingy around since bash don't run easily on billware.
3190 * A mallocator system has been added. [Christophe Thiery]
3191 Mallocators allow you to recycle your unused objects instead of freeing them
3192 and allocating new ones.
3194 Documentation update:
3195 * FAQ reworking + New FAQs:
3196 - "Valgrind spits tons of errors!" [Mt]
3197 - "How to repport bugs" [Mt]
3198 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3199 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3200 - Communication time measurement within MSG [AL]
3201 - I experience weird communication times when I change the latency [AL]
3202 * GRAS tutorial [Mt]
3204 - an introduction to the framework and to the used communication model
3205 - an initiatic tour introducing the most proheminent features:
3207 . Lesson 0: Installing GRAS
3208 . Lesson 1: Setting up your own project
3209 o Part 2: Message passing
3210 . Lesson 2: Exchanging simple messages
3211 . Lesson 3: Passing arguments to the processes (in SG)
3212 . Lesson 4: Attaching callbacks to messages
3213 . Lesson 5: Using globals in processes
3214 . Lesson 6: Logging information properly
3215 . Lesson 7: Using internal timers
3216 . Lesson 8: Handling errors through exceptions
3217 . Lesson 9: Exchanging simple data
3218 . Lesson 10: Remote Procedure Calling (RPC)
3219 . Lesson 11: Explicitly waiting for messages
3220 . Recapping of message passing features in GRAS
3221 - A HOWTO section containing:
3222 o HOWTO design a GRAS application
3223 More are due, of course. They will come latter. In the meanwhile, you can
3224 check the examples which are still here.
3226 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3228 SimGrid (3.1) stable; urgency=high
3232 There was a stack corruption somewhere, visible only when optimizing
3233 with these versions. [Vince]
3236 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3237 module that enabled to play with Directed Acyclic Graphs. It is built
3238 directly on top of SURF and provides an API rather close to the old
3239 SG. Some old codes using SG are currently under rewrite to check that
3240 all needful functions are provided. [Christophe Thiery]
3243 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3244 extension of the classical CLM03 model that gracefully handles
3245 failures. This is now the default model for MSG and GRAS. It doesn't
3246 handle parallel tasks yet though. [AL]
3247 * Bug fix: Weights were not correctly set in the network part.
3248 WARNING: This may have resulted in incorrect results with simulations
3249 where there are more than one flow on a given link. [AL]
3252 * After a (long ?) discussion on simgrid-devel, we have decided that the
3253 convention we had on units was stupid. That is why it has been decided
3254 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3255 WARNING: This means that all previous platform files will not work as
3256 such with this version! A warning is issued to ask users to update
3258 A conversion script can be found in the contrib module of the CVS, under
3259 the name contrib/platform_generation/surfxml_update.pl [MQ]
3262 * Bug fix: Processes were started in reverse order, wrt deployment file.
3263 WARNING: if your code relies on this bug, please fix it. [AL]
3264 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3265 being executed on two different locations. [AL]
3266 * Bug fix: Failures are now better supported thanks to Derrick's tests
3267 (there was many failure situations I hadn't thought of and that weren't
3268 correctly handled). [AL]
3269 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3273 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3274 Exception occurring on server-side are propagated back to client (!).
3276 API CHANGE: the callback changed their prototype. Change:
3277 int my_handler(gras_socket_t expeditor, void *payload_data) {
3279 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3280 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3282 * New! function: gras_msg_handleall to deal with all messages arriving
3283 within a given period.
3284 * New! function: gras_socket_server_range to get a server socket in a
3285 range of port numbers (ease to avoid port number conflicts) [MQ]
3286 * New! gras processes display their backtrace when they get a SIGUSR1
3287 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3288 Sweet to debug RL processes [MQ]
3292 - Do not force experiment sizes to be expressed in kb, or it becomes
3293 impossible to measure the latency this way (needs one byte-long tests)
3294 WARNING: this changes the amok_bw_* function semantic. [MQ]
3295 - Implements the link saturation stuff. [MQ]
3296 * Peer management module:
3297 New! module factorizing code that we wrote over and over [MQ].
3300 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3301 - Test units are placed directly into the library code, they get extracted
3302 automatically and placed into the src/testall binary.
3303 - Convert most of the XBT tests to this system.
3304 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3305 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3306 live in the same C code. This required to change a little bit the API
3307 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3308 * New module: sparse graph structure with basic algorithms (this is work
3309 in progress and the API is not considered to be frozen yet). [AL]
3310 * Display more information on backtraces: source line & function names are
3311 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3312 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3313 * Reworked a little bit some #include statements to load only required
3314 headers. Some user code that relied on SimGrid to include stdlib or
3315 stdio may need to include it by themselves. [AL]
3316 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3318 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3319 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3320 but I guess that almost nobody used those parts. [MQ]
3322 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3324 SimGrid (3.0.1) stable; urgency=low
3327 * Unfortunately, I had missed 5 misnamed functions:
3328 xbt_fifo_item_t xbt_fifo_newitem(void);
3329 void xbt_fifo_freeitem(xbt_fifo_item_t);
3330 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3331 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3332 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3333 They're now deprecated. Please use their new versions:
3334 xbt_fifo_item_t xbt_fifo_new_item(void);
3335 void xbt_fifo_free_item(xbt_fifo_item_t);
3336 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3337 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3338 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3340 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3341 * Documentation: xbt_log module unmercifully reworked [MQ]
3342 * Bugfix: there was a problem with the ending of contexts with
3343 the pthread backend. It caused some weird deadlock or behavior
3344 depending on the pthread implementation. [AL]
3345 * Bugfix: get the exceptions raised in the simulator repport where
3346 and why they come from when they are not caught in time [AL, MQ]
3349 * Bugfix: Do repport the error when two non-connected hosts try to
3350 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3353 * Add additionnal checkings on communications. Assert that two
3354 communicating hosts are connected by a set of links... [AL]
3357 * Add additionnal checkings on channel values in communication [AL]
3358 * New: MSG_task_get_source to see on which host a task was generated [HC]
3359 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3360 the number of tasks waiting to be received on channel and sent
3362 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3363 waits for the first task coming from a given host.. [AL]
3365 GRAS new functionnalities: [MQ]
3366 * Enhance the parsing macro to allow the size of multidimentional objects
3367 to be given thru annotations.
3368 * New example (and documentation): Matrix Multiplication a la RPC
3369 (as when I was young!) and fix a bunch of bugs found on the way.
3371 GRAS performance improvements: [MQ]
3373 * Reduce the amount of cbps creation/destruction by making it static to
3374 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3377 * Change libdata to a set so that we can search for stuff by ID (and thus
3378 reduce the insane amount of dict lookups)
3381 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3382 RL communication on top of SG since it's so uneffective.
3383 It may also be used for inter-thread communication in RL, one day.
3384 * Use gras_datadesc_copy() to exchange messages on top of SG
3386 - improve message exchange performance on top of SG
3387 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3388 * Don't exchange on the network the size of the used part of buffer,
3389 instead, specify the possible buffer size to read().
3391 - reduces the amount of read/write calls (one pair per exchange)
3392 - reduces the amount of exchanged data (the size)
3393 - allows to retrieve all arrived data on receiver side, if we don't need
3394 it right now (subsequent read will peek the buffer)
3395 - allows the receiver to proceed with the beginning of the stream before
3396 everything is arrived
3397 - make it possible to build an iov transport (using readv/writev)
3399 - take care of the data with non-stable storage (like stacked data),
3401 * If possible, TCP send uses vector I/O (when writev() is here)
3402 - Don't use it for receive since we send data sizes and data on the
3403 same stream, so we wouldn't be able to chain large amount of chunks
3404 before having to flush the stuff to read the size.
3405 * Rework the transport plugin mechanism to simplify it and reduce the
3406 amount of pointer dereferencement when searching for the right function
3409 * I guess that now, we do almost as few system calls as possible while
3410 doing as few data copy as possible.
3412 To improve it further, we could try to send all the sizes first and then
3413 all the data (to use iov on receiving size), but it's only a partial
3414 solution: when you have 2 dimensional data, the sizes of the second
3415 dimension is data of the first dimension, so you need 3 streams.
3417 I'm not sure the potential performance gains justify the coding burden.
3419 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3421 SimGrid (3.00) stable; urgency=high
3424 * New! Give the possibility to hijack the surf parser and thus bypass
3425 MSG_create_environment and MSG_launch_application. Have a look at
3426 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3428 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3430 SimGrid (2.96) unstable; urgency=low
3435 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3436 This deprecates the xbt_error_t mechanisms.
3437 It modifies (simplifies) all XBT and GRAS API.
3438 MSG API keeps unchanged (exceptions raised by XBT are caught from
3439 within MSG and masked with existing error handling facilities)
3442 * New! Add a FATPIPE model. [AL]
3443 * New! Add a parallel task model. [AL]
3444 * New! Add automatically a loopback interface (in the default
3445 network model) if none was precised.
3448 * Bugfix: MSG_process_resume now works with the current running process.
3450 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3451 * Modification of MSG_task_get_compute_duration. Once a task has been
3452 processed, the value returned by this function is now equal to 0. [AL]
3453 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3454 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3455 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3456 of a m_task. For now, MSG_task_cancel only works with computation
3458 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3459 of the processor (in Mflop/s) regardless of the current load on the
3461 * API Change: use proper naming convention for MSG_getClock and
3462 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3464 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3465 This function changes the priority of a computation task. This priority
3466 doesn't affect the transfer rate. A priority of 2 will make a task
3467 receive two times more cpu power than the other ones. This function
3468 has been added to suit the needs of Nguyen The Loc and hasn't been that
3469 much tested yet. So if it fails, please report it and send me your code.
3471 * API Change: removed all functions and types that were marked "deprecated"
3472 since many months. Renamed MSG_global_init_args to MSG_global_init.
3474 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3476 SimGrid (2.95) unstable; urgency=low
3479 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3480 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3481 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3484 * Bugfix: Allow absolute paths for platform description files [MQ]
3485 * Bugfix: do free the variables after use. Leads to drastic performance
3487 * Implement max_duration (ie, timeouts) on resources [AL]
3490 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3492 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3493 the needed support (timeouts) [AL]
3496 * Implement measurement sockets. You can now get the bandwidth between two
3497 hosts thanks to AMOK (see below). [MQ]
3498 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3499 dynar over the network (yeah) [MQ]
3500 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3501 * simple/static token-ring example. [Alexandre Colucci and MQ]
3502 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3503 in place before [MQ]
3506 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3507 type parsing. "short" and "long" modifiers were also ignored; other
3508 modifier (such as reference level) are still ignored. [MQ]
3509 * Embeed the buffer size within the buffer itself on SG. [MQ]
3510 That way, send() are atomic and cannot get intermixed anymore (at least
3511 the ones which are less than 100k; bigger messages still have the issue)
3512 * Array size pushed by the field, not by the field type (or each
3513 and every long int will push stuff to the cbps) [MQ]
3514 * use select() to sleep since it allows to portably sleep less than one
3517 GRAS (minor cleanups)
3518 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3521 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3522 * type category 'ignored' killed as it was never used and were difficult
3524 * whether a type can cycle or not is now a flag, leaving room for more
3525 flags (as "ignored", if we feel the need one day ;)
3526 * Rename raw sockets to measurement sockets since "raw" has another
3527 meaning in networking community.
3530 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3531 and offers features not belonging to GRAS but that most applications
3532 need. One day, it may be a set of plugins loadable at runtime.
3533 * New module: bandwidth
3534 bandwidth measurement between arbitrary nodes running this module. [MQ]
3536 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3538 SimGrid (2.94) unstable; urgency=low
3540 The first beta release of SimGrid 3 !
3544 * Update the main page and the FAQ. Adding references to gforge.
3547 * Add a gras_os_getpid function.
3550 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3551 * Extend the logs so that they also print PID, hostname, date, ... if
3553 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3554 and kill the old version which were in testsuite/
3555 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3559 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3560 for users. Instead, each programming environment (GRAS, MSG,...) use it
3561 when needed to provide such a feature to users.
3562 Don't shortcut the mechanism or you will also shortcut the virtualization
3563 you need on the simulator.
3567 * Cleanups in configury with regard to compile optimization/warning flags.
3568 Also add -fno-loop-optimize to any powerpc since it's the optimization
3569 killing gcc (< 3.4.0).
3570 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3571 needed by MSG examples complications
3572 * Borrow configury beautifications from PHP
3575 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3576 warning (thanks loris for stumbling into this one).
3577 * Bugfix: stop loading private headers (gras_config.h) from the public
3581 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3582 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3583 RL would imply a DNS resolver, which is *hard* to do in a portable way
3584 (and therefore delayed).
3585 * Implement a real timer mechanism and use it in timing macros. This allows
3586 to avoid rounding errors and get a 0.000005 sec precision in timing
3587 macros. While I was at it, various cleanups:
3588 - allow to declare more than one timed section per file (fix a stupid bug)
3589 - move some private declaration to the right place
3590 - merge conditional execution and timing macros into emulation module
3591 - document the module
3592 - make sure the module cleanups its mess on gras_exit
3593 * Documentation improvements:
3594 - (new) how to compile applications using GRAS
3595 - (new) emulation support (timing macros)
3597 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3599 SimGrid (2.93) unstable; urgency=low
3601 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3604 - Use Paje properly where used. Still to be sanitized properly.
3605 - Portability fix: Add an implementation of the contexts using pthread
3609 - Add xbt_procname(): returns the name of the current process.
3610 Use it to show the current process's name in all logging.
3612 - fix detection of older flex version and the reaction, since we do
3613 depend on modern ones (we use lex_destroy)
3614 - Better separation of SG and RL in the libs: remove all simulation code
3615 from libgras. As a result, this lib is now only 200k when stripped.
3616 Some of the xbt modules may also be duplicated (two sets and such) and
3617 should be cleaned/killed before SG3.
3618 - Insist on using xlC on AIX because of weird problems involving gcc there.
3619 - Cleanup the make remote stuff. This is now done by scripts
3620 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3621 mainly for our private use, but we're working on changing them to user
3624 - Bugfix: flush the socket on close only if there is some *output*.
3625 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3627 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3630 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3632 SimGrid (2.92) unstable; urgency=low
3634 Alpha 3 on the path to SimGrid 3
3638 - New! First try of benchmarking macros.
3639 - New! First try so that gras_stub_generator generate deployment and
3640 remote compilation helpers.
3642 - Bugfix: Initialization fix in msg_test.
3646 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3648 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3650 - API Change: gras_os_sleep to take the amount of seconds as a double.
3651 Accepting an int was error prone since it was the only location where
3652 seconds were coded as such. It leaded to damn rounding errors.
3653 - Bugfix: Hard to belive that timers ever worked before this.
3655 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3657 SimGrid (2.91) unstable; urgency=low
3659 Alpha 2 on the path to SimGrid 3
3663 - Bug fix in the lmm_solver.
3665 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3666 through the function MSG_paje_output.
3667 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3668 - It is possible to bound the rate of a communication in MSG with
3669 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3670 to write it in the changelog).
3671 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3676 - Various cleanups to the autotools stuff
3677 - Begin to move Gras examples to examples/gras/
3678 - Let make distcheck work again (yeah!)
3680 - documentation overhauled using doxygen.
3681 gtk-doc-tools is dead in SimGrid now.
3682 - Automatically extract all existing logging categories, and add the list
3683 to the documentation (long standing one, to say the less)
3685 - Cleanup the known architecture table. Reorder the entries to group what
3686 should be, and use a more consistent naming scheme.
3687 (some of the test dataset are still to be regenerated)
3688 - New! Allow library to register globals on each process just as userdata
3690 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3691 lookup time (for now).
3692 Use it in msg and trp.
3693 This cleans a lot the internals and helps enforcing privacy of the
3694 headers between the gras components.
3695 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3696 - Bugfix: gras_os_time was delirious in RL.
3697 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3699 - Reenable GRAS now that it works.
3701 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3703 SimGrid (2.90) unstable; urgency=low
3705 Alpha 1 on the path to SimGrid 3
3707 * It is a long time since the last release of SimGrid. I'm sorry about
3708 that but as I had told you, I was rewriting a lot of things. I apologize
3709 to those who had been reporting bugs to me and that I had not answered.
3710 If your bug is still in the new version, please tell me. Here is a
3711 summary of the main changes.
3713 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3714 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3715 portability, speed, and a lot more but you'll figure it out later.
3716 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3717 I wanted to release the new MSG as soon as possible and I have
3718 broken GRAS, which is the reason why, for now, only MSG is fully
3719 functional. A laconic description of these projects is available
3720 in the documentation.
3722 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3723 kernel optimized for our needs (called SURF but only the developers
3724 should use it). Hence, MSG is now roughly 30 times faster and I think
3725 that by rewriting a little bit MSG, I could event speed it up a little
3726 bit more. Beside the gain in speed, it is also much easier to encode a
3727 new platform model with SURF than it was with SG. More to come...
3729 * REVOLUTION 3: I have tried to change a little as possible the API of
3730 MSG but a few things really had to disappear. The main differences
3731 with the previous version are:
3732 1) no more m_links_t and the corresponding functions. Platforms are
3733 directly read from a XML description and cannot be hard-coded
3734 anymore. The same format is used for application deployment
3735 description. The new format is described in the documentation.
3736 Have a look in tools/platform_generation. There is a tiny script
3737 that converts from the old platform format to the new one. Concerning
3738 the application deployment format, parsing the old one is tricky.
3739 I think most of you should however be able to convert your files. If
3740 it is really an issue, I can write a C code that does the conversion.
3742 2) the toolbox tbx does not exist anymore. We now have a library
3743 with much more data-structures but without the hash-tables (we have
3744 dictionaries that are much faster).
3746 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3748 *****************************************************************************
3749 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3750 * don't feel like converting the oldies. (Mt) *
3751 *****************************************************************************
3754 Version 2.90: "the long awaited one"
3755 - Finished rewriting and debugging MSG. Rewrote the documentation.
3756 - disable GRAS for now since it needs to be ported to the newest SG
3759 - Finish the port to windows (using mingw32 for cross-compile)
3762 - Main loop and datastructures of SURF. A cpu resource object is
3763 functional. Surf can thus be used to create cpu's with variable
3764 performance on which you can execute some actions.
3766 2004-11-15 Martin Quinson
3767 - Port to ARM. Simply added the alignment and size descriptions. Should
3768 work, but the ARM machines are so slow that I didn't had the opportunity
3769 to 'make check' over there yet.
3771 2004-11-15 Arnaud Legrand
3772 - Trace manager now written. It uses a heap structure and is therefore
3773 expected to be efficient. It may however be speeded up (particularly
3774 when many events occur at the same date) by using red and black
3775 trees. One day maybe...
3776 - Max-min linear system solver written. It uses a sparse matrix
3777 structure taking advantage of its expected use. Most operations are
3778 O(1) and free/calloc are called as few as possible. The computation of
3779 the minimum could however be improved by using a red and black tree
3782 2004-11-03 Arnaud Legrand
3783 - Rename every gras_* function that was in xbt/ to its xbt_
3785 - Add a heap and a doubly-linked list to xbt
3786 - Added a dichotomy to the dictionaries. make check works as well before
3787 so I assume that the patch is correct. I do not know however if things
3788 run effectively faster than before now. :)
3790 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3791 SimGrid, and the version number is bumped to 2.x
3793 2004-10-29 Martin Quinson
3794 - Introduction of the remote errors.
3795 They are the result of a RMI/RPC on the remote machine.
3796 ErrCodes being scalar values, you can't get the host on which those
3797 errors did happen. Extending the error mechanism as in Gnome is possible.
3798 No idea yet whether it is a good idea.
3800 2004-10-28 Martin Quinson
3801 - Interface revolution: the Starred Structure Eradication.
3802 I used to do typedef struct {} toto_t; and then handle *toto_t.
3803 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3804 - ???_t is a valid type (builded with typedef)
3805 - s_toto_t is a structure (access to fields with .)
3806 - s_toto is a structure needing 'struct' keyword to be used
3807 - e_toto_t is an enum
3808 - toto_t is an 'object' (struct*)
3810 typedef struct s_toto {} s_toto_t, *toto_t;
3811 typedef enum {} e_toto_t;
3812 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3813 s_toto_t) is private.
3815 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3816 so that it can set the variable to NULL. It was so for dicts and sets,
3817 it changed for dynars.
3819 - Fix a bunch of memleaks in dict_remove
3820 - Fix a bug in sg/server_socket opening: it failed all the time.
3822 2004-10-07 Martin Quinson
3823 - Speed up dynar lookup operation a bit.
3825 gras_dynar_get is dead.
3827 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3828 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3829 (which gives you the address of the stored data).
3831 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3832 copy of the data using an affectation to do the job and not a memcpy.
3834 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3835 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3837 It does not really speedup the dynar test because they are
3838 setting elements all the time (and look them seldom). But the dict does
3839 far more lookup than setting.
3841 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3843 2004-10-05 Martin Quinson
3844 - Allow to (en/dis)able the cycle detection at run time.
3846 Whether we should check for cycle or not is now a property of each
3847 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3848 datadesc_cycle_unset allow to remove this property when previously set.
3850 Note that the cycle detection is off by default since it impacts the
3851 performance. Watch the data you feed GRAS with ;)
3853 This property is hereditary. Any element embedded in a structure having it
3854 set have it set for the time of this data exchange.
3856 You should set it both on sender and receiver side. If you don't set it on
3857 sender side, it will enter an endless loop. If you forget on receiver
3858 side, the cycles won't be recreated after communication.
3860 - Header reorganization.
3861 Kill gras_private.h, each submodule must load the headers it needs.
3863 2004-10-04 Martin Quinson
3864 - Interface revolution: do not try to survive to malloc failure.
3866 Now, gras_malloc and friends call gras_abort() on failure.
3867 As a conclusion, malloc_error is not a valid error anymore, and all
3868 functions for which it was the only gras_error_t return value are
3869 changed. They now return void, or there result directly.
3870 This simplify the API a lot.
3872 2004-09-29 Martin Quinson
3873 - Re-enable raw sockets.
3874 Created by gras_socket_{client,server}_ext;
3875 Used with gras_raw_{send,recv}
3878 It should allow to kill the last bits of gras first version soon.
3880 This is not completely satisfactory yet (duplicate code with
3881 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3884 - Simplify transport plugin (internal) interface by not passing any
3885 argument to _server and _client, but embedding them in the socket
3888 2004-09-28 Martin Quinson
3889 - Finish the port to AIX.
3890 autoconf was my problem (segfault within the malloc replacement
3891 function. No idea why)
3893 2004-09-16 Martin Quinson
3894 - Fix some size_t madness on 64bit architectures.
3896 2004-09-08 Martin Quinson
3897 - Reduce the number of system headers loaded, overload some more system
3898 calls (such as malloc to cast the result of the system one, and work
3900 - Fix and reintroduce the config support
3902 2004-09-07 Martin Quinson
3903 - Source code reorganization to allow Arnaud to surf all over there.
3904 - Allow to document the logging categories.
3905 - Remove all uppercase from logging categories and useless cleanup in names.
3907 2004-08-18 Martin Quinson
3908 Version 0.6.2 (protocol not changed; API changed)
3909 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3910 gras_error_t), and NULL when not found. Functions expecting a msgtype
3911 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3912 hopefully usefull message.
3913 - Portability to prehistoric sparcs again
3915 2004-08-17 Martin Quinson
3916 Version 0.6.1 (protocol not changed; ABI not changed)
3917 - prealloc some buffers to speed things up
3919 2004-08-11 Martin Quinson
3920 Version 0.6 (protocol not changed; ABI expended)
3921 - The parsing macro can deal with the references, provided that you add
3922 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3924 2004-08-09 Martin Quinson
3925 Version 0.5 (protocol not changed; ABI changed)
3926 - Allow to off turn the cycle detection code in data exchange at
3927 compilation time. It should be at run time, but I'm short of time (and
3928 the config stuff is still broken). That way, we keep dict out of the
3929 critical path, which is good because the performance is poor:
3930 - search not dichotomial yet
3931 - dynar give no way to access their content and memcpy everytime
3932 - In composed data description (struct, ref and so on), stop foolness of
3933 keeping the subtype's ID, but store the type itself. This keeps sets out
3934 of the critical path, which is good since they rely on dynar and
3935 dictionaries. The only loose of that is that we cannot detect the
3936 redeclaration of a structure/union with another content (but I'm not sure
3937 the code detected well this error before anyway). We still can detect
3938 the redefinition discrepancy for the other types.
3939 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3940 it breaks the code because of type-punning used all over the place).
3941 This breaks on all non-gcc architectures (for now).
3943 All those changes (plus the buffer of last time) allow me to gain 2 order
3944 of magnitude on cruel tests consisting of 800000 array of integers on two
3945 level of a hierarchical structure (200 secondes -> 4 secondes)
3948 - the selector of reference must now return the type it points to, not
3949 the ID of this type.
3951 2004-08-06 Martin Quinson
3952 Version 0.4 (protocol changed; ABI not changed)
3953 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3954 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3955 - Implement a buffer transport, and use it by default (it relies on tcp in
3956 real life and on sg in simulation).
3957 That's a bit hackish since I had a new field to the structure to store
3958 its data without interfering with the subtype ones. Inheritance
3959 is tricky in C. And that's a kind of reverse inheritance with one class
3960 derivating two classes. Or maybe a game with java interfaces. Anyway,
3961 that's damn hard in C (at least).
3962 Moreover, I got tired while trying to ensure plugin separation and
3963 genericity in SG mode. MSG wants me to do weird things, so let's go for
3964 cruel hacks (temporarily of course ;).
3965 See comment in transport_private.h:71
3966 - do not include all the _interface headers in private but in the files
3967 which really need them (to cut the compilation time when they are
3970 2004-07-26 Martin Quinson
3971 Version 0.3 (protocol not changed; ABI changed)
3972 - Major overhault of the datadesc interface to simplify it:
3973 - shorted the function names:
3974 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3975 - add a trivial way to push/pop integers into the cbps without malloc.
3976 This allows to make really generic sub_type description, which simply
3977 pop their size of the stack.
3978 - add a function gras_datadesc_ref_pop_arr() which does what users want
3979 most of the time: Declare a dynamic array (which pops its size of the
3980 stack) and declare a reference to it. Poor name, but anyway.
3981 - kill the post-send callback, add a post-receive one
3983 2004-07-23 Martin Quinson
3984 Version 0.2 (protocol changed; ABI changed)
3985 - add some testing for cpbs in the test cases, and fix some more bugs.
3986 This invalidate again the little64 data file, since I cannot regenerate
3988 - remove an awfull optimization in the logging stuff, allowing me to:
3989 - understand it again
3990 - learn gcc how to check that the argument match the provided format
3991 - fix all errors revealed by gcc after that
3992 - internal keys of dict are not \0 terminated. Deal with it properly in
3993 loggings instead of segfaulting when the user want to see the logs :-/
3995 2004-07-22 Martin Quinson
3996 - Fix some stupid bug preventing cbps (callback postit) from working
3998 2004-07-21 Martin Quinson
3999 - Some documentation cleanups
4000 - remove the useless last argument of msgtype_declare
4001 - rename the Virtu functions to fit into the 'os' namespace
4002 - move headers src/include -> src/include/gras/ and stop fooling with
4003 gras -> . symbolic link
4004 - make distcheck is now successful
4006 2004-07-19 Martin Quinson
4008 - Build shared library also
4009 - Install html doc to the right location
4010 - stop removing maintainer files in make clean
4011 - build tests only on make check
4013 2004-07-13 Martin Quinson
4015 - No major issue in previous version => change versioning schema
4016 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4017 corresponding dataset.
4019 2004-07-11 Martin Quinson
4021 - Get it working with any kind of structure (we can compute the padding
4022 bytes remotely for all the architectures I have access to)
4023 - Implement the structure parsing macro (still not quite robust/complete)
4024 - Improvement to the remote testing toysuite
4026 2004-07-10 Martin Quinson
4027 [autoconf mechanism]
4028 - get ride of a bunch of deprecated macros
4029 - actually run the test for two-compliment, not only compile it :-/
4030 - test whether the structures get packed (and bail out if yes. Damn.
4031 Alignment is a serious matter)
4032 - test whether the structures get compacted (but respecting the alignment
4033 constraints of each types)
4034 - test whether the array fields of structures can straddle alignment boundaries
4036 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4037 (took me 2 hours to find that bug, looking at the wrong place)
4038 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4039 before sending/receiving and are used to compute the offsets of fields
4040 - Given that padding size depend even on compiler options, keep track of
4041 alignment and aligned_size only for the current architecture. Not a big
4042 deal since we send structure fields one after the other (seems
4044 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4045 program, let it work on linux/gcc/little32. portability todo.
4047 2004-07-08 Martin Quinson
4048 - import and improve remote compilation support from FAST
4049 - make sure make check works on half a dozen of machines out there
4051 2004-07-07 Martin Quinson
4052 Let's say it's version 0.0.3 ;)
4053 - Implement conversions (yuhu!)
4054 - Let it work on solaris (beside conversion, of course)
4055 - Stupid me, using rand() to generate the conversion datatests in not wise.
4057 2004-07-06 Martin Quinson
4058 - Let make dist work, since I'm gonna need it to compile on remote hosts
4059 - Let Tests/datadesc_usage write the architecture on which the file was
4060 generated as first byte.
4061 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4062 architecture descriptions.
4063 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4064 on those architectures.
4065 - Optimization: send/recv array of scalar in one shoot
4067 2004-07-05 Martin Quinson
4068 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4070 - Plug a whole bunch of memleaks
4071 - each process now have to call gras_{init,exit}. One day, their log
4072 settings will be separated
4073 - Continue the code factorisation between SG, RL and common in Transport.
4075 2004-07-04 Martin Quinson
4077 - Redistribution between SG and RL.
4078 We wanna have to accept in SG, so move accepted related parts of RL in
4079 the common part. (more precisely, the dynar of all known sockets is no
4080 more a static in transport.c, but part of the process_data)
4082 [gras_stub_generator]
4083 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4085 2004-07-03 Martin Quinson
4086 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4087 them all up in one shot)
4089 - Fix the ugly case of reference to dynamic array.
4090 - New (semi-public) function gras_datadesc_size to allow the messaging
4091 layer to malloc the needed space for the buffer.
4093 - gras_socket_close now expect the socket to close (and not its address to
4094 put NULL in it after it). This is because the socket passed to handlers
4095 is one of their argument (=> not writable).
4097 - propagate the interface cleanup from last week in datadesc, ie remove a
4098 superfluous level of indirection. User pass adress of variable
4099 containing data (both when sending and receiving), and not of a variable
4100 being a pointer to the data. Let's say that I like it better ;)
4101 The price for that is constructs like "int msg=*(int*)payload" in
4102 handlers, but it's a fine price, IMHO.
4104 - Let it work in RL (yuhu)
4106 2004-06-21 Martin Quinson
4108 - porting SG plugin and SG select to new standards (works almost).
4109 - plug memleaks and fix bugs around.
4112 - cleanup the prototype of data recv and force users to specify when they
4113 want to handle references to objects. Test case working even for cycles.
4114 - plug memleaks. Valgrind is perfectly ok with this.
4116 2004-06-12 Martin Quinson
4118 - cleanup the separation between plugin and main code in plugin creation
4120 2004-06-11 Martin Quinson
4122 - Reput hook for raw sockets, needed for BW experiments
4123 - kill a few lines of dead code
4124 [Data description] Interface cleanup
4125 - gras_datadesc_by_name returns the searched type or NULL.
4126 That way, no variable is needed to use a type desc once, which makes
4128 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4129 two parameters were strings (field name, type name), leading to
4131 [Dicos] Interface cleanup
4132 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4133 This is consistant with the dynar API.
4135 2004-04-21 Martin Quinson
4137 - Porting to new standards.
4139 - interface cleanup.
4140 There is no bag anymore, no need to take extra provision to mask the
4141 pointers behind "ID".
4142 Better splitup of functions between files create/exchange/convert.
4143 This is still a bit artificial since convert and receive are so
4144 interleaved, but anyway.
4146 - add a queued message list to procdata (the ones not matching criteria
4148 - factorize some more code between SG and RL wrt procdata
4150 - use gras_exit in example to track memleaks
4151 - get rid of gs_example now that GS is properly integrated into gras
4152 - update run_test to integrate the latest tests (datadesc)
4154 - rename WARNINGn macros to WARNn since it prooved error-prone
4156 2004-04-19 Martin Quinson
4158 - register init/exit functions within gras module mechanism
4159 - send/receive function.
4160 Convertion is not implemented, but short-cutted if not needed.
4161 struct/array elements are sent one by one (instead of block-wise), but
4162 nobody really cares (yet). Get a prototype before optimizing.
4163 - tests (using a file socket) for DD send/receive on:
4164 - base types: int, float
4165 - array: fixed size, string (ie ref to dynamic string)
4166 - structure: homogeneous, heterogeneous
4167 - chained list, graph with cycle
4168 Believe it or not, valgrind is not too unhappy with the results. The
4169 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4173 - Bugfix when using a filename explicitly (instead of '-')
4175 2004-04-09 Martin Quinson
4177 - factorize more code between RL and SG in socket creation
4178 - Complete the implementation and tests of:
4180 o file (only in RL, and mainly for debugging)
4182 I lost 3 days to design a portable address resolver, and then decided
4183 that the prototype mainly have to run on my box.
4184 Addressing portability too early may be like optimizing too early :-/
4186 - use gras_init in the Tests instead of the crappy parse_log_opt
4187 (the latter function is removed)
4188 [Conditional execution]
4189 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4190 [Code reorganisation]
4191 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4192 Build examples against the RL library, since there is no way to disable
4193 its creation for now.
4195 For information, the beginning of coding on GRAS was back in june
4196 2003. I guess that every line has been rewritten at least twice since