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.
13 - On some network models (L07, CM02, LV08), an implicit loopback link is created
14 for local communication. Default bandwidth was 500MBps and latency was 15µs,
15 which was rather bad. Default values have been changed to 10GBps and 0 latency.
16 Options network/loopback-lat and network/loopback-bw have been added to adapt
17 these values if needed. Network and routing models already providing separate
18 loopbacks (clusters and such) are not impacted by this change.
21 - LTO is now enabled for Intel/clang compilers.
22 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
23 Setting it to "auto" will use all available cores, while setting it to n will
24 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
27 - Many MSG tests were converted to the new S4U's interface in C, that
28 was extended for that.
31 - Remove obsolete option --cfg=contexts/parallel-threshold.
33 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
34 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
35 - FG#41: Add sg_actor_create C interface
36 - FG#43: xbt::random needs some care
37 - FG#48: The Impossible Did Happen (yet again)
38 - FG!24: Documentation and fix for xbt/random
39 - FG!35: Add a modeling hint for parallel links in doc
40 - GF#18137: Allow different stack sizes?
41 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
42 - GH#139: Allow pthread creation in SMPI
43 - GH#336: Packet-level simulation using SMPI?
44 - GH!337: Fix link_energy plugin for wifi platforms
45 - GH!339: Add Mailbox set_receiver method to python binding
47 ----------------------------------------------------------------------------
49 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
51 The Palindrom Day Release.
53 Important user-visible changes:
54 - Improve the Python usability (stability and documentation).
55 - A nasty synchronization bug was ironed out, see also below.
56 - Python's doc was organized as a separate tree, now integrated with C++.
57 - C bindings of S4U were not part of the doc.
58 - The C++ doc was also improved as methods are now split by theme.
59 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
60 - This is mandatory to use the Java bindings.
61 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
62 - SimDAG++: Automatic dependencies on S4U activities (experimental)
63 - Some features are already implemented but not all of them
64 - Cannot block an activity until it's scheduled on a resource
65 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
66 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
69 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
70 - Drop support for Viva/Triva (old visualization tools), hence removed
71 TRACE_get_node_types() and TRACE_get_edge_types() functions.
74 - Require pybind11 version 2.4.
75 - Greatly improve locking during context switches with Python.
76 - Declaring Python callbacks to this_actor::on_exit is now working.
77 - No more tests are failing currently, even on win32.
80 - Actor: Merge signals on_migration_start/end into on_host_change
81 - Actor: Rename migrate() into set_host()
82 - Disk: Allow users to get the read and write nominal bandwidth values
83 - Exec: Implement wait_for(timeout)
84 - Io: Implement wait_for(timeout)
87 - Parse errors now raise a simgrid::ParseError that you may want to catch.
90 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
91 been renamed to fix a spelling error. As usual, the previous definitions are
92 kept with a deprecation warning. However, to avoid runtime errors, it is now
93 forbidden to override these deprecated methods in a derived class. Please use
94 the new names immediately if you need to override them.
97 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
98 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
99 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
100 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
101 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
102 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
103 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
104 - New module: random, providing classical random numbers generators.
107 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
108 allocations' size above which they are considered "shared" by default (as if
109 they were performed through SMPI_SHARED_MALLOC macros).
110 Default = 0 = disabled feature.
111 Note : malloc, calloc and free are now overriden by smpicc/cxx by default.
112 This can cause some troubles if codes are already overriding these. If this
113 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
114 help, but will make this feature unusable.
115 - Input check is now performed much more consistently and outputs precise
116 warnings for each error found
117 - New supported calls : MPI_File_set_view, MPI_File_get_view
118 - MPI I/O now only supports disks, as storage is discontinued.
119 - MPI I/O allows opening files without specifying absolute path
120 - Note : SMPI configuration options are now initialized by smpi_init_options()
121 call, which should be called if SMPI is not being used through smpirun.
123 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
124 - FG#9: Python bindings crashing
125 - FG#39: Missing s4u::Comm::wait_any_for example
126 - FG#42: Add support for ThreadSanitizer (TSan)
127 - FG!19: Removing RngStream
128 - FG!20: A module for RNG calls
129 - FG!21: Choice between ad-hoc and standard distributions implementations
131 - FG!25: Fix link in Mutex doc
132 - FG!26: Fix links in SMPI interface doc
133 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
134 - GH#305: Unscheduled tasks are still excuted
135 - GH#313: smpirun: manual outdated w.r.t. --help
136 - GH#321: [S4U] Get task remaining work ratio
137 - GH#323: Crash when an actor turn off his physical host
138 - GH#335: Missing links on dragonfly example svg
139 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
140 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
142 ----------------------------------------------------------------------------
144 SimGrid (3.24) October 9. 2019
146 The Clean Disk Release.
149 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
150 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
151 thus becomes the default resource to execute I/O operations. Examples
152 and the file system plugin have been modified to use disks instead of
153 storage elements. Storage can still be used but is doomed to disappear.
154 Users are thus adviced to update their simulators accordingly.
155 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
156 for consistency with pthread_barrier_wait()
157 - Host::get_englobing_zone() returns the englobing netzone
158 - Actor::on_destruction is now called in the destructor
159 Actor::on_termination new signal called when the actor terminates
161 - Global signals are now part of the Engine:
162 - on_platform_creation: after config settings, before the XML parsing
163 - on_platform_created: right after the XML parsing
164 - on_time_advance: each time the clock advances
165 - on_simulation_end: after simulation, before cleanups
166 - on_deadlock: as the name implies.
168 - sg_{actor,host,link}_{data,data_set}() now all exist.
169 Use them to attach user data to the object and retrieve it.
172 - Introduce an experimental Wifi model. It sounds reasonable
173 according to the state of the art, but it still has to be properly
174 validated, at least against ns-3 if not against reality.
175 - Improved the usability of ns-3. Several bugs were ironed out.
176 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
177 It is now expressed as 'idle:epsilon:allCores' to properly model the
178 consumption of non-whole tasks on mono-core hosts. More info in the doc.
181 - convert a new set of functions to the S4U C interface and move the old MSG
182 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
185 - Fortran bindings for DVFS have been removed.
186 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
187 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
188 - SMPI can now be selected by cmake's find_module(MPI) with
189 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
190 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
191 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
192 - trace-call-location can be used with TI traces, and replayed, and is compatible
193 with smpi/comp-adjustment-file.
194 - sleep events are now correctly traced/replayed.
195 - Default for trace-call-location is now to use file names and not full paths.
196 To revert to previous behavior (in case of collision of filenames), option
197 "smpi/trace-call-use-absolute-path" can be set to yes.
200 - Use the included xxHash as an hash implem when C++14 is usable.
201 - Option model-checker/hash was removed. This is always activated now.
202 - New option smpi/buffering controls the MPI buffering in MC mode.
203 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
204 This is useful to check for MPI compliance.
207 - New section on plugins: how to define a new one, and existing ones.
210 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
211 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
214 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
215 and <mount> tags. The rationale is finish to decouple what should be handle
216 by SimGrid kernel (i.e., resources and activities) from plugins built in
217 user space (e.g., the file system plugin). This new tag currently accepts
218 only three attributes (id, read_bw, and write_bw). All the other information
219 that was declared with the storage related tags now has to be expressed as
220 properties. An example of platform using this new tag is available at
221 examples/platforms/hosts_with_disks.xml
224 - 'expect signal' can now accept more than one potential signal.
226 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
227 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
228 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
229 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
230 - FG#31: per-actor data
231 - FG#34: SG_BARRIER_SERIAL_THREAD?
232 - FG#35: model-checker does not like buster-produced binaries
233 - FG!13: MC: complete workaround in the error msg seen on modern systems
234 - FG!15: execute_flops now logs compute
235 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
236 - FG!17: ns-3: unblock the right number of communications + others issues
237 - FG!18: Improving the performance of the ns-3 bindings
238 - GH#189: Energy consumption of parallel tasks
239 - GH#219: Error in the throughput of TCP transfer
240 - GH!330: Minor polishing of English: Configuring_Simgrid
241 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
242 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
243 - GH!334: Rework energy plugin, again
245 ----------------------------------------------------------------------------
247 SimGrid (3.23.2) July 8. 2019
250 - Nicer introduction page.
251 - Migrate the "Deploy your application" page to the new doc.
252 - Move Java as a subtree of MSG.
255 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
258 - Drop xbt_dynar_sort_strings().
261 - Really fix FG#26: Turning off a link should raise NetworkFailureException
262 - FG#27: Wrong exception thrown to wait_any when link is turned off
263 - GH#328: Java: Canceling multiple tasks in a single vm/host
265 ----------------------------------------------------------------------------
267 SimGrid (3.23) June 25. 2019
269 The Exotic Solstice Release.
272 - SunOS and Haiku OS support. Because exotic platforms are fun.
273 - Stop setting random seed with srand48() at initialization.
274 - Use addr2line as a fallback for stacktraces when backtrace is not available.
275 - Build option -Denable_documentation is now OFF by default.
276 - Network model 'NS3' was renamed into 'ns-3'.
279 - Simgrid can now hopefully be installed with pip.
282 - wait_any can now be used for asynchronous executions too.
285 - New log appenders: stdout and stderr. Use stdout for xbt_help.
286 - Drop xbt_dict_dump.
289 - 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
290 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
291 - MPI interface is now const correct for input parameters
292 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
293 - Fortran bindings for async collectives.
294 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
297 - Remove option 'model-check/record': Paths are recorded in any cases now.
298 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
299 always sparse. That's an immense gain in memory, and can even be
300 faster because of cache effects. So there is no need to clutter the
301 code to allow the user to go for the unefficient mode.
304 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
305 models proved to be more accurate than these old experiments.
307 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
308 - FG#1: Broken link in error messages
309 - FG#2: missing installation documentation
310 - FG#3: missing documentation in smpirun
311 - FG#6: Python bindings not available on PyPI
312 - FG#7: simple cmake call requires doxygen
313 - FG#8: make python bindings an optional dependency
314 - FG#10: Can not use MSG_process_set_data from SMPI any more
315 - FG#11: Auto-restart actors forget their on_exit behavior
316 - FG#12: -Denable_lto=OFF doesn't disable LTO
317 - FG#13: Installs unstripped file 'bin/graphicator'
318 - FG#14: Installs the empty directory 'doc/simgrid/html'
319 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
320 - FG#17: Dead link in doc (pls_ns3)
321 - FG#20: 'tesh --help' should return 0
322 - FG#21: Documentation link on http://simgrid.org/ broken
323 - FG#22: Debian installation instruction are broken
324 - FG#26: Turning off a link should raise NetworkFailureException exceptions
325 - GH#133: Java: a process can run on a VM even if its host is off
326 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
327 - GH#326: Valgrind-detected error for join() when energy plugin is activated
329 ----------------------------------------------------------------------------
331 SimGrid (3.22) April 1. 2019
333 The Easter Christmas Release (this one is somewhat late).
336 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
337 - Not complete yet: asynchronous activities (amongst others) are still missing
338 - Still ongoing: the interface may change in the future. We need more
339 testers! Please report any glitches.
340 - No new project using Java should start now. Please switch to Python.
343 - Some of the internal cleanups may lead to speed improvements:
344 - The hard limitation on the amount of simulated actors with
345 Java+Mac was removed. Now, the available memory is the only limit.
346 - Our refcounting was tidyied, leading to 10% speedups in some cases.
347 - We are still working on making our code robust to the actor kills
348 and hosts' churn. Things are improving, but it's not perfect yet.
349 - Replaced our own code to display a backtrace (that was forking addr2line)
350 with the Boost.Stacktrace library.
351 You won't see your backtraces without this optional dependency.
352 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
353 - Stop setting random seed with srand() at initialization.
356 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
357 XML file version remains 4.2 since old files are still compatible.
360 - Process termination which was broken at version 3.21 has been repaired.
361 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
362 - Hide the examples into examples/deprecated. New users should use Python.
365 - Drop MSG_process_create_from_stdfunc() from the API.
366 This C++ function was a pimple in the C API, made necessary at some
367 point by the Java bindings. This is fixed now.
368 - Hide the examples into examples/deprecated. New users should use S4U.
369 - MSG_process_create and MSG_process_attach now crash if the host used
370 to run this process is off.
371 - Fix the protype of MSG_process_on_exit()
372 Now use (int,void*) callbacks instead of (void*,void*) ones.
373 The implementation was ways too messy to actually work, I guess.
376 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
377 too often and provide more accurate timings.
378 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
379 - MPI_Alltoallw support
380 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
381 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
382 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
383 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
385 - Basic implementation of generalized requests (SMPI doesn't
386 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
389 - Drop sg_cmdline. Please use xbt_cmdline instead.
390 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
391 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
392 OsSemaphore is implemented in a portable way with C++11 threads.
393 This should allow much more threads to be created at the same time,
394 allowing Mac and Java users to simulate many more actors.
395 - Implement the 'thread' factory with std::thread instead of xbt ones.
396 It is not possible to set the stack size with threads anymore, but
397 -fsplit-stack is the way to go nowadays when using threads.
398 - Drop the xbt_os_thread_t module (now unused)
399 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
400 - Drop xbt_str_join_array().
401 - Drop cunit, use Catch2 instead.
404 - Many cleanups in the kernel::activity namespace. This was long
405 overdue, and shall open the path to many future endeavors.
408 - #132: Java: a process can not shut down its own host
409 - #220: S4U: detached send causes exception if sender terminates soon after sending
410 - #261: Document the parameters of parallel execution's constructor
411 - #300: [s4u] BarrierPtr is missing
412 - #314: SMPI args internal cleanup
413 - #316: Fix a bug related to the CPU utilization of multi-core VM
414 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
415 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
416 - #325: Turning off a host has different behavior on sleeping actors and computing actors
418 ----------------------------------------------------------------------------
420 SimGrid (3.21) October 3. 2018
422 The Restarting Documentation (TRD) Release.
425 - Convert the user manual to Sphinx for improved usability.
426 Unstable version now hosted on https://simgrid.frama.io/simgrid
427 Many glitches remain; stay tuned.
428 - Restrict the Doxygen reference API to the installed header files.
429 (documenting all internals this way is overkill + doxygen scales badly)
430 - New tutorial on S4U (highly inspired from the old MSG one, but with
431 a git repository to fork as a starting point, and a docker image)
432 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
433 - The Developper manual is still to be converted (not compiled until then)
434 - Some parts are still missing in this conversion (such as the
435 platform and deployment sections) while others were blindly converted
436 and would need more love (such as the configuration flags).
437 Things will certainly further improve in the future releases.
440 - Fully reimplement auto_restart mechanism that was utterly broken
441 (fix #22, #131 and #281 that were all very related)
442 - Implement semaphores (fix #303 and #291)
443 - When creating an actor from a function and its parameters,
444 move-only parameters are not allowed anymore, as it would prevent
445 the actor to be restartable if its parameters are consumed on run.
446 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
447 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
448 IO operations that can be started, waited for, or canceled as a regular
450 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
454 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
455 - New host variable: 'core_count'
458 - Remove xbt_os_thread_specific features
459 - Remove portability wrapper to condition variables
460 - Remove xbt_os_thread_yield()
463 - MPICH collective selector now mimics MPICH 3.3b
464 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
465 - MPI_Init(NULL,NULL) is now allowed
466 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
467 - Command-line arguments should now be handled properly in Fortran simulations
468 - MPI Attributes and related callbacks should work in Fortran
469 - Apps using C + Fortran should now work
470 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
471 - Support MPI_Error_String
474 - Due to an internal bug, Msg.run() must now be your last line.
475 We hope to fix it in a future release, and we are sorry for the inconvenience.
478 - #22: Process autorestart seem to only work with CAS01 cpus
479 - #56: Feature request: dynamic SMPI replay
480 - #93: simgrid should not eat --help
481 - #111: How to change the loopback link characteristics?
482 - #116: Platform cluster radical
483 - #117: Platform cluster/host order
484 - #129: Replay_multiple on steroid
485 - #131: Java : autostart does not work
486 - #143: Setting a breakpoint at a given time
487 - #235: Network energy models should be integrated
488 - #264: Add ptask L07 resource tracing
489 - #271: Dynamic and manual replay of SMPI traces
490 - #279: Breakpoints option uses time deltas instead of absolute time
491 - #281: Daemonized actor no longer daemonized after an auto-restart
492 - #282: TIT convertor erase traces when using absolute path in the trace list file
493 - #285: segfault when a s4u actor kills itself with msg process tracing activated
494 - #286: Pajé traces are not exposing the number of cores
495 - #287: Command-line parsing should stop when encountering '--'
496 - #288: MPI_Init(NULL, NULL)
497 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
498 - #290: Method to check if a mailbox has a message ready to be consumed
499 - #291: [s4u] semaphores do not seem available yet
500 - #294: Alltoall collective implementation of mpich
501 - #295: Starting a Host as OFF is not supported
502 - #296: DTD too permissive
503 - #299: Add a s4u barrier example
505 ----------------------------------------------------------------------------
507 SimGrid (3.20) Released June 24. 2018
509 The proxy snake_case() release.
512 - The whole API is now using snake_case() instead of mixing it with camelCase.
513 Compatibility wrappers in place for one year.
514 - We took the opportunity to increase the consistency of the API.
515 Please report any remaining glitches.
516 - New function simgrid::s4u::Host::get_actor_count:
517 Returns the number of actors running on a specific host.
520 - SMPI is now tested with ~45 proxy apps from various sources, with none or
521 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
522 - Replay: The replay file has been re-written in C++.
523 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
524 supported. They are stored in the trace and used when replayed.
525 - Basic support of MPI_Cancel. Robustness not guaranteed.
526 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
527 (only for MPI_COMM_TYPE_SHARED).
528 - New option: smpi/privatize-libs, to add external shared libs to be privatized
529 by SMPI. They will be copied locally and loaded separately by each process.
530 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
531 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
532 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
533 with internal index optind with multiple processes. Only works if MPI_Init has
535 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
536 syntax without preprocessing tricks.
539 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
540 behavior tracing is no longer limited to MSG
541 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
542 this means that traces now need to be updated or re-obtained; both Irecv
543 and Isend lines in your traces have now in total 5 mandatory fields:
544 <rankid> <command> <to/from rankid> <tag> <size>
545 To update your traces, it suffices to add a 0 for the tag here.
546 - TIT now also supports waiting for a distinct request via MPI_Wait.
547 Wait/Test now wait for a specific request, not just the last one that was
548 issued. This unfortunately means another update, because we need to
549 identify which request you want to wait for. We do this via the
550 triplet (sender, receiver, tag), which needs to be added:
551 <rankid> <command> <sender> <receiver> <tag>
552 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
553 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
556 - The deprecation of MSG is ongoing (but this should not impact you).
557 Many MSG functions are now simple wrappers on the C API of S4U. If
558 you wish to convert your code to S4U, find the S4U counterparts of
559 your MSG calls in src/msg/msg_legacy.cpp.
560 - MSG can still be used, but won't evolve anymore.
563 - Allow to run the Link energy plugin from the command line with
564 --cfg=plugin:link_energy
565 - Rename Energy plugin into host_energy
566 - Rename Load plugin into host_load
569 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
570 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
571 - Improve performance when handling timeouts of simix synchros.
574 - Config: the C API is now deprecated (will be removed in 3.23), and
575 the C++ API has been slightly improved.
578 - Fix several build issues on OSX.
579 - Move simgrid_config.h to simgrid/config.h (old header still working)
582 - #143: Setting a breakpoint at a given time
583 - #150: Inconsistent event names in SMPI replay
584 - #258: daemonized actors hang after all non-daemonized actors have completed
585 - #267: Linker error on unit_tmgr
586 - #269: SMPI: tracing of MPI_Wait/all/any broken
587 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
589 ----------------------------------------------------------------------------
591 SimGrid (3.19.1) Released March 22 2018
593 The Fixed ABI Release.
595 As you may know, we are currently refactoring SimGrid in deep.
596 Upcoming SimGrid4 will be really different from SimGrid3: modular,
597 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
599 Our goal is to smooth this transition, with backward compatibility
600 and automatic update paths, while still progressing toward SimGrid4.
602 SimGrid remains open during works: The last pure SimGrid3 release was
603 v3.12 while all subsequent versions are usable alpha versions of
604 SimGrid4: Existing interfaces remain unchanged, but the new S4U
605 interface is budding and the internals are deeply reorganized.
607 Since 2015, we work hard to reduce the changes to public APIs. When
608 we need to rename a public library symbol in S4U, we let your
609 compiler issue an explicative warning when you use the deprecated
610 function. These messages remain for four releases, i.e. for one full
611 year, before turning into an error. Starting with v3.15, your can
612 also adapt to API changes with the SIMGRID_VERSION macro, that is
613 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
615 Starting with this v3.19.1, our commitment to reduce the changes to
616 the public interfaces is extended from the API to the ABI: a program
617 using only MSG or SimDag and compiled against a given version of
618 simgrid can probably be used with a later version of SimGrid without
619 recompilation. We will do our best... but don't expect too much of
620 it, that's a really difficult goal during such profund refactoring.
622 The difference between v3.19 and v3.19.1 is that the former was
623 accidentally breaking the ABI of MSG, while the later is restoring
626 S4U and kernel APIs will still evolve until SimGrid4, with one-year
627 deprecation warnings as currently. In fact, cleaning up these
628 interfaces and converting them to snake_case() is one release goal of
629 v3.20. But don't worry, we are working to smooth this upgrade path.
631 In summary, new projects should start with S4U to benefit of the
632 future, but old MSG projects should still be usable with no change.
634 ----------------------------------------------------------------------------
636 SimGrid (3.19) Released March 20 2018
638 The Moscovitly-cold Spring release.
640 SMPI user-visible improvements:
641 - Support of MPI_Type_create_subarray
642 - Unimplemented MPI calls now abort (instead of issuing a warning)
643 Some harmless calls around errhandlers still only issue a warning.
644 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
645 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
646 is now possible (mmap privatization was refactored to this end).
647 Adding new SMPI instances should also be technically doable now.
649 SMPI internal cleanups:
650 - Removed index notion from SMPI -- no more "getPid() - 1"!
651 In tracing files, actors are called rank-<PROCESS-ID> but these are
652 NOT the real ranks (tracing several communicators is not supported yet).
653 - Deprecate smpi_process_index() (will be removed in 3.22)
654 - Better testing on classical proxy apps, through an external project:
655 https://github.com/simgrid/SMPI-proxy-apps/
658 - Use a graphical TOC to make it easier to find the documentation you need
659 - Revamp the MSG tutorial
662 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
663 - Remove parameter of MSG_process_killall().
664 Resetting the PID was bogus anyway (several actors could have the same PID).
667 - Execution->setHost() can be called after start(), to migrate it.
668 - Comm::test_any() is now implemented.
669 - s4u::Actor now has onCreation() and onDestruction() signals.
670 - Install some kernel header files for the users' plugins and more.
671 Warning, their API is really not stable yet.
672 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
673 include/simgrid/kernel/routing/*Zone.hpp
676 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
678 Build System and other cleanups:
679 - Remove unused run-time parameter "tracing/onelink-only".
680 - SimGrid now works with intel compilers (MC still buggy)
681 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
682 - tesh kills subprocesses on timeout (but not on windows)
685 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
688 - #194: Feature request: simgrid::s4u::Comm::test_any()
689 - #245: migrating an actor does not migrate its execution
690 - #253: Feature Request: expose clusters as objects
691 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
692 - #255: Tesh broken on Windows
693 - #256: Modernize FindSimGrid.cmake
694 - #257: Fix (ab)use of CMake install
696 ----------------------------------------------------------------------------
698 SimGrid (3.18) Released December 24 2017
700 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
702 MSG IS NOW DEPRECATED. Please use S4U for new projects.
703 - Support for MSG should not be removed from SimGrid before 2020,
704 but future evolutions will be limited to the integration of
705 user-provided patches.
708 - s4u::onDeadlock() signal, run before stopping the simulation.
709 - s4u::Exec: asynchronous executions (abstraction of background
711 - s4u::Host->getLoad() returns the achieved speed in flops/s
714 - Link::name() is deprecated and will be removed in v3.21, use
715 Link::getCname() instead.
716 - Mailbox::getName() changed to return a std::string, use
717 Mailbox::getCname() instead to get a char*.
718 - Storage::getName() changed to return a std::string, use
719 Storage::getCname() instead to get a char*.
720 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
721 to have both allocation and deallocation of the map in user space.
724 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
725 The amount of flops remaining to do is a vector, not a scalar.
726 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
727 name implies on both sequential and parallel tasks.
728 - Both changes fix GitHub's #223 using PR #237 as a basis.
729 Thanks Michael Mercier.
730 - Most examples were converted to S4U and hidden elsewhere at tests.
731 Rationale: we still want MSG to work; we want newcomers to use S4U.
734 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
735 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
738 - Switch to the faster dlopen privatization mechanism by default
739 - Documentation improvments
742 - Live migration is getting moved to a plugin. Dirty page tracking is
743 the first part of this plugin. This imply that VM migration is now
744 only possible if one this function is called:
745 - C/MSG: MSG_vm_live_migration_plugin_init()
746 - C/C++: sg_vm_live_migration_plugin_init()
747 - Java: Msg.liveMigrationInit()
748 For C and C++, "simgrid/plugins/live_migration.h" also has to be
752 - Define class simgrid::xbt::Path to manage file names.
753 - Removed unused functions:
754 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
755 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
756 xbt_graph_export_graphviz()
757 - xbt/str.h: xbt_str_join()
758 - Remove unused datatypes:
759 - xbt/heap.h: use std::priority_queue or boost::heap instead
760 - xbt/swag.h: use boost::intrusive::list instead
763 - New link_energy plugin for the consumption of the links.
764 - All of the operations on files and storage contents have been
765 packaged into a plugin (src/plugins/file_system). The current
766 public interface can be found in
767 include/simgrid/plugins/file_system.h
768 To use these functions you now have to initialize the plugin by
769 calling MSG_storage_file_system_init() just after calling
770 MSG_init() or sg_storage_file_system_init() just after creating
774 - Remove the undocumented/untested tag <include>
777 - Remove viva specific tracing as the tool is no longer maintained
780 - #248: Kill of finished processes leads to segfault
781 - #240: xbt_cond_wait_timeout should gracefully return for C apps
782 - #239: please implement signal s4u::onDeadlock()
783 - #230: segfaults when exit() without run()
784 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
785 - #223: MSG_task_get_flops_amount() not working with parallel tasks
786 - #222: Actor::kill() doesn't really kill and segfaults
787 - #221: odd LMM warning when killing an actor
788 - #120: Memory leak when the processes are forcefully killed
790 ----------------------------------------------------------------------------
792 SimGrid (3.17) Released October 8 2017
794 The Drained Leaks release: (almost) no known leaks despite the tests.
796 Even very long simulations will preserve your memory: our testsuite
797 stresses SimGrid for over 45mn on fast machines for a coverage of
798 over 80%, and there is only one single known leak, of about 4kb.
802 - Comm.detach(): start and forget about asynchronous emission. A cleanup
803 handler may be given to free resources if the comm cannot be completed.
804 - this_actor::send(mailbox) is now mailbox->put()
805 - New: simgrid::s4u::Comm::wait_all()
806 - New: Host.execute() for remote executions.
809 - Unused option network/sender-gap is removed.
812 - Deprecate MSG_task_isend_with_matching(): This unused feature
813 really complicates our internals. Will be removed in v3.20.
816 - Improved context termination. It is now possible to free resources, even
817 when a process is forcibly killed.
818 - ContextBoost: add support for Boost versions above 1.61.
821 - Bring back run-time option --cfg=exception/cutpath to remove exception
823 - Removed unused functions:
824 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
825 xbt_str_rtrim(), xbt_str_trim().
826 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
829 - Removed header files obsolete since SimGrid 3.12:
830 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
832 - #3: SD_exit should be made optional
833 - #120: Memory leak when the processes are forcefully killed
834 - #159: Threading test regression in Actor refcounting
835 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
836 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
837 - #186: Actor::killAll() segfaults if some process is blocked on wait()
838 - #191: VM migration and pstate
839 - #192: Updating the energy consumptions of all hosts crashes with VMs
840 - #195: All actors have PID=0 in the logs
841 - #204: Sometimes segfault with thread contexts and mmap privatization
842 - #222: Actor::kill() doesn't really kill and segfaults
843 - #225: Actor::kill() doesn't really kill when victims are doing a join()
845 ----------------------------------------------------------------------------
847 SimGrid (3.16) Released June 22. 2017.
849 The Blooming Spring Release: developments are budding.
852 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
853 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
854 - Java JAR file should be finally fully working on Mac OSX too.
856 XML platforms: Switch to platform v4.1 format.
857 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
858 - <zone> can be used as a synonym for the now deprecated <as>
859 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
860 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
861 - <actor> can be used as a synonym for the now deprecated <process>
862 - state_file and avail_file periodicity is now easier to express
863 (check the documentation) the old behavior should still work.
864 * Storage: not backward compatible, but it seems that nobody ever used it
865 - Remove attribute 'content_type' of <storage_type>: was never used
866 - Make attribute 'model' of <storage_type> optional: for future usage
867 - Remove Bconnection model property: was never (in)validated. Replaced by
868 the maximum of the read and write bandwidth as a resource constraint.
871 - New and Backwards Compatibility break:
872 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
873 When one wants to get the list of tasks whose states have changed during a
874 simulation round, s/he has to allocate and free a dynar and use it as
875 argument to this function. The former SD_simulate (double how_long)
879 - Allow multicore VMs, with the correct sharing (unless you overcommit)
880 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
883 - New function to update the consumption of all hosts at once.
884 - Fix the model for multi-core hosts, linear in the amount of busy
885 cores with an abnormality for pIdle (see paper).
886 BUG: Still not satisfactory for ptask on mono-cores.
889 - The netzone are now available from the MSG API.
890 The old names still work, but are now deprecated.
893 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
894 - New option: smpi/keep-temps to not cleanup temp files
895 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
896 mallocs mode is used, allows to change the size of the fake file used
897 (default 1MB), to potentially limit the number of mappings for large runs.
898 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
899 - Fortran ifort and flang compilers support
900 - New RMA calls supported (experimental):
901 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
902 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
903 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
904 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
905 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
906 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
909 - New: this_actor::isMaestro() returns whether we are in kernel mode.
910 - Behind the scene, ActivityImpl are now automatically refcounted.
911 Removing this nasty bottleneck will greatly simplify our internals.
914 - Replay: New function xbt_replay_action_get():
915 Retrieve the function previously associated to an event type.
916 - DROPPED FUNCTION: xbt_str_varsubst()
917 - DROPPED MODULE: strbuff. We don't need it anymore.
918 - DROPPED MODULE: matrix. We don't need it anymore.
919 - DROPPED MODULE: lib. We don't need it anymore.
921 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
923 SimGrid (3.15) Released March 22. 2017
925 The Spring Release: continuous integration servers become green
926 We fixed even the transient bugs on all target architectures:
927 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
928 Windows, FreeBSD, NetBSD.
930 - Rename NetCards to NetPoints.
931 This was intended to help NS3 users, but that's not a netcard.
932 That's a point in the routing algorithm, let's avoid wrong simplifications.
935 - New: MSG_process_yield(). Stop and yield to other processes.
936 - New: MSG_process_daemon(). Daemon processes are automatically killed
937 when the last non-daemon process terminates
938 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
939 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
940 to make room for the upcoming network energy plugin.
941 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
944 - Ensure that an actor can kill itself with Process::exit()
945 - Kill the obscure NativeException. Nobody want to survive the issues
946 it denotes, so use JniException that is a RuntimeException (not to
947 be caught explicitly).
948 - Partial bug fix in initialization. SimGrid flags on command line were
949 consumed at C level but stayed in the original Java String[] args.
950 This could mess users' args[i] if SG flags were not put at the end of
952 The SimGrid flags are now removed from the Java arguments. However,
953 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
954 if args.length is greater than the number of YOUR OWN ARGUMENTS.
955 It might be if you have --log or --cfg flags in the command line.
956 - Fix numerous memleaks all around the place. In particular, around VMs.
960 - s4u::onPlatformCreated: right before the simulation starts
961 - s4u::onSimulationEnd: right after the main simulation loop
962 - s4u::onTimeAdvance: right after a clock change (time is discrete)
963 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
964 event from the availability_file changes the avail speed.
965 - Links are now usable from s4u
966 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
967 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
968 - New examples: The conversion of MSG examples to S4U has begun
969 - Actors: create, kill, migration, and suspend
970 - Applications: master-worker and token-ring
971 - Action replay: communications and storage
972 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
975 - Backwards Compatibility breaks
976 - The SD_route_get_list and SD_route_get_size functions have been removed.
977 They are replaced by the unique
978 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
979 The route from src to dst is built in the links dynar whose size can
980 be obtained with xbt_dynar_length.
981 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
982 sg_host_route_bandwidth and sg_host_route_latency.
983 Macros ensure the backwards compatibility, but you should fix your code
986 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
987 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
988 - Bump our claim of support from MPI 1.1 to MPI 2.2.
989 We don't support 100% of it, but it should be enough. Contact us if not.
990 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
991 naive and probably inaccurate)
992 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
993 mapped onto a unique small file using some system magic.
994 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
995 - Fortran: cleanups, fixes, support of user-added operations
996 - RMA: MPI_Accumulate are applied in correct order.
997 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
998 - Better support for MPI_IN_PLACE
999 - Support for MPI_Win attrs and keyvals.
1000 - Support MPI_Comm_group_create, MPI_Type_size_x
1001 - MPI ops should be properly applied to all allowed types
1002 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1003 algorithm by a non failing one from mpich.
1006 - Kill the fifo data container: we don't use it anymore.
1009 - A <cluster> can now be created with different speed values to represent pstates.
1010 see examples/platforms/energy_cluster.xml for an example.
1012 SimGrid (3.14.159) Released December 28. 2016
1014 The previous release was only a crude approximation of a Pi release;
1015 this one should be a bit better.
1017 - Revert a last minute change that broke on Mac OSX.
1018 - Fix the changelog and NEWS files.
1019 - Improve the documentation, in particular of the routing module.
1020 - Integrate some patches flying around in Debian.
1022 SimGrid (3.14) Released December 24. 2016
1025 * The whole documentation was reworked and reorganized. There is
1026 still much room for improvement here, but we're on it.
1029 * We now need python3 for our testing infrastructure.
1030 * Model checking needs libevent
1033 * Now works on FreeBSD
1036 * Privatization now works on FreeBSD
1038 * Privatization is now activated by default in smpirun
1039 You can turn it off with -no-privatize if something goes wrong.
1041 * Call-location tracing for SMPI.
1042 You can add the exact location (filename / linenumber) of an MPI call to
1043 your trace files and slow down or speed up the simulation between two
1044 consecutive calls by using an adjustment file (see the documentation).
1046 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1047 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1048 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1049 values were ignored in some cases. The timings of these functions can now
1050 be significantly different.
1052 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1053 smpi/running-power is renamed to smpi/host-speed
1055 * smpi/grow-injected-times option to enable or disable multiplication of the
1056 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1057 make simulation less precise (but also much faster).
1059 * smpirun script should be (much) faster for large deployments.
1061 * SMPI tracing: fixed issue with poor matching of send/receives.
1063 * Replay: Fix broken waitall
1065 New functions and features
1066 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1068 Dropped / renamed functions and features
1069 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1070 - MSG_mailbox_is_empty() -> Mailbox::empty()
1071 - MSG_mailbox_front() -> Mailbox::front()
1072 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1073 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1074 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1075 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1076 * Don't pass the free_f parameter to property related functions:
1077 - MSG_host_set_property_value()
1078 - MSG_as_router_set_property_value()
1079 - MSG_storage_set_property_value()
1080 * VM properties. Since msg_vm_t are msg_host_t, just use
1081 MSG_host_get_property_value() and friends
1082 * VM I/O related things:
1083 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1084 - Unimplemented save/restore methods
1085 * MSG_as_router_get_property_value() was redundant with
1086 MSG_environment_as_get_property_value().
1087 - Removed MSG_as_router_*propert*() functions
1088 - Added MSG_environment_as_set_property_value() for consistency
1089 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1090 Well, they are still there for now, but deprecated with a warning.
1091 Please switch to xbt_dict_new_homogeneous() before this is removed
1093 * Task affinity. Its intended behavior (that was very badly tested
1094 and probably not really working) was deceiving what most users
1095 would have hoped here.
1096 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1097 * Option network/coordinates is now useless and should be dropped.
1100 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1103 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1104 * Added Dragonfly topology support
1107 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1108 EiBps, EBps, Eibps, Ebps for bandwidth.
1109 They may become useful to some lucky ones.
1112 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1113 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1114 * ex/async_waitAll: new example, on asynchronous communications
1117 * Memory usage should be decreased for simulations with a large number
1118 of processes. This also helps for SMPI.
1120 SimGrid (3.13) stable; urgency=low
1122 The Half Release, a.k.a. the Zealous Easter Trim.
1124 - We removed half of the lines, that were mostly experimental cruft.
1125 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1126 (not counting blanks and comments -- according to openhub.net)
1127 - The internals are now compiled in C++ (and will soon be clean C++)
1128 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1129 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1131 * Backwards Compatibility breaks
1132 - Removed Lua simulation bindings (switch to C or Java for that).
1133 Lua can still be used to describe platforms
1134 - Removed Java kernel plug-ins.
1135 Will be reintroduced after the ongoing major internals reorg.
1137 - the following functions were removed.
1138 They were too specific and should be reimplemented in a generic
1139 way, with filter function.
1140 - MSG_task_listen_from_host
1141 - MSG_mailbox_get_count_host_waiting_tasks
1142 - MSG_mailbox_put_with_timeout was removed.
1143 Please use MSG_task_send_with_timeout instead.
1145 - the SD_application_reinit function was removed. It has been a noop for a while.
1146 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1147 replaced by a more flexible mechanism.
1148 - The following functions thus do not exist anymore
1149 - SD_workstation_get_access_mode
1150 - SD_workstation_set_access_mode
1151 - SD_workstation_get_current_task
1152 - Basic estimation functions have been removed but can easily be replaced
1153 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1154 - SD_workstation_get_computation_time => amount / sg_host_speed()
1156 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1157 Use VM.getSpeed()*load/100 for the legacy behavior.
1159 - option enable_tracing was removed. It was not doing anything for a while.
1160 - In the ModelChecker:
1161 - the model-checker now ptraces the model-checked process which means
1162 you cannot use a debugger on the latter anymore (we might make this
1163 optional in the feature);
1164 - removed soft-dirty page tracking;
1165 - remove model-checked side snapshot management,
1166 MC_snapshot() and MC_compare_snapshot();
1167 - keep the MC_cut() function as a stub (it was not really working
1168 in the previous release).
1171 * All options are consistently in kebab-case. Old names are kept as alias.
1174 * Switch to platform v4 format.
1175 - Rename from 'power' to 'speed' the attributes describing the amount of
1176 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1177 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1178 - In <host> and <link>, attributes availability and state are gone.
1179 It was redundant with state and availability traces, and with peak values.
1180 - In <cluster>, attributes availability_file and state_file are gone.
1181 It was too complex and unused.
1182 - Kill <gpu>. Was not doing anything.
1183 - The DOCTYPE points to the right URL:
1184 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1185 (the file at this address now documents the changelog since its v1)
1186 - A warning is emitted for unit-less values (they are still accepted).
1187 - speed. Default: 'f' or 'flops'. Also defined:
1188 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1189 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1190 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1191 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1192 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1193 - latency. Default: 's' second. Also defined:
1194 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1196 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1197 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1200 * s4u::Host is now the preferred public interface to the Host features.
1201 sg_host_* functions are C bindings to the exact same behavior
1202 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1205 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1206 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1207 They were deprecated since a few years, and probably did not even compile anymore.
1210 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1211 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1212 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1213 encouraged to update their codes. A list of the modified functions can be found at the end of
1214 include/simgrid/simdag.h
1217 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1218 the lambda or closure passed as a parameter will run in kernel mode.
1219 All the callback functions should be rewritten to that interface at some point.
1222 * Reorganizing and cleaning the internals all around the place.
1225 * Remove old default barrier/bcast buggy algorithms (see #18407)
1226 * Various bug fixes to handle more codes
1227 * Remove the need for the --foreground option of smpirun (it is still
1228 accepted for backward compatibility).
1231 * Kill the setset data container: MC don't use it anymore.
1232 * Kill the queue data container: it made more sense with GRAS.
1233 * Kill the xbt_peer_t data type: it's useless without GRAS.
1234 * Kill rm_cb feature of config sets: it was never useful.
1235 * Kill graphxml parsing feature. It was not used.
1236 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1238 - xbt_dynar_sort_strings(), when the content is char*
1239 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1240 They throw exceptions on invalid input;
1241 * C++ support for declaring CLI flags (simgrid::config::Flag);
1242 * class for abstracting different signal backends (simgrid::xbt::signal).
1243 with no external dependencies (we need very simple signals).
1246 * refactoring and cleanup of the code;
1247 * ongoing process to cleanly separate the model-checking algorithms
1248 from the code model-checking support.
1250 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1252 SimGrid (3.12) stable; urgency=low
1254 The Facelift Release.
1257 * Require g++ v4.7 at least to not speak prehistorical C++.
1258 * Require Boost 1.48 (for signal2 component).
1259 * Java must be version 7 at least when activated.
1260 * Builds on Windows again (including Java bindings).
1261 * Tracing is now always enabled (no way to turn it out)
1262 * Remove GTNetS. It was not working anyway.
1263 * Various cleanups in the cmake scripts.
1264 * Move headers around to sort them out on installed systems:
1265 - instr/instr.h -> simgrid/instr.h
1266 - instr/jedule/* -> simgrid/jedule
1267 - simdag/datatypes.h was removed
1268 - simdag/simdag.h -> simgrid/simdag.h
1269 - msg/datatypes.h was removed
1270 - msg/msg.h -> simgrid/msg.h
1273 * Interface improvement:
1274 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1275 - Sanitize the interface in MSG_task_ module:
1276 - Merge two functions that were close enough but misleading:
1277 set_compute_duration(t) -> set_flops_amount(t)
1278 get_remaining_computation(t) -> get_flops_amount(t)
1279 - set_data_size(t) -> set_bytes_amount(t)
1280 get_data_size(t) -> get_bytes_amount(t)
1281 - Massive cleanups in the functions related to the energy
1282 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1283 - New: MSG_host_get_pstate()
1285 - msg/energy/onoff: switching hosts on and off
1287 * Interface improvement:
1288 - Rename Host.isAvail() to Host.isOn()
1289 - Rename Process.currentProcess() to Process.getCurrentProcess()
1290 - Rename Task.setDataSize() to Task.setBytesAmount()
1291 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1293 - #18874: Actually allows the GC to reclaim tasks
1297 - SIMIX_process_throw: raises an exception in a remote process
1298 * Refactoring: Separate sub-modules
1299 - libsmx: the public interface, as libc in a real system
1300 - popping: the strange dance that converts a user request into a kernel handling
1301 - smx_context_*: the virtualization mechanisms that embed the user code
1302 - smx_*: the handling of each simcalls
1303 * Interface cleanup:
1304 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1305 * Rename smx_action_t into smx_synchro_t, making explicit that these
1306 things are used to synchronize processes with their environment.
1307 For example, a communication is a sort of synchronization involving
1308 the communicating processes (that may block until the exchange) and
1309 the platform. The same can be said from computations, etc.
1311 - #18888: segfault when a process ends before its kill_time
1315 - 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.
1316 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1317 - MPI_Type_set_name, MPI_Type_get_name
1318 - MPI_*_c2f and MPI_*_f2c functions
1319 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1320 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1321 - Activate a lot of new tests from the mpich 3 testsuite
1323 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1324 - InfiniBand network model added: Based on the works of Jerome Vienne
1325 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1326 - When smpi/display_timing is set, also display global simulation time and application times
1327 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1328 * Collective communications
1329 - 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
1330 - 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/).
1331 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1333 - Replay now uses algorithms from wanted collective selector
1334 - Replay can be used with SMP-aware algorithms
1335 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1336 - 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.
1338 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1339 - Correctly use loopback on fat-tree clusters
1340 - Asynchronous small messages shouldn't trigger deadlocks anymore
1341 * Energy/DVFS cleanup and improvement
1342 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1343 - new: smpi_get_host_pstate()
1346 - "Full" network optimization flag was broken since Surf++
1347 - Better handling of precision flags in maxmin
1348 - Fix bug causing sometimes "Impossible" errors
1349 - Properly pass cluster properties to included hosts
1350 * Improvement of the Energy plugin.
1351 - Always update the consumption before returning that value
1352 - New property: watt_off to denote the dissipation when the host is off
1353 - New functions getWattMinAt and getWattMaxAt to retrieve the
1354 dissipation of pstates that we are not currently at.
1355 * Java: class NetworkLink renamed to Link
1356 * New function: simcall_process_get_kill_time()
1357 * Massive rename s/workstation/host/
1358 - That's intrusive, but that's good for the project consistency. Sorry.
1359 - Change config option "workstation/model" into "host/model"
1362 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1363 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1364 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1365 - xbt_dirname and xbt_basename for non-POSIX systems
1367 * The model checker now runs as a separate process.
1368 * The model checker runs must now be launched with the new simgrid-mc program.
1369 * Record/Replay: the MC can display a textual representation of a path in the
1370 execution graph. It can then be replayed outside of the model checker.
1372 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1374 SimGrid (3.11) stable; urgency=low
1379 * Normalizing pointers addresses tool for better diff between logs
1382 * Add cloud examples using new VMs
1383 - examples/msg/cloud/two_tasks_vm.tesh
1384 - examples/msg/cloud/simple_vm.tesh
1385 - examples/java/cloud/cloud.tesh
1386 - examples/java/cloud/migration/migration.tesh
1387 * Add java surf examples:
1388 - examples/java/surfPlugin/surf_plugin.tesh
1389 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1390 - examples/java/surfCpuModel/surf_cpu_model.tesh
1391 * Add SMPI+MSG example:
1392 - examples/smpi/smpi_msg_masterslave/
1397 - msg task destroy cancel test
1398 - msg_host on/off test
1399 * Move all tests in testsuite to teshsuite (adding tesh files)
1400 * Restructure teshsuites
1401 - one folder for each kind of test
1402 * Restructure AddTests.cmake
1404 - structure the order of tests (with sections)
1407 * Add virtual machine
1408 - creation of a VM on a PM
1409 - migration of a VM from a PM to another PM
1411 - MSG_process_join(msg_process_t process, double timeout)
1412 - msg_bar_t MSG_barrier_init(unsigned int count)
1413 - int MSG_barrier_wait(msg_bar_t barrier)
1414 - void MSG_barrier_destroy(msg_bar_t barrier)
1415 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1416 * New option "msg/debug_multiple_use" to help debugging when a task is used
1419 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1421 - Increase code coverage in test suites
1425 * Protect context stacks against stack overflow. The number of protected memory
1426 pages allocated on the top of each stack (1 by default) can be configured
1427 with the new command line option --cfg=contexts/guard_size:<value>.
1428 * Simcalls are now generated by a python script that
1429 - generates files included by SimGrid
1430 - checks that all the functions exist, and proposes prototypes
1432 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1433 host_set_data, host_get_data
1435 - simcall_process_join(smx_process_t process, double timeout)
1436 * Fix bug where sleeping processing could not be suspended.
1439 * Translate surf models from C to C++
1440 - Generic classes for all models: Model, Resource, Action
1441 - A generic interface for each kind of model (CPU, Network, Storage
1442 Workstation, WorkstationVM)
1444 * Translate surf routings from C to C++
1445 * Add callbacks using sigc++ or boost::signals2
1446 - Add callback functions for resource creation/destruction
1447 - Add callback functions for action state change
1448 - Handle Energy as a plugin
1449 * Replace swag by boost::intrusive
1450 * Add new routing models for clusters. For documentation, see
1451 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1452 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1453 parameters for cluster tag
1454 - Fat trees, with topology="FAT_TREE" and
1455 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1456 - see examples/platforms/torus_cluster.xml and
1457 examples/platforms/fat_tree_cluster.xml
1458 * More documentation
1461 * Hostfiles support host:nb_processes construct to deploy several processes on
1463 * Collective communication algorithms should not crash if used with
1464 improper number of nodes and report the error.
1465 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1466 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1467 MPI_Dims_create, MPI_Cart_sub are supported.
1468 * New interface to use SMPI programmatically (still depends on MSG for
1469 some parts, see examples/smpi/smpi_msg_masterslave):
1470 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1474 * Global variables privatization in MPI executables is now performed at runtime
1475 with the option smpi/privatize_global_variables (default:no).
1476 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1477 dynamic libraries loaded by the application are not privatized (static
1478 linking with these libraries is advised in this case)
1481 * Options defined in XML work correctly now.
1484 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1485 copied into simgrid.jar or not (ON by default). Use this option if you want
1486 to reduce the size of the installed simgrid.jar, *and* the native libraries
1487 are kept installed elsewhere.
1488 * Surf binding with SWIG (code generated in maintainer mode only):
1489 - plugin to handle callbacks
1490 - CPU model only for the moment
1493 * Supernovae build mode is definitively removed. It was used to improve
1494 inlining and inter-module optimizations. It is nowadays superseded by
1495 link-time optimizations commonly available in compilers.
1496 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1497 * Add boost dependency for surf++
1498 * Add new macro for tests
1499 - ADD_TESH(name <tesh_args>)
1500 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1504 - xbt_bar_t XBT_barrier_init(unsigned int count)
1505 - int XBT_barrier_wait(xbt_bar_t barrier)
1506 - void XBT_barrier_destroy(xbt_bar_t barrier)
1507 * Make the xbt_os_time module public
1509 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1511 SimGrid (3.10) stable; urgency=low
1513 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1516 * Reintegrate Java to the main archive as desynchronizing these
1517 package is not acceptable anymore (Java is now considered stable)
1518 * Add explicit synchronization facilities through semaphores
1519 * Bug fix: Task.setDataSize() only changed the C world, not the value
1520 cached in the Java world
1523 * Dramatically change the way files are handled. API and internals changed, but
1524 this part of MSG was not considered as production grade either.
1525 * Add explicit synchronization facilities through semaphores
1526 * Add a new function MSG_host_get_process_list()
1527 * Preliminary DVFS support (see examples/msg/energy/ for details)
1530 * SMPI is now included directly in the libsimgrid as the windows
1531 linker doesn't force us on splitting it anymore.
1532 * Improvements of the SMPI replay tool:
1533 - Most of the collective communications are now rooted in the same process as
1534 in the original application.
1535 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1536 used until now). Multiple data types can now be used in a trace.
1537 - The replay tool now supports traces produce either by TAU or a modified
1539 - Bug Fix: the compute part of the reduce action is now taken into account.
1540 - Gatherv collective is now supported
1541 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1542 of smpirun outputs time independent traces for the current run. One file
1543 is created per process. If too many processes are simulated, this behavior
1544 can be changed to one file for all processes by using the
1545 tracing/smpi/format/ti_one_file flag
1546 * smpirun generates the host file if needed (with given host count and platform)
1547 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1548 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1549 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1550 --cfg=smpi/coll_selector:(mpich/ompi)
1551 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1553 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1554 happening inside a collective SMPI call.
1555 * Fix the behavior of complex data types handling.
1556 * Make MPI_Wtime another synchronization point to take computations into
1558 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1559 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1560 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1561 (no privatization of global variables yet)
1562 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1565 * Verification of liveness properties is now available for SMPI applications
1566 (in addition to MSG applications)
1567 * Bugged examples using SMPI in examples/smpi/mc/
1568 * Add --cfg=model-check/visited option. Allows the verification of infinite
1569 programs. Detection of loops in the execution thanks to the system state
1570 comparison and reduction of the state space to explore. Can be combined with
1571 DPOR for safety properties.
1574 * Allow to change SimGrid configuration (see --help) within the code
1575 thanks to SD_config() as it can be done in MSG.
1576 * Add a new function SD_task_set_amount() upon user request.
1579 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1580 * Remove rule based routing (no more PCRE dependency)
1581 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1582 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1583 * Add a loopback_bw and loopback_lat options to cluster tag.
1584 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1585 routing. Now peers are encapsulated in an AS and have their own private
1586 router but this is transparent.
1589 * Our own implementation of getline is renamed xbt_getline, and gets
1590 used even if the OS provide a getline(). This should reduce the
1591 configuration complexity by using the same code on all platforms.
1592 * New type: xbt_cfg_elm_boolean.
1593 * Allow to use yes/no for boolean configuration options in the command line.
1594 * Allow to disable SimGrid cleanups at exit from command line option.
1595 There are situations where one may want a simulation to end with an exit.
1596 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1597 annoying when scripting around the simulator. Adding a
1598 --cfg=clean_atexit:no allows to circumvent this issue.
1601 * Lots of memory leaks were corrected in this release.
1602 There is no known memory leaks anymore, in all of our 600+ tests.
1603 * New command line option --version, to get SimGrid version information.
1604 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1606 * Supernovae builds are deprecated, and expected to be removed in the next
1609 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1611 SimGrid (3.9) stable; urgency=low
1613 The Grasgory release: GRAS is really dead now.
1615 * Complete overhaul of the internal host structures scheme.
1618 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1619 was considered as experimental and badly maintained since then.
1620 * Keeping it was thus a trap to our potential users, that could take
1621 it instead of MSG or SMPI by mistake despite is pity state.
1622 * GRAS seems to have very few users (if any), and no one volunteered
1623 to maintain it further. It also induces a lot of XBT code (for
1624 portability sake), that must be maintained too.
1625 * For all these reasons, we killed GRAS. If someone wants to revive it
1626 in the future, don't cry, our git history still remembers of GRAS.
1629 * Major overhaul. Merge our documentation again as time proved that
1630 splitting it was really not helping our users.
1631 * Further improve the developer documentation to help newcomers
1632 hacking on SimGrid itself. The user documentation (and in
1633 particular, the beginner documentation) is still in a sorry state.
1636 * Now works on Windows too!
1637 * Much more extensive test suite, from MPICH
1640 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1641 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1642 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1643 * Bug fix: let task be scheduled when the last dependency to be solved is
1644 a control dependency.
1645 * Remove SD_load_environment_script function.
1646 Use the C sg_platf function if you want to declare a platform
1650 * New function: MSG_process_get_number()
1651 * Old function documented: MSG_config()
1652 * Remove MSG_load_platform_script function
1653 Use the C sg_platf function if you want to declare a platform
1657 * Change the default value of the TCP_gamma constant (maximal size of TCP
1658 congestion window) to a more realistic 4MiB value. If you notice changes in
1659 your simulation results, you can fall back to the previous 20k tiny window
1660 by adding --cfg=network/TCP_gamma:20000 on command line.
1661 * (Hopefully) fix a bug wrt periodic availability and state traces
1662 * Bug fix: use default values at start when first event in availability/state
1663 trace is not at time 0.
1666 * remove the "new_" part of function name sg_platf_new_trace_connect
1667 (resulting in sg_platf_trace_connect), since it does not create
1671 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1672 use case where it's really mandatory, and maintaining it was a pain in
1674 * New: xbt_fifo_search(), search an item with a user-provided
1675 comparison function instead of dumb pointer comparison.
1678 * Fix the lua deployment:
1679 Use `simgrid.init_application()` before deployment instead of
1680 `simgrid.msg_register_application()` after.
1683 * Transfer the tracing files into the corresponding modules.
1685 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1687 SimGrid (3.8.1) stable; urgency=low
1689 The "we are told that some people want to also *install* the simgrid
1692 * Add missing file "tesh.1" to the archive.
1694 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1696 SimGrid (3.8) stable; urgency=low
1698 The Psssshiiiit release: SimGrid jumps into the Cloud.
1701 * Add an experimental interface to manipulate VMs. They are mainly
1702 process groups with very few intrinsic semantic, but they should
1703 allow you to build the semantic you want easily.
1704 * New function: MSG_host_set_property_value()
1705 * New function: MSG_process_on_exit(). To clean memory in all cases.
1706 * Bug fixes that made the host (and link) failures unusable.
1707 * Add a way to auto-restart process when the host in which they are
1708 executing comes back (ON_FAILURE="RESTART" on deployment file,
1709 MSG_process_auto_restart_set).
1710 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1711 please stop using the old ones, they are DEPRECATED.
1713 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1714 Please use MSG_init() instead. (reducing the amount of entry
1715 points in the library helps us).
1716 * Make it impossible to link against the wrong version of the lib
1717 * Deprecate MSG_clean(). No need to call it anymore.
1718 * Function MSG_get_host_number() is not deprecated anymore.
1721 * Split the doc into a user guide and a reference guide.
1722 * Start a developper guide to help people hacking on SimGrid.
1725 * Enable tracing by default. This modules rocks you should use it.
1726 * Remove option custom_flags. Now use environment variables CFLAGS
1728 * Use default cmake things to detect lua instead of home grown ones.
1729 * New option "enable_mallocators" to disable mallocators, for debugging
1730 purpose ("on" by default).
1733 * Bug fixes around the resource failures: don't let the processes
1734 survive the host they are running onto.
1735 * Add an interface to auto-restart processes when the host in which they are
1736 executing comes back.
1737 * Ensures that SIMIX_clean is called automatically. It's not part of
1738 the public interface anymore (bindings should be updated).
1741 * Bug fix for when SD_Simulate is called with a positive value: be careful
1742 when comparing doubles. Sometimes they are different for non significant
1744 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1745 parallel task whose initial work is distributed among host according
1746 to the Amdahl's law. Such tasks are created with a parameter alpha
1747 that corresponds to the non-parallelizable part of the computation.
1748 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1749 between two sets of workstations assuming a 1D block distribution (each
1750 workstation owns a similar share of data) on both sides.
1752 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1753 Data redistribution will be automatically scheduled once parent and child
1754 are both scheduled. The filling of computation_amount and
1755 communication_amount structures is now done seamlessly thanks to the chosen
1757 * New function SD_workstation_dump to display various information
1758 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1759 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1760 bandwidth on the route onto which the task is scheduled and the amount of
1762 To divide the nominal bandwidth by 2, the rate then has to be:
1763 rate = bandwidth/(2*amount)
1764 * Compute tasks that have failed can now be rescheduled and executed again
1765 (from their beginning)
1766 * Increasing source code coverage (src/simdag is now covered at 95.8%
1770 * Re-implement time-independent trace replay using SMPI (at the
1771 smpi_smp_* level) instead of MSG. This should replace
1772 examples/msg/actions/actions.c
1773 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1774 hindexed and structs)
1775 * Implement the exchange of non-contiguous data.
1776 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1777 * Correct behavior of smpi/sender_gap and set its default value to 0
1778 * Add option to asynchronously send small messages to allow better
1779 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1780 specifies the size in bytes under which messages will be asynchronously sent.
1781 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1782 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1785 * New C interface to define a platform: XML is now optional.
1786 For more info, please check include/simgrid/platf.h.
1787 * New interface to define random platforms from the C:
1788 For more info, please check include/simgrid/platf_generator.h and
1789 examples/msg/masterslave/masterslave_platfgen.c
1790 * Export a sg_cmdline dynar containing all the arguments we got from
1794 * Two new tracing options for adding comments to trace file so you
1795 can track your experiments (see --help-tracing for details).
1796 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1797 * Adding the SimGrid version that generated the trace file as a comment.
1798 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1799 * Fix to avoid key clashes on Paje links
1800 * Other minor fixes related to the Paje specification
1803 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1804 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1805 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1806 * Greatly improve the robustness of mmalloc to user errors (such as
1807 using an area after freeing it, or freeing it twice)
1809 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1811 SimGrid-java (3.8.1) stable; urgency=low
1813 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1814 Warning: all methods are visible, but only some of them are
1815 implemented so far. Check the source (src/jtrace.c)
1816 for further information.
1817 * New module: org.simgrid.msg.File (SimGrid File management functions)
1818 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1819 * Change the meaning of Process.restart: now restart the process from
1820 the beginning, like MSG_process_restart in C.
1821 * Add Process.setAutoRestart: handling of process restart when failed
1823 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1824 you to retrieve the properties of the processes/hosts
1825 * Deprecate Msg.clean(): you can just forget about it now.
1826 * New function Process.getCount(), that only works when compiling
1827 with the not yet released version 3.9 of the C library.
1829 * New context factory based on Coroutines. It mandates a modified JVM
1830 but then, the simulations run about five times faster, and there is
1831 no limit to the amount of processes (beside of the available memory).
1833 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1835 SimGrid (3.7.1) stable; urgency=low
1838 * Restore the prototype of MSG_process_create_with_environment() to
1839 the pre-3.7 situation by removing the kill_time argument.
1840 * Add a MSG_process_set_kill_time() function instead.
1843 * Fix weird behaviors when dealing with parallel tasks.
1846 * Simgrid is now built as a dll.
1847 * Simgrid-java now works on Windows.
1848 * Simgrid-Java is now included into Windows package.
1851 * First pre-build package for MacOSX.
1854 * Fix compilation when using MSG_USE_DEPRECATED.
1855 * Fix some compilation issues on Macs and Windows.
1856 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1857 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1859 We discovered that the Lua console is broken, but we are missing the
1860 manpower to fix it right now. The problem existed in 3.7 too, so we
1861 are not blocking the release for that. Sorry if you depended on this
1862 feature, any help would be really welcome.
1864 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1867 SimGrid-java (3.7.1) stable; urgency=low
1869 The "Java aint got to be bloated and slow" release
1872 * Various internal cleanups and performance improvement
1873 Simulations are expected to run up to twice faster or so
1874 * Make Process.kill(process) an instance method, not a static one
1875 * User processes are not java.lang.Thread subclasses.
1876 This breaks the compatibility (sorry), but previous API was
1877 brain-dead, making it impossible to have non-trivial
1878 initializations in the process constructor.
1879 * Require a full constructor per Process sub-class.
1880 Kinda breaks the compatibility (sorry), but this allows a much more
1881 efficient way to launch the processes at simulation startup.
1882 * Do not embeed our version of semaphores, java 1.5 can be considered
1883 as sufficiently prevalent for us to not dupplicate its features.
1887 * Add examples for almost every part of the API
1888 We spotted and fixed a lot of bugs in the process
1890 * New module: asynchronous communication API
1891 * New function: Process.sleep()
1892 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1893 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1894 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1896 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1898 SimGrid (3.7) stable; urgency=low
1900 The "spring cleanups (before the next Big Project kicks in)" release.
1903 * We can specify the SMPI latency/bandwidth factor with command line
1904 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1905 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1906 You can also use the "config tag" from platform file by adding this line
1907 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1908 (see "example/platforms/tag_config.xml" to use "config tag").
1909 Note that the command line supersedes the platform file configuration.
1910 * Change the correction factors used in LMM model, according to
1911 the latest experiments described in INRIA RR-7821.
1912 Accuracy should be improved this way.
1913 * Use the partial invalidation optimization by default for the
1914 network too. Should produce the exact same results, only faster.
1915 * Major cleanup in surf to merge models and split some optimization
1916 mechanisms from the core of the models. As a result you can now
1917 specify which model to use (e.g., --cfg=network/model:LV08
1918 --cfg=cpu/model:Cas01) and which optimization mode to use
1919 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1920 Incompatible combinations should err at initialization. See
1921 --help-models for the list of all models and optimization modes.
1922 * The CLM03 workstation model was dropped for simplicity because it
1923 used the deprecated CM02 network model. Use default instead.
1924 * Rename the TCP_gamma configuration option to network/TCP_gamma
1925 * Rename the coordinates configuration option to
1926 network/coordinates, and document it
1927 * Use now crosstraffic keyword instead of the terribly misleading
1928 fullduplex keyword. It is activated by default now in the current
1929 default model, use --cfg=network/crosstraffic:0 to turn it off.
1930 * Ongoing refactoring the model parsing to make XML files optional
1931 See include/simgrid/platf.h for details (still to be completed)
1934 * Major overhaul of the documentation. Almost instructive now :/
1935 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1936 functions and friends. This interface was considered as
1937 deprecated since over 2 years, it's time to inform our users that it is.
1938 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1939 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1940 use these (crappy) functions in your code.
1941 These functions will be removed soon. Stop using them now.
1942 * Deprecate MSG_get_host_{table,number}
1943 Implement MSG_hosts_as_dynar() instead.
1944 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1945 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1948 * Stabilize the parallel execution mode of user contexts
1949 * Introduce configuration variables to control parallel execution:
1950 - contexts/synchro: Synchronization mode to use when running
1951 contexts in parallel (either futex, posix or busy_wait)
1952 - contexts/parallel_threshold: Minimal number of user contexts
1953 that must be part of a scheduling round to switch to parallel
1954 execution mode (raw contexts only)
1955 * Fix bugs that prevented to use suspend/resume along with
1956 synchronization structures.
1957 * Fix bugs in process termination that lead to invalid memory access
1958 in very specific conditions.
1961 * Introduce a parallel mode for the models (controlled by surf/nthreads
1962 configuration item). In our tests, running the models in parallel
1963 never lead to any speedups because they are so fast that the gain
1964 of computing each model in parallel does not amortizes the
1965 synchronization costs, even when ultra fast futexes are used.
1966 This is released anyway because YMMV.
1969 * Performance boost by using a swag internally to compute the set of
1970 tasks that are finished and should constitute the return value of
1974 * Enable it by default now that it is considered rather stable.
1977 * Documentation of the tracing functions.
1978 * Performance gains when tracing categorized/uncategorized resource
1979 utilization by avoiding calls to get route when updating resource
1980 variables. LMM constraints are being used instead.
1981 * API changed to set task categories. Use MSG_task_set_category instead
1982 of TRACE_msg_set_task_category, and SD_task_set_category instead
1983 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1984 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1985 * Examples for link user variables added (see at examples/msg/tracing/)
1986 * Deprecated function TRACE_msg_set_process_category completely removed
1987 * Trace header updated according to the latest Paje file format
1988 * Tracing network lazy updates, no longer obligate users to use full updates
1989 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1990 * Experimental: let user code declare/set/push/pop application states for hosts
1991 * API changed to allow the manual creation of graph configuration files
1992 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1995 * Improve the API of Lua MSG bindings, using the Lua spirit.
1996 * Each simulated process now lives in its own Lua world (globals are
1997 automatically duplicated). It helps writing simulators. Will allow
1998 to run Splay programs within SimGrid in the future.
1999 * Add a Chord example in Lua, equivalent to the MSG one.
2002 * Start the implementation of a solution to express temporal
2003 properties, not only local assertions. This is still an
2004 experimental work in progress, stay clear from it to be safe.
2008 - Add new runtime parameters --help-logs and --help-log-categories
2009 to display information about supported logging parameters and
2011 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2012 don't exists anymore.
2013 * Mallocators: allow value NULL for the reset function.
2015 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2016 create homogeneous dictionaries, where all the elements share the
2017 same free function. Non homogeneous dictionaries will be
2018 deprecated in the next release.
2019 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2020 - Multi-level dictionaries are deprecated.
2022 - new function xbt_dynar_search_or_negative() that is useful when
2023 you have less than 2 million elements in your dynar and don't
2024 want of the extra complexity of catching exceptions when the
2025 element is not found.
2027 - Make xbt_os_thread module (for thread portability) public.
2028 Documentation is still to come, sorry.
2030 - Cleanups and simplifications to make it maintainable again.
2031 - Exotic features (such as memalign and valloc) were removed.
2032 - The metadata were extended and improved so that the
2033 model-checker becomes able to explore and inspect the heaps.
2034 - This may induce a performance drop when enable_model-checking is
2035 ON in cmake (even if it's not used in the simulation), but it is
2036 necessary at this point to get MC working.
2038 Turn model-checking OFF if simulation performance matters to you.
2039 Not enabling it at runtime is not enough, disable it in cmake.
2041 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2044 SimGrid (3.6.2) stable; urgency=low
2046 The "Not coding new stuff allows to polish old things" release.
2049 * New bindings to the NS3 packet level simulator (experimental)
2050 * Use the raw (efficient) execution contexts instead of the sysv
2051 (portable) ones when possible.
2052 * libpcre is now mandatory in any cases since not using it led to
2053 severe performance loss and possibly other issues
2054 * Update the XML platforms:
2055 - G5K: include the latest machine in Nancy
2056 - GridPP and LCG: new platforms
2057 * Documentation was partially updated, at least (more to come)
2059 Bug fixes, cosmetics and small improvements
2060 * Free terminated processes before the end of the simulation to avoid
2061 exhausting the memory of users having very dynamic amount of
2063 * Bug fix and cosmetics about canceling non-running tasks
2064 * Bug fix about the dot loader's issues when using libcgraph
2067 * Create an installer for windows with nsis (amd64 and win32)
2068 - Add an hello world project to illustrate simgrid project creation.
2069 - Embed libpcre into the Simgrid installer to avoid
2070 its compilation burden
2071 * The raw execution contexts should work on Apple now
2072 * Port to Windows 64 bits
2073 - Sysv contexts now have an implementation for this arch
2074 - GRAS communication features now support this arch
2075 * Drop support for borland compiler on windows
2076 - this code was not maintained, and we kinda depend on gcc nowadays
2077 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2078 * Fix portability issue on unstable ubuntu: linker became picky on
2081 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2083 SimGrid (3.6.1) stable; urgency=low
2085 The "Oops, we broke Macs too" release
2088 * Fixed contexts detection so that raw ones are used when possible
2089 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2090 leads to a strange error, with user code segfaulting sometimes when
2091 the generated code is not perfectly aligned (which is not
2092 controllable from the user side, depends on the amount of code)
2095 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2097 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2099 SimGrid-java (3.6) unstable; urgency=low
2102 * Split of every thing from simgrid v3.5 into a separate package.
2104 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2106 SimGrid (3.6) stable; urgency=medium
2108 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2111 * Bindings now constitute their own package, separated from the main one.
2112 Rationale: reduce our maintenance nightmare by reducing the module coupling
2113 They will soon be released on their own on gforge.
2115 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2116 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2118 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2119 * It's not quite deprecated for now because we have no replacement,
2120 but it may soon become the case.
2123 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2124 * Fortran: New user-level cache variable to store the rank of the running
2125 process. This improves performance by an order of magnitude.
2126 * C: New coccinelle script to automatically locate and modify global and
2127 local static variables.
2128 * Improved SMPI network model with a sender-side gap to account for multiple
2132 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2133 only say if a communication is finished, no matter whether it succeeded or
2134 failed. You can call MSG_comm_get_status() to know the status of a finished
2136 * New function MSG_task_dsend() to send a task and detach it. When a
2137 communication is detached, you are never notified of its success or failure
2138 and the memory is released automatically once it is finished. This function
2139 is useful when you don't care about the end nor the success of a
2141 * Change the prototypes of action replay. Sorry for inconvenience,
2142 but this is really more efficient this way (and to adapt your code,
2143 you just have to fix the initialization, that shouldn't be too long)
2144 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2145 ever managed to do anything useful with it.
2146 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2147 takes a dynar of communications. It returns immediately and gives the
2148 index of a finished communication (if any).
2149 * New example: a basic implementation of the Chord P2P algorithm.
2152 * New model for multi-core CPUs. You can now use the core attribute to
2153 precise the number of cores of a host. This is a basic model. Every
2154 process running on the host receives at most the power provided in
2155 the DTD (throughput<=power). Total throughput of process cannot exceed
2157 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2158 router linked by an up-link and a down-link (possibly asymmetrical).
2159 This kind of pattern allows to easily build last-mile model style platforms.
2160 Aggregating such patterns in a rule-based AS is thus the technique of
2161 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2162 * New model for Vivaldi routing. We transformed the Vivaldi network model
2163 into a Vivaldi routing model (based on the rule-based model). This allows to
2164 combine Vivaldi based latencies with last-mile platforms.
2167 * Added a check for NaN of IEEE754 infinite in the double entries of
2169 * Introduce a new context factory "raw", highly inspirated from the
2170 ucontext factory, but using manually crafted functions in assembly to
2171 do the work in an efficient manner.
2172 * Allow to change the used context factory at run time, not only at
2173 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2174 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2175 contexts at runtime (only with raw contexts or ucontexts).
2176 * Completely rewrote this module to allow parallel execution of user
2177 processes. Use --cfg=contexts/nthreads:N to execute user processes
2178 with N parallel threads (the default is 1, meaning no parallelism).
2179 * Allow to decide dynamically between sequential and parallel modes.
2180 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2181 processes in parallel only when their number is greater than or equal to P
2183 * Added a check for NaN of IEEE754 infinite in the double entries of
2187 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2188 won't output the state of processes when interrupted with Ctrl-C
2189 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2190 NULL-terminated array. This may solve backward compatibility issues
2191 due to the change to return type of SD_simulate. See also:
2192 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2193 * Add new macros with variable number of arguments.
2194 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2195 - in xbt/asserts.h: xbt_assert
2196 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2197 - in xbt/ex.h: THROWF and RETHROWF.
2198 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2200 * Change xbt_die() to accept a format string with arguments, just like printf.
2201 * New data structure: xbt_lib_t, like a dict but more general and with better
2205 * New configuration options
2206 Options triva/categorized and triva/uncategorized can be used to generate
2207 graph configuration files for Triva visualization tool.
2208 * Configuration option tracing/platform is renamed to tracing/categorized
2209 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2210 * New TRACE_user_link_variable interface:
2211 User provides the name of the link and the tracing variable to attach to it
2212 * the declaration of tracing categories must be done after the environment
2214 * simpler tracing interface, just one way to declare categories
2215 TRACE_category or TRACE_category_with_color, it is up to you
2216 * links in the trace file are again identified by their names
2217 * trace contains the full platform hierarchy exactly as declared using the ASes
2218 * Options tracing/msg/[task|process]:1 groups the process by hosts
2219 for both cases, tasks and processes must have names that are unique during the simulation
2220 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2221 * The experimental option tracing/msg/volume is deprecated
2222 its functionality may be reincorporated if needed
2224 The tracing generates a trace file with unordered timestamped events,
2225 because of the way the core simulator (surf) works. A script available
2226 at the tools directory (fix-paje-trace.sh) can be used to put the events
2227 in order. We have changed the tracing so it can generate ordered timestamped
2228 events in the final trace, but depending on the simulator (and how much time
2229 is simulated) that can lead to a huge memory utilization. It is deactivated
2230 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2232 Build Infrastructure
2233 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2234 - We are trying hard to keep the API stable, but it may happen that
2235 some things change (we're a research project after all, not a
2236 nuclear plant operating system). If such things should happen, you
2237 could rely on that macro to adapt.
2238 - current value: 30600 for 3.06.00, aka 3.6
2239 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2240 a number that can be compared with SIMGRID_VERSION.
2241 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2242 assertions and verbose/debug logging events are disabled at compile time.
2244 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2246 SimGrid (3.5) stable; urgency=medium
2249 * New feature to any SimGrid-based simulator: Model-Checking
2250 Check SIN#1 for more details.
2253 * New Model SMPI (three-interval linear regression for correction factors)
2254 See RR-7426, available at http://hal.inria.fr/inria-00527150
2255 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2256 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2257 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2258 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2259 * See also src/smpi/README
2264 - Tracing API changes: TRACE_start and TRACE_end should not be called
2265 by user-code. They are automatically called by simulators created
2266 with SimDAG, MSG and SMPI if the toolkit is compiled with
2267 tracing_enabled=ON. Categories declaration and utilization remain the
2268 same for MSG and SimDag.
2269 - A function was added to the tracing API to declare categories with
2271 - TRACE_category_with_color (char *category, char *color)
2272 where color must be in the following format
2273 "%f %f %f", red, green, blue
2274 and red, green, blue are float values in the interval [0, 1]
2275 - User can specify NULL as color parameter, or continue calling
2276 TRACE_category (cat)
2277 On that case, the tracing system will define random colors
2278 - The following command-line options are supported:
2279 --cfg=tracing/filename:msg.trace
2280 --cfg=tracing:1 (activate tracing, needed to use others)
2281 --cfg=tracing/platform:1 (categorized resource use)
2282 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2283 --cfg=tracing/msg/task:1 (task creation)
2284 --cfg=tracing/msg/process:1 (process creation, migration)
2285 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2286 --cfg=tracing/smpi:1 (SMPI interface tracing)
2287 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2288 - examples of examples/msg/tracing updated
2290 - DAXLoader and DOTLoader functions can generate tasks with categories
2291 - A new function to attribute a category to SD tasks:
2292 TRACE_sd_set_task_category (SD_task_t task, char *category)
2293 * Tracing the MPI interface implemented by SMPI
2294 - Collective operations are traced with states
2295 - Point-to-Point operations are traced with states/links
2296 - Tracing activated by a parameter "-trace filename" passed
2297 to smpirun during execution (considering that simgrid
2298 is compiled with tracing enabled)
2299 - To run the simulation with gdb, the simulator
2300 accepts --cfg=tracing/smpi:1 to trace SMPI
2301 - tesh files to check if smpi tracing is ok
2302 - See examples/smpi/NAS/DT-trace
2303 * GTNetS tracing re-worked
2304 - adaptation to the tracing system of GTNets to cope
2305 with modifications regarding the fullduplex mode
2306 - new tesh files to check if gtnets tracing is ok
2309 * Asynchronous communications through the functions:
2310 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2311 * New function: MSG_load_platform_script()
2312 to make possible using a lua script instead of XML files to set up platforms
2313 * New function: MSG_set_function
2314 to associate functions to processes, used when bypassing the parser
2315 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2317 Platforms: Add some more examples in examples/platforms
2318 * Grid'5000: see www.grid5000.fr
2319 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2322 * Change the XML format. This is a very important modification. SimGrid
2323 3.5 introduces a new hierarchical format based on the notion of
2324 Autonomous Systems. Compatibility with old format is ensured
2325 through the perl script provided in the install bin directory
2326 bin/simgrid_update_xml.
2327 It is now possible to build platforms with specific routing
2328 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2329 connect several platforms together. We will try to provide soon
2330 set of realistic platforms exploiting these properties (have a
2331 look at examples/platforms/ for the moment).
2332 * Take the opportunity of the XML format change to be a good XML citizen:
2333 rename link:ctn to link_ctn and similar changes (also dealt with by
2335 * Add a new routing scheme (rule-based) using regular expressions. It
2336 enables to have an extremely low memory footprint when the
2337 underlying routing is simple and can be compactly described. You
2338 need to have libpcre4-dev (perl regular expressions) installed if
2339 you want to use this routing scheme.
2340 * Revive the cluster TAG and allow to easily and efficiently (both in
2341 term of memory and speed) connect clusters together. Have a look
2342 at teshsuite/simdag/platforms/ to see how this can be done. With
2343 this tag, you can create clusters with thousands of tasks at no
2344 cost (have a look at examples/platforms/).
2345 Note: clusters are implemented as ASes, so there is no need for an
2346 enclosing AS tag if you have only one cluster in your platform.
2347 * Add new generic functions in the public interface that allows the user
2348 to call SURF 'create_resource' methods from your code (same
2349 functionality as the XML bypass mechanism but with a much lighter
2351 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2352 that uses a piecewise linear approximation to produce better
2353 results when exchanging small messages.
2354 * Add a new parameter to handle correctly full duplex link and account
2355 for interferences between uplink and downlink communications
2356 (activate with --cfg=fullduplex:1).
2359 * Rename the SD_READY (all dependencies are satisfied and task is
2360 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2361 dependencies are satisfied) state.
2362 This prevents a confusion between the notion of "ready to schedule"
2363 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2364 simulated" (SD_RUNNABLE) used by the simulation kernel.
2365 * Change the way a task is considered as ready. Instead of removing
2366 dependencies when a task is done, a counter is decreased. This way,
2367 it is always possible to reach ancestors thanks to the
2368 SD_taks_get_parents function (even after the end of the simulation.)
2369 * Change the return type of SD_Simulate from (SD_task_t*) into
2370 xbt_dynar_t. This function was in handling a dynar internally and
2371 converted it into a NULL terminated array for historical reasons.
2372 * New function SD_dotload(char*) to load a DAG described in dot
2373 format. This loader and the corresponding examples require the
2374 installation of the graphviz library.
2375 * Fix a bug in the management of tasks of size 0 in the surf network
2376 models. This problem was only visible with SIMDAG and you should
2377 thus disregard results produced with earlier versions if you
2378 relied on this feature (some tasks were blocked because of this).
2379 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2380 even though your applications were doing only point-to-point
2381 communications and sequential computations. Now you can really use any
2382 model you want (of course, if you create real parallel tasks, which are
2383 not implemented in most models beside ptaskL07, this will abort).
2384 * Add an example that schedules a DAX on an heterogeneous platform
2385 using a Min-Min strategy.
2386 * New function SD_workstation_get_current_task() that returns the kind
2387 of task currently running on a workstation in the sequential access
2389 * Raise some warnings when unexecuted tasks remains at the end of the
2390 simulation. This is usually caused by cycles in the DAG.
2393 * New function: SIMIX_process_set_function() called by MSG_set_function
2394 * Change the underlying waiting queue in semaphores so that a process
2395 can wait on several of them simultaneously (as in waitany).
2396 * Fix the way to handle tokens in semaphores so that all access patterns
2397 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2398 * kill the dirty pimple SIMIX_message_sizes_output()
2399 Please use (proper) visualization instead
2402 * New data container: setset (set of sets of elements)
2403 * New module: mmalloc (mapped malloc, allowing to have several
2404 independent segments of malloc)
2405 * New function: xbt_dict_cursor_set_data()
2406 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2407 * New function: xbt_dynar_is_empty()
2408 * New function: xbt_fifo_get_last_item()
2409 * Fix xbt_dynar_shrink(): use the right element size.
2410 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2411 uninitialized areas during expand.
2412 * Fix semaphores: previous implementation was severely broken.
2413 * Use library init/fini functions for our initialization.
2414 - you can use logs and other feature as soon as you want in your
2415 code (even before the xbt_init / MSG_init)
2416 - xbt_exit is now a no-op and produce a warning when used.
2419 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2420 benefit from the latest improvement to the simulation kernel.
2421 * Kill measurement sockets for now. If you rely on them, sorry. This
2422 release is not for you. This feature will be reintroduced in the
2423 future, but we cannot delay the release any further.
2424 * New function: gras_msgtype_get_name().
2425 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2426 * Fix (at last) the pmm example: it should not randomly fail anymore.
2428 Build chain: bug fixes and overall polishing
2429 * Cmake is now stable enough. Hence, we killed the autotools.
2430 * Port to windows ( TM :)
2431 * Fix the 'make install' target.
2432 No need to use 'make install-simgrid' anymore
2433 * Introduce a 'make dist' target compiling a *source* archive
2434 'make package' compiles a binary archive
2435 * Compile java files only on need
2436 * Add --cd and --setenv command line options to tesh
2437 * Out of source builds are not fully supported yet, but we are close
2438 * Enable supernovae and optimization flags by default for our users
2441 * Add layer to set up environment directly from lua, without XML.
2442 * The effect of gras_stub_generator can be achieved through
2443 lua too (check examples/gras/console/ping_generator.lua)
2445 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2447 SimGrid (3.4.1) stable; urgency=low
2449 The "Polishing easter eggs is probably a good idea" release.
2450 This is a bug fixes release only.
2454 * Fix a bug preventing the tasks from begin garbage collected.
2457 * Fix a bug occurring when a host involved in a communication fails.
2458 This was not detected properly by the other peer involved in the
2459 communication. Now, it's reported as a network error.
2462 * Warn the user about loop dependencies in data flow of DAX files
2463 * Obey the control-flow dependencies of DAX files
2466 * Add option "enable_smpi" allowing to not compile SMPI.
2467 Probably useful for the (Mac) users experiencing a build error here
2468 * Improve the detection of lua5.1 and ruby1.8
2470 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2472 SimGrid (3.4) stable; urgency=low
2474 The "Easter in Cargese" release. Also known as (major changes):
2476 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2477 ~> bindings were greatly improved
2478 ~> new tracing infrastructure for better visualization introduced
2480 * the "Welcome to configury modernity" release.
2481 ~> we switched from autotools to cmake, and improved our cdash
2484 A more detailled list of changes follow (full detail in svn log).
2486 Java Bindings: Various Cleanups
2487 * (install java-gcj-compat-dev on debian-like to use them)
2488 * Remove put/get: no need to export deprecated interface in Java
2489 Use send/receive instead.
2490 * Cleanup the examples and add a README per directory
2491 * Remove example autoDestination (that's the only way to go now)
2492 * Remove example explicitDestination (was a plain copy of basic)
2493 * Make JniException a runtime exception, so that there is no need to
2494 declare the fact that you may encounter such a beast. I guess that
2495 nobody will ever want to survive such error.
2496 * Create specific errors for each MSG case of failure:
2497 host failure, transfer failure, timeout, task cancelled
2498 * Cleanup the exceptions that may get thrown by each function
2499 * Other internal cleanups in Java bindings. Performance still bad :/
2500 Ruby and Lua Bindings: create them
2501 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2502 * That's new and great, you should try them out.
2503 Same functionalities than Java bindings, only even less polished
2505 * Kill the useless "rate" argument of SD_task_get_execution_time()
2506 Everyone used to provide -1 as a value, it was not used, and the
2507 semantic of a possible use wasn't even clear.
2508 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2509 either as comm costs or compute costs to mean that there is no
2510 such thing for that specific task.
2511 * Add a SD_task_set_name() function
2512 * Fix SD_task_unschedule() on typed tasks
2513 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2514 * In DAX loader, accept useless 'level' attributes to <job> since
2515 LIGO DAGs have them (seem to be to ease graphical representation).
2517 * Add an example masterslave_mailbox.c using send/receive and not
2518 the deprecated put/get interface.
2519 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2520 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2522 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2523 (declare MSG_USE_DEPRECATED to still have the old name)
2524 * Add a MSG_task_set_data() function
2525 * About trace replay (see examples/msg/actions):
2527 - Allow to work with splitted trace files for each process
2528 Give the specific trace file as argument of each process,
2529 and call MSG_action_trace_run(NULL)
2530 You can still have one merged file for all processes.
2531 - Fix implementation of collective operations
2532 * Allow task_execute() on 0-sized tasks (closes #10063)
2534 * This is the first release of SimGrid where SMPI is not considered
2535 beta anymore (even if some corners should still be improved)
2536 * Port over the new SIMIX_network submodule (internal refactoring)
2537 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2538 * Implement more missing elements of the standard:
2540 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2542 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2543 MPI_Scan() MPI_Get_processor_name()
2544 - Added implementation of missing case for Alltoall (warning: it's
2545 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2546 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2547 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2548 SMPI_MPI_Allgatherv()
2549 * Bug fixes include:
2550 - MPI_Waitsome() was broken
2551 - Allow relative includes in smpicc
2552 - Command line cfg argument 'reference_speed' was ignored...
2553 - Some functions did not properly lead to auto-benching of user code
2554 - smpicc passes -O2 by default (just like openmpi one)
2556 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2557 * Bug fixes about timeouts during communications
2558 * add SIMIX_message_sizes_output() as a pimple to write to file the
2559 amount of messages per size. Use gnuplot to get histogram.
2560 Pimple because that's the only user-visible function of simix,
2561 defined directly in xbt.h (irk, sorry)
2563 - Add a SIMIX_sem_get_capacity() function
2564 - Fix interactions with processe resume/suspende
2565 - release_forever() was stupidly broken
2566 - Fix SIMIX_display_process_status() for processes in a semaphore
2567 - Make SIMIX_sem_block_onto() user-visible
2568 * Refactoring context stuff:
2569 - Use pseudo-OOP for better modularity
2570 - reimplement SIMIX_process_kill() without process_schedule() so
2571 that the latter can take as invariant that it is called from
2573 - Merge context_start into context_new for sake of simplicity
2575 * Add a Vivaldi network model, coded live during SUD'10 ;)
2576 * Rename configuration variables to start a hierarchy:
2577 o cpu_model -> cpu/model
2578 o network_model -> network/model
2579 o workstation_model -> workstation/model
2580 * New configuration variables:
2581 o network/bandwidth_factor: correction to bandwidth
2582 o network/latency_factor: correction to latency
2583 o netwotk/weight_S: correction to the weight of competing streams
2584 * Add a long description to the models, that users can see with such
2585 argument on the command line: --cfg=cpu/model:help
2586 * --help-models display the long description of all known models
2588 * config: add the ability to set a default value after registration
2589 Does not override any previously set value (e.g. from cmd line)
2590 * dict: allow to have integer key and data.
2591 When so, you need to use the following functions
2592 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2593 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2594 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2595 In contrary to regular dicts, the key is not malloced before copy.
2596 Mixing scalar and regular elements in the same dict is not tested
2598 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2599 Tracing for Visualization:
2600 * SimGrid is now instrumented in order to generate a trace file for
2601 visualization analysis: to use it, need to compile SimGrid with the
2602 "tracing" option enabled, and instrument the program using SimGrid with
2603 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2604 (among other functions).
2605 * The instrumentation only traces the platform utilization for now
2606 * Documentation to use the tracing functions and how to analyze the
2607 traces with the Triva tool is written.
2608 * More information about: SimGrid FAQ (in the section Tracing Simulations
2611 * We moved to cmake as default build system. Autotools support will
2612 be dropped soon. Check the FAQ for more info about how to use it.
2613 * Greatly improved our cdash/ctest interactions
2614 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2615 * Added memory checking tests with valgrind; lot of memleak fixing.
2616 This may be the first release of simgrid with so few memory issues
2617 * Added code coverage tests.
2618 Our coverage is still improvable, but at least we see it on cdash.
2620 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2622 SimGrid (3.3.4) stable; urgency=low
2624 The "Desktop Grid needs love too" release (also called Xmas release).
2626 Models improvements:
2627 * Major speedup in the maxmin system solving by using lazy evaluation
2628 Instead of solving completely the maxmin system at each iteration,
2629 only invalidate (and recompute) the modified parts.
2630 This new feature is enabled in default models but you can try to
2631 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2632 * Cas01 IMproved as default CPU model
2633 This CPU model is the same Cas01 model, but it uses the
2634 maxmin-selective-update flag and a heap structure to manage
2635 actions on SURF kernel.
2636 It reduces the complexity to find the next action to finish and,
2637 consequently, it's faster than the old Cas01.
2638 This is the new default CPU model (Cas01).
2639 * Rename the old Cas01 model to Cas01_fullupdate
2640 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2641 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2642 * CpuTI (CPU Trace Integration)
2643 A new CPU model whose objective is simulate faster when using
2644 availability trace files.
2645 Instead of using a full featured, over engineered maxmin system for
2646 CPU modeling, this model does the pre-integration of traces files
2647 to calculate the amount of CPU power available, and so, executes
2648 faster than the old CPU models.
2649 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2650 * Use LV08 as default network model since it gives better accuracy
2651 for small messages and shouldn't change things for big ones.
2652 Use --cfg=network_model:CM02 to get the previous behavior.
2655 ******************************************
2656 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2657 ******************************************
2658 * The new CPU model may changes simulations!
2659 The point is that events occurring at the exact same timestamp
2660 are not scheduled in the same order with the old and new
2661 version. This may be enough to completely change the execution
2662 of simulations in some cases.
2663 * The new network model will change simulations!
2664 This new model is more realistic than the previous one, so you
2665 should consider redoing your old experiments with this model.
2666 Sorry for the inconvenience.
2669 * Introduce the supernovae compilation mode
2670 When compiled that way, the whole SimGrid (or almost) is put in a
2671 single compilation unit and compiled in one shoot.
2672 This is to help gcc which has difficulties to inline stuff from one
2674 The speedup seem to be above 15%, althrough more tests are needed on
2675 amd64 to confirm that gain.
2678 * Port of MSG's mailbox on top of SIMIX network
2679 The put/get mechanism was greatly simplified on the way.
2682 * New SIMIX network module. Provides:
2683 - Mailbox: rendez-vous mechanism to find with who you want to speak
2684 - Synchronous send/recv: easier and hopefully faster since the
2685 logic is handled in the maestro process directly now
2686 - Asynchronous send/recv: you dreamt of it? It's here now
2687 Too bad that nobody cared enough to propagate the change to MSG.
2688 * Add semaphores as SIMIX synchronization mechanism.
2691 * new function SD_daxload(char*) to load a DAX file
2692 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2693 * Introduce typed tasks. Specify its kind and cost at creation.
2694 At scheduling, just give where it should be placed, and the cost
2695 for each involved resource is automatically computed.
2696 Existing constructors so far (more to come of course):
2697 - SD_task_create_comm_e2e() for end-to-end communication
2698 - SD_task_create_comp_seq() for sequential computation
2699 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2700 * new function SD_task_dump() for debugging display
2701 * new function SD_task_dotty(task,FILE*) writing to file the info
2702 about the task in dotty format
2703 * SD_task_dependency_exists() can now cope with having one of its
2704 arguments NULL. If so, it tests whether the other argument has any
2706 * Add getters on list of preceding/following tasks:
2707 SD_task_get_parents(task) and SD_task_get_children(task)
2708 * Add getters on amount of workstations and list:
2709 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2710 * Add getter on task kind: SD_task_get_kind(task)
2711 * Update the start_time and finish_time of tasks on completion/failure
2712 * Bugfix: Remove task from state swags when destroyed
2715 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2718 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2719 dynar with one separate thread per value of the dynar.
2720 * Change the prototype of xbt_thread_create(), sorry.
2721 Added a boolean parameter indicating whether we want to join this
2722 thread (used in SG only for now)
2723 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2726 * GTNetS wrappers should now be usable again (and betterly tested too)
2727 * Fix a major regression from 3.2 where the timeout provided to
2728 MSG_task_put_with_timeout() was used as absolute time before which
2729 the comm should be done.
2730 * Start to fix the <cluster> tag.
2731 - Internal links should be good now (beside of the loopback, which
2732 use the private link instead)
2733 - paths to the external world is still rather broken
2734 - the <route:multi> tag is just broken. Actually that's brain-dead.
2735 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2736 to make it less stupid
2737 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2738 * Fix a source-level compatibility glitch from 3.2: after defining
2739 MSG_USE_DEPRECATED, you can use the old name
2740 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2741 * Allow to compile from the SVN with automake 1.11
2742 * Fix some problems when using the "start_time" tag in deployment XMLs.
2743 * Fix #8569: XBT/synchro.h has redundant declarations
2744 * Fix #8563: MSG return values and exceptions
2745 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2746 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2747 Thanks to Robson Peixoto.
2748 * Fix a vicious bug in dictionaries inducing that some elements were
2749 not freed on xbt_dict_free()
2751 Portability report of this version:
2752 * Main portability targets:
2753 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2754 on (amd64/i386/ia64)
2755 - mac leopard on i386
2756 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2757 but nothing critical.
2758 * Other platforms: windows, AIX and others were not tested for this release
2760 Timing report of this version:
2761 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2762 scenario parameters). From 8h to a few seconds in desktop grid settings.
2763 * Supernovae brings about 25% speedup on i386.
2765 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2767 SimGrid (3.3.3) stable; urgency=low
2769 The "Need for Speed" release.
2771 The timings done to validate the 3.3.2 were faulty.
2772 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2774 The problem was a conversion from a manually handled vector to
2775 xbt_dynar_t on the critical path.
2776 xbt_dynar_foreach calls functions, inducing stack management crap.
2778 We inlined these functions and xbt_dynar_foreach is now breath taking.
2779 We also inlined xbt_swag_belong on the way.
2781 Here are some approximate speedup measurements (on master/slaves
2782 simulations lasting between 10s and 20s each):
2783 3.3.1 -> 3.3.2: about same performance
2784 3.3.2 -> 3.3.3: 40% speedup
2785 3.3.1 -> 3.3.3: 40% speedup
2786 3.3.1 with inline patch -> 3.3.3: 30% speedup
2788 Our reading is that the refactoring which occurred in 3.3.2 made us
2789 suffer much more from the xbt_dynar_foreach low performance, but
2790 once we solved this, this refactoring proved to be very performance
2791 effective. From the 40% speedup, somehow, 10% are due to the
2792 inlining and 30% to the refactoring.
2794 That's a pitty that gcc cannot inline functions placed in other files
2795 alone. We have to choose between:
2796 - break the encapsulation (by putting private data structures and
2797 accessors in headers files to help gcc)
2798 - live with low performance
2799 - switch to a decent compiler such as icc (not quite possible).
2801 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2803 SimGrid (3.3.2) stable; urgency=low
2805 The "Simplicity does not preceed complexity, but follows it" release.
2807 The main contributors of this release were (lexical order):
2808 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2811 * Extract the routing logic into its own object.
2812 (was dupplicated in network.c and workstation_LV07.c;
2813 Allows to implement other ways of storing that info)
2814 => kill now useless network_card concept
2815 - Use dynar to represent routes (instead of void** + int*)
2816 - kill link_set (use surf_network_model->resource_set instead)
2817 - Add a command-line option to choose the routing schema to use
2818 - Add three new models:
2819 * Floyd (shortest path computed at initialization)
2820 * Dijikstra (shortest path recomputed all the time)
2821 * Cached Dijikstra (shortest path computed on need)
2822 All these models where contributed by Silas De Munck, and are
2823 described in his ICCS09 paper.
2825 * Simplify model declaration
2826 (less redirections, less function to write when defining a model)
2827 - Factorize stuff between models:
2830 surf_model_resource_set(model)
2831 surf_model_resource_by_name(model, name)
2832 - Unify the types of models in s_surf_model_t (using an union)
2833 - Embeed fields of common_public directly into s_surf_model_t
2834 - Rename model methods:
2835 action_free ~> action_unref
2836 action_change_state ~> action_state_set
2837 action_get_state ~> action_state_get
2838 - Change model methods into functions:
2839 (model)->common_public->action_use ~> surf_action_ref
2841 * Implement a generic resource; use it as ancestor to specific ones
2842 (allows to kill duplicated code in models)
2843 Drawback: timer command don't need no name nor properties;
2844 workstation_CLM03 don't need no properties
2845 (but I guess we can live with those few bytes wasted)
2847 * Improve the action object model
2848 - implement a constructor avoiding dupplicated code about field
2849 initialization in generic_action part.
2851 * Kill the SDP model: it has an external dependency, is deprecated
2852 in flavor of modern lmm models, and didn't compile since a while
2855 * Relocation of the context module from XBT to SIMIX.
2856 (the context were decoupled from the simix processes, duplicating a lot of code)
2857 => a lot of code was factorized
2858 - less overhead is introduced during scheduling
2859 - simpler API for the context factory
2860 - the logic for process creation,destruction and manipulation was simplified
2861 * Simplification of the s_smx_process_t data structure.
2862 => accesing the simix level data associated to a process is faster now,
2863 and the code is a lot more readable.
2866 * Implement some more MPI primitives:
2867 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2868 -implementation: Bcast: flat or 2-ary tree (default),
2869 Barrier: 4-ary tree,
2871 Allreduce: Reduce then Bcast
2872 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2873 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2874 Alltoallv: flat tree, like ompi
2876 * Add support for optimized collectives (Bcast is now binomial by default)
2877 * Port smpirun and smpicc to OS X
2880 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2881 Shout out if you used it.
2884 * Display the list of still queued messages in SG mode when existing
2888 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2889 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2890 * Add xbt_str_from_file(FILE*)
2891 * Add xbt_dict_get_key achieving a linear reverse search
2892 * Remove the context module
2894 Portability report of this version:
2895 * Main portability targets:
2896 - Linux(debian)/x86/context
2897 - Linux(debian)/x86/pthreads
2898 - Linux(debian)/amd64/context
2899 - Linux(debian)/amd64/pthreads
2900 On these, we still have the eratic breakages of gras/pmm and
2901 amok/saturate_sg reported in previous version. We still think
2902 that the tests are the cause of the fault, not the tested code.
2904 - Mac OSX Leopard/x86/context
2905 Still false negative in tesh autotesting.
2906 Smpi still fails, but this time because readlink does not accept -f
2907 Everything seems to work properly beside of that.
2910 - AIX version 5.3 (only tested contexts this time)
2911 Smpi still fails there because mktemp is not installed.
2912 Everything seems to work properly beside of that.
2914 I managed to compile it for the first time, but several breakages.
2915 Won't delay the release for this exotic platform.
2917 * Windows: it's still lagging behind. If you want to help, please
2920 Timing report of this version:
2921 This version seem to be more than 5% faster than 3.3.1 (on linux
2922 64bits with contextes). The gain is less than expected, we are
2923 investigating this for next release.
2925 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2927 SimGrid (3.3.1) stable; urgency=low
2930 * Implement a --cfg-help to show existing configuration variables
2931 * Build chain do not require doxygen in maintainer mode
2934 * fix a bug on struct sizeof computation, which prevented the
2935 exchange of arrays of structs in some conditions
2936 - added a regression test about this in datadesc_usage
2937 * Allow the exchange of 0-long dynamic vectors.
2938 - for that, use -1 as indicator of dynamic size instead of 0
2939 - This implied to change any size from unsigned long to long,
2940 reducing a bit communication abilities, but I guess that with
2941 64bits being quite common, this is more than enough.
2942 - This also induce a protocol change, thus bumping network protocol
2943 version from 0 to 1 (if we have external users, we have to get
2944 clean on that point too ;)
2945 - added two regression tests about this in datadesc_usage
2946 * Be more verbose when propagating local exceptions
2947 This helps debugging.
2948 * Display the status of simulated processes when receiving SIGINT in
2952 * Allow to control the simulation from a trace file.
2953 New functions MSG_action_register() and MSG_action_trace_run()
2954 The first one allows to associate a function execution to each
2955 kind of action while the second one parses a trace file and
2956 triggers the corresponding actions within the system.
2957 For now, only a toy example is provided in examples/msg/actions
2958 * Add an example of process migration in examples/msg/migration
2959 * Fix a bug in task exchange which broke MSG_task_get_sender()
2960 Add a teshsuite regression test for that.
2961 [Bug: if MSG_task_get_sender() is called after sender exit,
2963 * Fix a bug which prevented suspend/resume to work properly
2964 * Display the status of simulated processes when receiving SIGINT
2965 This fixes a regression of v3.3. due to the introduction of SIMIX
2966 * Bug fixing in failure management:
2967 - trace could not start by a failure at time 0
2968 - failure during communications were not working
2971 * Add SIMIX_process_set_name() to change the name of the current
2972 process in the log messages.
2973 * Store smx_hosts in a dict since we only retrieve them by name
2974 * Move the configuration infrastructure to surf
2977 * Move the configuration infrastructure to surf
2980 * Massive internal cleanups:
2981 - Store internal structures on processes instead of hosts (allows
2982 to have more than one process per host, in addition of being more
2984 - Cleanup the initialization/finalization process
2985 - Kill a whole bunch of unneeded synchronization:
2986 processes run in exclusive manner within the simulator
2987 - Move queues from global tables to process data fields
2989 - now accept -platform and -hostfile arguments
2990 - Pass the right rank value to processes according to the hostfile
2991 * Compile the examples by default, and use them as regression tests
2992 * Implement MPI_Wtime()
2993 * Change the reference speed to a command line option
2996 * TCP_gamma can now be specified as command line option using
2997 --cfg=TCP_gamma:10000000.0
2998 * Change the --surf-path cmd line option into --cfg=path:
3001 * Also include strbuff from xbt.h public header
3002 * xbt_ex_display(): do not free the exception after displaying
3003 This allows to do more with the given exception afterward.
3004 Users should call xbt_ex_free() themselves.
3008 Portability report of this version:
3009 * Main portability targets:
3010 - Linux(debian)/x86/context
3011 - Linux(debian)/x86/pthreads
3012 - Linux(debian)/amd64/context
3013 - Linux(debian)/amd64/pthreads
3014 These targets fail about 1/10 of times on gras/pmm, but we believe
3015 that this is because of the test, not because of simgrid.
3016 amok/saturate_sg fails even more rarely, and the test may not be
3019 - Mac OSX Leopard/x86/context
3020 The test suite still spits tons of errors because some obscure
3021 force prevents us from removing the temporary directories
3022 arguing that they still contain some metadata I've never heard of.
3023 Smpi fails because seq is not installed.
3024 Everything seems to work properly beside of that.
3027 - AIX version 5.3 (both contexts and pthread)
3028 Smpi still fails there because mktemp is not installed.
3029 XML inclusions seems rosty on AIX.
3031 * Windows: it's still lagging behind. If you want to help, please
3034 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3036 SimGrid (3.3) stable; urgency=high
3040 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3041 [Malek Cherier & Mt]
3043 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3046 Until now, GRAS were using MSG as an interface to SURF. It was
3047 quite difficult because both interface have several differences
3048 (MSG channels vs GRAS sockets were the most notable point).
3050 This also opens the gate to SMPI (which should occur soon) and speed
3051 up simulations by to 40% (even if it were not the main goal).
3053 **************************************
3054 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3055 **************************************
3056 The point is that events occurring at the exact same timestamp are
3057 not scheduled in the same order with the old and new version. This
3058 may be enough to completely change the execution of simulations in
3059 some cases. Sorry for the inconvenience.
3061 * Cleanup and upgrade the XML format to push further scalability
3062 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3064 * Improve the testing infrastructure with tesh. Now a very large part of
3065 the code is tested not only by being run but also by checking that the
3066 output match an expected output [Mt].
3068 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3069 is really less memory-demanding, which should allow you to use
3070 larger files in SimGrid [AL].
3072 * Inform valgrind about our contextes, so that it becomes usable
3073 with the default (and more effecient) version of SimGrid
3074 [contributed by Sékou Diakite, many thanks]
3077 * Introduce a listener thread in charge of receiving incoming
3078 messages from the network. It allows to overlap communication and
3079 computation but most notably, it removes some stupid deadlocks due
3080 to the fact that so far, a process could not send and receive at
3081 the same time. This made most non trivial communication schema
3083 * Convert the PIDs from long int to int to match the MSG ones (and
3084 linux ones too) [Mt]
3085 * New function: gras_agent_spawn() to launch a new process on
3086 current host. Only working in simulation for now. [Mt]
3087 * New function: gras_os_hostport() returning a constant form (ie,
3088 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3091 * Make the backtrace of exceptions more human readable [Mt]
3092 * New module: xbt/str [Mt]
3093 a ton of string utility functions (split, join, printf to a newly
3094 allocated buffer, trim, etc)
3095 * New module: xbt/hash [Mt]
3096 SHA1 hashing algorithm (more to come if needed)
3097 * New module: xbt/synchro [Mt]
3098 synchronization tools (mutex and conditions) working the same way
3099 in simulation and in real life (mainly useful for GRAS, but not
3101 * New module: xbt/queue [Mt]
3102 classical producer/consumer synchronization scheme
3103 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3104 (using the classical functions will get serialized) [Mt]
3105 * Make dictionary internal table dynamic. No need to specify its size
3106 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3108 * Make sure the log channels are organized as a tree under windows
3109 (because of ANSI C compatibility issue, any channel were child of
3110 root directly) [Mt].
3113 * Cleaned many thing in surf and fixed a few bugs [AL].
3114 * Add a nice command line configuration mechanism to compose models [AL].
3115 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3116 the previous one (KCCFLN05). It relies on something that looks like
3117 a max-min sharing mechanism but cannot be written as such. A new solver
3118 was thus designed [AL].
3119 * Add a new solver to lmm. Based on Lagrange optimization and
3120 gradient-based descent, it enables to efficiently maximise systems s.a
3122 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3124 This solver enables to propose two new network models for TCP Reno and
3125 TCP Vegas based on Low's work. These models still need to be fully
3126 tested though [Pedro Velho].
3129 * Bug fix in SD_simulate. Now the time bound given as argument is
3131 * Use the new parallel task model (ptask_L07) as default.
3132 * Use the SURF command line configuration mechanism.
3133 * 0-size tasks (for synchronization) should now work.
3135 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3137 SimGrid (3.2) stable; urgency=high
3141 We still experience issues on this platform, but we believe that at
3142 least MSG is usable.
3144 GRAS API BREAKAGE (for simplification purpose, sorry):
3145 * the gras_msgtype_by_name is not used anymore. Instead of
3146 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3147 you can write (and must)
3148 gras_msg_send(toserver, "request", &request);
3149 - If you still want to pass a gras_msgtype_t to the function (to cache
3150 the type and avoid the lookup time), use the gras_msg_send_() variant.
3151 - Impacted functions:
3152 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3153 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3154 * The callbacks are now expected to return 0 when everything went well
3155 (just like the main() function)
3157 GRAS new features and improvements:
3158 * New module mechanism where user code can use per process globals [Mt]
3159 This is similar to gras_userdata_*() functions, but for libraries. It
3160 factorize some code developped over and over in the examples and AMOK.
3161 It has still to be documented and used (only amok/peermanagement is
3163 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3164 loss when they were small enough to fit into the buffer and sent quickly
3165 enough so that they can all get received in one shoot.
3166 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3167 if not found. Use the *_or_null() variant for the old semantic.
3168 * In gras_msg_handle, do not discard messages without callback.
3169 They are probably messages to be explicitly awaited later (ie, proofs of
3170 mis-synchronization in userland since they are sent before being awaited)
3171 No big deal usually.
3172 * gras_socket_meas_send/recv: semantic changed!
3173 The numerical arguments used to be (1) the total amount of data to send
3174 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3175 messages. This was need for the fool willing to send more than MAXINT
3176 bytes on quite fat pipes.
3179 * Do really rename the hostmanagement module to peermanagement. [Mt]
3180 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3181 but this is rather new and this was documented in the module
3182 documentation (poor excuses, I admit)
3183 * Bandwidth measurement semantic changed! This follows the changes to
3184 gras_socket_meas_send/recv explained above.
3187 * A sequential mode has been added to the workstations. When a workstation
3188 is in sequential mode, it can execute only one task, and the other tasks
3189 are waiting in a FIFO. [Christophe Thiery]
3192 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3193 model for SIMDAG. [Christophe Thiery]
3194 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3195 equal to 0, which caused some bad behaviors in
3196 saturated_constraint_set_update. I now use a threshold mechanism like in
3200 * When running manually src/testall, you select specific units [Mt]
3201 testall is the result of our cunit mechanism, and should replace all
3202 the scripty thingy around since bash don't run easily on billware.
3204 * A mallocator system has been added. [Christophe Thiery]
3205 Mallocators allow you to recycle your unused objects instead of freeing them
3206 and allocating new ones.
3208 Documentation update:
3209 * FAQ reworking + New FAQs:
3210 - "Valgrind spits tons of errors!" [Mt]
3211 - "How to repport bugs" [Mt]
3212 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3213 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3214 - Communication time measurement within MSG [AL]
3215 - I experience weird communication times when I change the latency [AL]
3216 * GRAS tutorial [Mt]
3218 - an introduction to the framework and to the used communication model
3219 - an initiatic tour introducing the most proheminent features:
3221 . Lesson 0: Installing GRAS
3222 . Lesson 1: Setting up your own project
3223 o Part 2: Message passing
3224 . Lesson 2: Exchanging simple messages
3225 . Lesson 3: Passing arguments to the processes (in SG)
3226 . Lesson 4: Attaching callbacks to messages
3227 . Lesson 5: Using globals in processes
3228 . Lesson 6: Logging information properly
3229 . Lesson 7: Using internal timers
3230 . Lesson 8: Handling errors through exceptions
3231 . Lesson 9: Exchanging simple data
3232 . Lesson 10: Remote Procedure Calling (RPC)
3233 . Lesson 11: Explicitly waiting for messages
3234 . Recapping of message passing features in GRAS
3235 - A HOWTO section containing:
3236 o HOWTO design a GRAS application
3237 More are due, of course. They will come latter. In the meanwhile, you can
3238 check the examples which are still here.
3240 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3242 SimGrid (3.1) stable; urgency=high
3246 There was a stack corruption somewhere, visible only when optimizing
3247 with these versions. [Vince]
3250 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3251 module that enabled to play with Directed Acyclic Graphs. It is built
3252 directly on top of SURF and provides an API rather close to the old
3253 SG. Some old codes using SG are currently under rewrite to check that
3254 all needful functions are provided. [Christophe Thiery]
3257 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3258 extension of the classical CLM03 model that gracefully handles
3259 failures. This is now the default model for MSG and GRAS. It doesn't
3260 handle parallel tasks yet though. [AL]
3261 * Bug fix: Weights were not correctly set in the network part.
3262 WARNING: This may have resulted in incorrect results with simulations
3263 where there are more than one flow on a given link. [AL]
3266 * After a (long ?) discussion on simgrid-devel, we have decided that the
3267 convention we had on units was stupid. That is why it has been decided
3268 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3269 WARNING: This means that all previous platform files will not work as
3270 such with this version! A warning is issued to ask users to update
3272 A conversion script can be found in the contrib module of the CVS, under
3273 the name contrib/platform_generation/surfxml_update.pl [MQ]
3276 * Bug fix: Processes were started in reverse order, wrt deployment file.
3277 WARNING: if your code relies on this bug, please fix it. [AL]
3278 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3279 being executed on two different locations. [AL]
3280 * Bug fix: Failures are now better supported thanks to Derrick's tests
3281 (there was many failure situations I hadn't thought of and that weren't
3282 correctly handled). [AL]
3283 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3287 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3288 Exception occurring on server-side are propagated back to client (!).
3290 API CHANGE: the callback changed their prototype. Change:
3291 int my_handler(gras_socket_t expeditor, void *payload_data) {
3293 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3294 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3296 * New! function: gras_msg_handleall to deal with all messages arriving
3297 within a given period.
3298 * New! function: gras_socket_server_range to get a server socket in a
3299 range of port numbers (ease to avoid port number conflicts) [MQ]
3300 * New! gras processes display their backtrace when they get a SIGUSR1
3301 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3302 Sweet to debug RL processes [MQ]
3306 - Do not force experiment sizes to be expressed in kb, or it becomes
3307 impossible to measure the latency this way (needs one byte-long tests)
3308 WARNING: this changes the amok_bw_* function semantic. [MQ]
3309 - Implements the link saturation stuff. [MQ]
3310 * Peer management module:
3311 New! module factorizing code that we wrote over and over [MQ].
3314 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3315 - Test units are placed directly into the library code, they get extracted
3316 automatically and placed into the src/testall binary.
3317 - Convert most of the XBT tests to this system.
3318 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3319 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3320 live in the same C code. This required to change a little bit the API
3321 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3322 * New module: sparse graph structure with basic algorithms (this is work
3323 in progress and the API is not considered to be frozen yet). [AL]
3324 * Display more information on backtraces: source line & function names are
3325 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3326 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3327 * Reworked a little bit some #include statements to load only required
3328 headers. Some user code that relied on SimGrid to include stdlib or
3329 stdio may need to include it by themselves. [AL]
3330 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3332 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3333 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3334 but I guess that almost nobody used those parts. [MQ]
3336 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3338 SimGrid (3.0.1) stable; urgency=low
3341 * Unfortunately, I had missed 5 misnamed functions:
3342 xbt_fifo_item_t xbt_fifo_newitem(void);
3343 void xbt_fifo_freeitem(xbt_fifo_item_t);
3344 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3345 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3346 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3347 They're now deprecated. Please use their new versions:
3348 xbt_fifo_item_t xbt_fifo_new_item(void);
3349 void xbt_fifo_free_item(xbt_fifo_item_t);
3350 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3351 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3352 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3354 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3355 * Documentation: xbt_log module unmercifully reworked [MQ]
3356 * Bugfix: there was a problem with the ending of contexts with
3357 the pthread backend. It caused some weird deadlock or behavior
3358 depending on the pthread implementation. [AL]
3359 * Bugfix: get the exceptions raised in the simulator repport where
3360 and why they come from when they are not caught in time [AL, MQ]
3363 * Bugfix: Do repport the error when two non-connected hosts try to
3364 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3367 * Add additionnal checkings on communications. Assert that two
3368 communicating hosts are connected by a set of links... [AL]
3371 * Add additionnal checkings on channel values in communication [AL]
3372 * New: MSG_task_get_source to see on which host a task was generated [HC]
3373 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3374 the number of tasks waiting to be received on channel and sent
3376 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3377 waits for the first task coming from a given host.. [AL]
3379 GRAS new functionnalities: [MQ]
3380 * Enhance the parsing macro to allow the size of multidimentional objects
3381 to be given thru annotations.
3382 * New example (and documentation): Matrix Multiplication a la RPC
3383 (as when I was young!) and fix a bunch of bugs found on the way.
3385 GRAS performance improvements: [MQ]
3387 * Reduce the amount of cbps creation/destruction by making it static to
3388 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3391 * Change libdata to a set so that we can search for stuff by ID (and thus
3392 reduce the insane amount of dict lookups)
3395 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3396 RL communication on top of SG since it's so uneffective.
3397 It may also be used for inter-thread communication in RL, one day.
3398 * Use gras_datadesc_copy() to exchange messages on top of SG
3400 - improve message exchange performance on top of SG
3401 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3402 * Don't exchange on the network the size of the used part of buffer,
3403 instead, specify the possible buffer size to read().
3405 - reduces the amount of read/write calls (one pair per exchange)
3406 - reduces the amount of exchanged data (the size)
3407 - allows to retrieve all arrived data on receiver side, if we don't need
3408 it right now (subsequent read will peek the buffer)
3409 - allows the receiver to proceed with the beginning of the stream before
3410 everything is arrived
3411 - make it possible to build an iov transport (using readv/writev)
3413 - take care of the data with non-stable storage (like stacked data),
3415 * If possible, TCP send uses vector I/O (when writev() is here)
3416 - Don't use it for receive since we send data sizes and data on the
3417 same stream, so we wouldn't be able to chain large amount of chunks
3418 before having to flush the stuff to read the size.
3419 * Rework the transport plugin mechanism to simplify it and reduce the
3420 amount of pointer dereferencement when searching for the right function
3423 * I guess that now, we do almost as few system calls as possible while
3424 doing as few data copy as possible.
3426 To improve it further, we could try to send all the sizes first and then
3427 all the data (to use iov on receiving size), but it's only a partial
3428 solution: when you have 2 dimensional data, the sizes of the second
3429 dimension is data of the first dimension, so you need 3 streams.
3431 I'm not sure the potential performance gains justify the coding burden.
3433 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3435 SimGrid (3.00) stable; urgency=high
3438 * New! Give the possibility to hijack the surf parser and thus bypass
3439 MSG_create_environment and MSG_launch_application. Have a look at
3440 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3442 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3444 SimGrid (2.96) unstable; urgency=low
3449 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3450 This deprecates the xbt_error_t mechanisms.
3451 It modifies (simplifies) all XBT and GRAS API.
3452 MSG API keeps unchanged (exceptions raised by XBT are caught from
3453 within MSG and masked with existing error handling facilities)
3456 * New! Add a FATPIPE model. [AL]
3457 * New! Add a parallel task model. [AL]
3458 * New! Add automatically a loopback interface (in the default
3459 network model) if none was precised.
3462 * Bugfix: MSG_process_resume now works with the current running process.
3464 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3465 * Modification of MSG_task_get_compute_duration. Once a task has been
3466 processed, the value returned by this function is now equal to 0. [AL]
3467 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3468 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3469 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3470 of a m_task. For now, MSG_task_cancel only works with computation
3472 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3473 of the processor (in Mflop/s) regardless of the current load on the
3475 * API Change: use proper naming convention for MSG_getClock and
3476 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3478 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3479 This function changes the priority of a computation task. This priority
3480 doesn't affect the transfer rate. A priority of 2 will make a task
3481 receive two times more cpu power than the other ones. This function
3482 has been added to suit the needs of Nguyen The Loc and hasn't been that
3483 much tested yet. So if it fails, please report it and send me your code.
3485 * API Change: removed all functions and types that were marked "deprecated"
3486 since many months. Renamed MSG_global_init_args to MSG_global_init.
3488 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3490 SimGrid (2.95) unstable; urgency=low
3493 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3494 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3495 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3498 * Bugfix: Allow absolute paths for platform description files [MQ]
3499 * Bugfix: do free the variables after use. Leads to drastic performance
3501 * Implement max_duration (ie, timeouts) on resources [AL]
3504 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3506 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3507 the needed support (timeouts) [AL]
3510 * Implement measurement sockets. You can now get the bandwidth between two
3511 hosts thanks to AMOK (see below). [MQ]
3512 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3513 dynar over the network (yeah) [MQ]
3514 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3515 * simple/static token-ring example. [Alexandre Colucci and MQ]
3516 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3517 in place before [MQ]
3520 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3521 type parsing. "short" and "long" modifiers were also ignored; other
3522 modifier (such as reference level) are still ignored. [MQ]
3523 * Embeed the buffer size within the buffer itself on SG. [MQ]
3524 That way, send() are atomic and cannot get intermixed anymore (at least
3525 the ones which are less than 100k; bigger messages still have the issue)
3526 * Array size pushed by the field, not by the field type (or each
3527 and every long int will push stuff to the cbps) [MQ]
3528 * use select() to sleep since it allows to portably sleep less than one
3531 GRAS (minor cleanups)
3532 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3535 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3536 * type category 'ignored' killed as it was never used and were difficult
3538 * whether a type can cycle or not is now a flag, leaving room for more
3539 flags (as "ignored", if we feel the need one day ;)
3540 * Rename raw sockets to measurement sockets since "raw" has another
3541 meaning in networking community.
3544 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3545 and offers features not belonging to GRAS but that most applications
3546 need. One day, it may be a set of plugins loadable at runtime.
3547 * New module: bandwidth
3548 bandwidth measurement between arbitrary nodes running this module. [MQ]
3550 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3552 SimGrid (2.94) unstable; urgency=low
3554 The first beta release of SimGrid 3 !
3558 * Update the main page and the FAQ. Adding references to gforge.
3561 * Add a gras_os_getpid function.
3564 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3565 * Extend the logs so that they also print PID, hostname, date, ... if
3567 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3568 and kill the old version which were in testsuite/
3569 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3573 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3574 for users. Instead, each programming environment (GRAS, MSG,...) use it
3575 when needed to provide such a feature to users.
3576 Don't shortcut the mechanism or you will also shortcut the virtualization
3577 you need on the simulator.
3581 * Cleanups in configury with regard to compile optimization/warning flags.
3582 Also add -fno-loop-optimize to any powerpc since it's the optimization
3583 killing gcc (< 3.4.0).
3584 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3585 needed by MSG examples complications
3586 * Borrow configury beautifications from PHP
3589 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3590 warning (thanks loris for stumbling into this one).
3591 * Bugfix: stop loading private headers (gras_config.h) from the public
3595 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3596 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3597 RL would imply a DNS resolver, which is *hard* to do in a portable way
3598 (and therefore delayed).
3599 * Implement a real timer mechanism and use it in timing macros. This allows
3600 to avoid rounding errors and get a 0.000005 sec precision in timing
3601 macros. While I was at it, various cleanups:
3602 - allow to declare more than one timed section per file (fix a stupid bug)
3603 - move some private declaration to the right place
3604 - merge conditional execution and timing macros into emulation module
3605 - document the module
3606 - make sure the module cleanups its mess on gras_exit
3607 * Documentation improvements:
3608 - (new) how to compile applications using GRAS
3609 - (new) emulation support (timing macros)
3611 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3613 SimGrid (2.93) unstable; urgency=low
3615 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3618 - Use Paje properly where used. Still to be sanitized properly.
3619 - Portability fix: Add an implementation of the contexts using pthread
3623 - Add xbt_procname(): returns the name of the current process.
3624 Use it to show the current process's name in all logging.
3626 - fix detection of older flex version and the reaction, since we do
3627 depend on modern ones (we use lex_destroy)
3628 - Better separation of SG and RL in the libs: remove all simulation code
3629 from libgras. As a result, this lib is now only 200k when stripped.
3630 Some of the xbt modules may also be duplicated (two sets and such) and
3631 should be cleaned/killed before SG3.
3632 - Insist on using xlC on AIX because of weird problems involving gcc there.
3633 - Cleanup the make remote stuff. This is now done by scripts
3634 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3635 mainly for our private use, but we're working on changing them to user
3638 - Bugfix: flush the socket on close only if there is some *output*.
3639 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3641 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3644 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3646 SimGrid (2.92) unstable; urgency=low
3648 Alpha 3 on the path to SimGrid 3
3652 - New! First try of benchmarking macros.
3653 - New! First try so that gras_stub_generator generate deployment and
3654 remote compilation helpers.
3656 - Bugfix: Initialization fix in msg_test.
3660 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3662 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3664 - API Change: gras_os_sleep to take the amount of seconds as a double.
3665 Accepting an int was error prone since it was the only location where
3666 seconds were coded as such. It leaded to damn rounding errors.
3667 - Bugfix: Hard to belive that timers ever worked before this.
3669 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3671 SimGrid (2.91) unstable; urgency=low
3673 Alpha 2 on the path to SimGrid 3
3677 - Bug fix in the lmm_solver.
3679 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3680 through the function MSG_paje_output.
3681 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3682 - It is possible to bound the rate of a communication in MSG with
3683 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3684 to write it in the changelog).
3685 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3690 - Various cleanups to the autotools stuff
3691 - Begin to move Gras examples to examples/gras/
3692 - Let make distcheck work again (yeah!)
3694 - documentation overhauled using doxygen.
3695 gtk-doc-tools is dead in SimGrid now.
3696 - Automatically extract all existing logging categories, and add the list
3697 to the documentation (long standing one, to say the less)
3699 - Cleanup the known architecture table. Reorder the entries to group what
3700 should be, and use a more consistent naming scheme.
3701 (some of the test dataset are still to be regenerated)
3702 - New! Allow library to register globals on each process just as userdata
3704 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3705 lookup time (for now).
3706 Use it in msg and trp.
3707 This cleans a lot the internals and helps enforcing privacy of the
3708 headers between the gras components.
3709 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3710 - Bugfix: gras_os_time was delirious in RL.
3711 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3713 - Reenable GRAS now that it works.
3715 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3717 SimGrid (2.90) unstable; urgency=low
3719 Alpha 1 on the path to SimGrid 3
3721 * It is a long time since the last release of SimGrid. I'm sorry about
3722 that but as I had told you, I was rewriting a lot of things. I apologize
3723 to those who had been reporting bugs to me and that I had not answered.
3724 If your bug is still in the new version, please tell me. Here is a
3725 summary of the main changes.
3727 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3728 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3729 portability, speed, and a lot more but you'll figure it out later.
3730 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3731 I wanted to release the new MSG as soon as possible and I have
3732 broken GRAS, which is the reason why, for now, only MSG is fully
3733 functional. A laconic description of these projects is available
3734 in the documentation.
3736 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3737 kernel optimized for our needs (called SURF but only the developers
3738 should use it). Hence, MSG is now roughly 30 times faster and I think
3739 that by rewriting a little bit MSG, I could event speed it up a little
3740 bit more. Beside the gain in speed, it is also much easier to encode a
3741 new platform model with SURF than it was with SG. More to come...
3743 * REVOLUTION 3: I have tried to change a little as possible the API of
3744 MSG but a few things really had to disappear. The main differences
3745 with the previous version are:
3746 1) no more m_links_t and the corresponding functions. Platforms are
3747 directly read from a XML description and cannot be hard-coded
3748 anymore. The same format is used for application deployment
3749 description. The new format is described in the documentation.
3750 Have a look in tools/platform_generation. There is a tiny script
3751 that converts from the old platform format to the new one. Concerning
3752 the application deployment format, parsing the old one is tricky.
3753 I think most of you should however be able to convert your files. If
3754 it is really an issue, I can write a C code that does the conversion.
3756 2) the toolbox tbx does not exist anymore. We now have a library
3757 with much more data-structures but without the hash-tables (we have
3758 dictionaries that are much faster).
3760 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3762 *****************************************************************************
3763 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3764 * don't feel like converting the oldies. (Mt) *
3765 *****************************************************************************
3768 Version 2.90: "the long awaited one"
3769 - Finished rewriting and debugging MSG. Rewrote the documentation.
3770 - disable GRAS for now since it needs to be ported to the newest SG
3773 - Finish the port to windows (using mingw32 for cross-compile)
3776 - Main loop and datastructures of SURF. A cpu resource object is
3777 functional. Surf can thus be used to create cpu's with variable
3778 performance on which you can execute some actions.
3780 2004-11-15 Martin Quinson
3781 - Port to ARM. Simply added the alignment and size descriptions. Should
3782 work, but the ARM machines are so slow that I didn't had the opportunity
3783 to 'make check' over there yet.
3785 2004-11-15 Arnaud Legrand
3786 - Trace manager now written. It uses a heap structure and is therefore
3787 expected to be efficient. It may however be speeded up (particularly
3788 when many events occur at the same date) by using red and black
3789 trees. One day maybe...
3790 - Max-min linear system solver written. It uses a sparse matrix
3791 structure taking advantage of its expected use. Most operations are
3792 O(1) and free/calloc are called as few as possible. The computation of
3793 the minimum could however be improved by using a red and black tree
3796 2004-11-03 Arnaud Legrand
3797 - Rename every gras_* function that was in xbt/ to its xbt_
3799 - Add a heap and a doubly-linked list to xbt
3800 - Added a dichotomy to the dictionaries. make check works as well before
3801 so I assume that the patch is correct. I do not know however if things
3802 run effectively faster than before now. :)
3804 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3805 SimGrid, and the version number is bumped to 2.x
3807 2004-10-29 Martin Quinson
3808 - Introduction of the remote errors.
3809 They are the result of a RMI/RPC on the remote machine.
3810 ErrCodes being scalar values, you can't get the host on which those
3811 errors did happen. Extending the error mechanism as in Gnome is possible.
3812 No idea yet whether it is a good idea.
3814 2004-10-28 Martin Quinson
3815 - Interface revolution: the Starred Structure Eradication.
3816 I used to do typedef struct {} toto_t; and then handle *toto_t.
3817 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3818 - ???_t is a valid type (builded with typedef)
3819 - s_toto_t is a structure (access to fields with .)
3820 - s_toto is a structure needing 'struct' keyword to be used
3821 - e_toto_t is an enum
3822 - toto_t is an 'object' (struct*)
3824 typedef struct s_toto {} s_toto_t, *toto_t;
3825 typedef enum {} e_toto_t;
3826 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3827 s_toto_t) is private.
3829 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3830 so that it can set the variable to NULL. It was so for dicts and sets,
3831 it changed for dynars.
3833 - Fix a bunch of memleaks in dict_remove
3834 - Fix a bug in sg/server_socket opening: it failed all the time.
3836 2004-10-07 Martin Quinson
3837 - Speed up dynar lookup operation a bit.
3839 gras_dynar_get is dead.
3841 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3842 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3843 (which gives you the address of the stored data).
3845 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3846 copy of the data using an affectation to do the job and not a memcpy.
3848 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3849 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3851 It does not really speedup the dynar test because they are
3852 setting elements all the time (and look them seldom). But the dict does
3853 far more lookup than setting.
3855 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3857 2004-10-05 Martin Quinson
3858 - Allow to (en/dis)able the cycle detection at run time.
3860 Whether we should check for cycle or not is now a property of each
3861 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3862 datadesc_cycle_unset allow to remove this property when previously set.
3864 Note that the cycle detection is off by default since it impacts the
3865 performance. Watch the data you feed GRAS with ;)
3867 This property is hereditary. Any element embedded in a structure having it
3868 set have it set for the time of this data exchange.
3870 You should set it both on sender and receiver side. If you don't set it on
3871 sender side, it will enter an endless loop. If you forget on receiver
3872 side, the cycles won't be recreated after communication.
3874 - Header reorganization.
3875 Kill gras_private.h, each submodule must load the headers it needs.
3877 2004-10-04 Martin Quinson
3878 - Interface revolution: do not try to survive to malloc failure.
3880 Now, gras_malloc and friends call gras_abort() on failure.
3881 As a conclusion, malloc_error is not a valid error anymore, and all
3882 functions for which it was the only gras_error_t return value are
3883 changed. They now return void, or there result directly.
3884 This simplify the API a lot.
3886 2004-09-29 Martin Quinson
3887 - Re-enable raw sockets.
3888 Created by gras_socket_{client,server}_ext;
3889 Used with gras_raw_{send,recv}
3892 It should allow to kill the last bits of gras first version soon.
3894 This is not completely satisfactory yet (duplicate code with
3895 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3898 - Simplify transport plugin (internal) interface by not passing any
3899 argument to _server and _client, but embedding them in the socket
3902 2004-09-28 Martin Quinson
3903 - Finish the port to AIX.
3904 autoconf was my problem (segfault within the malloc replacement
3905 function. No idea why)
3907 2004-09-16 Martin Quinson
3908 - Fix some size_t madness on 64bit architectures.
3910 2004-09-08 Martin Quinson
3911 - Reduce the number of system headers loaded, overload some more system
3912 calls (such as malloc to cast the result of the system one, and work
3914 - Fix and reintroduce the config support
3916 2004-09-07 Martin Quinson
3917 - Source code reorganization to allow Arnaud to surf all over there.
3918 - Allow to document the logging categories.
3919 - Remove all uppercase from logging categories and useless cleanup in names.
3921 2004-08-18 Martin Quinson
3922 Version 0.6.2 (protocol not changed; API changed)
3923 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3924 gras_error_t), and NULL when not found. Functions expecting a msgtype
3925 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3926 hopefully usefull message.
3927 - Portability to prehistoric sparcs again
3929 2004-08-17 Martin Quinson
3930 Version 0.6.1 (protocol not changed; ABI not changed)
3931 - prealloc some buffers to speed things up
3933 2004-08-11 Martin Quinson
3934 Version 0.6 (protocol not changed; ABI expended)
3935 - The parsing macro can deal with the references, provided that you add
3936 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3938 2004-08-09 Martin Quinson
3939 Version 0.5 (protocol not changed; ABI changed)
3940 - Allow to off turn the cycle detection code in data exchange at
3941 compilation time. It should be at run time, but I'm short of time (and
3942 the config stuff is still broken). That way, we keep dict out of the
3943 critical path, which is good because the performance is poor:
3944 - search not dichotomial yet
3945 - dynar give no way to access their content and memcpy everytime
3946 - In composed data description (struct, ref and so on), stop foolness of
3947 keeping the subtype's ID, but store the type itself. This keeps sets out
3948 of the critical path, which is good since they rely on dynar and
3949 dictionaries. The only loose of that is that we cannot detect the
3950 redeclaration of a structure/union with another content (but I'm not sure
3951 the code detected well this error before anyway). We still can detect
3952 the redefinition discrepancy for the other types.
3953 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3954 it breaks the code because of type-punning used all over the place).
3955 This breaks on all non-gcc architectures (for now).
3957 All those changes (plus the buffer of last time) allow me to gain 2 order
3958 of magnitude on cruel tests consisting of 800000 array of integers on two
3959 level of a hierarchical structure (200 secondes -> 4 secondes)
3962 - the selector of reference must now return the type it points to, not
3963 the ID of this type.
3965 2004-08-06 Martin Quinson
3966 Version 0.4 (protocol changed; ABI not changed)
3967 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3968 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3969 - Implement a buffer transport, and use it by default (it relies on tcp in
3970 real life and on sg in simulation).
3971 That's a bit hackish since I had a new field to the structure to store
3972 its data without interfering with the subtype ones. Inheritance
3973 is tricky in C. And that's a kind of reverse inheritance with one class
3974 derivating two classes. Or maybe a game with java interfaces. Anyway,
3975 that's damn hard in C (at least).
3976 Moreover, I got tired while trying to ensure plugin separation and
3977 genericity in SG mode. MSG wants me to do weird things, so let's go for
3978 cruel hacks (temporarily of course ;).
3979 See comment in transport_private.h:71
3980 - do not include all the _interface headers in private but in the files
3981 which really need them (to cut the compilation time when they are
3984 2004-07-26 Martin Quinson
3985 Version 0.3 (protocol not changed; ABI changed)
3986 - Major overhault of the datadesc interface to simplify it:
3987 - shorted the function names:
3988 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3989 - add a trivial way to push/pop integers into the cbps without malloc.
3990 This allows to make really generic sub_type description, which simply
3991 pop their size of the stack.
3992 - add a function gras_datadesc_ref_pop_arr() which does what users want
3993 most of the time: Declare a dynamic array (which pops its size of the
3994 stack) and declare a reference to it. Poor name, but anyway.
3995 - kill the post-send callback, add a post-receive one
3997 2004-07-23 Martin Quinson
3998 Version 0.2 (protocol changed; ABI changed)
3999 - add some testing for cpbs in the test cases, and fix some more bugs.
4000 This invalidate again the little64 data file, since I cannot regenerate
4002 - remove an awfull optimization in the logging stuff, allowing me to:
4003 - understand it again
4004 - learn gcc how to check that the argument match the provided format
4005 - fix all errors revealed by gcc after that
4006 - internal keys of dict are not \0 terminated. Deal with it properly in
4007 loggings instead of segfaulting when the user want to see the logs :-/
4009 2004-07-22 Martin Quinson
4010 - Fix some stupid bug preventing cbps (callback postit) from working
4012 2004-07-21 Martin Quinson
4013 - Some documentation cleanups
4014 - remove the useless last argument of msgtype_declare
4015 - rename the Virtu functions to fit into the 'os' namespace
4016 - move headers src/include -> src/include/gras/ and stop fooling with
4017 gras -> . symbolic link
4018 - make distcheck is now successful
4020 2004-07-19 Martin Quinson
4022 - Build shared library also
4023 - Install html doc to the right location
4024 - stop removing maintainer files in make clean
4025 - build tests only on make check
4027 2004-07-13 Martin Quinson
4029 - No major issue in previous version => change versioning schema
4030 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4031 corresponding dataset.
4033 2004-07-11 Martin Quinson
4035 - Get it working with any kind of structure (we can compute the padding
4036 bytes remotely for all the architectures I have access to)
4037 - Implement the structure parsing macro (still not quite robust/complete)
4038 - Improvement to the remote testing toysuite
4040 2004-07-10 Martin Quinson
4041 [autoconf mechanism]
4042 - get ride of a bunch of deprecated macros
4043 - actually run the test for two-compliment, not only compile it :-/
4044 - test whether the structures get packed (and bail out if yes. Damn.
4045 Alignment is a serious matter)
4046 - test whether the structures get compacted (but respecting the alignment
4047 constraints of each types)
4048 - test whether the array fields of structures can straddle alignment boundaries
4050 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4051 (took me 2 hours to find that bug, looking at the wrong place)
4052 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4053 before sending/receiving and are used to compute the offsets of fields
4054 - Given that padding size depend even on compiler options, keep track of
4055 alignment and aligned_size only for the current architecture. Not a big
4056 deal since we send structure fields one after the other (seems
4058 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4059 program, let it work on linux/gcc/little32. portability todo.
4061 2004-07-08 Martin Quinson
4062 - import and improve remote compilation support from FAST
4063 - make sure make check works on half a dozen of machines out there
4065 2004-07-07 Martin Quinson
4066 Let's say it's version 0.0.3 ;)
4067 - Implement conversions (yuhu!)
4068 - Let it work on solaris (beside conversion, of course)
4069 - Stupid me, using rand() to generate the conversion datatests in not wise.
4071 2004-07-06 Martin Quinson
4072 - Let make dist work, since I'm gonna need it to compile on remote hosts
4073 - Let Tests/datadesc_usage write the architecture on which the file was
4074 generated as first byte.
4075 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4076 architecture descriptions.
4077 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4078 on those architectures.
4079 - Optimization: send/recv array of scalar in one shoot
4081 2004-07-05 Martin Quinson
4082 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4084 - Plug a whole bunch of memleaks
4085 - each process now have to call gras_{init,exit}. One day, their log
4086 settings will be separated
4087 - Continue the code factorisation between SG, RL and common in Transport.
4089 2004-07-04 Martin Quinson
4091 - Redistribution between SG and RL.
4092 We wanna have to accept in SG, so move accepted related parts of RL in
4093 the common part. (more precisely, the dynar of all known sockets is no
4094 more a static in transport.c, but part of the process_data)
4096 [gras_stub_generator]
4097 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4099 2004-07-03 Martin Quinson
4100 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4101 them all up in one shot)
4103 - Fix the ugly case of reference to dynamic array.
4104 - New (semi-public) function gras_datadesc_size to allow the messaging
4105 layer to malloc the needed space for the buffer.
4107 - gras_socket_close now expect the socket to close (and not its address to
4108 put NULL in it after it). This is because the socket passed to handlers
4109 is one of their argument (=> not writable).
4111 - propagate the interface cleanup from last week in datadesc, ie remove a
4112 superfluous level of indirection. User pass adress of variable
4113 containing data (both when sending and receiving), and not of a variable
4114 being a pointer to the data. Let's say that I like it better ;)
4115 The price for that is constructs like "int msg=*(int*)payload" in
4116 handlers, but it's a fine price, IMHO.
4118 - Let it work in RL (yuhu)
4120 2004-06-21 Martin Quinson
4122 - porting SG plugin and SG select to new standards (works almost).
4123 - plug memleaks and fix bugs around.
4126 - cleanup the prototype of data recv and force users to specify when they
4127 want to handle references to objects. Test case working even for cycles.
4128 - plug memleaks. Valgrind is perfectly ok with this.
4130 2004-06-12 Martin Quinson
4132 - cleanup the separation between plugin and main code in plugin creation
4134 2004-06-11 Martin Quinson
4136 - Reput hook for raw sockets, needed for BW experiments
4137 - kill a few lines of dead code
4138 [Data description] Interface cleanup
4139 - gras_datadesc_by_name returns the searched type or NULL.
4140 That way, no variable is needed to use a type desc once, which makes
4142 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4143 two parameters were strings (field name, type name), leading to
4145 [Dicos] Interface cleanup
4146 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4147 This is consistant with the dynar API.
4149 2004-04-21 Martin Quinson
4151 - Porting to new standards.
4153 - interface cleanup.
4154 There is no bag anymore, no need to take extra provision to mask the
4155 pointers behind "ID".
4156 Better splitup of functions between files create/exchange/convert.
4157 This is still a bit artificial since convert and receive are so
4158 interleaved, but anyway.
4160 - add a queued message list to procdata (the ones not matching criteria
4162 - factorize some more code between SG and RL wrt procdata
4164 - use gras_exit in example to track memleaks
4165 - get rid of gs_example now that GS is properly integrated into gras
4166 - update run_test to integrate the latest tests (datadesc)
4168 - rename WARNINGn macros to WARNn since it prooved error-prone
4170 2004-04-19 Martin Quinson
4172 - register init/exit functions within gras module mechanism
4173 - send/receive function.
4174 Convertion is not implemented, but short-cutted if not needed.
4175 struct/array elements are sent one by one (instead of block-wise), but
4176 nobody really cares (yet). Get a prototype before optimizing.
4177 - tests (using a file socket) for DD send/receive on:
4178 - base types: int, float
4179 - array: fixed size, string (ie ref to dynamic string)
4180 - structure: homogeneous, heterogeneous
4181 - chained list, graph with cycle
4182 Believe it or not, valgrind is not too unhappy with the results. The
4183 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4187 - Bugfix when using a filename explicitly (instead of '-')
4189 2004-04-09 Martin Quinson
4191 - factorize more code between RL and SG in socket creation
4192 - Complete the implementation and tests of:
4194 o file (only in RL, and mainly for debugging)
4196 I lost 3 days to design a portable address resolver, and then decided
4197 that the prototype mainly have to run on my box.
4198 Addressing portability too early may be like optimizing too early :-/
4200 - use gras_init in the Tests instead of the crappy parse_log_opt
4201 (the latter function is removed)
4202 [Conditional execution]
4203 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4204 [Code reorganisation]
4205 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4206 Build examples against the RL library, since there is no way to disable
4207 its creation for now.
4209 For information, the beginning of coding on GRAS was back in june
4210 2003. I guess that every line has been rewritten at least twice since