1 ----------------------------------------------------------------------------
3 SimGrid (3.24.1) NOT RELEASED YET (v3.25 expected December 22. 2029, 04:19 UTC)
6 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
7 - Drop support for Viva/Triva (old visualization tools), hence removed
8 TRACE_get_node_types() and TRACE_get_edge_types() functions.
11 - Actor: Merge signals on_migration_start/end into on_host_change
12 - Actor: Rename migrate() into set_host()
13 - Disk: Allow users to get the read and write nominal bandwidth values
14 - Exec: Implement wait_for(timeout)
15 - Io: Implement wait_for(timeout)
18 - Parse errors now raise a simgrid::ParseError that you may want to catch.
21 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
22 been renamed to fix a spelling error. As usual, the previous definitions are
23 kept with a deprecation warning. However, to avoid runtime errors, it is now
24 forbidden to override these deprecated methods in a derived class. Please use
25 the new names immediately if you need to override them.
28 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
29 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
30 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
31 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
32 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
33 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
34 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
35 - New module: random, providing classical random numbers generators.
38 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
39 allocations' size above which they are considered "shared" by default (as if
40 they were performed through SMPI_SHARED_MALLOC macros).
41 Default = 0 = disabled feature.
42 Note : malloc, calloc and free are now overriden by smpicc/cxx by default.
43 This can cause some troubles if codes are already overriding these. If this
44 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
45 help, but will make this feature unusable.
46 - Input check is now performed much more consistently and outputs precise
47 warnings for each error found
48 - New supported calls : MPI_File_set_view, MPI_File_get_view
49 - MPI I/O now only supports disks, as storage is discontinued.
50 - MPI I/O allows opening files without specifying absolute path
51 - Note : SMPI configuration options are now initialized by smpi_init_options()
52 call, which should be called if SMPI is not being used through smpirun.
54 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
55 - FG#39: Missing s4u::Comm::wait_any_for example
56 - FG#42: Add support for ThreadSanitizer (TSan)
57 - FG!19: Removing RngStream
58 - FG!20: A module for RNG calls
59 - FG!21: Choice between ad-hoc and standard distributions implementations
61 - FG!25: Fix link in Mutex doc
62 - FG!26: Fix links in SMPI interface doc
63 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
64 - GH#305: Unscheduled tasks are still excuted
65 - GH#313: smpirun: manual outdated w.r.t. --help
66 - GH#321: [S4U] Get task remaining work ratio
67 - GH#323: Crash when an actor turn off his physical host
68 - GH#335: Missing links on dragonfly example svg
69 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
70 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
72 ----------------------------------------------------------------------------
74 SimGrid (3.24) October 9. 2019
76 The Clean Disk Release.
79 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
80 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
81 thus becomes the default resource to execute I/O operations. Examples
82 and the file system plugin have been modified to use disks instead of
83 storage elements. Storage can still be used but is doomed to disappear.
84 Users are thus adviced to update their simulators accordingly.
85 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
86 for consistency with pthread_barrier_wait()
87 - Host::get_englobing_zone() returns the englobing netzone
88 - Actor::on_destruction is now called in the destructor
89 Actor::on_termination new signal called when the actor terminates
91 - Global signals are now part of the Engine:
92 - on_platform_creation: after config settings, before the XML parsing
93 - on_platform_created: right after the XML parsing
94 - on_time_advance: each time the clock advances
95 - on_simulation_end: after simulation, before cleanups
96 - on_deadlock: as the name implies.
98 - sg_{actor,host,link}_{data,data_set}() now all exist.
99 Use them to attach user data to the object and retrieve it.
102 - Introduce an experimental Wifi model. It sounds reasonable
103 according to the state of the art, but it still has to be properly
104 validated, at least against ns-3 if not against reality.
105 - Improved the usability of ns-3. Several bugs were ironed out.
106 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
107 It is now expressed as 'idle:epsilon:allCores' to properly model the
108 consumption of non-whole tasks on mono-core hosts. More info in the doc.
111 - convert a new set of functions to the S4U C interface and move the old MSG
112 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
115 - Fortran bindings for DVFS have been removed.
116 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
117 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
118 - SMPI can now be selected by cmake's find_module(MPI) with
119 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
120 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
121 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
122 - trace-call-location can be used with TI traces, and replayed, and is compatible
123 with smpi/comp-adjustment-file.
124 - sleep events are now correctly traced/replayed.
125 - Default for trace-call-location is now to use file names and not full paths.
126 To revert to previous behavior (in case of collision of filenames), option
127 "smpi/trace-call-use-absolute-path" can be set to yes.
130 - Use the included xxHash as an hash implem when C++14 is usable.
131 - Option model-checker/hash was removed. This is always activated now.
132 - New option smpi/buffering controls the MPI buffering in MC mode.
133 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
134 This is useful to check for MPI compliance.
137 - New section on plugins: how to define a new one, and existing ones.
140 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
141 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
144 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
145 and <mount> tags. The rationale is finish to decouple what should be handle
146 by SimGrid kernel (i.e., resources and activities) from plugins built in
147 user space (e.g., the file system plugin). This new tag currently accepts
148 only three attributes (id, read_bw, and write_bw). All the other information
149 that was declared with the storage related tags now has to be expressed as
150 properties. An example of platform using this new tag is available at
151 examples/platforms/hosts_with_disks.xml
154 - 'expect signal' can now accept more than one potential signal.
156 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
157 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
158 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
159 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
160 - FG#31: per-actor data
161 - FG#34: SG_BARRIER_SERIAL_THREAD?
162 - FG#35: model-checker does not like buster-produced binaries
163 - FG!13: MC: complete workaround in the error msg seen on modern systems
164 - FG!15: execute_flops now logs compute
165 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
166 - FG!17: ns-3: unblock the right number of communications + others issues
167 - FG!18: Improving the performance of the ns-3 bindings
168 - GH#189: Energy consumption of parallel tasks
169 - GH#219: Error in the throughput of TCP transfer
170 - GH!330: Minor polishing of English: Configuring_Simgrid
171 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
172 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
173 - GH!334: Rework energy plugin, again
175 ----------------------------------------------------------------------------
177 SimGrid (3.23.2) July 8. 2019
180 - Nicer introduction page.
181 - Migrate the "Deploy your application" page to the new doc.
182 - Move Java as a subtree of MSG.
185 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
188 - Drop xbt_dynar_sort_strings().
191 - Really fix FG#26: Turning off a link should raise NetworkFailureException
192 - FG#27: Wrong exception thrown to wait_any when link is turned off
193 - GH#328: Java: Canceling multiple tasks in a single vm/host
195 ----------------------------------------------------------------------------
197 SimGrid (3.23) June 25. 2019
199 The Exotic Solstice Release.
202 - SunOS and Haiku OS support. Because exotic platforms are fun.
203 - Stop setting random seed with srand48() at initialization.
204 - Use addr2line as a fallback for stacktraces when backtrace is not available.
205 - Build option -Denable_documentation is now OFF by default.
206 - Network model 'NS3' was renamed into 'ns-3'.
209 - Simgrid can now hopefully be installed with pip.
212 - wait_any can now be used for asynchronous executions too.
215 - New log appenders: stdout and stderr. Use stdout for xbt_help.
216 - Drop xbt_dict_dump.
219 - 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
220 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
221 - MPI interface is now const correct for input parameters
222 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
223 - Fortran bindings for async collectives.
224 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
227 - Remove option 'model-check/record': Paths are recorded in any cases now.
228 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
229 always sparse. That's an immense gain in memory, and can even be
230 faster because of cache effects. So there is no need to clutter the
231 code to allow the user to go for the unefficient mode.
234 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
235 models proved to be more accurate than these old experiments.
237 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
238 - FG#1: Broken link in error messages
239 - FG#2: missing installation documentation
240 - FG#3: missing documentation in smpirun
241 - FG#6: Python bindings not available on PyPI
242 - FG#7: simple cmake call requires doxygen
243 - FG#8: make python bindings an optional dependency
244 - FG#10: Can not use MSG_process_set_data from SMPI any more
245 - FG#11: Auto-restart actors forget their on_exit behavior
246 - FG#12: -Denable_lto=OFF doesn't disable LTO
247 - FG#13: Installs unstripped file 'bin/graphicator'
248 - FG#14: Installs the empty directory 'doc/simgrid/html'
249 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
250 - FG#17: Dead link in doc (pls_ns3)
251 - FG#20: 'tesh --help' should return 0
252 - FG#21: Documentation link on http://simgrid.org/ broken
253 - FG#22: Debian installation instruction are broken
254 - FG#26: Turning off a link should raise NetworkFailureException exceptions
255 - GH#133: Java: a process can run on a VM even if its host is off
256 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
257 - GH#326: Valgrind-detected error for join() when energy plugin is activated
259 ----------------------------------------------------------------------------
261 SimGrid (3.22) April 1. 2019
263 The Easter Christmas Release (this one is somewhat late).
266 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
267 - Not complete yet: asynchronous activities (amongst others) are still missing
268 - Still ongoing: the interface may change in the future. We need more
269 testers! Please report any glitches.
270 - No new project using Java should start now. Please switch to Python.
273 - Some of the internal cleanups may lead to speed improvements:
274 - The hard limitation on the amount of simulated actors with
275 Java+Mac was removed. Now, the available memory is the only limit.
276 - Our refcounting was tidyied, leading to 10% speedups in some cases.
277 - We are still working on making our code robust to the actor kills
278 and hosts' churn. Things are improving, but it's not perfect yet.
279 - Replaced our own code to display a backtrace (that was forking addr2line)
280 with the Boost.Stacktrace library.
281 You won't see your backtraces without this optional dependency.
282 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
283 - Stop setting random seed with srand() at initialization.
286 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
287 XML file version remains 4.2 since old files are still compatible.
290 - Process termination which was broken at version 3.21 has been repaired.
291 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
292 - Hide the examples into examples/deprecated. New users should use Python.
295 - Drop MSG_process_create_from_stdfunc() from the API.
296 This C++ function was a pimple in the C API, made necessary at some
297 point by the Java bindings. This is fixed now.
298 - Hide the examples into examples/deprecated. New users should use S4U.
299 - MSG_process_create and MSG_process_attach now crash if the host used
300 to run this process is off.
301 - Fix the protype of MSG_process_on_exit()
302 Now use (int,void*) callbacks instead of (void*,void*) ones.
303 The implementation was ways too messy to actually work, I guess.
306 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
307 too often and provide more accurate timings.
308 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
309 - MPI_Alltoallw support
310 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
311 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
312 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
313 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
315 - Basic implementation of generalized requests (SMPI doesn't
316 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
319 - Drop sg_cmdline. Please use xbt_cmdline instead.
320 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
321 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
322 OsSemaphore is implemented in a portable way with C++11 threads.
323 This should allow much more threads to be created at the same time,
324 allowing Mac and Java users to simulate many more actors.
325 - Implement the 'thread' factory with std::thread instead of xbt ones.
326 It is not possible to set the stack size with threads anymore, but
327 -fsplit-stack is the way to go nowadays when using threads.
328 - Drop the xbt_os_thread_t module (now unused)
329 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
330 - Drop xbt_str_join_array().
331 - Drop cunit, use Catch2 instead.
334 - Many cleanups in the kernel::activity namespace. This was long
335 overdue, and shall open the path to many future endeavors.
338 - #132: Java: a process can not shut down its own host
339 - #220: S4U: detached send causes exception if sender terminates soon after sending
340 - #261: Document the parameters of parallel execution's constructor
341 - #300: [s4u] BarrierPtr is missing
342 - #314: SMPI args internal cleanup
343 - #316: Fix a bug related to the CPU utilization of multi-core VM
344 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
345 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
346 - #325: Turning off a host has different behavior on sleeping actors and computing actors
348 ----------------------------------------------------------------------------
350 SimGrid (3.21) October 3. 2018
352 The Restarting Documentation (TRD) Release.
355 - Convert the user manual to Sphinx for improved usability.
356 Unstable version now hosted on https://simgrid.frama.io/simgrid
357 Many glitches remain; stay tuned.
358 - Restrict the Doxygen reference API to the installed header files.
359 (documenting all internals this way is overkill + doxygen scales badly)
360 - New tutorial on S4U (highly inspired from the old MSG one, but with
361 a git repository to fork as a starting point, and a docker image)
362 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
363 - The Developper manual is still to be converted (not compiled until then)
364 - Some parts are still missing in this conversion (such as the
365 platform and deployment sections) while others were blindly converted
366 and would need more love (such as the configuration flags).
367 Things will certainly further improve in the future releases.
370 - Fully reimplement auto_restart mechanism that was utterly broken
371 (fix #22, #131 and #281 that were all very related)
372 - Implement semaphores (fix #303 and #291)
373 - When creating an actor from a function and its parameters,
374 move-only parameters are not allowed anymore, as it would prevent
375 the actor to be restartable if its parameters are consumed on run.
376 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
377 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
378 IO operations that can be started, waited for, or canceled as a regular
380 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
384 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
385 - New host variable: 'core_count'
388 - Remove xbt_os_thread_specific features
389 - Remove portability wrapper to condition variables
390 - Remove xbt_os_thread_yield()
393 - MPICH collective selector now mimics MPICH 3.3b
394 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
395 - MPI_Init(NULL,NULL) is now allowed
396 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
397 - Command-line arguments should now be handled properly in Fortran simulations
398 - MPI Attributes and related callbacks should work in Fortran
399 - Apps using C + Fortran should now work
400 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
401 - Support MPI_Error_String
404 - Due to an internal bug, Msg.run() must now be your last line.
405 We hope to fix it in a future release, and we are sorry for the inconvenience.
408 - #22: Process autorestart seem to only work with CAS01 cpus
409 - #56: Feature request: dynamic SMPI replay
410 - #93: simgrid should not eat --help
411 - #111: How to change the loopback link characteristics?
412 - #116: Platform cluster radical
413 - #117: Platform cluster/host order
414 - #129: Replay_multiple on steroid
415 - #131: Java : autostart does not work
416 - #143: Setting a breakpoint at a given time
417 - #235: Network energy models should be integrated
418 - #264: Add ptask L07 resource tracing
419 - #271: Dynamic and manual replay of SMPI traces
420 - #279: Breakpoints option uses time deltas instead of absolute time
421 - #281: Daemonized actor no longer daemonized after an auto-restart
422 - #282: TIT convertor erase traces when using absolute path in the trace list file
423 - #285: segfault when a s4u actor kills itself with msg process tracing activated
424 - #286: Pajé traces are not exposing the number of cores
425 - #287: Command-line parsing should stop when encountering '--'
426 - #288: MPI_Init(NULL, NULL)
427 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
428 - #290: Method to check if a mailbox has a message ready to be consumed
429 - #291: [s4u] semaphores do not seem available yet
430 - #294: Alltoall collective implementation of mpich
431 - #295: Starting a Host as OFF is not supported
432 - #296: DTD too permissive
433 - #299: Add a s4u barrier example
435 ----------------------------------------------------------------------------
437 SimGrid (3.20) Released June 24. 2018
439 The proxy snake_case() release.
442 - The whole API is now using snake_case() instead of mixing it with camelCase.
443 Compatibility wrappers in place for one year.
444 - We took the opportunity to increase the consistency of the API.
445 Please report any remaining glitches.
446 - New function simgrid::s4u::Host::get_actor_count:
447 Returns the number of actors running on a specific host.
450 - SMPI is now tested with ~45 proxy apps from various sources, with none or
451 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
452 - Replay: The replay file has been re-written in C++.
453 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
454 supported. They are stored in the trace and used when replayed.
455 - Basic support of MPI_Cancel. Robustness not guaranteed.
456 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
457 (only for MPI_COMM_TYPE_SHARED).
458 - New option: smpi/privatize-libs, to add external shared libs to be privatized
459 by SMPI. They will be copied locally and loaded separately by each process.
460 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
461 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
462 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
463 with internal index optind with multiple processes. Only works if MPI_Init has
465 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
466 syntax without preprocessing tricks.
469 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
470 behavior tracing is no longer limited to MSG
471 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
472 this means that traces now need to be updated or re-obtained; both Irecv
473 and Isend lines in your traces have now in total 5 mandatory fields:
474 <rankid> <command> <to/from rankid> <tag> <size>
475 To update your traces, it suffices to add a 0 for the tag here.
476 - TIT now also supports waiting for a distinct request via MPI_Wait.
477 Wait/Test now wait for a specific request, not just the last one that was
478 issued. This unfortunately means another update, because we need to
479 identify which request you want to wait for. We do this via the
480 triplet (sender, receiver, tag), which needs to be added:
481 <rankid> <command> <sender> <receiver> <tag>
482 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
483 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
486 - The deprecation of MSG is ongoing (but this should not impact you).
487 Many MSG functions are now simple wrappers on the C API of S4U. If
488 you wish to convert your code to S4U, find the S4U counterparts of
489 your MSG calls in src/msg/msg_legacy.cpp.
490 - MSG can still be used, but won't evolve anymore.
493 - Allow to run the Link energy plugin from the command line with
494 --cfg=plugin:link_energy
495 - Rename Energy plugin into host_energy
496 - Rename Load plugin into host_load
499 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
500 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
501 - Improve performance when handling timeouts of simix synchros.
504 - Config: the C API is now deprecated (will be removed in 3.23), and
505 the C++ API has been slightly improved.
508 - Fix several build issues on OSX.
509 - Move simgrid_config.h to simgrid/config.h (old header still working)
512 - #143: Setting a breakpoint at a given time
513 - #150: Inconsistent event names in SMPI replay
514 - #258: daemonized actors hang after all non-daemonized actors have completed
515 - #267: Linker error on unit_tmgr
516 - #269: SMPI: tracing of MPI_Wait/all/any broken
517 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
519 ----------------------------------------------------------------------------
521 SimGrid (3.19.1) Released March 22 2018
523 The Fixed ABI Release.
525 As you may know, we are currently refactoring SimGrid in deep.
526 Upcoming SimGrid4 will be really different from SimGrid3: modular,
527 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
529 Our goal is to smooth this transition, with backward compatibility
530 and automatic update paths, while still progressing toward SimGrid4.
532 SimGrid remains open during works: The last pure SimGrid3 release was
533 v3.12 while all subsequent versions are usable alpha versions of
534 SimGrid4: Existing interfaces remain unchanged, but the new S4U
535 interface is budding and the internals are deeply reorganized.
537 Since 2015, we work hard to reduce the changes to public APIs. When
538 we need to rename a public library symbol in S4U, we let your
539 compiler issue an explicative warning when you use the deprecated
540 function. These messages remain for four releases, i.e. for one full
541 year, before turning into an error. Starting with v3.15, your can
542 also adapt to API changes with the SIMGRID_VERSION macro, that is
543 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
545 Starting with this v3.19.1, our commitment to reduce the changes to
546 the public interfaces is extended from the API to the ABI: a program
547 using only MSG or SimDag and compiled against a given version of
548 simgrid can probably be used with a later version of SimGrid without
549 recompilation. We will do our best... but don't expect too much of
550 it, that's a really difficult goal during such profund refactoring.
552 The difference between v3.19 and v3.19.1 is that the former was
553 accidentally breaking the ABI of MSG, while the later is restoring
556 S4U and kernel APIs will still evolve until SimGrid4, with one-year
557 deprecation warnings as currently. In fact, cleaning up these
558 interfaces and converting them to snake_case() is one release goal of
559 v3.20. But don't worry, we are working to smooth this upgrade path.
561 In summary, new projects should start with S4U to benefit of the
562 future, but old MSG projects should still be usable with no change.
564 ----------------------------------------------------------------------------
566 SimGrid (3.19) Released March 20 2018
568 The Moscovitly-cold Spring release.
570 SMPI user-visible improvements:
571 - Support of MPI_Type_create_subarray
572 - Unimplemented MPI calls now abort (instead of issuing a warning)
573 Some harmless calls around errhandlers still only issue a warning.
574 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
575 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
576 is now possible (mmap privatization was refactored to this end).
577 Adding new SMPI instances should also be technically doable now.
579 SMPI internal cleanups:
580 - Removed index notion from SMPI -- no more "getPid() - 1"!
581 In tracing files, actors are called rank-<PROCESS-ID> but these are
582 NOT the real ranks (tracing several communicators is not supported yet).
583 - Deprecate smpi_process_index() (will be removed in 3.22)
584 - Better testing on classical proxy apps, through an external project:
585 https://github.com/simgrid/SMPI-proxy-apps/
588 - Use a graphical TOC to make it easier to find the documentation you need
589 - Revamp the MSG tutorial
592 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
593 - Remove parameter of MSG_process_killall().
594 Resetting the PID was bogus anyway (several actors could have the same PID).
597 - Execution->setHost() can be called after start(), to migrate it.
598 - Comm::test_any() is now implemented.
599 - s4u::Actor now has onCreation() and onDestruction() signals.
600 - Install some kernel header files for the users' plugins and more.
601 Warning, their API is really not stable yet.
602 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
603 include/simgrid/kernel/routing/*Zone.hpp
606 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
608 Build System and other cleanups:
609 - Remove unused run-time parameter "tracing/onelink-only".
610 - SimGrid now works with intel compilers (MC still buggy)
611 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
612 - tesh kills subprocesses on timeout (but not on windows)
615 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
618 - #194: Feature request: simgrid::s4u::Comm::test_any()
619 - #245: migrating an actor does not migrate its execution
620 - #253: Feature Request: expose clusters as objects
621 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
622 - #255: Tesh broken on Windows
623 - #256: Modernize FindSimGrid.cmake
624 - #257: Fix (ab)use of CMake install
626 ----------------------------------------------------------------------------
628 SimGrid (3.18) Released December 24 2017
630 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
632 MSG IS NOW DEPRECATED. Please use S4U for new projects.
633 - Support for MSG should not be removed from SimGrid before 2020,
634 but future evolutions will be limited to the integration of
635 user-provided patches.
638 - s4u::onDeadlock() signal, run before stopping the simulation.
639 - s4u::Exec: asynchronous executions (abstraction of background
641 - s4u::Host->getLoad() returns the achieved speed in flops/s
644 - Link::name() is deprecated and will be removed in v3.21, use
645 Link::getCname() instead.
646 - Mailbox::getName() changed to return a std::string, use
647 Mailbox::getCname() instead to get a char*.
648 - Storage::getName() changed to return a std::string, use
649 Storage::getCname() instead to get a char*.
650 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
651 to have both allocation and deallocation of the map in user space.
654 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
655 The amount of flops remaining to do is a vector, not a scalar.
656 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
657 name implies on both sequential and parallel tasks.
658 - Both changes fix GitHub's #223 using PR #237 as a basis.
659 Thanks Michael Mercier.
660 - Most examples were converted to S4U and hidden elsewhere at tests.
661 Rationale: we still want MSG to work; we want newcomers to use S4U.
664 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
665 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
668 - Switch to the faster dlopen privatization mechanism by default
669 - Documentation improvments
672 - Live migration is getting moved to a plugin. Dirty page tracking is
673 the first part of this plugin. This imply that VM migration is now
674 only possible if one this function is called:
675 - C/MSG: MSG_vm_live_migration_plugin_init()
676 - C/C++: sg_vm_live_migration_plugin_init()
677 - Java: Msg.liveMigrationInit()
678 For C and C++, "simgrid/plugins/live_migration.h" also has to be
682 - Define class simgrid::xbt::Path to manage file names.
683 - Removed unused functions:
684 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
685 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
686 xbt_graph_export_graphviz()
687 - xbt/str.h: xbt_str_join()
688 - Remove unused datatypes:
689 - xbt/heap.h: use std::priority_queue or boost::heap instead
690 - xbt/swag.h: use boost::intrusive::list instead
693 - New link_energy plugin for the consumption of the links.
694 - All of the operations on files and storage contents have been
695 packaged into a plugin (src/plugins/file_system). The current
696 public interface can be found in
697 include/simgrid/plugins/file_system.h
698 To use these functions you now have to initialize the plugin by
699 calling MSG_storage_file_system_init() just after calling
700 MSG_init() or sg_storage_file_system_init() just after creating
704 - Remove the undocumented/untested tag <include>
707 - Remove viva specific tracing as the tool is no longer maintained
710 - #248: Kill of finished processes leads to segfault
711 - #240: xbt_cond_wait_timeout should gracefully return for C apps
712 - #239: please implement signal s4u::onDeadlock()
713 - #230: segfaults when exit() without run()
714 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
715 - #223: MSG_task_get_flops_amount() not working with parallel tasks
716 - #222: Actor::kill() doesn't really kill and segfaults
717 - #221: odd LMM warning when killing an actor
718 - #120: Memory leak when the processes are forcefully killed
720 ----------------------------------------------------------------------------
722 SimGrid (3.17) Released October 8 2017
724 The Drained Leaks release: (almost) no known leaks despite the tests.
726 Even very long simulations will preserve your memory: our testsuite
727 stresses SimGrid for over 45mn on fast machines for a coverage of
728 over 80%, and there is only one single known leak, of about 4kb.
732 - Comm.detach(): start and forget about asynchronous emission. A cleanup
733 handler may be given to free resources if the comm cannot be completed.
734 - this_actor::send(mailbox) is now mailbox->put()
735 - New: simgrid::s4u::Comm::wait_all()
736 - New: Host.execute() for remote executions.
739 - Unused option network/sender-gap is removed.
742 - Deprecate MSG_task_isend_with_matching(): This unused feature
743 really complicates our internals. Will be removed in v3.20.
746 - Improved context termination. It is now possible to free resources, even
747 when a process is forcibly killed.
748 - ContextBoost: add support for Boost versions above 1.61.
751 - Bring back run-time option --cfg=exception/cutpath to remove exception
753 - Removed unused functions:
754 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
755 xbt_str_rtrim(), xbt_str_trim().
756 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
759 - Removed header files obsolete since SimGrid 3.12:
760 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
762 - #3: SD_exit should be made optional
763 - #120: Memory leak when the processes are forcefully killed
764 - #159: Threading test regression in Actor refcounting
765 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
766 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
767 - #186: Actor::killAll() segfaults if some process is blocked on wait()
768 - #191: VM migration and pstate
769 - #192: Updating the energy consumptions of all hosts crashes with VMs
770 - #195: All actors have PID=0 in the logs
771 - #204: Sometimes segfault with thread contexts and mmap privatization
772 - #222: Actor::kill() doesn't really kill and segfaults
773 - #225: Actor::kill() doesn't really kill when victims are doing a join()
775 ----------------------------------------------------------------------------
777 SimGrid (3.16) Released June 22. 2017.
779 The Blooming Spring Release: developments are budding.
782 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
783 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
784 - Java JAR file should be finally fully working on Mac OSX too.
786 XML platforms: Switch to platform v4.1 format.
787 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
788 - <zone> can be used as a synonym for the now deprecated <as>
789 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
790 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
791 - <actor> can be used as a synonym for the now deprecated <process>
792 - state_file and avail_file periodicity is now easier to express
793 (check the documentation) the old behavior should still work.
794 * Storage: not backward compatible, but it seems that nobody ever used it
795 - Remove attribute 'content_type' of <storage_type>: was never used
796 - Make attribute 'model' of <storage_type> optional: for future usage
797 - Remove Bconnection model property: was never (in)validated. Replaced by
798 the maximum of the read and write bandwidth as a resource constraint.
801 - New and Backwards Compatibility break:
802 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
803 When one wants to get the list of tasks whose states have changed during a
804 simulation round, s/he has to allocate and free a dynar and use it as
805 argument to this function. The former SD_simulate (double how_long)
809 - Allow multicore VMs, with the correct sharing (unless you overcommit)
810 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
813 - New function to update the consumption of all hosts at once.
814 - Fix the model for multi-core hosts, linear in the amount of busy
815 cores with an abnormality for pIdle (see paper).
816 BUG: Still not satisfactory for ptask on mono-cores.
819 - The netzone are now available from the MSG API.
820 The old names still work, but are now deprecated.
823 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
824 - New option: smpi/keep-temps to not cleanup temp files
825 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
826 mallocs mode is used, allows to change the size of the fake file used
827 (default 1MB), to potentially limit the number of mappings for large runs.
828 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
829 - Fortran ifort and flang compilers support
830 - New RMA calls supported (experimental):
831 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
832 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
833 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
834 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
835 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
836 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
839 - New: this_actor::isMaestro() returns whether we are in kernel mode.
840 - Behind the scene, ActivityImpl are now automatically refcounted.
841 Removing this nasty bottleneck will greatly simplify our internals.
844 - Replay: New function xbt_replay_action_get():
845 Retrieve the function previously associated to an event type.
846 - DROPPED FUNCTION: xbt_str_varsubst()
847 - DROPPED MODULE: strbuff. We don't need it anymore.
848 - DROPPED MODULE: matrix. We don't need it anymore.
849 - DROPPED MODULE: lib. We don't need it anymore.
851 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
853 SimGrid (3.15) Released March 22. 2017
855 The Spring Release: continuous integration servers become green
856 We fixed even the transient bugs on all target architectures:
857 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
858 Windows, FreeBSD, NetBSD.
860 - Rename NetCards to NetPoints.
861 This was intended to help NS3 users, but that's not a netcard.
862 That's a point in the routing algorithm, let's avoid wrong simplifications.
865 - New: MSG_process_yield(). Stop and yield to other processes.
866 - New: MSG_process_daemon(). Daemon processes are automatically killed
867 when the last non-daemon process terminates
868 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
869 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
870 to make room for the upcoming network energy plugin.
871 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
874 - Ensure that an actor can kill itself with Process::exit()
875 - Kill the obscure NativeException. Nobody want to survive the issues
876 it denotes, so use JniException that is a RuntimeException (not to
877 be caught explicitly).
878 - Partial bug fix in initialization. SimGrid flags on command line were
879 consumed at C level but stayed in the original Java String[] args.
880 This could mess users' args[i] if SG flags were not put at the end of
882 The SimGrid flags are now removed from the Java arguments. However,
883 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
884 if args.length is greater than the number of YOUR OWN ARGUMENTS.
885 It might be if you have --log or --cfg flags in the command line.
886 - Fix numerous memleaks all around the place. In particular, around VMs.
890 - s4u::onPlatformCreated: right before the simulation starts
891 - s4u::onSimulationEnd: right after the main simulation loop
892 - s4u::onTimeAdvance: right after a clock change (time is discrete)
893 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
894 event from the availability_file changes the avail speed.
895 - Links are now usable from s4u
896 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
897 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
898 - New examples: The conversion of MSG examples to S4U has begun
899 - Actors: create, kill, migration, and suspend
900 - Applications: master-worker and token-ring
901 - Action replay: communications and storage
902 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
905 - Backwards Compatibility breaks
906 - The SD_route_get_list and SD_route_get_size functions have been removed.
907 They are replaced by the unique
908 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
909 The route from src to dst is built in the links dynar whose size can
910 be obtained with xbt_dynar_length.
911 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
912 sg_host_route_bandwidth and sg_host_route_latency.
913 Macros ensure the backwards compatibility, but you should fix your code
916 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
917 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
918 - Bump our claim of support from MPI 1.1 to MPI 2.2.
919 We don't support 100% of it, but it should be enough. Contact us if not.
920 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
921 naive and probably inaccurate)
922 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
923 mapped onto a unique small file using some system magic.
924 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
925 - Fortran: cleanups, fixes, support of user-added operations
926 - RMA: MPI_Accumulate are applied in correct order.
927 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
928 - Better support for MPI_IN_PLACE
929 - Support for MPI_Win attrs and keyvals.
930 - Support MPI_Comm_group_create, MPI_Type_size_x
931 - MPI ops should be properly applied to all allowed types
932 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
933 algorithm by a non failing one from mpich.
936 - Kill the fifo data container: we don't use it anymore.
939 - A <cluster> can now be created with different speed values to represent pstates.
940 see examples/platforms/energy_cluster.xml for an example.
942 SimGrid (3.14.159) Released December 28. 2016
944 The previous release was only a crude approximation of a Pi release;
945 this one should be a bit better.
947 - Revert a last minute change that broke on Mac OSX.
948 - Fix the changelog and NEWS files.
949 - Improve the documentation, in particular of the routing module.
950 - Integrate some patches flying around in Debian.
952 SimGrid (3.14) Released December 24. 2016
955 * The whole documentation was reworked and reorganized. There is
956 still much room for improvement here, but we're on it.
959 * We now need python3 for our testing infrastructure.
960 * Model checking needs libevent
963 * Now works on FreeBSD
966 * Privatization now works on FreeBSD
968 * Privatization is now activated by default in smpirun
969 You can turn it off with -no-privatize if something goes wrong.
971 * Call-location tracing for SMPI.
972 You can add the exact location (filename / linenumber) of an MPI call to
973 your trace files and slow down or speed up the simulation between two
974 consecutive calls by using an adjustment file (see the documentation).
976 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
977 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
978 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
979 values were ignored in some cases. The timings of these functions can now
980 be significantly different.
982 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
983 smpi/running-power is renamed to smpi/host-speed
985 * smpi/grow-injected-times option to enable or disable multiplication of the
986 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
987 make simulation less precise (but also much faster).
989 * smpirun script should be (much) faster for large deployments.
991 * SMPI tracing: fixed issue with poor matching of send/receives.
993 * Replay: Fix broken waitall
995 New functions and features
996 * MSG_parallel_task_execute_with_timeout, to timeout computations.
998 Dropped / renamed functions and features
999 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1000 - MSG_mailbox_is_empty() -> Mailbox::empty()
1001 - MSG_mailbox_front() -> Mailbox::front()
1002 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1003 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1004 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1005 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1006 * Don't pass the free_f parameter to property related functions:
1007 - MSG_host_set_property_value()
1008 - MSG_as_router_set_property_value()
1009 - MSG_storage_set_property_value()
1010 * VM properties. Since msg_vm_t are msg_host_t, just use
1011 MSG_host_get_property_value() and friends
1012 * VM I/O related things:
1013 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1014 - Unimplemented save/restore methods
1015 * MSG_as_router_get_property_value() was redundant with
1016 MSG_environment_as_get_property_value().
1017 - Removed MSG_as_router_*propert*() functions
1018 - Added MSG_environment_as_set_property_value() for consistency
1019 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1020 Well, they are still there for now, but deprecated with a warning.
1021 Please switch to xbt_dict_new_homogeneous() before this is removed
1023 * Task affinity. Its intended behavior (that was very badly tested
1024 and probably not really working) was deceiving what most users
1025 would have hoped here.
1026 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1027 * Option network/coordinates is now useless and should be dropped.
1030 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1033 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1034 * Added Dragonfly topology support
1037 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1038 EiBps, EBps, Eibps, Ebps for bandwidth.
1039 They may become useful to some lucky ones.
1042 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1043 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1044 * ex/async_waitAll: new example, on asynchronous communications
1047 * Memory usage should be decreased for simulations with a large number
1048 of processes. This also helps for SMPI.
1050 SimGrid (3.13) stable; urgency=low
1052 The Half Release, a.k.a. the Zealous Easter Trim.
1054 - We removed half of the lines, that were mostly experimental cruft.
1055 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1056 (not counting blanks and comments -- according to openhub.net)
1057 - The internals are now compiled in C++ (and will soon be clean C++)
1058 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1059 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1061 * Backwards Compatibility breaks
1062 - Removed Lua simulation bindings (switch to C or Java for that).
1063 Lua can still be used to describe platforms
1064 - Removed Java kernel plug-ins.
1065 Will be reintroduced after the ongoing major internals reorg.
1067 - the following functions were removed.
1068 They were too specific and should be reimplemented in a generic
1069 way, with filter function.
1070 - MSG_task_listen_from_host
1071 - MSG_mailbox_get_count_host_waiting_tasks
1072 - MSG_mailbox_put_with_timeout was removed.
1073 Please use MSG_task_send_with_timeout instead.
1075 - the SD_application_reinit function was removed. It has been a noop for a while.
1076 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1077 replaced by a more flexible mechanism.
1078 - The following functions thus do not exist anymore
1079 - SD_workstation_get_access_mode
1080 - SD_workstation_set_access_mode
1081 - SD_workstation_get_current_task
1082 - Basic estimation functions have been removed but can easily be replaced
1083 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1084 - SD_workstation_get_computation_time => amount / sg_host_speed()
1086 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1087 Use VM.getSpeed()*load/100 for the legacy behavior.
1089 - option enable_tracing was removed. It was not doing anything for a while.
1090 - In the ModelChecker:
1091 - the model-checker now ptraces the model-checked process which means
1092 you cannot use a debugger on the latter anymore (we might make this
1093 optional in the feature);
1094 - removed soft-dirty page tracking;
1095 - remove model-checked side snapshot management,
1096 MC_snapshot() and MC_compare_snapshot();
1097 - keep the MC_cut() function as a stub (it was not really working
1098 in the previous release).
1101 * All options are consistently in kebab-case. Old names are kept as alias.
1104 * Switch to platform v4 format.
1105 - Rename from 'power' to 'speed' the attributes describing the amount of
1106 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1107 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1108 - In <host> and <link>, attributes availability and state are gone.
1109 It was redundant with state and availability traces, and with peak values.
1110 - In <cluster>, attributes availability_file and state_file are gone.
1111 It was too complex and unused.
1112 - Kill <gpu>. Was not doing anything.
1113 - The DOCTYPE points to the right URL:
1114 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1115 (the file at this address now documents the changelog since its v1)
1116 - A warning is emitted for unit-less values (they are still accepted).
1117 - speed. Default: 'f' or 'flops'. Also defined:
1118 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1119 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1120 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1121 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1122 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1123 - latency. Default: 's' second. Also defined:
1124 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1126 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1127 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1130 * s4u::Host is now the preferred public interface to the Host features.
1131 sg_host_* functions are C bindings to the exact same behavior
1132 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1135 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1136 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1137 They were deprecated since a few years, and probably did not even compile anymore.
1140 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1141 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1142 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1143 encouraged to update their codes. A list of the modified functions can be found at the end of
1144 include/simgrid/simdag.h
1147 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1148 the lambda or closure passed as a parameter will run in kernel mode.
1149 All the callback functions should be rewritten to that interface at some point.
1152 * Reorganizing and cleaning the internals all around the place.
1155 * Remove old default barrier/bcast buggy algorithms (see #18407)
1156 * Various bug fixes to handle more codes
1157 * Remove the need for the --foreground option of smpirun (it is still
1158 accepted for backward compatibility).
1161 * Kill the setset data container: MC don't use it anymore.
1162 * Kill the queue data container: it made more sense with GRAS.
1163 * Kill the xbt_peer_t data type: it's useless without GRAS.
1164 * Kill rm_cb feature of config sets: it was never useful.
1165 * Kill graphxml parsing feature. It was not used.
1166 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1168 - xbt_dynar_sort_strings(), when the content is char*
1169 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1170 They throw exceptions on invalid input;
1171 * C++ support for declaring CLI flags (simgrid::config::Flag);
1172 * class for abstracting different signal backends (simgrid::xbt::signal).
1173 with no external dependencies (we need very simple signals).
1176 * refactoring and cleanup of the code;
1177 * ongoing process to cleanly separate the model-checking algorithms
1178 from the code model-checking support.
1180 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1182 SimGrid (3.12) stable; urgency=low
1184 The Facelift Release.
1187 * Require g++ v4.7 at least to not speak prehistorical C++.
1188 * Require Boost 1.48 (for signal2 component).
1189 * Java must be version 7 at least when activated.
1190 * Builds on Windows again (including Java bindings).
1191 * Tracing is now always enabled (no way to turn it out)
1192 * Remove GTNetS. It was not working anyway.
1193 * Various cleanups in the cmake scripts.
1194 * Move headers around to sort them out on installed systems:
1195 - instr/instr.h -> simgrid/instr.h
1196 - instr/jedule/* -> simgrid/jedule
1197 - simdag/datatypes.h was removed
1198 - simdag/simdag.h -> simgrid/simdag.h
1199 - msg/datatypes.h was removed
1200 - msg/msg.h -> simgrid/msg.h
1203 * Interface improvement:
1204 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1205 - Sanitize the interface in MSG_task_ module:
1206 - Merge two functions that were close enough but misleading:
1207 set_compute_duration(t) -> set_flops_amount(t)
1208 get_remaining_computation(t) -> get_flops_amount(t)
1209 - set_data_size(t) -> set_bytes_amount(t)
1210 get_data_size(t) -> get_bytes_amount(t)
1211 - Massive cleanups in the functions related to the energy
1212 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1213 - New: MSG_host_get_pstate()
1215 - msg/energy/onoff: switching hosts on and off
1217 * Interface improvement:
1218 - Rename Host.isAvail() to Host.isOn()
1219 - Rename Process.currentProcess() to Process.getCurrentProcess()
1220 - Rename Task.setDataSize() to Task.setBytesAmount()
1221 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1223 - #18874: Actually allows the GC to reclaim tasks
1227 - SIMIX_process_throw: raises an exception in a remote process
1228 * Refactoring: Separate sub-modules
1229 - libsmx: the public interface, as libc in a real system
1230 - popping: the strange dance that converts a user request into a kernel handling
1231 - smx_context_*: the virtualization mechanisms that embed the user code
1232 - smx_*: the handling of each simcalls
1233 * Interface cleanup:
1234 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1235 * Rename smx_action_t into smx_synchro_t, making explicit that these
1236 things are used to synchronize processes with their environment.
1237 For example, a communication is a sort of synchronization involving
1238 the communicating processes (that may block until the exchange) and
1239 the platform. The same can be said from computations, etc.
1241 - #18888: segfault when a process ends before its kill_time
1245 - 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.
1246 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1247 - MPI_Type_set_name, MPI_Type_get_name
1248 - MPI_*_c2f and MPI_*_f2c functions
1249 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1250 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1251 - Activate a lot of new tests from the mpich 3 testsuite
1253 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1254 - InfiniBand network model added: Based on the works of Jerome Vienne
1255 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1256 - When smpi/display_timing is set, also display global simulation time and application times
1257 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1258 * Collective communications
1259 - 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
1260 - 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/).
1261 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1263 - Replay now uses algorithms from wanted collective selector
1264 - Replay can be used with SMP-aware algorithms
1265 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1266 - 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.
1268 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1269 - Correctly use loopback on fat-tree clusters
1270 - Asynchronous small messages shouldn't trigger deadlocks anymore
1271 * Energy/DVFS cleanup and improvement
1272 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1273 - new: smpi_get_host_pstate()
1276 - "Full" network optimization flag was broken since Surf++
1277 - Better handling of precision flags in maxmin
1278 - Fix bug causing sometimes "Impossible" errors
1279 - Properly pass cluster properties to included hosts
1280 * Improvement of the Energy plugin.
1281 - Always update the consumption before returning that value
1282 - New property: watt_off to denote the dissipation when the host is off
1283 - New functions getWattMinAt and getWattMaxAt to retrieve the
1284 dissipation of pstates that we are not currently at.
1285 * Java: class NetworkLink renamed to Link
1286 * New function: simcall_process_get_kill_time()
1287 * Massive rename s/workstation/host/
1288 - That's intrusive, but that's good for the project consistency. Sorry.
1289 - Change config option "workstation/model" into "host/model"
1292 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1293 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1294 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1295 - xbt_dirname and xbt_basename for non-POSIX systems
1297 * The model checker now runs as a separate process.
1298 * The model checker runs must now be launched with the new simgrid-mc program.
1299 * Record/Replay: the MC can display a textual representation of a path in the
1300 execution graph. It can then be replayed outside of the model checker.
1302 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1304 SimGrid (3.11) stable; urgency=low
1309 * Normalizing pointers addresses tool for better diff between logs
1312 * Add cloud examples using new VMs
1313 - examples/msg/cloud/two_tasks_vm.tesh
1314 - examples/msg/cloud/simple_vm.tesh
1315 - examples/java/cloud/cloud.tesh
1316 - examples/java/cloud/migration/migration.tesh
1317 * Add java surf examples:
1318 - examples/java/surfPlugin/surf_plugin.tesh
1319 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1320 - examples/java/surfCpuModel/surf_cpu_model.tesh
1321 * Add SMPI+MSG example:
1322 - examples/smpi/smpi_msg_masterslave/
1327 - msg task destroy cancel test
1328 - msg_host on/off test
1329 * Move all tests in testsuite to teshsuite (adding tesh files)
1330 * Restructure teshsuites
1331 - one folder for each kind of test
1332 * Restructure AddTests.cmake
1334 - structure the order of tests (with sections)
1337 * Add virtual machine
1338 - creation of a VM on a PM
1339 - migration of a VM from a PM to another PM
1341 - MSG_process_join(msg_process_t process, double timeout)
1342 - msg_bar_t MSG_barrier_init(unsigned int count)
1343 - int MSG_barrier_wait(msg_bar_t barrier)
1344 - void MSG_barrier_destroy(msg_bar_t barrier)
1345 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1346 * New option "msg/debug_multiple_use" to help debugging when a task is used
1349 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1351 - Increase code coverage in test suites
1355 * Protect context stacks against stack overflow. The number of protected memory
1356 pages allocated on the top of each stack (1 by default) can be configured
1357 with the new command line option --cfg=contexts/guard_size:<value>.
1358 * Simcalls are now generated by a python script that
1359 - generates files included by SimGrid
1360 - checks that all the functions exist, and proposes prototypes
1362 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1363 host_set_data, host_get_data
1365 - simcall_process_join(smx_process_t process, double timeout)
1366 * Fix bug where sleeping processing could not be suspended.
1369 * Translate surf models from C to C++
1370 - Generic classes for all models: Model, Resource, Action
1371 - A generic interface for each kind of model (CPU, Network, Storage
1372 Workstation, WorkstationVM)
1374 * Translate surf routings from C to C++
1375 * Add callbacks using sigc++ or boost::signals2
1376 - Add callback functions for resource creation/destruction
1377 - Add callback functions for action state change
1378 - Handle Energy as a plugin
1379 * Replace swag by boost::intrusive
1380 * Add new routing models for clusters. For documentation, see
1381 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1382 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1383 parameters for cluster tag
1384 - Fat trees, with topology="FAT_TREE" and
1385 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1386 - see examples/platforms/torus_cluster.xml and
1387 examples/platforms/fat_tree_cluster.xml
1388 * More documentation
1391 * Hostfiles support host:nb_processes construct to deploy several processes on
1393 * Collective communication algorithms should not crash if used with
1394 improper number of nodes and report the error.
1395 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1396 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1397 MPI_Dims_create, MPI_Cart_sub are supported.
1398 * New interface to use SMPI programmatically (still depends on MSG for
1399 some parts, see examples/smpi/smpi_msg_masterslave):
1400 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1404 * Global variables privatization in MPI executables is now performed at runtime
1405 with the option smpi/privatize_global_variables (default:no).
1406 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1407 dynamic libraries loaded by the application are not privatized (static
1408 linking with these libraries is advised in this case)
1411 * Options defined in XML work correctly now.
1414 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1415 copied into simgrid.jar or not (ON by default). Use this option if you want
1416 to reduce the size of the installed simgrid.jar, *and* the native libraries
1417 are kept installed elsewhere.
1418 * Surf binding with SWIG (code generated in maintainer mode only):
1419 - plugin to handle callbacks
1420 - CPU model only for the moment
1423 * Supernovae build mode is definitively removed. It was used to improve
1424 inlining and inter-module optimizations. It is nowadays superseded by
1425 link-time optimizations commonly available in compilers.
1426 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1427 * Add boost dependency for surf++
1428 * Add new macro for tests
1429 - ADD_TESH(name <tesh_args>)
1430 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1434 - xbt_bar_t XBT_barrier_init(unsigned int count)
1435 - int XBT_barrier_wait(xbt_bar_t barrier)
1436 - void XBT_barrier_destroy(xbt_bar_t barrier)
1437 * Make the xbt_os_time module public
1439 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1441 SimGrid (3.10) stable; urgency=low
1443 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1446 * Reintegrate Java to the main archive as desynchronizing these
1447 package is not acceptable anymore (Java is now considered stable)
1448 * Add explicit synchronization facilities through semaphores
1449 * Bug fix: Task.setDataSize() only changed the C world, not the value
1450 cached in the Java world
1453 * Dramatically change the way files are handled. API and internals changed, but
1454 this part of MSG was not considered as production grade either.
1455 * Add explicit synchronization facilities through semaphores
1456 * Add a new function MSG_host_get_process_list()
1457 * Preliminary DVFS support (see examples/msg/energy/ for details)
1460 * SMPI is now included directly in the libsimgrid as the windows
1461 linker doesn't force us on splitting it anymore.
1462 * Improvements of the SMPI replay tool:
1463 - Most of the collective communications are now rooted in the same process as
1464 in the original application.
1465 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1466 used until now). Multiple data types can now be used in a trace.
1467 - The replay tool now supports traces produce either by TAU or a modified
1469 - Bug Fix: the compute part of the reduce action is now taken into account.
1470 - Gatherv collective is now supported
1471 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1472 of smpirun outputs time independent traces for the current run. One file
1473 is created per process. If too many processes are simulated, this behavior
1474 can be changed to one file for all processes by using the
1475 tracing/smpi/format/ti_one_file flag
1476 * smpirun generates the host file if needed (with given host count and platform)
1477 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1478 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1479 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1480 --cfg=smpi/coll_selector:(mpich/ompi)
1481 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1483 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1484 happening inside a collective SMPI call.
1485 * Fix the behavior of complex data types handling.
1486 * Make MPI_Wtime another synchronization point to take computations into
1488 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1489 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1490 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1491 (no privatization of global variables yet)
1492 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1495 * Verification of liveness properties is now available for SMPI applications
1496 (in addition to MSG applications)
1497 * Bugged examples using SMPI in examples/smpi/mc/
1498 * Add --cfg=model-check/visited option. Allows the verification of infinite
1499 programs. Detection of loops in the execution thanks to the system state
1500 comparison and reduction of the state space to explore. Can be combined with
1501 DPOR for safety properties.
1504 * Allow to change SimGrid configuration (see --help) within the code
1505 thanks to SD_config() as it can be done in MSG.
1506 * Add a new function SD_task_set_amount() upon user request.
1509 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1510 * Remove rule based routing (no more PCRE dependency)
1511 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1512 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1513 * Add a loopback_bw and loopback_lat options to cluster tag.
1514 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1515 routing. Now peers are encapsulated in an AS and have their own private
1516 router but this is transparent.
1519 * Our own implementation of getline is renamed xbt_getline, and gets
1520 used even if the OS provide a getline(). This should reduce the
1521 configuration complexity by using the same code on all platforms.
1522 * New type: xbt_cfg_elm_boolean.
1523 * Allow to use yes/no for boolean configuration options in the command line.
1524 * Allow to disable SimGrid cleanups at exit from command line option.
1525 There are situations where one may want a simulation to end with an exit.
1526 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1527 annoying when scripting around the simulator. Adding a
1528 --cfg=clean_atexit:no allows to circumvent this issue.
1531 * Lots of memory leaks were corrected in this release.
1532 There is no known memory leaks anymore, in all of our 600+ tests.
1533 * New command line option --version, to get SimGrid version information.
1534 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1536 * Supernovae builds are deprecated, and expected to be removed in the next
1539 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1541 SimGrid (3.9) stable; urgency=low
1543 The Grasgory release: GRAS is really dead now.
1545 * Complete overhaul of the internal host structures scheme.
1548 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1549 was considered as experimental and badly maintained since then.
1550 * Keeping it was thus a trap to our potential users, that could take
1551 it instead of MSG or SMPI by mistake despite is pity state.
1552 * GRAS seems to have very few users (if any), and no one volunteered
1553 to maintain it further. It also induces a lot of XBT code (for
1554 portability sake), that must be maintained too.
1555 * For all these reasons, we killed GRAS. If someone wants to revive it
1556 in the future, don't cry, our git history still remembers of GRAS.
1559 * Major overhaul. Merge our documentation again as time proved that
1560 splitting it was really not helping our users.
1561 * Further improve the developer documentation to help newcomers
1562 hacking on SimGrid itself. The user documentation (and in
1563 particular, the beginner documentation) is still in a sorry state.
1566 * Now works on Windows too!
1567 * Much more extensive test suite, from MPICH
1570 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1571 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1572 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1573 * Bug fix: let task be scheduled when the last dependency to be solved is
1574 a control dependency.
1575 * Remove SD_load_environment_script function.
1576 Use the C sg_platf function if you want to declare a platform
1580 * New function: MSG_process_get_number()
1581 * Old function documented: MSG_config()
1582 * Remove MSG_load_platform_script function
1583 Use the C sg_platf function if you want to declare a platform
1587 * Change the default value of the TCP_gamma constant (maximal size of TCP
1588 congestion window) to a more realistic 4MiB value. If you notice changes in
1589 your simulation results, you can fall back to the previous 20k tiny window
1590 by adding --cfg=network/TCP_gamma:20000 on command line.
1591 * (Hopefully) fix a bug wrt periodic availability and state traces
1592 * Bug fix: use default values at start when first event in availability/state
1593 trace is not at time 0.
1596 * remove the "new_" part of function name sg_platf_new_trace_connect
1597 (resulting in sg_platf_trace_connect), since it does not create
1601 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1602 use case where it's really mandatory, and maintaining it was a pain in
1604 * New: xbt_fifo_search(), search an item with a user-provided
1605 comparison function instead of dumb pointer comparison.
1608 * Fix the lua deployment:
1609 Use `simgrid.init_application()` before deployment instead of
1610 `simgrid.msg_register_application()` after.
1613 * Transfer the tracing files into the corresponding modules.
1615 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1617 SimGrid (3.8.1) stable; urgency=low
1619 The "we are told that some people want to also *install* the simgrid
1622 * Add missing file "tesh.1" to the archive.
1624 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1626 SimGrid (3.8) stable; urgency=low
1628 The Psssshiiiit release: SimGrid jumps into the Cloud.
1631 * Add an experimental interface to manipulate VMs. They are mainly
1632 process groups with very few intrinsic semantic, but they should
1633 allow you to build the semantic you want easily.
1634 * New function: MSG_host_set_property_value()
1635 * New function: MSG_process_on_exit(). To clean memory in all cases.
1636 * Bug fixes that made the host (and link) failures unusable.
1637 * Add a way to auto-restart process when the host in which they are
1638 executing comes back (ON_FAILURE="RESTART" on deployment file,
1639 MSG_process_auto_restart_set).
1640 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1641 please stop using the old ones, they are DEPRECATED.
1643 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1644 Please use MSG_init() instead. (reducing the amount of entry
1645 points in the library helps us).
1646 * Make it impossible to link against the wrong version of the lib
1647 * Deprecate MSG_clean(). No need to call it anymore.
1648 * Function MSG_get_host_number() is not deprecated anymore.
1651 * Split the doc into a user guide and a reference guide.
1652 * Start a developper guide to help people hacking on SimGrid.
1655 * Enable tracing by default. This modules rocks you should use it.
1656 * Remove option custom_flags. Now use environment variables CFLAGS
1658 * Use default cmake things to detect lua instead of home grown ones.
1659 * New option "enable_mallocators" to disable mallocators, for debugging
1660 purpose ("on" by default).
1663 * Bug fixes around the resource failures: don't let the processes
1664 survive the host they are running onto.
1665 * Add an interface to auto-restart processes when the host in which they are
1666 executing comes back.
1667 * Ensures that SIMIX_clean is called automatically. It's not part of
1668 the public interface anymore (bindings should be updated).
1671 * Bug fix for when SD_Simulate is called with a positive value: be careful
1672 when comparing doubles. Sometimes they are different for non significant
1674 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1675 parallel task whose initial work is distributed among host according
1676 to the Amdahl's law. Such tasks are created with a parameter alpha
1677 that corresponds to the non-parallelizable part of the computation.
1678 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1679 between two sets of workstations assuming a 1D block distribution (each
1680 workstation owns a similar share of data) on both sides.
1682 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1683 Data redistribution will be automatically scheduled once parent and child
1684 are both scheduled. The filling of computation_amount and
1685 communication_amount structures is now done seamlessly thanks to the chosen
1687 * New function SD_workstation_dump to display various information
1688 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1689 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1690 bandwidth on the route onto which the task is scheduled and the amount of
1692 To divide the nominal bandwidth by 2, the rate then has to be:
1693 rate = bandwidth/(2*amount)
1694 * Compute tasks that have failed can now be rescheduled and executed again
1695 (from their beginning)
1696 * Increasing source code coverage (src/simdag is now covered at 95.8%
1700 * Re-implement time-independent trace replay using SMPI (at the
1701 smpi_smp_* level) instead of MSG. This should replace
1702 examples/msg/actions/actions.c
1703 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1704 hindexed and structs)
1705 * Implement the exchange of non-contiguous data.
1706 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1707 * Correct behavior of smpi/sender_gap and set its default value to 0
1708 * Add option to asynchronously send small messages to allow better
1709 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1710 specifies the size in bytes under which messages will be asynchronously sent.
1711 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1712 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1715 * New C interface to define a platform: XML is now optional.
1716 For more info, please check include/simgrid/platf.h.
1717 * New interface to define random platforms from the C:
1718 For more info, please check include/simgrid/platf_generator.h and
1719 examples/msg/masterslave/masterslave_platfgen.c
1720 * Export a sg_cmdline dynar containing all the arguments we got from
1724 * Two new tracing options for adding comments to trace file so you
1725 can track your experiments (see --help-tracing for details).
1726 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1727 * Adding the SimGrid version that generated the trace file as a comment.
1728 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1729 * Fix to avoid key clashes on Paje links
1730 * Other minor fixes related to the Paje specification
1733 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1734 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1735 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1736 * Greatly improve the robustness of mmalloc to user errors (such as
1737 using an area after freeing it, or freeing it twice)
1739 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1741 SimGrid-java (3.8.1) stable; urgency=low
1743 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1744 Warning: all methods are visible, but only some of them are
1745 implemented so far. Check the source (src/jtrace.c)
1746 for further information.
1747 * New module: org.simgrid.msg.File (SimGrid File management functions)
1748 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1749 * Change the meaning of Process.restart: now restart the process from
1750 the beginning, like MSG_process_restart in C.
1751 * Add Process.setAutoRestart: handling of process restart when failed
1753 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1754 you to retrieve the properties of the processes/hosts
1755 * Deprecate Msg.clean(): you can just forget about it now.
1756 * New function Process.getCount(), that only works when compiling
1757 with the not yet released version 3.9 of the C library.
1759 * New context factory based on Coroutines. It mandates a modified JVM
1760 but then, the simulations run about five times faster, and there is
1761 no limit to the amount of processes (beside of the available memory).
1763 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1765 SimGrid (3.7.1) stable; urgency=low
1768 * Restore the prototype of MSG_process_create_with_environment() to
1769 the pre-3.7 situation by removing the kill_time argument.
1770 * Add a MSG_process_set_kill_time() function instead.
1773 * Fix weird behaviors when dealing with parallel tasks.
1776 * Simgrid is now built as a dll.
1777 * Simgrid-java now works on Windows.
1778 * Simgrid-Java is now included into Windows package.
1781 * First pre-build package for MacOSX.
1784 * Fix compilation when using MSG_USE_DEPRECATED.
1785 * Fix some compilation issues on Macs and Windows.
1786 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1787 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1789 We discovered that the Lua console is broken, but we are missing the
1790 manpower to fix it right now. The problem existed in 3.7 too, so we
1791 are not blocking the release for that. Sorry if you depended on this
1792 feature, any help would be really welcome.
1794 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1797 SimGrid-java (3.7.1) stable; urgency=low
1799 The "Java aint got to be bloated and slow" release
1802 * Various internal cleanups and performance improvement
1803 Simulations are expected to run up to twice faster or so
1804 * Make Process.kill(process) an instance method, not a static one
1805 * User processes are not java.lang.Thread subclasses.
1806 This breaks the compatibility (sorry), but previous API was
1807 brain-dead, making it impossible to have non-trivial
1808 initializations in the process constructor.
1809 * Require a full constructor per Process sub-class.
1810 Kinda breaks the compatibility (sorry), but this allows a much more
1811 efficient way to launch the processes at simulation startup.
1812 * Do not embeed our version of semaphores, java 1.5 can be considered
1813 as sufficiently prevalent for us to not dupplicate its features.
1817 * Add examples for almost every part of the API
1818 We spotted and fixed a lot of bugs in the process
1820 * New module: asynchronous communication API
1821 * New function: Process.sleep()
1822 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1823 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1824 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1826 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1828 SimGrid (3.7) stable; urgency=low
1830 The "spring cleanups (before the next Big Project kicks in)" release.
1833 * We can specify the SMPI latency/bandwidth factor with command line
1834 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1835 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1836 You can also use the "config tag" from platform file by adding this line
1837 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1838 (see "example/platforms/tag_config.xml" to use "config tag").
1839 Note that the command line supersedes the platform file configuration.
1840 * Change the correction factors used in LMM model, according to
1841 the latest experiments described in INRIA RR-7821.
1842 Accuracy should be improved this way.
1843 * Use the partial invalidation optimization by default for the
1844 network too. Should produce the exact same results, only faster.
1845 * Major cleanup in surf to merge models and split some optimization
1846 mechanisms from the core of the models. As a result you can now
1847 specify which model to use (e.g., --cfg=network/model:LV08
1848 --cfg=cpu/model:Cas01) and which optimization mode to use
1849 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1850 Incompatible combinations should err at initialization. See
1851 --help-models for the list of all models and optimization modes.
1852 * The CLM03 workstation model was dropped for simplicity because it
1853 used the deprecated CM02 network model. Use default instead.
1854 * Rename the TCP_gamma configuration option to network/TCP_gamma
1855 * Rename the coordinates configuration option to
1856 network/coordinates, and document it
1857 * Use now crosstraffic keyword instead of the terribly misleading
1858 fullduplex keyword. It is activated by default now in the current
1859 default model, use --cfg=network/crosstraffic:0 to turn it off.
1860 * Ongoing refactoring the model parsing to make XML files optional
1861 See include/simgrid/platf.h for details (still to be completed)
1864 * Major overhaul of the documentation. Almost instructive now :/
1865 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1866 functions and friends. This interface was considered as
1867 deprecated since over 2 years, it's time to inform our users that it is.
1868 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1869 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1870 use these (crappy) functions in your code.
1871 These functions will be removed soon. Stop using them now.
1872 * Deprecate MSG_get_host_{table,number}
1873 Implement MSG_hosts_as_dynar() instead.
1874 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1875 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1878 * Stabilize the parallel execution mode of user contexts
1879 * Introduce configuration variables to control parallel execution:
1880 - contexts/synchro: Synchronization mode to use when running
1881 contexts in parallel (either futex, posix or busy_wait)
1882 - contexts/parallel_threshold: Minimal number of user contexts
1883 that must be part of a scheduling round to switch to parallel
1884 execution mode (raw contexts only)
1885 * Fix bugs that prevented to use suspend/resume along with
1886 synchronization structures.
1887 * Fix bugs in process termination that lead to invalid memory access
1888 in very specific conditions.
1891 * Introduce a parallel mode for the models (controlled by surf/nthreads
1892 configuration item). In our tests, running the models in parallel
1893 never lead to any speedups because they are so fast that the gain
1894 of computing each model in parallel does not amortizes the
1895 synchronization costs, even when ultra fast futexes are used.
1896 This is released anyway because YMMV.
1899 * Performance boost by using a swag internally to compute the set of
1900 tasks that are finished and should constitute the return value of
1904 * Enable it by default now that it is considered rather stable.
1907 * Documentation of the tracing functions.
1908 * Performance gains when tracing categorized/uncategorized resource
1909 utilization by avoiding calls to get route when updating resource
1910 variables. LMM constraints are being used instead.
1911 * API changed to set task categories. Use MSG_task_set_category instead
1912 of TRACE_msg_set_task_category, and SD_task_set_category instead
1913 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1914 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1915 * Examples for link user variables added (see at examples/msg/tracing/)
1916 * Deprecated function TRACE_msg_set_process_category completely removed
1917 * Trace header updated according to the latest Paje file format
1918 * Tracing network lazy updates, no longer obligate users to use full updates
1919 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1920 * Experimental: let user code declare/set/push/pop application states for hosts
1921 * API changed to allow the manual creation of graph configuration files
1922 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1925 * Improve the API of Lua MSG bindings, using the Lua spirit.
1926 * Each simulated process now lives in its own Lua world (globals are
1927 automatically duplicated). It helps writing simulators. Will allow
1928 to run Splay programs within SimGrid in the future.
1929 * Add a Chord example in Lua, equivalent to the MSG one.
1932 * Start the implementation of a solution to express temporal
1933 properties, not only local assertions. This is still an
1934 experimental work in progress, stay clear from it to be safe.
1938 - Add new runtime parameters --help-logs and --help-log-categories
1939 to display information about supported logging parameters and
1941 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1942 don't exists anymore.
1943 * Mallocators: allow value NULL for the reset function.
1945 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1946 create homogeneous dictionaries, where all the elements share the
1947 same free function. Non homogeneous dictionaries will be
1948 deprecated in the next release.
1949 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1950 - Multi-level dictionaries are deprecated.
1952 - new function xbt_dynar_search_or_negative() that is useful when
1953 you have less than 2 million elements in your dynar and don't
1954 want of the extra complexity of catching exceptions when the
1955 element is not found.
1957 - Make xbt_os_thread module (for thread portability) public.
1958 Documentation is still to come, sorry.
1960 - Cleanups and simplifications to make it maintainable again.
1961 - Exotic features (such as memalign and valloc) were removed.
1962 - The metadata were extended and improved so that the
1963 model-checker becomes able to explore and inspect the heaps.
1964 - This may induce a performance drop when enable_model-checking is
1965 ON in cmake (even if it's not used in the simulation), but it is
1966 necessary at this point to get MC working.
1968 Turn model-checking OFF if simulation performance matters to you.
1969 Not enabling it at runtime is not enough, disable it in cmake.
1971 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1974 SimGrid (3.6.2) stable; urgency=low
1976 The "Not coding new stuff allows to polish old things" release.
1979 * New bindings to the NS3 packet level simulator (experimental)
1980 * Use the raw (efficient) execution contexts instead of the sysv
1981 (portable) ones when possible.
1982 * libpcre is now mandatory in any cases since not using it led to
1983 severe performance loss and possibly other issues
1984 * Update the XML platforms:
1985 - G5K: include the latest machine in Nancy
1986 - GridPP and LCG: new platforms
1987 * Documentation was partially updated, at least (more to come)
1989 Bug fixes, cosmetics and small improvements
1990 * Free terminated processes before the end of the simulation to avoid
1991 exhausting the memory of users having very dynamic amount of
1993 * Bug fix and cosmetics about canceling non-running tasks
1994 * Bug fix about the dot loader's issues when using libcgraph
1997 * Create an installer for windows with nsis (amd64 and win32)
1998 - Add an hello world project to illustrate simgrid project creation.
1999 - Embed libpcre into the Simgrid installer to avoid
2000 its compilation burden
2001 * The raw execution contexts should work on Apple now
2002 * Port to Windows 64 bits
2003 - Sysv contexts now have an implementation for this arch
2004 - GRAS communication features now support this arch
2005 * Drop support for borland compiler on windows
2006 - this code was not maintained, and we kinda depend on gcc nowadays
2007 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2008 * Fix portability issue on unstable ubuntu: linker became picky on
2011 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2013 SimGrid (3.6.1) stable; urgency=low
2015 The "Oops, we broke Macs too" release
2018 * Fixed contexts detection so that raw ones are used when possible
2019 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2020 leads to a strange error, with user code segfaulting sometimes when
2021 the generated code is not perfectly aligned (which is not
2022 controllable from the user side, depends on the amount of code)
2025 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2027 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2029 SimGrid-java (3.6) unstable; urgency=low
2032 * Split of every thing from simgrid v3.5 into a separate package.
2034 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2036 SimGrid (3.6) stable; urgency=medium
2038 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2041 * Bindings now constitute their own package, separated from the main one.
2042 Rationale: reduce our maintenance nightmare by reducing the module coupling
2043 They will soon be released on their own on gforge.
2045 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2046 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2048 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2049 * It's not quite deprecated for now because we have no replacement,
2050 but it may soon become the case.
2053 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2054 * Fortran: New user-level cache variable to store the rank of the running
2055 process. This improves performance by an order of magnitude.
2056 * C: New coccinelle script to automatically locate and modify global and
2057 local static variables.
2058 * Improved SMPI network model with a sender-side gap to account for multiple
2062 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2063 only say if a communication is finished, no matter whether it succeeded or
2064 failed. You can call MSG_comm_get_status() to know the status of a finished
2066 * New function MSG_task_dsend() to send a task and detach it. When a
2067 communication is detached, you are never notified of its success or failure
2068 and the memory is released automatically once it is finished. This function
2069 is useful when you don't care about the end nor the success of a
2071 * Change the prototypes of action replay. Sorry for inconvenience,
2072 but this is really more efficient this way (and to adapt your code,
2073 you just have to fix the initialization, that shouldn't be too long)
2074 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2075 ever managed to do anything useful with it.
2076 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2077 takes a dynar of communications. It returns immediately and gives the
2078 index of a finished communication (if any).
2079 * New example: a basic implementation of the Chord P2P algorithm.
2082 * New model for multi-core CPUs. You can now use the core attribute to
2083 precise the number of cores of a host. This is a basic model. Every
2084 process running on the host receives at most the power provided in
2085 the DTD (throughput<=power). Total throughput of process cannot exceed
2087 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2088 router linked by an up-link and a down-link (possibly asymmetrical).
2089 This kind of pattern allows to easily build last-mile model style platforms.
2090 Aggregating such patterns in a rule-based AS is thus the technique of
2091 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2092 * New model for Vivaldi routing. We transformed the Vivaldi network model
2093 into a Vivaldi routing model (based on the rule-based model). This allows to
2094 combine Vivaldi based latencies with last-mile platforms.
2097 * Added a check for NaN of IEEE754 infinite in the double entries of
2099 * Introduce a new context factory "raw", highly inspirated from the
2100 ucontext factory, but using manually crafted functions in assembly to
2101 do the work in an efficient manner.
2102 * Allow to change the used context factory at run time, not only at
2103 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2104 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2105 contexts at runtime (only with raw contexts or ucontexts).
2106 * Completely rewrote this module to allow parallel execution of user
2107 processes. Use --cfg=contexts/nthreads:N to execute user processes
2108 with N parallel threads (the default is 1, meaning no parallelism).
2109 * Allow to decide dynamically between sequential and parallel modes.
2110 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2111 processes in parallel only when their number is greater than or equal to P
2113 * Added a check for NaN of IEEE754 infinite in the double entries of
2117 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2118 won't output the state of processes when interrupted with Ctrl-C
2119 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2120 NULL-terminated array. This may solve backward compatibility issues
2121 due to the change to return type of SD_simulate. See also:
2122 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2123 * Add new macros with variable number of arguments.
2124 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2125 - in xbt/asserts.h: xbt_assert
2126 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2127 - in xbt/ex.h: THROWF and RETHROWF.
2128 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2130 * Change xbt_die() to accept a format string with arguments, just like printf.
2131 * New data structure: xbt_lib_t, like a dict but more general and with better
2135 * New configuration options
2136 Options triva/categorized and triva/uncategorized can be used to generate
2137 graph configuration files for Triva visualization tool.
2138 * Configuration option tracing/platform is renamed to tracing/categorized
2139 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2140 * New TRACE_user_link_variable interface:
2141 User provides the name of the link and the tracing variable to attach to it
2142 * the declaration of tracing categories must be done after the environment
2144 * simpler tracing interface, just one way to declare categories
2145 TRACE_category or TRACE_category_with_color, it is up to you
2146 * links in the trace file are again identified by their names
2147 * trace contains the full platform hierarchy exactly as declared using the ASes
2148 * Options tracing/msg/[task|process]:1 groups the process by hosts
2149 for both cases, tasks and processes must have names that are unique during the simulation
2150 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2151 * The experimental option tracing/msg/volume is deprecated
2152 its functionality may be reincorporated if needed
2154 The tracing generates a trace file with unordered timestamped events,
2155 because of the way the core simulator (surf) works. A script available
2156 at the tools directory (fix-paje-trace.sh) can be used to put the events
2157 in order. We have changed the tracing so it can generate ordered timestamped
2158 events in the final trace, but depending on the simulator (and how much time
2159 is simulated) that can lead to a huge memory utilization. It is deactivated
2160 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2162 Build Infrastructure
2163 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2164 - We are trying hard to keep the API stable, but it may happen that
2165 some things change (we're a research project after all, not a
2166 nuclear plant operating system). If such things should happen, you
2167 could rely on that macro to adapt.
2168 - current value: 30600 for 3.06.00, aka 3.6
2169 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2170 a number that can be compared with SIMGRID_VERSION.
2171 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2172 assertions and verbose/debug logging events are disabled at compile time.
2174 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2176 SimGrid (3.5) stable; urgency=medium
2179 * New feature to any SimGrid-based simulator: Model-Checking
2180 Check SIN#1 for more details.
2183 * New Model SMPI (three-interval linear regression for correction factors)
2184 See RR-7426, available at http://hal.inria.fr/inria-00527150
2185 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2186 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2187 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2188 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2189 * See also src/smpi/README
2194 - Tracing API changes: TRACE_start and TRACE_end should not be called
2195 by user-code. They are automatically called by simulators created
2196 with SimDAG, MSG and SMPI if the toolkit is compiled with
2197 tracing_enabled=ON. Categories declaration and utilization remain the
2198 same for MSG and SimDag.
2199 - A function was added to the tracing API to declare categories with
2201 - TRACE_category_with_color (char *category, char *color)
2202 where color must be in the following format
2203 "%f %f %f", red, green, blue
2204 and red, green, blue are float values in the interval [0, 1]
2205 - User can specify NULL as color parameter, or continue calling
2206 TRACE_category (cat)
2207 On that case, the tracing system will define random colors
2208 - The following command-line options are supported:
2209 --cfg=tracing/filename:msg.trace
2210 --cfg=tracing:1 (activate tracing, needed to use others)
2211 --cfg=tracing/platform:1 (categorized resource use)
2212 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2213 --cfg=tracing/msg/task:1 (task creation)
2214 --cfg=tracing/msg/process:1 (process creation, migration)
2215 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2216 --cfg=tracing/smpi:1 (SMPI interface tracing)
2217 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2218 - examples of examples/msg/tracing updated
2220 - DAXLoader and DOTLoader functions can generate tasks with categories
2221 - A new function to attribute a category to SD tasks:
2222 TRACE_sd_set_task_category (SD_task_t task, char *category)
2223 * Tracing the MPI interface implemented by SMPI
2224 - Collective operations are traced with states
2225 - Point-to-Point operations are traced with states/links
2226 - Tracing activated by a parameter "-trace filename" passed
2227 to smpirun during execution (considering that simgrid
2228 is compiled with tracing enabled)
2229 - To run the simulation with gdb, the simulator
2230 accepts --cfg=tracing/smpi:1 to trace SMPI
2231 - tesh files to check if smpi tracing is ok
2232 - See examples/smpi/NAS/DT-trace
2233 * GTNetS tracing re-worked
2234 - adaptation to the tracing system of GTNets to cope
2235 with modifications regarding the fullduplex mode
2236 - new tesh files to check if gtnets tracing is ok
2239 * Asynchronous communications through the functions:
2240 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2241 * New function: MSG_load_platform_script()
2242 to make possible using a lua script instead of XML files to set up platforms
2243 * New function: MSG_set_function
2244 to associate functions to processes, used when bypassing the parser
2245 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2247 Platforms: Add some more examples in examples/platforms
2248 * Grid'5000: see www.grid5000.fr
2249 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2252 * Change the XML format. This is a very important modification. SimGrid
2253 3.5 introduces a new hierarchical format based on the notion of
2254 Autonomous Systems. Compatibility with old format is ensured
2255 through the perl script provided in the install bin directory
2256 bin/simgrid_update_xml.
2257 It is now possible to build platforms with specific routing
2258 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2259 connect several platforms together. We will try to provide soon
2260 set of realistic platforms exploiting these properties (have a
2261 look at examples/platforms/ for the moment).
2262 * Take the opportunity of the XML format change to be a good XML citizen:
2263 rename link:ctn to link_ctn and similar changes (also dealt with by
2265 * Add a new routing scheme (rule-based) using regular expressions. It
2266 enables to have an extremely low memory footprint when the
2267 underlying routing is simple and can be compactly described. You
2268 need to have libpcre4-dev (perl regular expressions) installed if
2269 you want to use this routing scheme.
2270 * Revive the cluster TAG and allow to easily and efficiently (both in
2271 term of memory and speed) connect clusters together. Have a look
2272 at teshsuite/simdag/platforms/ to see how this can be done. With
2273 this tag, you can create clusters with thousands of tasks at no
2274 cost (have a look at examples/platforms/).
2275 Note: clusters are implemented as ASes, so there is no need for an
2276 enclosing AS tag if you have only one cluster in your platform.
2277 * Add new generic functions in the public interface that allows the user
2278 to call SURF 'create_resource' methods from your code (same
2279 functionality as the XML bypass mechanism but with a much lighter
2281 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2282 that uses a piecewise linear approximation to produce better
2283 results when exchanging small messages.
2284 * Add a new parameter to handle correctly full duplex link and account
2285 for interferences between uplink and downlink communications
2286 (activate with --cfg=fullduplex:1).
2289 * Rename the SD_READY (all dependencies are satisfied and task is
2290 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2291 dependencies are satisfied) state.
2292 This prevents a confusion between the notion of "ready to schedule"
2293 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2294 simulated" (SD_RUNNABLE) used by the simulation kernel.
2295 * Change the way a task is considered as ready. Instead of removing
2296 dependencies when a task is done, a counter is decreased. This way,
2297 it is always possible to reach ancestors thanks to the
2298 SD_taks_get_parents function (even after the end of the simulation.)
2299 * Change the return type of SD_Simulate from (SD_task_t*) into
2300 xbt_dynar_t. This function was in handling a dynar internally and
2301 converted it into a NULL terminated array for historical reasons.
2302 * New function SD_dotload(char*) to load a DAG described in dot
2303 format. This loader and the corresponding examples require the
2304 installation of the graphviz library.
2305 * Fix a bug in the management of tasks of size 0 in the surf network
2306 models. This problem was only visible with SIMDAG and you should
2307 thus disregard results produced with earlier versions if you
2308 relied on this feature (some tasks were blocked because of this).
2309 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2310 even though your applications were doing only point-to-point
2311 communications and sequential computations. Now you can really use any
2312 model you want (of course, if you create real parallel tasks, which are
2313 not implemented in most models beside ptaskL07, this will abort).
2314 * Add an example that schedules a DAX on an heterogeneous platform
2315 using a Min-Min strategy.
2316 * New function SD_workstation_get_current_task() that returns the kind
2317 of task currently running on a workstation in the sequential access
2319 * Raise some warnings when unexecuted tasks remains at the end of the
2320 simulation. This is usually caused by cycles in the DAG.
2323 * New function: SIMIX_process_set_function() called by MSG_set_function
2324 * Change the underlying waiting queue in semaphores so that a process
2325 can wait on several of them simultaneously (as in waitany).
2326 * Fix the way to handle tokens in semaphores so that all access patterns
2327 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2328 * kill the dirty pimple SIMIX_message_sizes_output()
2329 Please use (proper) visualization instead
2332 * New data container: setset (set of sets of elements)
2333 * New module: mmalloc (mapped malloc, allowing to have several
2334 independent segments of malloc)
2335 * New function: xbt_dict_cursor_set_data()
2336 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2337 * New function: xbt_dynar_is_empty()
2338 * New function: xbt_fifo_get_last_item()
2339 * Fix xbt_dynar_shrink(): use the right element size.
2340 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2341 uninitialized areas during expand.
2342 * Fix semaphores: previous implementation was severely broken.
2343 * Use library init/fini functions for our initialization.
2344 - you can use logs and other feature as soon as you want in your
2345 code (even before the xbt_init / MSG_init)
2346 - xbt_exit is now a no-op and produce a warning when used.
2349 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2350 benefit from the latest improvement to the simulation kernel.
2351 * Kill measurement sockets for now. If you rely on them, sorry. This
2352 release is not for you. This feature will be reintroduced in the
2353 future, but we cannot delay the release any further.
2354 * New function: gras_msgtype_get_name().
2355 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2356 * Fix (at last) the pmm example: it should not randomly fail anymore.
2358 Build chain: bug fixes and overall polishing
2359 * Cmake is now stable enough. Hence, we killed the autotools.
2360 * Port to windows ( TM :)
2361 * Fix the 'make install' target.
2362 No need to use 'make install-simgrid' anymore
2363 * Introduce a 'make dist' target compiling a *source* archive
2364 'make package' compiles a binary archive
2365 * Compile java files only on need
2366 * Add --cd and --setenv command line options to tesh
2367 * Out of source builds are not fully supported yet, but we are close
2368 * Enable supernovae and optimization flags by default for our users
2371 * Add layer to set up environment directly from lua, without XML.
2372 * The effect of gras_stub_generator can be achieved through
2373 lua too (check examples/gras/console/ping_generator.lua)
2375 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2377 SimGrid (3.4.1) stable; urgency=low
2379 The "Polishing easter eggs is probably a good idea" release.
2380 This is a bug fixes release only.
2384 * Fix a bug preventing the tasks from begin garbage collected.
2387 * Fix a bug occurring when a host involved in a communication fails.
2388 This was not detected properly by the other peer involved in the
2389 communication. Now, it's reported as a network error.
2392 * Warn the user about loop dependencies in data flow of DAX files
2393 * Obey the control-flow dependencies of DAX files
2396 * Add option "enable_smpi" allowing to not compile SMPI.
2397 Probably useful for the (Mac) users experiencing a build error here
2398 * Improve the detection of lua5.1 and ruby1.8
2400 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2402 SimGrid (3.4) stable; urgency=low
2404 The "Easter in Cargese" release. Also known as (major changes):
2406 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2407 ~> bindings were greatly improved
2408 ~> new tracing infrastructure for better visualization introduced
2410 * the "Welcome to configury modernity" release.
2411 ~> we switched from autotools to cmake, and improved our cdash
2414 A more detailled list of changes follow (full detail in svn log).
2416 Java Bindings: Various Cleanups
2417 * (install java-gcj-compat-dev on debian-like to use them)
2418 * Remove put/get: no need to export deprecated interface in Java
2419 Use send/receive instead.
2420 * Cleanup the examples and add a README per directory
2421 * Remove example autoDestination (that's the only way to go now)
2422 * Remove example explicitDestination (was a plain copy of basic)
2423 * Make JniException a runtime exception, so that there is no need to
2424 declare the fact that you may encounter such a beast. I guess that
2425 nobody will ever want to survive such error.
2426 * Create specific errors for each MSG case of failure:
2427 host failure, transfer failure, timeout, task cancelled
2428 * Cleanup the exceptions that may get thrown by each function
2429 * Other internal cleanups in Java bindings. Performance still bad :/
2430 Ruby and Lua Bindings: create them
2431 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2432 * That's new and great, you should try them out.
2433 Same functionalities than Java bindings, only even less polished
2435 * Kill the useless "rate" argument of SD_task_get_execution_time()
2436 Everyone used to provide -1 as a value, it was not used, and the
2437 semantic of a possible use wasn't even clear.
2438 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2439 either as comm costs or compute costs to mean that there is no
2440 such thing for that specific task.
2441 * Add a SD_task_set_name() function
2442 * Fix SD_task_unschedule() on typed tasks
2443 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2444 * In DAX loader, accept useless 'level' attributes to <job> since
2445 LIGO DAGs have them (seem to be to ease graphical representation).
2447 * Add an example masterslave_mailbox.c using send/receive and not
2448 the deprecated put/get interface.
2449 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2450 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2452 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2453 (declare MSG_USE_DEPRECATED to still have the old name)
2454 * Add a MSG_task_set_data() function
2455 * About trace replay (see examples/msg/actions):
2457 - Allow to work with splitted trace files for each process
2458 Give the specific trace file as argument of each process,
2459 and call MSG_action_trace_run(NULL)
2460 You can still have one merged file for all processes.
2461 - Fix implementation of collective operations
2462 * Allow task_execute() on 0-sized tasks (closes #10063)
2464 * This is the first release of SimGrid where SMPI is not considered
2465 beta anymore (even if some corners should still be improved)
2466 * Port over the new SIMIX_network submodule (internal refactoring)
2467 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2468 * Implement more missing elements of the standard:
2470 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2472 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2473 MPI_Scan() MPI_Get_processor_name()
2474 - Added implementation of missing case for Alltoall (warning: it's
2475 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2476 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2477 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2478 SMPI_MPI_Allgatherv()
2479 * Bug fixes include:
2480 - MPI_Waitsome() was broken
2481 - Allow relative includes in smpicc
2482 - Command line cfg argument 'reference_speed' was ignored...
2483 - Some functions did not properly lead to auto-benching of user code
2484 - smpicc passes -O2 by default (just like openmpi one)
2486 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2487 * Bug fixes about timeouts during communications
2488 * add SIMIX_message_sizes_output() as a pimple to write to file the
2489 amount of messages per size. Use gnuplot to get histogram.
2490 Pimple because that's the only user-visible function of simix,
2491 defined directly in xbt.h (irk, sorry)
2493 - Add a SIMIX_sem_get_capacity() function
2494 - Fix interactions with processe resume/suspende
2495 - release_forever() was stupidly broken
2496 - Fix SIMIX_display_process_status() for processes in a semaphore
2497 - Make SIMIX_sem_block_onto() user-visible
2498 * Refactoring context stuff:
2499 - Use pseudo-OOP for better modularity
2500 - reimplement SIMIX_process_kill() without process_schedule() so
2501 that the latter can take as invariant that it is called from
2503 - Merge context_start into context_new for sake of simplicity
2505 * Add a Vivaldi network model, coded live during SUD'10 ;)
2506 * Rename configuration variables to start a hierarchy:
2507 o cpu_model -> cpu/model
2508 o network_model -> network/model
2509 o workstation_model -> workstation/model
2510 * New configuration variables:
2511 o network/bandwidth_factor: correction to bandwidth
2512 o network/latency_factor: correction to latency
2513 o netwotk/weight_S: correction to the weight of competing streams
2514 * Add a long description to the models, that users can see with such
2515 argument on the command line: --cfg=cpu/model:help
2516 * --help-models display the long description of all known models
2518 * config: add the ability to set a default value after registration
2519 Does not override any previously set value (e.g. from cmd line)
2520 * dict: allow to have integer key and data.
2521 When so, you need to use the following functions
2522 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2523 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2524 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2525 In contrary to regular dicts, the key is not malloced before copy.
2526 Mixing scalar and regular elements in the same dict is not tested
2528 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2529 Tracing for Visualization:
2530 * SimGrid is now instrumented in order to generate a trace file for
2531 visualization analysis: to use it, need to compile SimGrid with the
2532 "tracing" option enabled, and instrument the program using SimGrid with
2533 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2534 (among other functions).
2535 * The instrumentation only traces the platform utilization for now
2536 * Documentation to use the tracing functions and how to analyze the
2537 traces with the Triva tool is written.
2538 * More information about: SimGrid FAQ (in the section Tracing Simulations
2541 * We moved to cmake as default build system. Autotools support will
2542 be dropped soon. Check the FAQ for more info about how to use it.
2543 * Greatly improved our cdash/ctest interactions
2544 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2545 * Added memory checking tests with valgrind; lot of memleak fixing.
2546 This may be the first release of simgrid with so few memory issues
2547 * Added code coverage tests.
2548 Our coverage is still improvable, but at least we see it on cdash.
2550 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2552 SimGrid (3.3.4) stable; urgency=low
2554 The "Desktop Grid needs love too" release (also called Xmas release).
2556 Models improvements:
2557 * Major speedup in the maxmin system solving by using lazy evaluation
2558 Instead of solving completely the maxmin system at each iteration,
2559 only invalidate (and recompute) the modified parts.
2560 This new feature is enabled in default models but you can try to
2561 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2562 * Cas01 IMproved as default CPU model
2563 This CPU model is the same Cas01 model, but it uses the
2564 maxmin-selective-update flag and a heap structure to manage
2565 actions on SURF kernel.
2566 It reduces the complexity to find the next action to finish and,
2567 consequently, it's faster than the old Cas01.
2568 This is the new default CPU model (Cas01).
2569 * Rename the old Cas01 model to Cas01_fullupdate
2570 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2571 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2572 * CpuTI (CPU Trace Integration)
2573 A new CPU model whose objective is simulate faster when using
2574 availability trace files.
2575 Instead of using a full featured, over engineered maxmin system for
2576 CPU modeling, this model does the pre-integration of traces files
2577 to calculate the amount of CPU power available, and so, executes
2578 faster than the old CPU models.
2579 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2580 * Use LV08 as default network model since it gives better accuracy
2581 for small messages and shouldn't change things for big ones.
2582 Use --cfg=network_model:CM02 to get the previous behavior.
2585 ******************************************
2586 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2587 ******************************************
2588 * The new CPU model may changes simulations!
2589 The point is that events occurring at the exact same timestamp
2590 are not scheduled in the same order with the old and new
2591 version. This may be enough to completely change the execution
2592 of simulations in some cases.
2593 * The new network model will change simulations!
2594 This new model is more realistic than the previous one, so you
2595 should consider redoing your old experiments with this model.
2596 Sorry for the inconvenience.
2599 * Introduce the supernovae compilation mode
2600 When compiled that way, the whole SimGrid (or almost) is put in a
2601 single compilation unit and compiled in one shoot.
2602 This is to help gcc which has difficulties to inline stuff from one
2604 The speedup seem to be above 15%, althrough more tests are needed on
2605 amd64 to confirm that gain.
2608 * Port of MSG's mailbox on top of SIMIX network
2609 The put/get mechanism was greatly simplified on the way.
2612 * New SIMIX network module. Provides:
2613 - Mailbox: rendez-vous mechanism to find with who you want to speak
2614 - Synchronous send/recv: easier and hopefully faster since the
2615 logic is handled in the maestro process directly now
2616 - Asynchronous send/recv: you dreamt of it? It's here now
2617 Too bad that nobody cared enough to propagate the change to MSG.
2618 * Add semaphores as SIMIX synchronization mechanism.
2621 * new function SD_daxload(char*) to load a DAX file
2622 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2623 * Introduce typed tasks. Specify its kind and cost at creation.
2624 At scheduling, just give where it should be placed, and the cost
2625 for each involved resource is automatically computed.
2626 Existing constructors so far (more to come of course):
2627 - SD_task_create_comm_e2e() for end-to-end communication
2628 - SD_task_create_comp_seq() for sequential computation
2629 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2630 * new function SD_task_dump() for debugging display
2631 * new function SD_task_dotty(task,FILE*) writing to file the info
2632 about the task in dotty format
2633 * SD_task_dependency_exists() can now cope with having one of its
2634 arguments NULL. If so, it tests whether the other argument has any
2636 * Add getters on list of preceding/following tasks:
2637 SD_task_get_parents(task) and SD_task_get_children(task)
2638 * Add getters on amount of workstations and list:
2639 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2640 * Add getter on task kind: SD_task_get_kind(task)
2641 * Update the start_time and finish_time of tasks on completion/failure
2642 * Bugfix: Remove task from state swags when destroyed
2645 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2648 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2649 dynar with one separate thread per value of the dynar.
2650 * Change the prototype of xbt_thread_create(), sorry.
2651 Added a boolean parameter indicating whether we want to join this
2652 thread (used in SG only for now)
2653 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2656 * GTNetS wrappers should now be usable again (and betterly tested too)
2657 * Fix a major regression from 3.2 where the timeout provided to
2658 MSG_task_put_with_timeout() was used as absolute time before which
2659 the comm should be done.
2660 * Start to fix the <cluster> tag.
2661 - Internal links should be good now (beside of the loopback, which
2662 use the private link instead)
2663 - paths to the external world is still rather broken
2664 - the <route:multi> tag is just broken. Actually that's brain-dead.
2665 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2666 to make it less stupid
2667 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2668 * Fix a source-level compatibility glitch from 3.2: after defining
2669 MSG_USE_DEPRECATED, you can use the old name
2670 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2671 * Allow to compile from the SVN with automake 1.11
2672 * Fix some problems when using the "start_time" tag in deployment XMLs.
2673 * Fix #8569: XBT/synchro.h has redundant declarations
2674 * Fix #8563: MSG return values and exceptions
2675 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2676 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2677 Thanks to Robson Peixoto.
2678 * Fix a vicious bug in dictionaries inducing that some elements were
2679 not freed on xbt_dict_free()
2681 Portability report of this version:
2682 * Main portability targets:
2683 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2684 on (amd64/i386/ia64)
2685 - mac leopard on i386
2686 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2687 but nothing critical.
2688 * Other platforms: windows, AIX and others were not tested for this release
2690 Timing report of this version:
2691 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2692 scenario parameters). From 8h to a few seconds in desktop grid settings.
2693 * Supernovae brings about 25% speedup on i386.
2695 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2697 SimGrid (3.3.3) stable; urgency=low
2699 The "Need for Speed" release.
2701 The timings done to validate the 3.3.2 were faulty.
2702 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2704 The problem was a conversion from a manually handled vector to
2705 xbt_dynar_t on the critical path.
2706 xbt_dynar_foreach calls functions, inducing stack management crap.
2708 We inlined these functions and xbt_dynar_foreach is now breath taking.
2709 We also inlined xbt_swag_belong on the way.
2711 Here are some approximate speedup measurements (on master/slaves
2712 simulations lasting between 10s and 20s each):
2713 3.3.1 -> 3.3.2: about same performance
2714 3.3.2 -> 3.3.3: 40% speedup
2715 3.3.1 -> 3.3.3: 40% speedup
2716 3.3.1 with inline patch -> 3.3.3: 30% speedup
2718 Our reading is that the refactoring which occurred in 3.3.2 made us
2719 suffer much more from the xbt_dynar_foreach low performance, but
2720 once we solved this, this refactoring proved to be very performance
2721 effective. From the 40% speedup, somehow, 10% are due to the
2722 inlining and 30% to the refactoring.
2724 That's a pitty that gcc cannot inline functions placed in other files
2725 alone. We have to choose between:
2726 - break the encapsulation (by putting private data structures and
2727 accessors in headers files to help gcc)
2728 - live with low performance
2729 - switch to a decent compiler such as icc (not quite possible).
2731 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2733 SimGrid (3.3.2) stable; urgency=low
2735 The "Simplicity does not preceed complexity, but follows it" release.
2737 The main contributors of this release were (lexical order):
2738 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2741 * Extract the routing logic into its own object.
2742 (was dupplicated in network.c and workstation_LV07.c;
2743 Allows to implement other ways of storing that info)
2744 => kill now useless network_card concept
2745 - Use dynar to represent routes (instead of void** + int*)
2746 - kill link_set (use surf_network_model->resource_set instead)
2747 - Add a command-line option to choose the routing schema to use
2748 - Add three new models:
2749 * Floyd (shortest path computed at initialization)
2750 * Dijikstra (shortest path recomputed all the time)
2751 * Cached Dijikstra (shortest path computed on need)
2752 All these models where contributed by Silas De Munck, and are
2753 described in his ICCS09 paper.
2755 * Simplify model declaration
2756 (less redirections, less function to write when defining a model)
2757 - Factorize stuff between models:
2760 surf_model_resource_set(model)
2761 surf_model_resource_by_name(model, name)
2762 - Unify the types of models in s_surf_model_t (using an union)
2763 - Embeed fields of common_public directly into s_surf_model_t
2764 - Rename model methods:
2765 action_free ~> action_unref
2766 action_change_state ~> action_state_set
2767 action_get_state ~> action_state_get
2768 - Change model methods into functions:
2769 (model)->common_public->action_use ~> surf_action_ref
2771 * Implement a generic resource; use it as ancestor to specific ones
2772 (allows to kill duplicated code in models)
2773 Drawback: timer command don't need no name nor properties;
2774 workstation_CLM03 don't need no properties
2775 (but I guess we can live with those few bytes wasted)
2777 * Improve the action object model
2778 - implement a constructor avoiding dupplicated code about field
2779 initialization in generic_action part.
2781 * Kill the SDP model: it has an external dependency, is deprecated
2782 in flavor of modern lmm models, and didn't compile since a while
2785 * Relocation of the context module from XBT to SIMIX.
2786 (the context were decoupled from the simix processes, duplicating a lot of code)
2787 => a lot of code was factorized
2788 - less overhead is introduced during scheduling
2789 - simpler API for the context factory
2790 - the logic for process creation,destruction and manipulation was simplified
2791 * Simplification of the s_smx_process_t data structure.
2792 => accesing the simix level data associated to a process is faster now,
2793 and the code is a lot more readable.
2796 * Implement some more MPI primitives:
2797 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2798 -implementation: Bcast: flat or 2-ary tree (default),
2799 Barrier: 4-ary tree,
2801 Allreduce: Reduce then Bcast
2802 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2803 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2804 Alltoallv: flat tree, like ompi
2806 * Add support for optimized collectives (Bcast is now binomial by default)
2807 * Port smpirun and smpicc to OS X
2810 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2811 Shout out if you used it.
2814 * Display the list of still queued messages in SG mode when existing
2818 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2819 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2820 * Add xbt_str_from_file(FILE*)
2821 * Add xbt_dict_get_key achieving a linear reverse search
2822 * Remove the context module
2824 Portability report of this version:
2825 * Main portability targets:
2826 - Linux(debian)/x86/context
2827 - Linux(debian)/x86/pthreads
2828 - Linux(debian)/amd64/context
2829 - Linux(debian)/amd64/pthreads
2830 On these, we still have the eratic breakages of gras/pmm and
2831 amok/saturate_sg reported in previous version. We still think
2832 that the tests are the cause of the fault, not the tested code.
2834 - Mac OSX Leopard/x86/context
2835 Still false negative in tesh autotesting.
2836 Smpi still fails, but this time because readlink does not accept -f
2837 Everything seems to work properly beside of that.
2840 - AIX version 5.3 (only tested contexts this time)
2841 Smpi still fails there because mktemp is not installed.
2842 Everything seems to work properly beside of that.
2844 I managed to compile it for the first time, but several breakages.
2845 Won't delay the release for this exotic platform.
2847 * Windows: it's still lagging behind. If you want to help, please
2850 Timing report of this version:
2851 This version seem to be more than 5% faster than 3.3.1 (on linux
2852 64bits with contextes). The gain is less than expected, we are
2853 investigating this for next release.
2855 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2857 SimGrid (3.3.1) stable; urgency=low
2860 * Implement a --cfg-help to show existing configuration variables
2861 * Build chain do not require doxygen in maintainer mode
2864 * fix a bug on struct sizeof computation, which prevented the
2865 exchange of arrays of structs in some conditions
2866 - added a regression test about this in datadesc_usage
2867 * Allow the exchange of 0-long dynamic vectors.
2868 - for that, use -1 as indicator of dynamic size instead of 0
2869 - This implied to change any size from unsigned long to long,
2870 reducing a bit communication abilities, but I guess that with
2871 64bits being quite common, this is more than enough.
2872 - This also induce a protocol change, thus bumping network protocol
2873 version from 0 to 1 (if we have external users, we have to get
2874 clean on that point too ;)
2875 - added two regression tests about this in datadesc_usage
2876 * Be more verbose when propagating local exceptions
2877 This helps debugging.
2878 * Display the status of simulated processes when receiving SIGINT in
2882 * Allow to control the simulation from a trace file.
2883 New functions MSG_action_register() and MSG_action_trace_run()
2884 The first one allows to associate a function execution to each
2885 kind of action while the second one parses a trace file and
2886 triggers the corresponding actions within the system.
2887 For now, only a toy example is provided in examples/msg/actions
2888 * Add an example of process migration in examples/msg/migration
2889 * Fix a bug in task exchange which broke MSG_task_get_sender()
2890 Add a teshsuite regression test for that.
2891 [Bug: if MSG_task_get_sender() is called after sender exit,
2893 * Fix a bug which prevented suspend/resume to work properly
2894 * Display the status of simulated processes when receiving SIGINT
2895 This fixes a regression of v3.3. due to the introduction of SIMIX
2896 * Bug fixing in failure management:
2897 - trace could not start by a failure at time 0
2898 - failure during communications were not working
2901 * Add SIMIX_process_set_name() to change the name of the current
2902 process in the log messages.
2903 * Store smx_hosts in a dict since we only retrieve them by name
2904 * Move the configuration infrastructure to surf
2907 * Move the configuration infrastructure to surf
2910 * Massive internal cleanups:
2911 - Store internal structures on processes instead of hosts (allows
2912 to have more than one process per host, in addition of being more
2914 - Cleanup the initialization/finalization process
2915 - Kill a whole bunch of unneeded synchronization:
2916 processes run in exclusive manner within the simulator
2917 - Move queues from global tables to process data fields
2919 - now accept -platform and -hostfile arguments
2920 - Pass the right rank value to processes according to the hostfile
2921 * Compile the examples by default, and use them as regression tests
2922 * Implement MPI_Wtime()
2923 * Change the reference speed to a command line option
2926 * TCP_gamma can now be specified as command line option using
2927 --cfg=TCP_gamma:10000000.0
2928 * Change the --surf-path cmd line option into --cfg=path:
2931 * Also include strbuff from xbt.h public header
2932 * xbt_ex_display(): do not free the exception after displaying
2933 This allows to do more with the given exception afterward.
2934 Users should call xbt_ex_free() themselves.
2938 Portability report of this version:
2939 * Main portability targets:
2940 - Linux(debian)/x86/context
2941 - Linux(debian)/x86/pthreads
2942 - Linux(debian)/amd64/context
2943 - Linux(debian)/amd64/pthreads
2944 These targets fail about 1/10 of times on gras/pmm, but we believe
2945 that this is because of the test, not because of simgrid.
2946 amok/saturate_sg fails even more rarely, and the test may not be
2949 - Mac OSX Leopard/x86/context
2950 The test suite still spits tons of errors because some obscure
2951 force prevents us from removing the temporary directories
2952 arguing that they still contain some metadata I've never heard of.
2953 Smpi fails because seq is not installed.
2954 Everything seems to work properly beside of that.
2957 - AIX version 5.3 (both contexts and pthread)
2958 Smpi still fails there because mktemp is not installed.
2959 XML inclusions seems rosty on AIX.
2961 * Windows: it's still lagging behind. If you want to help, please
2964 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2966 SimGrid (3.3) stable; urgency=high
2970 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2971 [Malek Cherier & Mt]
2973 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2976 Until now, GRAS were using MSG as an interface to SURF. It was
2977 quite difficult because both interface have several differences
2978 (MSG channels vs GRAS sockets were the most notable point).
2980 This also opens the gate to SMPI (which should occur soon) and speed
2981 up simulations by to 40% (even if it were not the main goal).
2983 **************************************
2984 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2985 **************************************
2986 The point is that events occurring at the exact same timestamp are
2987 not scheduled in the same order with the old and new version. This
2988 may be enough to completely change the execution of simulations in
2989 some cases. Sorry for the inconvenience.
2991 * Cleanup and upgrade the XML format to push further scalability
2992 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2994 * Improve the testing infrastructure with tesh. Now a very large part of
2995 the code is tested not only by being run but also by checking that the
2996 output match an expected output [Mt].
2998 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2999 is really less memory-demanding, which should allow you to use
3000 larger files in SimGrid [AL].
3002 * Inform valgrind about our contextes, so that it becomes usable
3003 with the default (and more effecient) version of SimGrid
3004 [contributed by Sékou Diakite, many thanks]
3007 * Introduce a listener thread in charge of receiving incoming
3008 messages from the network. It allows to overlap communication and
3009 computation but most notably, it removes some stupid deadlocks due
3010 to the fact that so far, a process could not send and receive at
3011 the same time. This made most non trivial communication schema
3013 * Convert the PIDs from long int to int to match the MSG ones (and
3014 linux ones too) [Mt]
3015 * New function: gras_agent_spawn() to launch a new process on
3016 current host. Only working in simulation for now. [Mt]
3017 * New function: gras_os_hostport() returning a constant form (ie,
3018 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3021 * Make the backtrace of exceptions more human readable [Mt]
3022 * New module: xbt/str [Mt]
3023 a ton of string utility functions (split, join, printf to a newly
3024 allocated buffer, trim, etc)
3025 * New module: xbt/hash [Mt]
3026 SHA1 hashing algorithm (more to come if needed)
3027 * New module: xbt/synchro [Mt]
3028 synchronization tools (mutex and conditions) working the same way
3029 in simulation and in real life (mainly useful for GRAS, but not
3031 * New module: xbt/queue [Mt]
3032 classical producer/consumer synchronization scheme
3033 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3034 (using the classical functions will get serialized) [Mt]
3035 * Make dictionary internal table dynamic. No need to specify its size
3036 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3038 * Make sure the log channels are organized as a tree under windows
3039 (because of ANSI C compatibility issue, any channel were child of
3040 root directly) [Mt].
3043 * Cleaned many thing in surf and fixed a few bugs [AL].
3044 * Add a nice command line configuration mechanism to compose models [AL].
3045 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3046 the previous one (KCCFLN05). It relies on something that looks like
3047 a max-min sharing mechanism but cannot be written as such. A new solver
3048 was thus designed [AL].
3049 * Add a new solver to lmm. Based on Lagrange optimization and
3050 gradient-based descent, it enables to efficiently maximise systems s.a
3052 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3054 This solver enables to propose two new network models for TCP Reno and
3055 TCP Vegas based on Low's work. These models still need to be fully
3056 tested though [Pedro Velho].
3059 * Bug fix in SD_simulate. Now the time bound given as argument is
3061 * Use the new parallel task model (ptask_L07) as default.
3062 * Use the SURF command line configuration mechanism.
3063 * 0-size tasks (for synchronization) should now work.
3065 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3067 SimGrid (3.2) stable; urgency=high
3071 We still experience issues on this platform, but we believe that at
3072 least MSG is usable.
3074 GRAS API BREAKAGE (for simplification purpose, sorry):
3075 * the gras_msgtype_by_name is not used anymore. Instead of
3076 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3077 you can write (and must)
3078 gras_msg_send(toserver, "request", &request);
3079 - If you still want to pass a gras_msgtype_t to the function (to cache
3080 the type and avoid the lookup time), use the gras_msg_send_() variant.
3081 - Impacted functions:
3082 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3083 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3084 * The callbacks are now expected to return 0 when everything went well
3085 (just like the main() function)
3087 GRAS new features and improvements:
3088 * New module mechanism where user code can use per process globals [Mt]
3089 This is similar to gras_userdata_*() functions, but for libraries. It
3090 factorize some code developped over and over in the examples and AMOK.
3091 It has still to be documented and used (only amok/peermanagement is
3093 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3094 loss when they were small enough to fit into the buffer and sent quickly
3095 enough so that they can all get received in one shoot.
3096 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3097 if not found. Use the *_or_null() variant for the old semantic.
3098 * In gras_msg_handle, do not discard messages without callback.
3099 They are probably messages to be explicitly awaited later (ie, proofs of
3100 mis-synchronization in userland since they are sent before being awaited)
3101 No big deal usually.
3102 * gras_socket_meas_send/recv: semantic changed!
3103 The numerical arguments used to be (1) the total amount of data to send
3104 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3105 messages. This was need for the fool willing to send more than MAXINT
3106 bytes on quite fat pipes.
3109 * Do really rename the hostmanagement module to peermanagement. [Mt]
3110 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3111 but this is rather new and this was documented in the module
3112 documentation (poor excuses, I admit)
3113 * Bandwidth measurement semantic changed! This follows the changes to
3114 gras_socket_meas_send/recv explained above.
3117 * A sequential mode has been added to the workstations. When a workstation
3118 is in sequential mode, it can execute only one task, and the other tasks
3119 are waiting in a FIFO. [Christophe Thiery]
3122 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3123 model for SIMDAG. [Christophe Thiery]
3124 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3125 equal to 0, which caused some bad behaviors in
3126 saturated_constraint_set_update. I now use a threshold mechanism like in
3130 * When running manually src/testall, you select specific units [Mt]
3131 testall is the result of our cunit mechanism, and should replace all
3132 the scripty thingy around since bash don't run easily on billware.
3134 * A mallocator system has been added. [Christophe Thiery]
3135 Mallocators allow you to recycle your unused objects instead of freeing them
3136 and allocating new ones.
3138 Documentation update:
3139 * FAQ reworking + New FAQs:
3140 - "Valgrind spits tons of errors!" [Mt]
3141 - "How to repport bugs" [Mt]
3142 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3143 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3144 - Communication time measurement within MSG [AL]
3145 - I experience weird communication times when I change the latency [AL]
3146 * GRAS tutorial [Mt]
3148 - an introduction to the framework and to the used communication model
3149 - an initiatic tour introducing the most proheminent features:
3151 . Lesson 0: Installing GRAS
3152 . Lesson 1: Setting up your own project
3153 o Part 2: Message passing
3154 . Lesson 2: Exchanging simple messages
3155 . Lesson 3: Passing arguments to the processes (in SG)
3156 . Lesson 4: Attaching callbacks to messages
3157 . Lesson 5: Using globals in processes
3158 . Lesson 6: Logging information properly
3159 . Lesson 7: Using internal timers
3160 . Lesson 8: Handling errors through exceptions
3161 . Lesson 9: Exchanging simple data
3162 . Lesson 10: Remote Procedure Calling (RPC)
3163 . Lesson 11: Explicitly waiting for messages
3164 . Recapping of message passing features in GRAS
3165 - A HOWTO section containing:
3166 o HOWTO design a GRAS application
3167 More are due, of course. They will come latter. In the meanwhile, you can
3168 check the examples which are still here.
3170 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3172 SimGrid (3.1) stable; urgency=high
3176 There was a stack corruption somewhere, visible only when optimizing
3177 with these versions. [Vince]
3180 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3181 module that enabled to play with Directed Acyclic Graphs. It is built
3182 directly on top of SURF and provides an API rather close to the old
3183 SG. Some old codes using SG are currently under rewrite to check that
3184 all needful functions are provided. [Christophe Thiery]
3187 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3188 extension of the classical CLM03 model that gracefully handles
3189 failures. This is now the default model for MSG and GRAS. It doesn't
3190 handle parallel tasks yet though. [AL]
3191 * Bug fix: Weights were not correctly set in the network part.
3192 WARNING: This may have resulted in incorrect results with simulations
3193 where there are more than one flow on a given link. [AL]
3196 * After a (long ?) discussion on simgrid-devel, we have decided that the
3197 convention we had on units was stupid. That is why it has been decided
3198 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3199 WARNING: This means that all previous platform files will not work as
3200 such with this version! A warning is issued to ask users to update
3202 A conversion script can be found in the contrib module of the CVS, under
3203 the name contrib/platform_generation/surfxml_update.pl [MQ]
3206 * Bug fix: Processes were started in reverse order, wrt deployment file.
3207 WARNING: if your code relies on this bug, please fix it. [AL]
3208 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3209 being executed on two different locations. [AL]
3210 * Bug fix: Failures are now better supported thanks to Derrick's tests
3211 (there was many failure situations I hadn't thought of and that weren't
3212 correctly handled). [AL]
3213 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3217 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3218 Exception occurring on server-side are propagated back to client (!).
3220 API CHANGE: the callback changed their prototype. Change:
3221 int my_handler(gras_socket_t expeditor, void *payload_data) {
3223 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3224 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3226 * New! function: gras_msg_handleall to deal with all messages arriving
3227 within a given period.
3228 * New! function: gras_socket_server_range to get a server socket in a
3229 range of port numbers (ease to avoid port number conflicts) [MQ]
3230 * New! gras processes display their backtrace when they get a SIGUSR1
3231 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3232 Sweet to debug RL processes [MQ]
3236 - Do not force experiment sizes to be expressed in kb, or it becomes
3237 impossible to measure the latency this way (needs one byte-long tests)
3238 WARNING: this changes the amok_bw_* function semantic. [MQ]
3239 - Implements the link saturation stuff. [MQ]
3240 * Peer management module:
3241 New! module factorizing code that we wrote over and over [MQ].
3244 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3245 - Test units are placed directly into the library code, they get extracted
3246 automatically and placed into the src/testall binary.
3247 - Convert most of the XBT tests to this system.
3248 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3249 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3250 live in the same C code. This required to change a little bit the API
3251 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3252 * New module: sparse graph structure with basic algorithms (this is work
3253 in progress and the API is not considered to be frozen yet). [AL]
3254 * Display more information on backtraces: source line & function names are
3255 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3256 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3257 * Reworked a little bit some #include statements to load only required
3258 headers. Some user code that relied on SimGrid to include stdlib or
3259 stdio may need to include it by themselves. [AL]
3260 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3262 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3263 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3264 but I guess that almost nobody used those parts. [MQ]
3266 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3268 SimGrid (3.0.1) stable; urgency=low
3271 * Unfortunately, I had missed 5 misnamed functions:
3272 xbt_fifo_item_t xbt_fifo_newitem(void);
3273 void xbt_fifo_freeitem(xbt_fifo_item_t);
3274 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3275 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3276 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3277 They're now deprecated. Please use their new versions:
3278 xbt_fifo_item_t xbt_fifo_new_item(void);
3279 void xbt_fifo_free_item(xbt_fifo_item_t);
3280 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3281 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3282 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3284 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3285 * Documentation: xbt_log module unmercifully reworked [MQ]
3286 * Bugfix: there was a problem with the ending of contexts with
3287 the pthread backend. It caused some weird deadlock or behavior
3288 depending on the pthread implementation. [AL]
3289 * Bugfix: get the exceptions raised in the simulator repport where
3290 and why they come from when they are not caught in time [AL, MQ]
3293 * Bugfix: Do repport the error when two non-connected hosts try to
3294 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3297 * Add additionnal checkings on communications. Assert that two
3298 communicating hosts are connected by a set of links... [AL]
3301 * Add additionnal checkings on channel values in communication [AL]
3302 * New: MSG_task_get_source to see on which host a task was generated [HC]
3303 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3304 the number of tasks waiting to be received on channel and sent
3306 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3307 waits for the first task coming from a given host.. [AL]
3309 GRAS new functionnalities: [MQ]
3310 * Enhance the parsing macro to allow the size of multidimentional objects
3311 to be given thru annotations.
3312 * New example (and documentation): Matrix Multiplication a la RPC
3313 (as when I was young!) and fix a bunch of bugs found on the way.
3315 GRAS performance improvements: [MQ]
3317 * Reduce the amount of cbps creation/destruction by making it static to
3318 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3321 * Change libdata to a set so that we can search for stuff by ID (and thus
3322 reduce the insane amount of dict lookups)
3325 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3326 RL communication on top of SG since it's so uneffective.
3327 It may also be used for inter-thread communication in RL, one day.
3328 * Use gras_datadesc_copy() to exchange messages on top of SG
3330 - improve message exchange performance on top of SG
3331 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3332 * Don't exchange on the network the size of the used part of buffer,
3333 instead, specify the possible buffer size to read().
3335 - reduces the amount of read/write calls (one pair per exchange)
3336 - reduces the amount of exchanged data (the size)
3337 - allows to retrieve all arrived data on receiver side, if we don't need
3338 it right now (subsequent read will peek the buffer)
3339 - allows the receiver to proceed with the beginning of the stream before
3340 everything is arrived
3341 - make it possible to build an iov transport (using readv/writev)
3343 - take care of the data with non-stable storage (like stacked data),
3345 * If possible, TCP send uses vector I/O (when writev() is here)
3346 - Don't use it for receive since we send data sizes and data on the
3347 same stream, so we wouldn't be able to chain large amount of chunks
3348 before having to flush the stuff to read the size.
3349 * Rework the transport plugin mechanism to simplify it and reduce the
3350 amount of pointer dereferencement when searching for the right function
3353 * I guess that now, we do almost as few system calls as possible while
3354 doing as few data copy as possible.
3356 To improve it further, we could try to send all the sizes first and then
3357 all the data (to use iov on receiving size), but it's only a partial
3358 solution: when you have 2 dimensional data, the sizes of the second
3359 dimension is data of the first dimension, so you need 3 streams.
3361 I'm not sure the potential performance gains justify the coding burden.
3363 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3365 SimGrid (3.00) stable; urgency=high
3368 * New! Give the possibility to hijack the surf parser and thus bypass
3369 MSG_create_environment and MSG_launch_application. Have a look at
3370 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3372 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3374 SimGrid (2.96) unstable; urgency=low
3379 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3380 This deprecates the xbt_error_t mechanisms.
3381 It modifies (simplifies) all XBT and GRAS API.
3382 MSG API keeps unchanged (exceptions raised by XBT are caught from
3383 within MSG and masked with existing error handling facilities)
3386 * New! Add a FATPIPE model. [AL]
3387 * New! Add a parallel task model. [AL]
3388 * New! Add automatically a loopback interface (in the default
3389 network model) if none was precised.
3392 * Bugfix: MSG_process_resume now works with the current running process.
3394 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3395 * Modification of MSG_task_get_compute_duration. Once a task has been
3396 processed, the value returned by this function is now equal to 0. [AL]
3397 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3398 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3399 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3400 of a m_task. For now, MSG_task_cancel only works with computation
3402 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3403 of the processor (in Mflop/s) regardless of the current load on the
3405 * API Change: use proper naming convention for MSG_getClock and
3406 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3408 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3409 This function changes the priority of a computation task. This priority
3410 doesn't affect the transfer rate. A priority of 2 will make a task
3411 receive two times more cpu power than the other ones. This function
3412 has been added to suit the needs of Nguyen The Loc and hasn't been that
3413 much tested yet. So if it fails, please report it and send me your code.
3415 * API Change: removed all functions and types that were marked "deprecated"
3416 since many months. Renamed MSG_global_init_args to MSG_global_init.
3418 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3420 SimGrid (2.95) unstable; urgency=low
3423 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3424 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3425 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3428 * Bugfix: Allow absolute paths for platform description files [MQ]
3429 * Bugfix: do free the variables after use. Leads to drastic performance
3431 * Implement max_duration (ie, timeouts) on resources [AL]
3434 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3436 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3437 the needed support (timeouts) [AL]
3440 * Implement measurement sockets. You can now get the bandwidth between two
3441 hosts thanks to AMOK (see below). [MQ]
3442 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3443 dynar over the network (yeah) [MQ]
3444 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3445 * simple/static token-ring example. [Alexandre Colucci and MQ]
3446 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3447 in place before [MQ]
3450 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3451 type parsing. "short" and "long" modifiers were also ignored; other
3452 modifier (such as reference level) are still ignored. [MQ]
3453 * Embeed the buffer size within the buffer itself on SG. [MQ]
3454 That way, send() are atomic and cannot get intermixed anymore (at least
3455 the ones which are less than 100k; bigger messages still have the issue)
3456 * Array size pushed by the field, not by the field type (or each
3457 and every long int will push stuff to the cbps) [MQ]
3458 * use select() to sleep since it allows to portably sleep less than one
3461 GRAS (minor cleanups)
3462 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3465 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3466 * type category 'ignored' killed as it was never used and were difficult
3468 * whether a type can cycle or not is now a flag, leaving room for more
3469 flags (as "ignored", if we feel the need one day ;)
3470 * Rename raw sockets to measurement sockets since "raw" has another
3471 meaning in networking community.
3474 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3475 and offers features not belonging to GRAS but that most applications
3476 need. One day, it may be a set of plugins loadable at runtime.
3477 * New module: bandwidth
3478 bandwidth measurement between arbitrary nodes running this module. [MQ]
3480 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3482 SimGrid (2.94) unstable; urgency=low
3484 The first beta release of SimGrid 3 !
3488 * Update the main page and the FAQ. Adding references to gforge.
3491 * Add a gras_os_getpid function.
3494 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3495 * Extend the logs so that they also print PID, hostname, date, ... if
3497 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3498 and kill the old version which were in testsuite/
3499 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3503 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3504 for users. Instead, each programming environment (GRAS, MSG,...) use it
3505 when needed to provide such a feature to users.
3506 Don't shortcut the mechanism or you will also shortcut the virtualization
3507 you need on the simulator.
3511 * Cleanups in configury with regard to compile optimization/warning flags.
3512 Also add -fno-loop-optimize to any powerpc since it's the optimization
3513 killing gcc (< 3.4.0).
3514 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3515 needed by MSG examples complications
3516 * Borrow configury beautifications from PHP
3519 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3520 warning (thanks loris for stumbling into this one).
3521 * Bugfix: stop loading private headers (gras_config.h) from the public
3525 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3526 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3527 RL would imply a DNS resolver, which is *hard* to do in a portable way
3528 (and therefore delayed).
3529 * Implement a real timer mechanism and use it in timing macros. This allows
3530 to avoid rounding errors and get a 0.000005 sec precision in timing
3531 macros. While I was at it, various cleanups:
3532 - allow to declare more than one timed section per file (fix a stupid bug)
3533 - move some private declaration to the right place
3534 - merge conditional execution and timing macros into emulation module
3535 - document the module
3536 - make sure the module cleanups its mess on gras_exit
3537 * Documentation improvements:
3538 - (new) how to compile applications using GRAS
3539 - (new) emulation support (timing macros)
3541 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3543 SimGrid (2.93) unstable; urgency=low
3545 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3548 - Use Paje properly where used. Still to be sanitized properly.
3549 - Portability fix: Add an implementation of the contexts using pthread
3553 - Add xbt_procname(): returns the name of the current process.
3554 Use it to show the current process's name in all logging.
3556 - fix detection of older flex version and the reaction, since we do
3557 depend on modern ones (we use lex_destroy)
3558 - Better separation of SG and RL in the libs: remove all simulation code
3559 from libgras. As a result, this lib is now only 200k when stripped.
3560 Some of the xbt modules may also be duplicated (two sets and such) and
3561 should be cleaned/killed before SG3.
3562 - Insist on using xlC on AIX because of weird problems involving gcc there.
3563 - Cleanup the make remote stuff. This is now done by scripts
3564 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3565 mainly for our private use, but we're working on changing them to user
3568 - Bugfix: flush the socket on close only if there is some *output*.
3569 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3571 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3574 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3576 SimGrid (2.92) unstable; urgency=low
3578 Alpha 3 on the path to SimGrid 3
3582 - New! First try of benchmarking macros.
3583 - New! First try so that gras_stub_generator generate deployment and
3584 remote compilation helpers.
3586 - Bugfix: Initialization fix in msg_test.
3590 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3592 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3594 - API Change: gras_os_sleep to take the amount of seconds as a double.
3595 Accepting an int was error prone since it was the only location where
3596 seconds were coded as such. It leaded to damn rounding errors.
3597 - Bugfix: Hard to belive that timers ever worked before this.
3599 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3601 SimGrid (2.91) unstable; urgency=low
3603 Alpha 2 on the path to SimGrid 3
3607 - Bug fix in the lmm_solver.
3609 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3610 through the function MSG_paje_output.
3611 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3612 - It is possible to bound the rate of a communication in MSG with
3613 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3614 to write it in the changelog).
3615 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3620 - Various cleanups to the autotools stuff
3621 - Begin to move Gras examples to examples/gras/
3622 - Let make distcheck work again (yeah!)
3624 - documentation overhauled using doxygen.
3625 gtk-doc-tools is dead in SimGrid now.
3626 - Automatically extract all existing logging categories, and add the list
3627 to the documentation (long standing one, to say the less)
3629 - Cleanup the known architecture table. Reorder the entries to group what
3630 should be, and use a more consistent naming scheme.
3631 (some of the test dataset are still to be regenerated)
3632 - New! Allow library to register globals on each process just as userdata
3634 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3635 lookup time (for now).
3636 Use it in msg and trp.
3637 This cleans a lot the internals and helps enforcing privacy of the
3638 headers between the gras components.
3639 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3640 - Bugfix: gras_os_time was delirious in RL.
3641 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3643 - Reenable GRAS now that it works.
3645 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3647 SimGrid (2.90) unstable; urgency=low
3649 Alpha 1 on the path to SimGrid 3
3651 * It is a long time since the last release of SimGrid. I'm sorry about
3652 that but as I had told you, I was rewriting a lot of things. I apologize
3653 to those who had been reporting bugs to me and that I had not answered.
3654 If your bug is still in the new version, please tell me. Here is a
3655 summary of the main changes.
3657 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3658 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3659 portability, speed, and a lot more but you'll figure it out later.
3660 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3661 I wanted to release the new MSG as soon as possible and I have
3662 broken GRAS, which is the reason why, for now, only MSG is fully
3663 functional. A laconic description of these projects is available
3664 in the documentation.
3666 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3667 kernel optimized for our needs (called SURF but only the developers
3668 should use it). Hence, MSG is now roughly 30 times faster and I think
3669 that by rewriting a little bit MSG, I could event speed it up a little
3670 bit more. Beside the gain in speed, it is also much easier to encode a
3671 new platform model with SURF than it was with SG. More to come...
3673 * REVOLUTION 3: I have tried to change a little as possible the API of
3674 MSG but a few things really had to disappear. The main differences
3675 with the previous version are:
3676 1) no more m_links_t and the corresponding functions. Platforms are
3677 directly read from a XML description and cannot be hard-coded
3678 anymore. The same format is used for application deployment
3679 description. The new format is described in the documentation.
3680 Have a look in tools/platform_generation. There is a tiny script
3681 that converts from the old platform format to the new one. Concerning
3682 the application deployment format, parsing the old one is tricky.
3683 I think most of you should however be able to convert your files. If
3684 it is really an issue, I can write a C code that does the conversion.
3686 2) the toolbox tbx does not exist anymore. We now have a library
3687 with much more data-structures but without the hash-tables (we have
3688 dictionaries that are much faster).
3690 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3692 *****************************************************************************
3693 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3694 * don't feel like converting the oldies. (Mt) *
3695 *****************************************************************************
3698 Version 2.90: "the long awaited one"
3699 - Finished rewriting and debugging MSG. Rewrote the documentation.
3700 - disable GRAS for now since it needs to be ported to the newest SG
3703 - Finish the port to windows (using mingw32 for cross-compile)
3706 - Main loop and datastructures of SURF. A cpu resource object is
3707 functional. Surf can thus be used to create cpu's with variable
3708 performance on which you can execute some actions.
3710 2004-11-15 Martin Quinson
3711 - Port to ARM. Simply added the alignment and size descriptions. Should
3712 work, but the ARM machines are so slow that I didn't had the opportunity
3713 to 'make check' over there yet.
3715 2004-11-15 Arnaud Legrand
3716 - Trace manager now written. It uses a heap structure and is therefore
3717 expected to be efficient. It may however be speeded up (particularly
3718 when many events occur at the same date) by using red and black
3719 trees. One day maybe...
3720 - Max-min linear system solver written. It uses a sparse matrix
3721 structure taking advantage of its expected use. Most operations are
3722 O(1) and free/calloc are called as few as possible. The computation of
3723 the minimum could however be improved by using a red and black tree
3726 2004-11-03 Arnaud Legrand
3727 - Rename every gras_* function that was in xbt/ to its xbt_
3729 - Add a heap and a doubly-linked list to xbt
3730 - Added a dichotomy to the dictionaries. make check works as well before
3731 so I assume that the patch is correct. I do not know however if things
3732 run effectively faster than before now. :)
3734 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3735 SimGrid, and the version number is bumped to 2.x
3737 2004-10-29 Martin Quinson
3738 - Introduction of the remote errors.
3739 They are the result of a RMI/RPC on the remote machine.
3740 ErrCodes being scalar values, you can't get the host on which those
3741 errors did happen. Extending the error mechanism as in Gnome is possible.
3742 No idea yet whether it is a good idea.
3744 2004-10-28 Martin Quinson
3745 - Interface revolution: the Starred Structure Eradication.
3746 I used to do typedef struct {} toto_t; and then handle *toto_t.
3747 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3748 - ???_t is a valid type (builded with typedef)
3749 - s_toto_t is a structure (access to fields with .)
3750 - s_toto is a structure needing 'struct' keyword to be used
3751 - e_toto_t is an enum
3752 - toto_t is an 'object' (struct*)
3754 typedef struct s_toto {} s_toto_t, *toto_t;
3755 typedef enum {} e_toto_t;
3756 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3757 s_toto_t) is private.
3759 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3760 so that it can set the variable to NULL. It was so for dicts and sets,
3761 it changed for dynars.
3763 - Fix a bunch of memleaks in dict_remove
3764 - Fix a bug in sg/server_socket opening: it failed all the time.
3766 2004-10-07 Martin Quinson
3767 - Speed up dynar lookup operation a bit.
3769 gras_dynar_get is dead.
3771 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3772 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3773 (which gives you the address of the stored data).
3775 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3776 copy of the data using an affectation to do the job and not a memcpy.
3778 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3779 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3781 It does not really speedup the dynar test because they are
3782 setting elements all the time (and look them seldom). But the dict does
3783 far more lookup than setting.
3785 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3787 2004-10-05 Martin Quinson
3788 - Allow to (en/dis)able the cycle detection at run time.
3790 Whether we should check for cycle or not is now a property of each
3791 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3792 datadesc_cycle_unset allow to remove this property when previously set.
3794 Note that the cycle detection is off by default since it impacts the
3795 performance. Watch the data you feed GRAS with ;)
3797 This property is hereditary. Any element embedded in a structure having it
3798 set have it set for the time of this data exchange.
3800 You should set it both on sender and receiver side. If you don't set it on
3801 sender side, it will enter an endless loop. If you forget on receiver
3802 side, the cycles won't be recreated after communication.
3804 - Header reorganization.
3805 Kill gras_private.h, each submodule must load the headers it needs.
3807 2004-10-04 Martin Quinson
3808 - Interface revolution: do not try to survive to malloc failure.
3810 Now, gras_malloc and friends call gras_abort() on failure.
3811 As a conclusion, malloc_error is not a valid error anymore, and all
3812 functions for which it was the only gras_error_t return value are
3813 changed. They now return void, or there result directly.
3814 This simplify the API a lot.
3816 2004-09-29 Martin Quinson
3817 - Re-enable raw sockets.
3818 Created by gras_socket_{client,server}_ext;
3819 Used with gras_raw_{send,recv}
3822 It should allow to kill the last bits of gras first version soon.
3824 This is not completely satisfactory yet (duplicate code with
3825 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3828 - Simplify transport plugin (internal) interface by not passing any
3829 argument to _server and _client, but embedding them in the socket
3832 2004-09-28 Martin Quinson
3833 - Finish the port to AIX.
3834 autoconf was my problem (segfault within the malloc replacement
3835 function. No idea why)
3837 2004-09-16 Martin Quinson
3838 - Fix some size_t madness on 64bit architectures.
3840 2004-09-08 Martin Quinson
3841 - Reduce the number of system headers loaded, overload some more system
3842 calls (such as malloc to cast the result of the system one, and work
3844 - Fix and reintroduce the config support
3846 2004-09-07 Martin Quinson
3847 - Source code reorganization to allow Arnaud to surf all over there.
3848 - Allow to document the logging categories.
3849 - Remove all uppercase from logging categories and useless cleanup in names.
3851 2004-08-18 Martin Quinson
3852 Version 0.6.2 (protocol not changed; API changed)
3853 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3854 gras_error_t), and NULL when not found. Functions expecting a msgtype
3855 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3856 hopefully usefull message.
3857 - Portability to prehistoric sparcs again
3859 2004-08-17 Martin Quinson
3860 Version 0.6.1 (protocol not changed; ABI not changed)
3861 - prealloc some buffers to speed things up
3863 2004-08-11 Martin Quinson
3864 Version 0.6 (protocol not changed; ABI expended)
3865 - The parsing macro can deal with the references, provided that you add
3866 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3868 2004-08-09 Martin Quinson
3869 Version 0.5 (protocol not changed; ABI changed)
3870 - Allow to off turn the cycle detection code in data exchange at
3871 compilation time. It should be at run time, but I'm short of time (and
3872 the config stuff is still broken). That way, we keep dict out of the
3873 critical path, which is good because the performance is poor:
3874 - search not dichotomial yet
3875 - dynar give no way to access their content and memcpy everytime
3876 - In composed data description (struct, ref and so on), stop foolness of
3877 keeping the subtype's ID, but store the type itself. This keeps sets out
3878 of the critical path, which is good since they rely on dynar and
3879 dictionaries. The only loose of that is that we cannot detect the
3880 redeclaration of a structure/union with another content (but I'm not sure
3881 the code detected well this error before anyway). We still can detect
3882 the redefinition discrepancy for the other types.
3883 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3884 it breaks the code because of type-punning used all over the place).
3885 This breaks on all non-gcc architectures (for now).
3887 All those changes (plus the buffer of last time) allow me to gain 2 order
3888 of magnitude on cruel tests consisting of 800000 array of integers on two
3889 level of a hierarchical structure (200 secondes -> 4 secondes)
3892 - the selector of reference must now return the type it points to, not
3893 the ID of this type.
3895 2004-08-06 Martin Quinson
3896 Version 0.4 (protocol changed; ABI not changed)
3897 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3898 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3899 - Implement a buffer transport, and use it by default (it relies on tcp in
3900 real life and on sg in simulation).
3901 That's a bit hackish since I had a new field to the structure to store
3902 its data without interfering with the subtype ones. Inheritance
3903 is tricky in C. And that's a kind of reverse inheritance with one class
3904 derivating two classes. Or maybe a game with java interfaces. Anyway,
3905 that's damn hard in C (at least).
3906 Moreover, I got tired while trying to ensure plugin separation and
3907 genericity in SG mode. MSG wants me to do weird things, so let's go for
3908 cruel hacks (temporarily of course ;).
3909 See comment in transport_private.h:71
3910 - do not include all the _interface headers in private but in the files
3911 which really need them (to cut the compilation time when they are
3914 2004-07-26 Martin Quinson
3915 Version 0.3 (protocol not changed; ABI changed)
3916 - Major overhault of the datadesc interface to simplify it:
3917 - shorted the function names:
3918 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3919 - add a trivial way to push/pop integers into the cbps without malloc.
3920 This allows to make really generic sub_type description, which simply
3921 pop their size of the stack.
3922 - add a function gras_datadesc_ref_pop_arr() which does what users want
3923 most of the time: Declare a dynamic array (which pops its size of the
3924 stack) and declare a reference to it. Poor name, but anyway.
3925 - kill the post-send callback, add a post-receive one
3927 2004-07-23 Martin Quinson
3928 Version 0.2 (protocol changed; ABI changed)
3929 - add some testing for cpbs in the test cases, and fix some more bugs.
3930 This invalidate again the little64 data file, since I cannot regenerate
3932 - remove an awfull optimization in the logging stuff, allowing me to:
3933 - understand it again
3934 - learn gcc how to check that the argument match the provided format
3935 - fix all errors revealed by gcc after that
3936 - internal keys of dict are not \0 terminated. Deal with it properly in
3937 loggings instead of segfaulting when the user want to see the logs :-/
3939 2004-07-22 Martin Quinson
3940 - Fix some stupid bug preventing cbps (callback postit) from working
3942 2004-07-21 Martin Quinson
3943 - Some documentation cleanups
3944 - remove the useless last argument of msgtype_declare
3945 - rename the Virtu functions to fit into the 'os' namespace
3946 - move headers src/include -> src/include/gras/ and stop fooling with
3947 gras -> . symbolic link
3948 - make distcheck is now successful
3950 2004-07-19 Martin Quinson
3952 - Build shared library also
3953 - Install html doc to the right location
3954 - stop removing maintainer files in make clean
3955 - build tests only on make check
3957 2004-07-13 Martin Quinson
3959 - No major issue in previous version => change versioning schema
3960 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3961 corresponding dataset.
3963 2004-07-11 Martin Quinson
3965 - Get it working with any kind of structure (we can compute the padding
3966 bytes remotely for all the architectures I have access to)
3967 - Implement the structure parsing macro (still not quite robust/complete)
3968 - Improvement to the remote testing toysuite
3970 2004-07-10 Martin Quinson
3971 [autoconf mechanism]
3972 - get ride of a bunch of deprecated macros
3973 - actually run the test for two-compliment, not only compile it :-/
3974 - test whether the structures get packed (and bail out if yes. Damn.
3975 Alignment is a serious matter)
3976 - test whether the structures get compacted (but respecting the alignment
3977 constraints of each types)
3978 - test whether the array fields of structures can straddle alignment boundaries
3980 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3981 (took me 2 hours to find that bug, looking at the wrong place)
3982 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3983 before sending/receiving and are used to compute the offsets of fields
3984 - Given that padding size depend even on compiler options, keep track of
3985 alignment and aligned_size only for the current architecture. Not a big
3986 deal since we send structure fields one after the other (seems
3988 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3989 program, let it work on linux/gcc/little32. portability todo.
3991 2004-07-08 Martin Quinson
3992 - import and improve remote compilation support from FAST
3993 - make sure make check works on half a dozen of machines out there
3995 2004-07-07 Martin Quinson
3996 Let's say it's version 0.0.3 ;)
3997 - Implement conversions (yuhu!)
3998 - Let it work on solaris (beside conversion, of course)
3999 - Stupid me, using rand() to generate the conversion datatests in not wise.
4001 2004-07-06 Martin Quinson
4002 - Let make dist work, since I'm gonna need it to compile on remote hosts
4003 - Let Tests/datadesc_usage write the architecture on which the file was
4004 generated as first byte.
4005 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4006 architecture descriptions.
4007 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4008 on those architectures.
4009 - Optimization: send/recv array of scalar in one shoot
4011 2004-07-05 Martin Quinson
4012 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4014 - Plug a whole bunch of memleaks
4015 - each process now have to call gras_{init,exit}. One day, their log
4016 settings will be separated
4017 - Continue the code factorisation between SG, RL and common in Transport.
4019 2004-07-04 Martin Quinson
4021 - Redistribution between SG and RL.
4022 We wanna have to accept in SG, so move accepted related parts of RL in
4023 the common part. (more precisely, the dynar of all known sockets is no
4024 more a static in transport.c, but part of the process_data)
4026 [gras_stub_generator]
4027 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4029 2004-07-03 Martin Quinson
4030 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4031 them all up in one shot)
4033 - Fix the ugly case of reference to dynamic array.
4034 - New (semi-public) function gras_datadesc_size to allow the messaging
4035 layer to malloc the needed space for the buffer.
4037 - gras_socket_close now expect the socket to close (and not its address to
4038 put NULL in it after it). This is because the socket passed to handlers
4039 is one of their argument (=> not writable).
4041 - propagate the interface cleanup from last week in datadesc, ie remove a
4042 superfluous level of indirection. User pass adress of variable
4043 containing data (both when sending and receiving), and not of a variable
4044 being a pointer to the data. Let's say that I like it better ;)
4045 The price for that is constructs like "int msg=*(int*)payload" in
4046 handlers, but it's a fine price, IMHO.
4048 - Let it work in RL (yuhu)
4050 2004-06-21 Martin Quinson
4052 - porting SG plugin and SG select to new standards (works almost).
4053 - plug memleaks and fix bugs around.
4056 - cleanup the prototype of data recv and force users to specify when they
4057 want to handle references to objects. Test case working even for cycles.
4058 - plug memleaks. Valgrind is perfectly ok with this.
4060 2004-06-12 Martin Quinson
4062 - cleanup the separation between plugin and main code in plugin creation
4064 2004-06-11 Martin Quinson
4066 - Reput hook for raw sockets, needed for BW experiments
4067 - kill a few lines of dead code
4068 [Data description] Interface cleanup
4069 - gras_datadesc_by_name returns the searched type or NULL.
4070 That way, no variable is needed to use a type desc once, which makes
4072 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4073 two parameters were strings (field name, type name), leading to
4075 [Dicos] Interface cleanup
4076 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4077 This is consistant with the dynar API.
4079 2004-04-21 Martin Quinson
4081 - Porting to new standards.
4083 - interface cleanup.
4084 There is no bag anymore, no need to take extra provision to mask the
4085 pointers behind "ID".
4086 Better splitup of functions between files create/exchange/convert.
4087 This is still a bit artificial since convert and receive are so
4088 interleaved, but anyway.
4090 - add a queued message list to procdata (the ones not matching criteria
4092 - factorize some more code between SG and RL wrt procdata
4094 - use gras_exit in example to track memleaks
4095 - get rid of gs_example now that GS is properly integrated into gras
4096 - update run_test to integrate the latest tests (datadesc)
4098 - rename WARNINGn macros to WARNn since it prooved error-prone
4100 2004-04-19 Martin Quinson
4102 - register init/exit functions within gras module mechanism
4103 - send/receive function.
4104 Convertion is not implemented, but short-cutted if not needed.
4105 struct/array elements are sent one by one (instead of block-wise), but
4106 nobody really cares (yet). Get a prototype before optimizing.
4107 - tests (using a file socket) for DD send/receive on:
4108 - base types: int, float
4109 - array: fixed size, string (ie ref to dynamic string)
4110 - structure: homogeneous, heterogeneous
4111 - chained list, graph with cycle
4112 Believe it or not, valgrind is not too unhappy with the results. The
4113 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4117 - Bugfix when using a filename explicitly (instead of '-')
4119 2004-04-09 Martin Quinson
4121 - factorize more code between RL and SG in socket creation
4122 - Complete the implementation and tests of:
4124 o file (only in RL, and mainly for debugging)
4126 I lost 3 days to design a portable address resolver, and then decided
4127 that the prototype mainly have to run on my box.
4128 Addressing portability too early may be like optimizing too early :-/
4130 - use gras_init in the Tests instead of the crappy parse_log_opt
4131 (the latter function is removed)
4132 [Conditional execution]
4133 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4134 [Code reorganisation]
4135 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4136 Build examples against the RL library, since there is no way to disable
4137 its creation for now.
4139 For information, the beginning of coding on GRAS was back in june
4140 2003. I guess that every line has been rewritten at least twice since