1 ----------------------------------------------------------------------------
3 SimGrid (3.25.1) UNRELEASED (target: March 20 2020 or v3.26 on 3/26/20)
5 Important user-visible changes:
6 - SimGrid now requires a compiler with C++14 support.
7 Sibling projects should upgrade their FindSimgrid.cmake
8 - Surf precision default value is now 1e-9, instead of 1e-5. This was changed as
9 several users had difficult to understand issues when using high bandwidth or
10 small latency events. The new value was already the default for SMPI and
11 should not cause too much performance hit. It can change some simulation
12 timings if simulation was skipping a lot of small events (which was actually
13 an issue of said simulation). The value can still be changed back using
14 --cfg=surf/precision:1e-5.
15 - On some network models (L07, CM02, LV08), an implicit loopback link is created
16 for local communication. Default bandwidth was 500MBps and latency was 15µs,
17 which was rather bad. Default values have been changed to 10GBps and 0 latency.
18 Options network/loopback-lat and network/loopback-bw have been added to adapt
19 these values if needed. Network and routing models already providing separate
20 loopbacks (clusters and such) are not impacted by this change.
21 - SMPI now adds 16 bytes of header to each message sent, in order to simulate
22 metadata that are added to each MPI message by MPI libraries (tag,
23 src/dest, and communicator information). This may change simulation times
24 for very small message sizes, but should be more realistic.
27 - New routing 'WiFi' that makes it easier to specify a wifi network zone.
28 - New plugins: wifi_energy and link_load. See the associated examples.
29 - New examples: wifi networks, and communication suspend.
32 - LTO is now enabled for Intel/clang compilers.
33 - LTO behavior on GCC can be parameterized using LTO_EXTRA_FLAG in cmake.
34 Setting it to "auto" will use all available cores, while setting it to n will
35 use n cores to speedup link step (usage: cmake -DLTO_EXTRA_FLAG=4).
36 - Remove obsolete option --cfg=contexts/parallel-threshold.
37 - Further improve the documentation.
40 - the on_exit() of each actor is also executed when the simulation deadlocks.
41 - New functions: s4u::Activity:: suspend(), resume() and is_suspended()
42 An example is provided for s4u::Comm but it should work with Exec and Io.
45 - Update proxy apps coverage of new ECP apps: 60+ apps now tested nightly on
46 https://framagit.org/simgrid/SMPI-proxy-apps
47 - MPI/IO: supports names starting with "./"
48 - Fortran: many MPI/IO bindings, support for MPI_UB/LB, C MPI_INT/FLOAT/DOUBLE
50 - smpicc/cxx/ff/f90 now will actually perform definition checks at link time. When
51 building shared libraries, this may cause issues, so environment variable
52 SMPI_NO_UNDEFINED_CHECK can be added to disable this.
54 C binding and interface:
55 - The return type of the sg_actor_on_exit() callbacks is now 'void'
56 (int value was previously ignored)
57 - Many MSG tests were converted to the new S4U's interface in C, that
58 was extended for that.
60 Fixed bugs (FG#.. -> FramaGit bugs; FG!.. -> FG merge requests)
61 (FG: issues on Framagit; GF: issues on GForge; GH: issues on GitHub)
62 - FG#41: Add sg_actor_create C interface
63 - FG#43: xbt::random needs some care
64 - FG#48: The Impossible Did Happen (yet again)
65 - FG#50: Suspending an actor executed at the current timestamp fails
66 - FG#54: How to suspend a comm?
67 - FG!24: Documentation and fix for xbt/random
68 - FG!35: Add a modeling hint for parallel links in doc
69 - FG!36: [xbt/random] Read/Write the state of the RNG
70 - FG!42: Link cumulated load plugin
71 - FG!44: Test condition variable wait_until+notify_all
72 - GF#18137: Allow different stack sizes?
73 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
74 - GH#139: Allow pthread creation in SMPI
75 - GH#336: Packet-level simulation using SMPI?
76 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
77 - GH!337: Fix link_energy plugin for wifi platforms
78 - GH!339: Add Mailbox set_receiver method to python binding
79 - GH!344: Cast hugepages macros parameters to int64
81 ----------------------------------------------------------------------------
83 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
85 The Palindrom Day Release.
87 Important user-visible changes:
88 - Improve the Python usability (stability and documentation).
89 - A nasty synchronization bug was ironed out, see also below.
90 - Python's doc was organized as a separate tree, now integrated with C++.
91 - C bindings of S4U were not part of the doc.
92 - The C++ doc was also improved as methods are now split by theme.
93 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
94 - This is mandatory to use the Java bindings.
95 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
96 - SimDAG++: Automatic dependencies on S4U activities (experimental)
97 - Some features are already implemented but not all of them
98 - Cannot block an activity until it's scheduled on a resource
99 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
100 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
103 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
104 - Drop support for Viva/Triva (old visualization tools), hence removed
105 TRACE_get_node_types() and TRACE_get_edge_types() functions.
108 - Require pybind11 version 2.4.
109 - Greatly improve locking during context switches with Python.
110 - Declaring Python callbacks to this_actor::on_exit is now working.
111 - No more tests are failing currently, even on win32.
114 - Actor: Merge signals on_migration_start/end into on_host_change
115 - Actor: Rename migrate() into set_host()
116 - Disk: Allow users to get the read and write nominal bandwidth values
117 - Exec: Implement wait_for(timeout)
118 - Io: Implement wait_for(timeout)
121 - Parse errors now raise a simgrid::ParseError that you may want to catch.
124 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
125 been renamed to fix a spelling error. As usual, the previous definitions are
126 kept with a deprecation warning. However, to avoid runtime errors, it is now
127 forbidden to override these deprecated methods in a derived class. Please use
128 the new names immediately if you need to override them.
131 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
132 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
133 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
134 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
135 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
136 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
137 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
138 - New module: random, providing classical random numbers generators.
141 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
142 allocations' size above which they are considered "shared" by default (as if
143 they were performed through SMPI_SHARED_MALLOC macros).
144 Default = 0 = disabled feature.
145 Note : malloc, calloc and free are now overriden by smpicc/cxx by default.
146 This can cause some troubles if codes are already overriding these. If this
147 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
148 help, but will make this feature unusable.
149 - Input check is now performed much more consistently and outputs precise
150 warnings for each error found
151 - New supported calls : MPI_File_set_view, MPI_File_get_view
152 - MPI I/O now only supports disks, as storage is discontinued.
153 - MPI I/O allows opening files without specifying absolute path
154 - Note : SMPI configuration options are now initialized by smpi_init_options()
155 call, which should be called if SMPI is not being used through smpirun.
157 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
158 - FG#9: Python bindings crashing
159 - FG#39: Missing s4u::Comm::wait_any_for example
160 - FG#42: Add support for ThreadSanitizer (TSan)
161 - FG!19: Removing RngStream
162 - FG!20: A module for RNG calls
163 - FG!21: Choice between ad-hoc and standard distributions implementations
165 - FG!25: Fix link in Mutex doc
166 - FG!26: Fix links in SMPI interface doc
167 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
168 - GH#305: Unscheduled tasks are still excuted
169 - GH#313: smpirun: manual outdated w.r.t. --help
170 - GH#321: [S4U] Get task remaining work ratio
171 - GH#323: Crash when an actor turn off his physical host
172 - GH#335: Missing links on dragonfly example svg
173 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
174 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
176 ----------------------------------------------------------------------------
178 SimGrid (3.24) October 9. 2019
180 The Clean Disk Release.
183 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
184 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
185 thus becomes the default resource to execute I/O operations. Examples
186 and the file system plugin have been modified to use disks instead of
187 storage elements. Storage can still be used but is doomed to disappear.
188 Users are thus adviced to update their simulators accordingly.
189 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
190 for consistency with pthread_barrier_wait()
191 - Host::get_englobing_zone() returns the englobing netzone
192 - Actor::on_destruction is now called in the destructor
193 Actor::on_termination new signal called when the actor terminates
195 - Global signals are now part of the Engine:
196 - on_platform_creation: after config settings, before the XML parsing
197 - on_platform_created: right after the XML parsing
198 - on_time_advance: each time the clock advances
199 - on_simulation_end: after simulation, before cleanups
200 - on_deadlock: as the name implies.
202 - sg_{actor,host,link}_{data,data_set}() now all exist.
203 Use them to attach user data to the object and retrieve it.
206 - Introduce an experimental Wifi model. It sounds reasonable
207 according to the state of the art, but it still has to be properly
208 validated, at least against ns-3 if not against reality.
209 - Improved the usability of ns-3. Several bugs were ironed out.
210 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
211 It is now expressed as 'idle:epsilon:allCores' to properly model the
212 consumption of non-whole tasks on mono-core hosts. More info in the doc.
215 - convert a new set of functions to the S4U C interface and move the old MSG
216 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
219 - Fortran bindings for DVFS have been removed.
220 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
221 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
222 - SMPI can now be selected by cmake's find_module(MPI) with
223 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
224 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
225 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
226 - trace-call-location can be used with TI traces, and replayed, and is compatible
227 with smpi/comp-adjustment-file.
228 - sleep events are now correctly traced/replayed.
229 - Default for trace-call-location is now to use file names and not full paths.
230 To revert to previous behavior (in case of collision of filenames), option
231 "smpi/trace-call-use-absolute-path" can be set to yes.
234 - Use the included xxHash as an hash implem when C++14 is usable.
235 - Option model-checker/hash was removed. This is always activated now.
236 - New option smpi/buffering controls the MPI buffering in MC mode.
237 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
238 This is useful to check for MPI compliance.
241 - New section on plugins: how to define a new one, and existing ones.
244 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
245 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
248 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
249 and <mount> tags. The rationale is finish to decouple what should be handle
250 by SimGrid kernel (i.e., resources and activities) from plugins built in
251 user space (e.g., the file system plugin). This new tag currently accepts
252 only three attributes (id, read_bw, and write_bw). All the other information
253 that was declared with the storage related tags now has to be expressed as
254 properties. An example of platform using this new tag is available at
255 examples/platforms/hosts_with_disks.xml
258 - 'expect signal' can now accept more than one potential signal.
260 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
261 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
262 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
263 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
264 - FG#31: per-actor data
265 - FG#34: SG_BARRIER_SERIAL_THREAD?
266 - FG#35: model-checker does not like buster-produced binaries
267 - FG!13: MC: complete workaround in the error msg seen on modern systems
268 - FG!15: execute_flops now logs compute
269 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
270 - FG!17: ns-3: unblock the right number of communications + others issues
271 - FG!18: Improving the performance of the ns-3 bindings
272 - GH#189: Energy consumption of parallel tasks
273 - GH#219: Error in the throughput of TCP transfer
274 - GH!330: Minor polishing of English: Configuring_Simgrid
275 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
276 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
277 - GH!334: Rework energy plugin, again
279 ----------------------------------------------------------------------------
281 SimGrid (3.23.2) July 8. 2019
284 - Nicer introduction page.
285 - Migrate the "Deploy your application" page to the new doc.
286 - Move Java as a subtree of MSG.
289 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
292 - Drop xbt_dynar_sort_strings().
295 - Really fix FG#26: Turning off a link should raise NetworkFailureException
296 - FG#27: Wrong exception thrown to wait_any when link is turned off
297 - GH#328: Java: Canceling multiple tasks in a single vm/host
299 ----------------------------------------------------------------------------
301 SimGrid (3.23) June 25. 2019
303 The Exotic Solstice Release.
306 - SunOS and Haiku OS support. Because exotic platforms are fun.
307 - Stop setting random seed with srand48() at initialization.
308 - Use addr2line as a fallback for stacktraces when backtrace is not available.
309 - Build option -Denable_documentation is now OFF by default.
310 - Network model 'NS3' was renamed into 'ns-3'.
313 - Simgrid can now hopefully be installed with pip.
316 - wait_any can now be used for asynchronous executions too.
319 - New log appenders: stdout and stderr. Use stdout for xbt_help.
320 - Drop xbt_dict_dump.
323 - 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
324 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
325 - MPI interface is now const correct for input parameters
326 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
327 - Fortran bindings for async collectives.
328 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
331 - Remove option 'model-check/record': Paths are recorded in any cases now.
332 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
333 always sparse. That's an immense gain in memory, and can even be
334 faster because of cache effects. So there is no need to clutter the
335 code to allow the user to go for the unefficient mode.
338 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
339 models proved to be more accurate than these old experiments.
341 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
342 - FG#1: Broken link in error messages
343 - FG#2: missing installation documentation
344 - FG#3: missing documentation in smpirun
345 - FG#6: Python bindings not available on PyPI
346 - FG#7: simple cmake call requires doxygen
347 - FG#8: make python bindings an optional dependency
348 - FG#10: Can not use MSG_process_set_data from SMPI any more
349 - FG#11: Auto-restart actors forget their on_exit behavior
350 - FG#12: -Denable_lto=OFF doesn't disable LTO
351 - FG#13: Installs unstripped file 'bin/graphicator'
352 - FG#14: Installs the empty directory 'doc/simgrid/html'
353 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
354 - FG#17: Dead link in doc (pls_ns3)
355 - FG#20: 'tesh --help' should return 0
356 - FG#21: Documentation link on http://simgrid.org/ broken
357 - FG#22: Debian installation instruction are broken
358 - FG#26: Turning off a link should raise NetworkFailureException exceptions
359 - GH#133: Java: a process can run on a VM even if its host is off
360 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
361 - GH#326: Valgrind-detected error for join() when energy plugin is activated
363 ----------------------------------------------------------------------------
365 SimGrid (3.22) April 1. 2019
367 The Easter Christmas Release (this one is somewhat late).
370 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
371 - Not complete yet: asynchronous activities (amongst others) are still missing
372 - Still ongoing: the interface may change in the future. We need more
373 testers! Please report any glitches.
374 - No new project using Java should start now. Please switch to Python.
377 - Some of the internal cleanups may lead to speed improvements:
378 - The hard limitation on the amount of simulated actors with
379 Java+Mac was removed. Now, the available memory is the only limit.
380 - Our refcounting was tidyied, leading to 10% speedups in some cases.
381 - We are still working on making our code robust to the actor kills
382 and hosts' churn. Things are improving, but it's not perfect yet.
383 - Replaced our own code to display a backtrace (that was forking addr2line)
384 with the Boost.Stacktrace library.
385 You won't see your backtraces without this optional dependency.
386 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
387 - Stop setting random seed with srand() at initialization.
390 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
391 XML file version remains 4.2 since old files are still compatible.
394 - Process termination which was broken at version 3.21 has been repaired.
395 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
396 - Hide the examples into examples/deprecated. New users should use Python.
399 - Drop MSG_process_create_from_stdfunc() from the API.
400 This C++ function was a pimple in the C API, made necessary at some
401 point by the Java bindings. This is fixed now.
402 - Hide the examples into examples/deprecated. New users should use S4U.
403 - MSG_process_create and MSG_process_attach now crash if the host used
404 to run this process is off.
405 - Fix the protype of MSG_process_on_exit()
406 Now use (int,void*) callbacks instead of (void*,void*) ones.
407 The implementation was ways too messy to actually work, I guess.
410 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
411 too often and provide more accurate timings.
412 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
413 - MPI_Alltoallw support
414 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
415 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
416 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
417 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
419 - Basic implementation of generalized requests (SMPI doesn't
420 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
423 - Drop sg_cmdline. Please use xbt_cmdline instead.
424 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
425 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
426 OsSemaphore is implemented in a portable way with C++11 threads.
427 This should allow much more threads to be created at the same time,
428 allowing Mac and Java users to simulate many more actors.
429 - Implement the 'thread' factory with std::thread instead of xbt ones.
430 It is not possible to set the stack size with threads anymore, but
431 -fsplit-stack is the way to go nowadays when using threads.
432 - Drop the xbt_os_thread_t module (now unused)
433 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
434 - Drop xbt_str_join_array().
435 - Drop cunit, use Catch2 instead.
438 - Many cleanups in the kernel::activity namespace. This was long
439 overdue, and shall open the path to many future endeavors.
442 - #132: Java: a process can not shut down its own host
443 - #220: S4U: detached send causes exception if sender terminates soon after sending
444 - #261: Document the parameters of parallel execution's constructor
445 - #300: [s4u] BarrierPtr is missing
446 - #314: SMPI args internal cleanup
447 - #316: Fix a bug related to the CPU utilization of multi-core VM
448 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
449 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
450 - #325: Turning off a host has different behavior on sleeping actors and computing actors
452 ----------------------------------------------------------------------------
454 SimGrid (3.21) October 3. 2018
456 The Restarting Documentation (TRD) Release.
459 - Convert the user manual to Sphinx for improved usability.
460 Unstable version now hosted on https://simgrid.frama.io/simgrid
461 Many glitches remain; stay tuned.
462 - Restrict the Doxygen reference API to the installed header files.
463 (documenting all internals this way is overkill + doxygen scales badly)
464 - New tutorial on S4U (highly inspired from the old MSG one, but with
465 a git repository to fork as a starting point, and a docker image)
466 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
467 - The Developper manual is still to be converted (not compiled until then)
468 - Some parts are still missing in this conversion (such as the
469 platform and deployment sections) while others were blindly converted
470 and would need more love (such as the configuration flags).
471 Things will certainly further improve in the future releases.
474 - Fully reimplement auto_restart mechanism that was utterly broken
475 (fix #22, #131 and #281 that were all very related)
476 - Implement semaphores (fix #303 and #291)
477 - When creating an actor from a function and its parameters,
478 move-only parameters are not allowed anymore, as it would prevent
479 the actor to be restartable if its parameters are consumed on run.
480 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
481 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
482 IO operations that can be started, waited for, or canceled as a regular
484 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
488 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
489 - New host variable: 'core_count'
492 - Remove xbt_os_thread_specific features
493 - Remove portability wrapper to condition variables
494 - Remove xbt_os_thread_yield()
497 - MPICH collective selector now mimics MPICH 3.3b
498 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
499 - MPI_Init(NULL,NULL) is now allowed
500 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
501 - Command-line arguments should now be handled properly in Fortran simulations
502 - MPI Attributes and related callbacks should work in Fortran
503 - Apps using C + Fortran should now work
504 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
505 - Support MPI_Error_String
508 - Due to an internal bug, Msg.run() must now be your last line.
509 We hope to fix it in a future release, and we are sorry for the inconvenience.
512 - #22: Process autorestart seem to only work with CAS01 cpus
513 - #56: Feature request: dynamic SMPI replay
514 - #93: simgrid should not eat --help
515 - #111: How to change the loopback link characteristics?
516 - #116: Platform cluster radical
517 - #117: Platform cluster/host order
518 - #129: Replay_multiple on steroid
519 - #131: Java : autostart does not work
520 - #143: Setting a breakpoint at a given time
521 - #235: Network energy models should be integrated
522 - #264: Add ptask L07 resource tracing
523 - #271: Dynamic and manual replay of SMPI traces
524 - #279: Breakpoints option uses time deltas instead of absolute time
525 - #281: Daemonized actor no longer daemonized after an auto-restart
526 - #282: TIT convertor erase traces when using absolute path in the trace list file
527 - #285: segfault when a s4u actor kills itself with msg process tracing activated
528 - #286: Pajé traces are not exposing the number of cores
529 - #287: Command-line parsing should stop when encountering '--'
530 - #288: MPI_Init(NULL, NULL)
531 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
532 - #290: Method to check if a mailbox has a message ready to be consumed
533 - #291: [s4u] semaphores do not seem available yet
534 - #294: Alltoall collective implementation of mpich
535 - #295: Starting a Host as OFF is not supported
536 - #296: DTD too permissive
537 - #299: Add a s4u barrier example
539 ----------------------------------------------------------------------------
541 SimGrid (3.20) Released June 24. 2018
543 The proxy snake_case() release.
546 - The whole API is now using snake_case() instead of mixing it with camelCase.
547 Compatibility wrappers in place for one year.
548 - We took the opportunity to increase the consistency of the API.
549 Please report any remaining glitches.
550 - New function simgrid::s4u::Host::get_actor_count:
551 Returns the number of actors running on a specific host.
554 - SMPI is now tested with ~45 proxy apps from various sources, with none or
555 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
556 - Replay: The replay file has been re-written in C++.
557 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
558 supported. They are stored in the trace and used when replayed.
559 - Basic support of MPI_Cancel. Robustness not guaranteed.
560 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
561 (only for MPI_COMM_TYPE_SHARED).
562 - New option: smpi/privatize-libs, to add external shared libs to be privatized
563 by SMPI. They will be copied locally and loaded separately by each process.
564 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
565 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
566 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
567 with internal index optind with multiple processes. Only works if MPI_Init has
569 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
570 syntax without preprocessing tricks.
573 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
574 behavior tracing is no longer limited to MSG
575 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
576 this means that traces now need to be updated or re-obtained; both Irecv
577 and Isend lines in your traces have now in total 5 mandatory fields:
578 <rankid> <command> <to/from rankid> <tag> <size>
579 To update your traces, it suffices to add a 0 for the tag here.
580 - TIT now also supports waiting for a distinct request via MPI_Wait.
581 Wait/Test now wait for a specific request, not just the last one that was
582 issued. This unfortunately means another update, because we need to
583 identify which request you want to wait for. We do this via the
584 triplet (sender, receiver, tag), which needs to be added:
585 <rankid> <command> <sender> <receiver> <tag>
586 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
587 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
590 - The deprecation of MSG is ongoing (but this should not impact you).
591 Many MSG functions are now simple wrappers on the C API of S4U. If
592 you wish to convert your code to S4U, find the S4U counterparts of
593 your MSG calls in src/msg/msg_legacy.cpp.
594 - MSG can still be used, but won't evolve anymore.
597 - Allow to run the Link energy plugin from the command line with
598 --cfg=plugin:link_energy
599 - Rename Energy plugin into host_energy
600 - Rename Load plugin into host_load
603 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
604 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
605 - Improve performance when handling timeouts of simix synchros.
608 - Config: the C API is now deprecated (will be removed in 3.23), and
609 the C++ API has been slightly improved.
612 - Fix several build issues on OSX.
613 - Move simgrid_config.h to simgrid/config.h (old header still working)
616 - #143: Setting a breakpoint at a given time
617 - #150: Inconsistent event names in SMPI replay
618 - #258: daemonized actors hang after all non-daemonized actors have completed
619 - #267: Linker error on unit_tmgr
620 - #269: SMPI: tracing of MPI_Wait/all/any broken
621 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
623 ----------------------------------------------------------------------------
625 SimGrid (3.19.1) Released March 22 2018
627 The Fixed ABI Release.
629 As you may know, we are currently refactoring SimGrid in deep.
630 Upcoming SimGrid4 will be really different from SimGrid3: modular,
631 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
633 Our goal is to smooth this transition, with backward compatibility
634 and automatic update paths, while still progressing toward SimGrid4.
636 SimGrid remains open during works: The last pure SimGrid3 release was
637 v3.12 while all subsequent versions are usable alpha versions of
638 SimGrid4: Existing interfaces remain unchanged, but the new S4U
639 interface is budding and the internals are deeply reorganized.
641 Since 2015, we work hard to reduce the changes to public APIs. When
642 we need to rename a public library symbol in S4U, we let your
643 compiler issue an explicative warning when you use the deprecated
644 function. These messages remain for four releases, i.e. for one full
645 year, before turning into an error. Starting with v3.15, your can
646 also adapt to API changes with the SIMGRID_VERSION macro, that is
647 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
649 Starting with this v3.19.1, our commitment to reduce the changes to
650 the public interfaces is extended from the API to the ABI: a program
651 using only MSG or SimDag and compiled against a given version of
652 simgrid can probably be used with a later version of SimGrid without
653 recompilation. We will do our best... but don't expect too much of
654 it, that's a really difficult goal during such profund refactoring.
656 The difference between v3.19 and v3.19.1 is that the former was
657 accidentally breaking the ABI of MSG, while the later is restoring
660 S4U and kernel APIs will still evolve until SimGrid4, with one-year
661 deprecation warnings as currently. In fact, cleaning up these
662 interfaces and converting them to snake_case() is one release goal of
663 v3.20. But don't worry, we are working to smooth this upgrade path.
665 In summary, new projects should start with S4U to benefit of the
666 future, but old MSG projects should still be usable with no change.
668 ----------------------------------------------------------------------------
670 SimGrid (3.19) Released March 20 2018
672 The Moscovitly-cold Spring release.
674 SMPI user-visible improvements:
675 - Support of MPI_Type_create_subarray
676 - Unimplemented MPI calls now abort (instead of issuing a warning)
677 Some harmless calls around errhandlers still only issue a warning.
678 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
679 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
680 is now possible (mmap privatization was refactored to this end).
681 Adding new SMPI instances should also be technically doable now.
683 SMPI internal cleanups:
684 - Removed index notion from SMPI -- no more "getPid() - 1"!
685 In tracing files, actors are called rank-<PROCESS-ID> but these are
686 NOT the real ranks (tracing several communicators is not supported yet).
687 - Deprecate smpi_process_index() (will be removed in 3.22)
688 - Better testing on classical proxy apps, through an external project:
689 https://github.com/simgrid/SMPI-proxy-apps/
692 - Use a graphical TOC to make it easier to find the documentation you need
693 - Revamp the MSG tutorial
696 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
697 - Remove parameter of MSG_process_killall().
698 Resetting the PID was bogus anyway (several actors could have the same PID).
701 - Execution->setHost() can be called after start(), to migrate it.
702 - Comm::test_any() is now implemented.
703 - s4u::Actor now has onCreation() and onDestruction() signals.
704 - Install some kernel header files for the users' plugins and more.
705 Warning, their API is really not stable yet.
706 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
707 include/simgrid/kernel/routing/*Zone.hpp
710 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
712 Build System and other cleanups:
713 - Remove unused run-time parameter "tracing/onelink-only".
714 - SimGrid now works with intel compilers (MC still buggy)
715 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
716 - tesh kills subprocesses on timeout (but not on windows)
719 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
722 - #194: Feature request: simgrid::s4u::Comm::test_any()
723 - #245: migrating an actor does not migrate its execution
724 - #253: Feature Request: expose clusters as objects
725 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
726 - #255: Tesh broken on Windows
727 - #256: Modernize FindSimGrid.cmake
728 - #257: Fix (ab)use of CMake install
730 ----------------------------------------------------------------------------
732 SimGrid (3.18) Released December 24 2017
734 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
736 MSG IS NOW DEPRECATED. Please use S4U for new projects.
737 - Support for MSG should not be removed from SimGrid before 2020,
738 but future evolutions will be limited to the integration of
739 user-provided patches.
742 - s4u::onDeadlock() signal, run before stopping the simulation.
743 - s4u::Exec: asynchronous executions (abstraction of background
745 - s4u::Host->getLoad() returns the achieved speed in flops/s
748 - Link::name() is deprecated and will be removed in v3.21, use
749 Link::getCname() instead.
750 - Mailbox::getName() changed to return a std::string, use
751 Mailbox::getCname() instead to get a char*.
752 - Storage::getName() changed to return a std::string, use
753 Storage::getCname() instead to get a char*.
754 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
755 to have both allocation and deallocation of the map in user space.
758 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
759 The amount of flops remaining to do is a vector, not a scalar.
760 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
761 name implies on both sequential and parallel tasks.
762 - Both changes fix GitHub's #223 using PR #237 as a basis.
763 Thanks Michael Mercier.
764 - Most examples were converted to S4U and hidden elsewhere at tests.
765 Rationale: we still want MSG to work; we want newcomers to use S4U.
768 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
769 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
772 - Switch to the faster dlopen privatization mechanism by default
773 - Documentation improvments
776 - Live migration is getting moved to a plugin. Dirty page tracking is
777 the first part of this plugin. This imply that VM migration is now
778 only possible if one this function is called:
779 - C/MSG: MSG_vm_live_migration_plugin_init()
780 - C/C++: sg_vm_live_migration_plugin_init()
781 - Java: Msg.liveMigrationInit()
782 For C and C++, "simgrid/plugins/live_migration.h" also has to be
786 - Define class simgrid::xbt::Path to manage file names.
787 - Removed unused functions:
788 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
789 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
790 xbt_graph_export_graphviz()
791 - xbt/str.h: xbt_str_join()
792 - Remove unused datatypes:
793 - xbt/heap.h: use std::priority_queue or boost::heap instead
794 - xbt/swag.h: use boost::intrusive::list instead
797 - New link_energy plugin for the consumption of the links.
798 - All of the operations on files and storage contents have been
799 packaged into a plugin (src/plugins/file_system). The current
800 public interface can be found in
801 include/simgrid/plugins/file_system.h
802 To use these functions you now have to initialize the plugin by
803 calling MSG_storage_file_system_init() just after calling
804 MSG_init() or sg_storage_file_system_init() just after creating
808 - Remove the undocumented/untested tag <include>
811 - Remove viva specific tracing as the tool is no longer maintained
814 - #248: Kill of finished processes leads to segfault
815 - #240: xbt_cond_wait_timeout should gracefully return for C apps
816 - #239: please implement signal s4u::onDeadlock()
817 - #230: segfaults when exit() without run()
818 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
819 - #223: MSG_task_get_flops_amount() not working with parallel tasks
820 - #222: Actor::kill() doesn't really kill and segfaults
821 - #221: odd LMM warning when killing an actor
822 - #120: Memory leak when the processes are forcefully killed
824 ----------------------------------------------------------------------------
826 SimGrid (3.17) Released October 8 2017
828 The Drained Leaks release: (almost) no known leaks despite the tests.
830 Even very long simulations will preserve your memory: our testsuite
831 stresses SimGrid for over 45mn on fast machines for a coverage of
832 over 80%, and there is only one single known leak, of about 4kb.
836 - Comm.detach(): start and forget about asynchronous emission. A cleanup
837 handler may be given to free resources if the comm cannot be completed.
838 - this_actor::send(mailbox) is now mailbox->put()
839 - New: simgrid::s4u::Comm::wait_all()
840 - New: Host.execute() for remote executions.
843 - Unused option network/sender-gap is removed.
846 - Deprecate MSG_task_isend_with_matching(): This unused feature
847 really complicates our internals. Will be removed in v3.20.
850 - Improved context termination. It is now possible to free resources, even
851 when a process is forcibly killed.
852 - ContextBoost: add support for Boost versions above 1.61.
855 - Bring back run-time option --cfg=exception/cutpath to remove exception
857 - Removed unused functions:
858 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
859 xbt_str_rtrim(), xbt_str_trim().
860 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
863 - Removed header files obsolete since SimGrid 3.12:
864 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
866 - #3: SD_exit should be made optional
867 - #120: Memory leak when the processes are forcefully killed
868 - #159: Threading test regression in Actor refcounting
869 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
870 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
871 - #186: Actor::killAll() segfaults if some process is blocked on wait()
872 - #191: VM migration and pstate
873 - #192: Updating the energy consumptions of all hosts crashes with VMs
874 - #195: All actors have PID=0 in the logs
875 - #204: Sometimes segfault with thread contexts and mmap privatization
876 - #222: Actor::kill() doesn't really kill and segfaults
877 - #225: Actor::kill() doesn't really kill when victims are doing a join()
879 ----------------------------------------------------------------------------
881 SimGrid (3.16) Released June 22. 2017.
883 The Blooming Spring Release: developments are budding.
886 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
887 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
888 - Java JAR file should be finally fully working on Mac OSX too.
890 XML platforms: Switch to platform v4.1 format.
891 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
892 - <zone> can be used as a synonym for the now deprecated <as>
893 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
894 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
895 - <actor> can be used as a synonym for the now deprecated <process>
896 - state_file and avail_file periodicity is now easier to express
897 (check the documentation) the old behavior should still work.
898 * Storage: not backward compatible, but it seems that nobody ever used it
899 - Remove attribute 'content_type' of <storage_type>: was never used
900 - Make attribute 'model' of <storage_type> optional: for future usage
901 - Remove Bconnection model property: was never (in)validated. Replaced by
902 the maximum of the read and write bandwidth as a resource constraint.
905 - New and Backwards Compatibility break:
906 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
907 When one wants to get the list of tasks whose states have changed during a
908 simulation round, s/he has to allocate and free a dynar and use it as
909 argument to this function. The former SD_simulate (double how_long)
913 - Allow multicore VMs, with the correct sharing (unless you overcommit)
914 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
917 - New function to update the consumption of all hosts at once.
918 - Fix the model for multi-core hosts, linear in the amount of busy
919 cores with an abnormality for pIdle (see paper).
920 BUG: Still not satisfactory for ptask on mono-cores.
923 - The netzone are now available from the MSG API.
924 The old names still work, but are now deprecated.
927 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
928 - New option: smpi/keep-temps to not cleanup temp files
929 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
930 mallocs mode is used, allows to change the size of the fake file used
931 (default 1MB), to potentially limit the number of mappings for large runs.
932 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
933 - Fortran ifort and flang compilers support
934 - New RMA calls supported (experimental):
935 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
936 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
937 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
938 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
939 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
940 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
943 - New: this_actor::isMaestro() returns whether we are in kernel mode.
944 - Behind the scene, ActivityImpl are now automatically refcounted.
945 Removing this nasty bottleneck will greatly simplify our internals.
948 - Replay: New function xbt_replay_action_get():
949 Retrieve the function previously associated to an event type.
950 - DROPPED FUNCTION: xbt_str_varsubst()
951 - DROPPED MODULE: strbuff. We don't need it anymore.
952 - DROPPED MODULE: matrix. We don't need it anymore.
953 - DROPPED MODULE: lib. We don't need it anymore.
955 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
957 SimGrid (3.15) Released March 22. 2017
959 The Spring Release: continuous integration servers become green
960 We fixed even the transient bugs on all target architectures:
961 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
962 Windows, FreeBSD, NetBSD.
964 - Rename NetCards to NetPoints.
965 This was intended to help NS3 users, but that's not a netcard.
966 That's a point in the routing algorithm, let's avoid wrong simplifications.
969 - New: MSG_process_yield(). Stop and yield to other processes.
970 - New: MSG_process_daemon(). Daemon processes are automatically killed
971 when the last non-daemon process terminates
972 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
973 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
974 to make room for the upcoming network energy plugin.
975 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
978 - Ensure that an actor can kill itself with Process::exit()
979 - Kill the obscure NativeException. Nobody want to survive the issues
980 it denotes, so use JniException that is a RuntimeException (not to
981 be caught explicitly).
982 - Partial bug fix in initialization. SimGrid flags on command line were
983 consumed at C level but stayed in the original Java String[] args.
984 This could mess users' args[i] if SG flags were not put at the end of
986 The SimGrid flags are now removed from the Java arguments. However,
987 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
988 if args.length is greater than the number of YOUR OWN ARGUMENTS.
989 It might be if you have --log or --cfg flags in the command line.
990 - Fix numerous memleaks all around the place. In particular, around VMs.
994 - s4u::onPlatformCreated: right before the simulation starts
995 - s4u::onSimulationEnd: right after the main simulation loop
996 - s4u::onTimeAdvance: right after a clock change (time is discrete)
997 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
998 event from the availability_file changes the avail speed.
999 - Links are now usable from s4u
1000 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1001 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1002 - New examples: The conversion of MSG examples to S4U has begun
1003 - Actors: create, kill, migration, and suspend
1004 - Applications: master-worker and token-ring
1005 - Action replay: communications and storage
1006 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1009 - Backwards Compatibility breaks
1010 - The SD_route_get_list and SD_route_get_size functions have been removed.
1011 They are replaced by the unique
1012 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1013 The route from src to dst is built in the links dynar whose size can
1014 be obtained with xbt_dynar_length.
1015 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1016 sg_host_route_bandwidth and sg_host_route_latency.
1017 Macros ensure the backwards compatibility, but you should fix your code
1020 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1021 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1022 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1023 We don't support 100% of it, but it should be enough. Contact us if not.
1024 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1025 naive and probably inaccurate)
1026 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1027 mapped onto a unique small file using some system magic.
1028 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1029 - Fortran: cleanups, fixes, support of user-added operations
1030 - RMA: MPI_Accumulate are applied in correct order.
1031 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1032 - Better support for MPI_IN_PLACE
1033 - Support for MPI_Win attrs and keyvals.
1034 - Support MPI_Comm_group_create, MPI_Type_size_x
1035 - MPI ops should be properly applied to all allowed types
1036 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1037 algorithm by a non failing one from mpich.
1040 - Kill the fifo data container: we don't use it anymore.
1043 - A <cluster> can now be created with different speed values to represent pstates.
1044 see examples/platforms/energy_cluster.xml for an example.
1046 SimGrid (3.14.159) Released December 28. 2016
1048 The previous release was only a crude approximation of a Pi release;
1049 this one should be a bit better.
1051 - Revert a last minute change that broke on Mac OSX.
1052 - Fix the changelog and NEWS files.
1053 - Improve the documentation, in particular of the routing module.
1054 - Integrate some patches flying around in Debian.
1056 SimGrid (3.14) Released December 24. 2016
1059 * The whole documentation was reworked and reorganized. There is
1060 still much room for improvement here, but we're on it.
1063 * We now need python3 for our testing infrastructure.
1064 * Model checking needs libevent
1067 * Now works on FreeBSD
1070 * Privatization now works on FreeBSD
1072 * Privatization is now activated by default in smpirun
1073 You can turn it off with -no-privatize if something goes wrong.
1075 * Call-location tracing for SMPI.
1076 You can add the exact location (filename / linenumber) of an MPI call to
1077 your trace files and slow down or speed up the simulation between two
1078 consecutive calls by using an adjustment file (see the documentation).
1080 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1081 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1082 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1083 values were ignored in some cases. The timings of these functions can now
1084 be significantly different.
1086 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1087 smpi/running-power is renamed to smpi/host-speed
1089 * smpi/grow-injected-times option to enable or disable multiplication of the
1090 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1091 make simulation less precise (but also much faster).
1093 * smpirun script should be (much) faster for large deployments.
1095 * SMPI tracing: fixed issue with poor matching of send/receives.
1097 * Replay: Fix broken waitall
1099 New functions and features
1100 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1102 Dropped / renamed functions and features
1103 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1104 - MSG_mailbox_is_empty() -> Mailbox::empty()
1105 - MSG_mailbox_front() -> Mailbox::front()
1106 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1107 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1108 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1109 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1110 * Don't pass the free_f parameter to property related functions:
1111 - MSG_host_set_property_value()
1112 - MSG_as_router_set_property_value()
1113 - MSG_storage_set_property_value()
1114 * VM properties. Since msg_vm_t are msg_host_t, just use
1115 MSG_host_get_property_value() and friends
1116 * VM I/O related things:
1117 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1118 - Unimplemented save/restore methods
1119 * MSG_as_router_get_property_value() was redundant with
1120 MSG_environment_as_get_property_value().
1121 - Removed MSG_as_router_*propert*() functions
1122 - Added MSG_environment_as_set_property_value() for consistency
1123 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1124 Well, they are still there for now, but deprecated with a warning.
1125 Please switch to xbt_dict_new_homogeneous() before this is removed
1127 * Task affinity. Its intended behavior (that was very badly tested
1128 and probably not really working) was deceiving what most users
1129 would have hoped here.
1130 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1131 * Option network/coordinates is now useless and should be dropped.
1134 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1137 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1138 * Added Dragonfly topology support
1141 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1142 EiBps, EBps, Eibps, Ebps for bandwidth.
1143 They may become useful to some lucky ones.
1146 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1147 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1148 * ex/async_waitAll: new example, on asynchronous communications
1151 * Memory usage should be decreased for simulations with a large number
1152 of processes. This also helps for SMPI.
1154 SimGrid (3.13) stable; urgency=low
1156 The Half Release, a.k.a. the Zealous Easter Trim.
1158 - We removed half of the lines, that were mostly experimental cruft.
1159 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1160 (not counting blanks and comments -- according to openhub.net)
1161 - The internals are now compiled in C++ (and will soon be clean C++)
1162 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1163 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1165 * Backwards Compatibility breaks
1166 - Removed Lua simulation bindings (switch to C or Java for that).
1167 Lua can still be used to describe platforms
1168 - Removed Java kernel plug-ins.
1169 Will be reintroduced after the ongoing major internals reorg.
1171 - the following functions were removed.
1172 They were too specific and should be reimplemented in a generic
1173 way, with filter function.
1174 - MSG_task_listen_from_host
1175 - MSG_mailbox_get_count_host_waiting_tasks
1176 - MSG_mailbox_put_with_timeout was removed.
1177 Please use MSG_task_send_with_timeout instead.
1179 - the SD_application_reinit function was removed. It has been a noop for a while.
1180 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1181 replaced by a more flexible mechanism.
1182 - The following functions thus do not exist anymore
1183 - SD_workstation_get_access_mode
1184 - SD_workstation_set_access_mode
1185 - SD_workstation_get_current_task
1186 - Basic estimation functions have been removed but can easily be replaced
1187 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1188 - SD_workstation_get_computation_time => amount / sg_host_speed()
1190 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1191 Use VM.getSpeed()*load/100 for the legacy behavior.
1193 - option enable_tracing was removed. It was not doing anything for a while.
1194 - In the ModelChecker:
1195 - the model-checker now ptraces the model-checked process which means
1196 you cannot use a debugger on the latter anymore (we might make this
1197 optional in the feature);
1198 - removed soft-dirty page tracking;
1199 - remove model-checked side snapshot management,
1200 MC_snapshot() and MC_compare_snapshot();
1201 - keep the MC_cut() function as a stub (it was not really working
1202 in the previous release).
1205 * All options are consistently in kebab-case. Old names are kept as alias.
1208 * Switch to platform v4 format.
1209 - Rename from 'power' to 'speed' the attributes describing the amount of
1210 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1211 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1212 - In <host> and <link>, attributes availability and state are gone.
1213 It was redundant with state and availability traces, and with peak values.
1214 - In <cluster>, attributes availability_file and state_file are gone.
1215 It was too complex and unused.
1216 - Kill <gpu>. Was not doing anything.
1217 - The DOCTYPE points to the right URL:
1218 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1219 (the file at this address now documents the changelog since its v1)
1220 - A warning is emitted for unit-less values (they are still accepted).
1221 - speed. Default: 'f' or 'flops'. Also defined:
1222 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1223 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1224 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1225 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1226 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1227 - latency. Default: 's' second. Also defined:
1228 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1230 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1231 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1234 * s4u::Host is now the preferred public interface to the Host features.
1235 sg_host_* functions are C bindings to the exact same behavior
1236 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1239 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1240 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1241 They were deprecated since a few years, and probably did not even compile anymore.
1244 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1245 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1246 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1247 encouraged to update their codes. A list of the modified functions can be found at the end of
1248 include/simgrid/simdag.h
1251 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1252 the lambda or closure passed as a parameter will run in kernel mode.
1253 All the callback functions should be rewritten to that interface at some point.
1256 * Reorganizing and cleaning the internals all around the place.
1259 * Remove old default barrier/bcast buggy algorithms (see #18407)
1260 * Various bug fixes to handle more codes
1261 * Remove the need for the --foreground option of smpirun (it is still
1262 accepted for backward compatibility).
1265 * Kill the setset data container: MC don't use it anymore.
1266 * Kill the queue data container: it made more sense with GRAS.
1267 * Kill the xbt_peer_t data type: it's useless without GRAS.
1268 * Kill rm_cb feature of config sets: it was never useful.
1269 * Kill graphxml parsing feature. It was not used.
1270 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1272 - xbt_dynar_sort_strings(), when the content is char*
1273 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1274 They throw exceptions on invalid input;
1275 * C++ support for declaring CLI flags (simgrid::config::Flag);
1276 * class for abstracting different signal backends (simgrid::xbt::signal).
1277 with no external dependencies (we need very simple signals).
1280 * refactoring and cleanup of the code;
1281 * ongoing process to cleanly separate the model-checking algorithms
1282 from the code model-checking support.
1284 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1286 SimGrid (3.12) stable; urgency=low
1288 The Facelift Release.
1291 * Require g++ v4.7 at least to not speak prehistorical C++.
1292 * Require Boost 1.48 (for signal2 component).
1293 * Java must be version 7 at least when activated.
1294 * Builds on Windows again (including Java bindings).
1295 * Tracing is now always enabled (no way to turn it out)
1296 * Remove GTNetS. It was not working anyway.
1297 * Various cleanups in the cmake scripts.
1298 * Move headers around to sort them out on installed systems:
1299 - instr/instr.h -> simgrid/instr.h
1300 - instr/jedule/* -> simgrid/jedule
1301 - simdag/datatypes.h was removed
1302 - simdag/simdag.h -> simgrid/simdag.h
1303 - msg/datatypes.h was removed
1304 - msg/msg.h -> simgrid/msg.h
1307 * Interface improvement:
1308 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1309 - Sanitize the interface in MSG_task_ module:
1310 - Merge two functions that were close enough but misleading:
1311 set_compute_duration(t) -> set_flops_amount(t)
1312 get_remaining_computation(t) -> get_flops_amount(t)
1313 - set_data_size(t) -> set_bytes_amount(t)
1314 get_data_size(t) -> get_bytes_amount(t)
1315 - Massive cleanups in the functions related to the energy
1316 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1317 - New: MSG_host_get_pstate()
1319 - msg/energy/onoff: switching hosts on and off
1321 * Interface improvement:
1322 - Rename Host.isAvail() to Host.isOn()
1323 - Rename Process.currentProcess() to Process.getCurrentProcess()
1324 - Rename Task.setDataSize() to Task.setBytesAmount()
1325 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1327 - #18874: Actually allows the GC to reclaim tasks
1331 - SIMIX_process_throw: raises an exception in a remote process
1332 * Refactoring: Separate sub-modules
1333 - libsmx: the public interface, as libc in a real system
1334 - popping: the strange dance that converts a user request into a kernel handling
1335 - smx_context_*: the virtualization mechanisms that embed the user code
1336 - smx_*: the handling of each simcalls
1337 * Interface cleanup:
1338 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1339 * Rename smx_action_t into smx_synchro_t, making explicit that these
1340 things are used to synchronize processes with their environment.
1341 For example, a communication is a sort of synchronization involving
1342 the communicating processes (that may block until the exchange) and
1343 the platform. The same can be said from computations, etc.
1345 - #18888: segfault when a process ends before its kill_time
1349 - 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.
1350 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1351 - MPI_Type_set_name, MPI_Type_get_name
1352 - MPI_*_c2f and MPI_*_f2c functions
1353 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1354 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1355 - Activate a lot of new tests from the mpich 3 testsuite
1357 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1358 - InfiniBand network model added: Based on the works of Jerome Vienne
1359 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1360 - When smpi/display_timing is set, also display global simulation time and application times
1361 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1362 * Collective communications
1363 - 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
1364 - 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/).
1365 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1367 - Replay now uses algorithms from wanted collective selector
1368 - Replay can be used with SMP-aware algorithms
1369 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1370 - 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.
1372 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1373 - Correctly use loopback on fat-tree clusters
1374 - Asynchronous small messages shouldn't trigger deadlocks anymore
1375 * Energy/DVFS cleanup and improvement
1376 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1377 - new: smpi_get_host_pstate()
1380 - "Full" network optimization flag was broken since Surf++
1381 - Better handling of precision flags in maxmin
1382 - Fix bug causing sometimes "Impossible" errors
1383 - Properly pass cluster properties to included hosts
1384 * Improvement of the Energy plugin.
1385 - Always update the consumption before returning that value
1386 - New property: watt_off to denote the dissipation when the host is off
1387 - New functions getWattMinAt and getWattMaxAt to retrieve the
1388 dissipation of pstates that we are not currently at.
1389 * Java: class NetworkLink renamed to Link
1390 * New function: simcall_process_get_kill_time()
1391 * Massive rename s/workstation/host/
1392 - That's intrusive, but that's good for the project consistency. Sorry.
1393 - Change config option "workstation/model" into "host/model"
1396 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1397 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1398 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1399 - xbt_dirname and xbt_basename for non-POSIX systems
1401 * The model checker now runs as a separate process.
1402 * The model checker runs must now be launched with the new simgrid-mc program.
1403 * Record/Replay: the MC can display a textual representation of a path in the
1404 execution graph. It can then be replayed outside of the model checker.
1406 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1408 SimGrid (3.11) stable; urgency=low
1413 * Normalizing pointers addresses tool for better diff between logs
1416 * Add cloud examples using new VMs
1417 - examples/msg/cloud/two_tasks_vm.tesh
1418 - examples/msg/cloud/simple_vm.tesh
1419 - examples/java/cloud/cloud.tesh
1420 - examples/java/cloud/migration/migration.tesh
1421 * Add java surf examples:
1422 - examples/java/surfPlugin/surf_plugin.tesh
1423 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1424 - examples/java/surfCpuModel/surf_cpu_model.tesh
1425 * Add SMPI+MSG example:
1426 - examples/smpi/smpi_msg_masterslave/
1431 - msg task destroy cancel test
1432 - msg_host on/off test
1433 * Move all tests in testsuite to teshsuite (adding tesh files)
1434 * Restructure teshsuites
1435 - one folder for each kind of test
1436 * Restructure AddTests.cmake
1438 - structure the order of tests (with sections)
1441 * Add virtual machine
1442 - creation of a VM on a PM
1443 - migration of a VM from a PM to another PM
1445 - MSG_process_join(msg_process_t process, double timeout)
1446 - msg_bar_t MSG_barrier_init(unsigned int count)
1447 - int MSG_barrier_wait(msg_bar_t barrier)
1448 - void MSG_barrier_destroy(msg_bar_t barrier)
1449 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1450 * New option "msg/debug_multiple_use" to help debugging when a task is used
1453 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1455 - Increase code coverage in test suites
1459 * Protect context stacks against stack overflow. The number of protected memory
1460 pages allocated on the top of each stack (1 by default) can be configured
1461 with the new command line option --cfg=contexts/guard_size:<value>.
1462 * Simcalls are now generated by a python script that
1463 - generates files included by SimGrid
1464 - checks that all the functions exist, and proposes prototypes
1466 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1467 host_set_data, host_get_data
1469 - simcall_process_join(smx_process_t process, double timeout)
1470 * Fix bug where sleeping processing could not be suspended.
1473 * Translate surf models from C to C++
1474 - Generic classes for all models: Model, Resource, Action
1475 - A generic interface for each kind of model (CPU, Network, Storage
1476 Workstation, WorkstationVM)
1478 * Translate surf routings from C to C++
1479 * Add callbacks using sigc++ or boost::signals2
1480 - Add callback functions for resource creation/destruction
1481 - Add callback functions for action state change
1482 - Handle Energy as a plugin
1483 * Replace swag by boost::intrusive
1484 * Add new routing models for clusters. For documentation, see
1485 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1486 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1487 parameters for cluster tag
1488 - Fat trees, with topology="FAT_TREE" and
1489 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1490 - see examples/platforms/torus_cluster.xml and
1491 examples/platforms/fat_tree_cluster.xml
1492 * More documentation
1495 * Hostfiles support host:nb_processes construct to deploy several processes on
1497 * Collective communication algorithms should not crash if used with
1498 improper number of nodes and report the error.
1499 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1500 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1501 MPI_Dims_create, MPI_Cart_sub are supported.
1502 * New interface to use SMPI programmatically (still depends on MSG for
1503 some parts, see examples/smpi/smpi_msg_masterslave):
1504 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1508 * Global variables privatization in MPI executables is now performed at runtime
1509 with the option smpi/privatize_global_variables (default:no).
1510 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1511 dynamic libraries loaded by the application are not privatized (static
1512 linking with these libraries is advised in this case)
1515 * Options defined in XML work correctly now.
1518 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1519 copied into simgrid.jar or not (ON by default). Use this option if you want
1520 to reduce the size of the installed simgrid.jar, *and* the native libraries
1521 are kept installed elsewhere.
1522 * Surf binding with SWIG (code generated in maintainer mode only):
1523 - plugin to handle callbacks
1524 - CPU model only for the moment
1527 * Supernovae build mode is definitively removed. It was used to improve
1528 inlining and inter-module optimizations. It is nowadays superseded by
1529 link-time optimizations commonly available in compilers.
1530 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1531 * Add boost dependency for surf++
1532 * Add new macro for tests
1533 - ADD_TESH(name <tesh_args>)
1534 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1538 - xbt_bar_t XBT_barrier_init(unsigned int count)
1539 - int XBT_barrier_wait(xbt_bar_t barrier)
1540 - void XBT_barrier_destroy(xbt_bar_t barrier)
1541 * Make the xbt_os_time module public
1543 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1545 SimGrid (3.10) stable; urgency=low
1547 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1550 * Reintegrate Java to the main archive as desynchronizing these
1551 package is not acceptable anymore (Java is now considered stable)
1552 * Add explicit synchronization facilities through semaphores
1553 * Bug fix: Task.setDataSize() only changed the C world, not the value
1554 cached in the Java world
1557 * Dramatically change the way files are handled. API and internals changed, but
1558 this part of MSG was not considered as production grade either.
1559 * Add explicit synchronization facilities through semaphores
1560 * Add a new function MSG_host_get_process_list()
1561 * Preliminary DVFS support (see examples/msg/energy/ for details)
1564 * SMPI is now included directly in the libsimgrid as the windows
1565 linker doesn't force us on splitting it anymore.
1566 * Improvements of the SMPI replay tool:
1567 - Most of the collective communications are now rooted in the same process as
1568 in the original application.
1569 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1570 used until now). Multiple data types can now be used in a trace.
1571 - The replay tool now supports traces produce either by TAU or a modified
1573 - Bug Fix: the compute part of the reduce action is now taken into account.
1574 - Gatherv collective is now supported
1575 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1576 of smpirun outputs time independent traces for the current run. One file
1577 is created per process. If too many processes are simulated, this behavior
1578 can be changed to one file for all processes by using the
1579 tracing/smpi/format/ti_one_file flag
1580 * smpirun generates the host file if needed (with given host count and platform)
1581 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1582 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1583 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1584 --cfg=smpi/coll_selector:(mpich/ompi)
1585 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1587 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1588 happening inside a collective SMPI call.
1589 * Fix the behavior of complex data types handling.
1590 * Make MPI_Wtime another synchronization point to take computations into
1592 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1593 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1594 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1595 (no privatization of global variables yet)
1596 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1599 * Verification of liveness properties is now available for SMPI applications
1600 (in addition to MSG applications)
1601 * Bugged examples using SMPI in examples/smpi/mc/
1602 * Add --cfg=model-check/visited option. Allows the verification of infinite
1603 programs. Detection of loops in the execution thanks to the system state
1604 comparison and reduction of the state space to explore. Can be combined with
1605 DPOR for safety properties.
1608 * Allow to change SimGrid configuration (see --help) within the code
1609 thanks to SD_config() as it can be done in MSG.
1610 * Add a new function SD_task_set_amount() upon user request.
1613 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1614 * Remove rule based routing (no more PCRE dependency)
1615 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1616 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1617 * Add a loopback_bw and loopback_lat options to cluster tag.
1618 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1619 routing. Now peers are encapsulated in an AS and have their own private
1620 router but this is transparent.
1623 * Our own implementation of getline is renamed xbt_getline, and gets
1624 used even if the OS provide a getline(). This should reduce the
1625 configuration complexity by using the same code on all platforms.
1626 * New type: xbt_cfg_elm_boolean.
1627 * Allow to use yes/no for boolean configuration options in the command line.
1628 * Allow to disable SimGrid cleanups at exit from command line option.
1629 There are situations where one may want a simulation to end with an exit.
1630 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1631 annoying when scripting around the simulator. Adding a
1632 --cfg=clean_atexit:no allows to circumvent this issue.
1635 * Lots of memory leaks were corrected in this release.
1636 There is no known memory leaks anymore, in all of our 600+ tests.
1637 * New command line option --version, to get SimGrid version information.
1638 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1640 * Supernovae builds are deprecated, and expected to be removed in the next
1643 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1645 SimGrid (3.9) stable; urgency=low
1647 The Grasgory release: GRAS is really dead now.
1649 * Complete overhaul of the internal host structures scheme.
1652 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1653 was considered as experimental and badly maintained since then.
1654 * Keeping it was thus a trap to our potential users, that could take
1655 it instead of MSG or SMPI by mistake despite is pity state.
1656 * GRAS seems to have very few users (if any), and no one volunteered
1657 to maintain it further. It also induces a lot of XBT code (for
1658 portability sake), that must be maintained too.
1659 * For all these reasons, we killed GRAS. If someone wants to revive it
1660 in the future, don't cry, our git history still remembers of GRAS.
1663 * Major overhaul. Merge our documentation again as time proved that
1664 splitting it was really not helping our users.
1665 * Further improve the developer documentation to help newcomers
1666 hacking on SimGrid itself. The user documentation (and in
1667 particular, the beginner documentation) is still in a sorry state.
1670 * Now works on Windows too!
1671 * Much more extensive test suite, from MPICH
1674 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1675 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1676 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1677 * Bug fix: let task be scheduled when the last dependency to be solved is
1678 a control dependency.
1679 * Remove SD_load_environment_script function.
1680 Use the C sg_platf function if you want to declare a platform
1684 * New function: MSG_process_get_number()
1685 * Old function documented: MSG_config()
1686 * Remove MSG_load_platform_script function
1687 Use the C sg_platf function if you want to declare a platform
1691 * Change the default value of the TCP_gamma constant (maximal size of TCP
1692 congestion window) to a more realistic 4MiB value. If you notice changes in
1693 your simulation results, you can fall back to the previous 20k tiny window
1694 by adding --cfg=network/TCP_gamma:20000 on command line.
1695 * (Hopefully) fix a bug wrt periodic availability and state traces
1696 * Bug fix: use default values at start when first event in availability/state
1697 trace is not at time 0.
1700 * remove the "new_" part of function name sg_platf_new_trace_connect
1701 (resulting in sg_platf_trace_connect), since it does not create
1705 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1706 use case where it's really mandatory, and maintaining it was a pain in
1708 * New: xbt_fifo_search(), search an item with a user-provided
1709 comparison function instead of dumb pointer comparison.
1712 * Fix the lua deployment:
1713 Use `simgrid.init_application()` before deployment instead of
1714 `simgrid.msg_register_application()` after.
1717 * Transfer the tracing files into the corresponding modules.
1719 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1721 SimGrid (3.8.1) stable; urgency=low
1723 The "we are told that some people want to also *install* the simgrid
1726 * Add missing file "tesh.1" to the archive.
1728 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1730 SimGrid (3.8) stable; urgency=low
1732 The Psssshiiiit release: SimGrid jumps into the Cloud.
1735 * Add an experimental interface to manipulate VMs. They are mainly
1736 process groups with very few intrinsic semantic, but they should
1737 allow you to build the semantic you want easily.
1738 * New function: MSG_host_set_property_value()
1739 * New function: MSG_process_on_exit(). To clean memory in all cases.
1740 * Bug fixes that made the host (and link) failures unusable.
1741 * Add a way to auto-restart process when the host in which they are
1742 executing comes back (ON_FAILURE="RESTART" on deployment file,
1743 MSG_process_auto_restart_set).
1744 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1745 please stop using the old ones, they are DEPRECATED.
1747 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1748 Please use MSG_init() instead. (reducing the amount of entry
1749 points in the library helps us).
1750 * Make it impossible to link against the wrong version of the lib
1751 * Deprecate MSG_clean(). No need to call it anymore.
1752 * Function MSG_get_host_number() is not deprecated anymore.
1755 * Split the doc into a user guide and a reference guide.
1756 * Start a developper guide to help people hacking on SimGrid.
1759 * Enable tracing by default. This modules rocks you should use it.
1760 * Remove option custom_flags. Now use environment variables CFLAGS
1762 * Use default cmake things to detect lua instead of home grown ones.
1763 * New option "enable_mallocators" to disable mallocators, for debugging
1764 purpose ("on" by default).
1767 * Bug fixes around the resource failures: don't let the processes
1768 survive the host they are running onto.
1769 * Add an interface to auto-restart processes when the host in which they are
1770 executing comes back.
1771 * Ensures that SIMIX_clean is called automatically. It's not part of
1772 the public interface anymore (bindings should be updated).
1775 * Bug fix for when SD_Simulate is called with a positive value: be careful
1776 when comparing doubles. Sometimes they are different for non significant
1778 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1779 parallel task whose initial work is distributed among host according
1780 to the Amdahl's law. Such tasks are created with a parameter alpha
1781 that corresponds to the non-parallelizable part of the computation.
1782 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1783 between two sets of workstations assuming a 1D block distribution (each
1784 workstation owns a similar share of data) on both sides.
1786 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1787 Data redistribution will be automatically scheduled once parent and child
1788 are both scheduled. The filling of computation_amount and
1789 communication_amount structures is now done seamlessly thanks to the chosen
1791 * New function SD_workstation_dump to display various information
1792 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1793 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1794 bandwidth on the route onto which the task is scheduled and the amount of
1796 To divide the nominal bandwidth by 2, the rate then has to be:
1797 rate = bandwidth/(2*amount)
1798 * Compute tasks that have failed can now be rescheduled and executed again
1799 (from their beginning)
1800 * Increasing source code coverage (src/simdag is now covered at 95.8%
1804 * Re-implement time-independent trace replay using SMPI (at the
1805 smpi_smp_* level) instead of MSG. This should replace
1806 examples/msg/actions/actions.c
1807 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1808 hindexed and structs)
1809 * Implement the exchange of non-contiguous data.
1810 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1811 * Correct behavior of smpi/sender_gap and set its default value to 0
1812 * Add option to asynchronously send small messages to allow better
1813 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1814 specifies the size in bytes under which messages will be asynchronously sent.
1815 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1816 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1819 * New C interface to define a platform: XML is now optional.
1820 For more info, please check include/simgrid/platf.h.
1821 * New interface to define random platforms from the C:
1822 For more info, please check include/simgrid/platf_generator.h and
1823 examples/msg/masterslave/masterslave_platfgen.c
1824 * Export a sg_cmdline dynar containing all the arguments we got from
1828 * Two new tracing options for adding comments to trace file so you
1829 can track your experiments (see --help-tracing for details).
1830 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1831 * Adding the SimGrid version that generated the trace file as a comment.
1832 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1833 * Fix to avoid key clashes on Paje links
1834 * Other minor fixes related to the Paje specification
1837 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1838 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1839 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1840 * Greatly improve the robustness of mmalloc to user errors (such as
1841 using an area after freeing it, or freeing it twice)
1843 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1845 SimGrid-java (3.8.1) stable; urgency=low
1847 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1848 Warning: all methods are visible, but only some of them are
1849 implemented so far. Check the source (src/jtrace.c)
1850 for further information.
1851 * New module: org.simgrid.msg.File (SimGrid File management functions)
1852 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1853 * Change the meaning of Process.restart: now restart the process from
1854 the beginning, like MSG_process_restart in C.
1855 * Add Process.setAutoRestart: handling of process restart when failed
1857 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1858 you to retrieve the properties of the processes/hosts
1859 * Deprecate Msg.clean(): you can just forget about it now.
1860 * New function Process.getCount(), that only works when compiling
1861 with the not yet released version 3.9 of the C library.
1863 * New context factory based on Coroutines. It mandates a modified JVM
1864 but then, the simulations run about five times faster, and there is
1865 no limit to the amount of processes (beside of the available memory).
1867 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1869 SimGrid (3.7.1) stable; urgency=low
1872 * Restore the prototype of MSG_process_create_with_environment() to
1873 the pre-3.7 situation by removing the kill_time argument.
1874 * Add a MSG_process_set_kill_time() function instead.
1877 * Fix weird behaviors when dealing with parallel tasks.
1880 * Simgrid is now built as a dll.
1881 * Simgrid-java now works on Windows.
1882 * Simgrid-Java is now included into Windows package.
1885 * First pre-build package for MacOSX.
1888 * Fix compilation when using MSG_USE_DEPRECATED.
1889 * Fix some compilation issues on Macs and Windows.
1890 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1891 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1893 We discovered that the Lua console is broken, but we are missing the
1894 manpower to fix it right now. The problem existed in 3.7 too, so we
1895 are not blocking the release for that. Sorry if you depended on this
1896 feature, any help would be really welcome.
1898 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1901 SimGrid-java (3.7.1) stable; urgency=low
1903 The "Java aint got to be bloated and slow" release
1906 * Various internal cleanups and performance improvement
1907 Simulations are expected to run up to twice faster or so
1908 * Make Process.kill(process) an instance method, not a static one
1909 * User processes are not java.lang.Thread subclasses.
1910 This breaks the compatibility (sorry), but previous API was
1911 brain-dead, making it impossible to have non-trivial
1912 initializations in the process constructor.
1913 * Require a full constructor per Process sub-class.
1914 Kinda breaks the compatibility (sorry), but this allows a much more
1915 efficient way to launch the processes at simulation startup.
1916 * Do not embeed our version of semaphores, java 1.5 can be considered
1917 as sufficiently prevalent for us to not dupplicate its features.
1921 * Add examples for almost every part of the API
1922 We spotted and fixed a lot of bugs in the process
1924 * New module: asynchronous communication API
1925 * New function: Process.sleep()
1926 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1927 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1928 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1930 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1932 SimGrid (3.7) stable; urgency=low
1934 The "spring cleanups (before the next Big Project kicks in)" release.
1937 * We can specify the SMPI latency/bandwidth factor with command line
1938 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1939 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1940 You can also use the "config tag" from platform file by adding this line
1941 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1942 (see "example/platforms/tag_config.xml" to use "config tag").
1943 Note that the command line supersedes the platform file configuration.
1944 * Change the correction factors used in LMM model, according to
1945 the latest experiments described in INRIA RR-7821.
1946 Accuracy should be improved this way.
1947 * Use the partial invalidation optimization by default for the
1948 network too. Should produce the exact same results, only faster.
1949 * Major cleanup in surf to merge models and split some optimization
1950 mechanisms from the core of the models. As a result you can now
1951 specify which model to use (e.g., --cfg=network/model:LV08
1952 --cfg=cpu/model:Cas01) and which optimization mode to use
1953 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1954 Incompatible combinations should err at initialization. See
1955 --help-models for the list of all models and optimization modes.
1956 * The CLM03 workstation model was dropped for simplicity because it
1957 used the deprecated CM02 network model. Use default instead.
1958 * Rename the TCP_gamma configuration option to network/TCP_gamma
1959 * Rename the coordinates configuration option to
1960 network/coordinates, and document it
1961 * Use now crosstraffic keyword instead of the terribly misleading
1962 fullduplex keyword. It is activated by default now in the current
1963 default model, use --cfg=network/crosstraffic:0 to turn it off.
1964 * Ongoing refactoring the model parsing to make XML files optional
1965 See include/simgrid/platf.h for details (still to be completed)
1968 * Major overhaul of the documentation. Almost instructive now :/
1969 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1970 functions and friends. This interface was considered as
1971 deprecated since over 2 years, it's time to inform our users that it is.
1972 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1973 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1974 use these (crappy) functions in your code.
1975 These functions will be removed soon. Stop using them now.
1976 * Deprecate MSG_get_host_{table,number}
1977 Implement MSG_hosts_as_dynar() instead.
1978 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1979 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1982 * Stabilize the parallel execution mode of user contexts
1983 * Introduce configuration variables to control parallel execution:
1984 - contexts/synchro: Synchronization mode to use when running
1985 contexts in parallel (either futex, posix or busy_wait)
1986 - contexts/parallel_threshold: Minimal number of user contexts
1987 that must be part of a scheduling round to switch to parallel
1988 execution mode (raw contexts only)
1989 * Fix bugs that prevented to use suspend/resume along with
1990 synchronization structures.
1991 * Fix bugs in process termination that lead to invalid memory access
1992 in very specific conditions.
1995 * Introduce a parallel mode for the models (controlled by surf/nthreads
1996 configuration item). In our tests, running the models in parallel
1997 never lead to any speedups because they are so fast that the gain
1998 of computing each model in parallel does not amortizes the
1999 synchronization costs, even when ultra fast futexes are used.
2000 This is released anyway because YMMV.
2003 * Performance boost by using a swag internally to compute the set of
2004 tasks that are finished and should constitute the return value of
2008 * Enable it by default now that it is considered rather stable.
2011 * Documentation of the tracing functions.
2012 * Performance gains when tracing categorized/uncategorized resource
2013 utilization by avoiding calls to get route when updating resource
2014 variables. LMM constraints are being used instead.
2015 * API changed to set task categories. Use MSG_task_set_category instead
2016 of TRACE_msg_set_task_category, and SD_task_set_category instead
2017 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2018 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2019 * Examples for link user variables added (see at examples/msg/tracing/)
2020 * Deprecated function TRACE_msg_set_process_category completely removed
2021 * Trace header updated according to the latest Paje file format
2022 * Tracing network lazy updates, no longer obligate users to use full updates
2023 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2024 * Experimental: let user code declare/set/push/pop application states for hosts
2025 * API changed to allow the manual creation of graph configuration files
2026 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2029 * Improve the API of Lua MSG bindings, using the Lua spirit.
2030 * Each simulated process now lives in its own Lua world (globals are
2031 automatically duplicated). It helps writing simulators. Will allow
2032 to run Splay programs within SimGrid in the future.
2033 * Add a Chord example in Lua, equivalent to the MSG one.
2036 * Start the implementation of a solution to express temporal
2037 properties, not only local assertions. This is still an
2038 experimental work in progress, stay clear from it to be safe.
2042 - Add new runtime parameters --help-logs and --help-log-categories
2043 to display information about supported logging parameters and
2045 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2046 don't exists anymore.
2047 * Mallocators: allow value NULL for the reset function.
2049 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2050 create homogeneous dictionaries, where all the elements share the
2051 same free function. Non homogeneous dictionaries will be
2052 deprecated in the next release.
2053 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2054 - Multi-level dictionaries are deprecated.
2056 - new function xbt_dynar_search_or_negative() that is useful when
2057 you have less than 2 million elements in your dynar and don't
2058 want of the extra complexity of catching exceptions when the
2059 element is not found.
2061 - Make xbt_os_thread module (for thread portability) public.
2062 Documentation is still to come, sorry.
2064 - Cleanups and simplifications to make it maintainable again.
2065 - Exotic features (such as memalign and valloc) were removed.
2066 - The metadata were extended and improved so that the
2067 model-checker becomes able to explore and inspect the heaps.
2068 - This may induce a performance drop when enable_model-checking is
2069 ON in cmake (even if it's not used in the simulation), but it is
2070 necessary at this point to get MC working.
2072 Turn model-checking OFF if simulation performance matters to you.
2073 Not enabling it at runtime is not enough, disable it in cmake.
2075 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2078 SimGrid (3.6.2) stable; urgency=low
2080 The "Not coding new stuff allows to polish old things" release.
2083 * New bindings to the NS3 packet level simulator (experimental)
2084 * Use the raw (efficient) execution contexts instead of the sysv
2085 (portable) ones when possible.
2086 * libpcre is now mandatory in any cases since not using it led to
2087 severe performance loss and possibly other issues
2088 * Update the XML platforms:
2089 - G5K: include the latest machine in Nancy
2090 - GridPP and LCG: new platforms
2091 * Documentation was partially updated, at least (more to come)
2093 Bug fixes, cosmetics and small improvements
2094 * Free terminated processes before the end of the simulation to avoid
2095 exhausting the memory of users having very dynamic amount of
2097 * Bug fix and cosmetics about canceling non-running tasks
2098 * Bug fix about the dot loader's issues when using libcgraph
2101 * Create an installer for windows with nsis (amd64 and win32)
2102 - Add an hello world project to illustrate simgrid project creation.
2103 - Embed libpcre into the Simgrid installer to avoid
2104 its compilation burden
2105 * The raw execution contexts should work on Apple now
2106 * Port to Windows 64 bits
2107 - Sysv contexts now have an implementation for this arch
2108 - GRAS communication features now support this arch
2109 * Drop support for borland compiler on windows
2110 - this code was not maintained, and we kinda depend on gcc nowadays
2111 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2112 * Fix portability issue on unstable ubuntu: linker became picky on
2115 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2117 SimGrid (3.6.1) stable; urgency=low
2119 The "Oops, we broke Macs too" release
2122 * Fixed contexts detection so that raw ones are used when possible
2123 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2124 leads to a strange error, with user code segfaulting sometimes when
2125 the generated code is not perfectly aligned (which is not
2126 controllable from the user side, depends on the amount of code)
2129 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2131 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2133 SimGrid-java (3.6) unstable; urgency=low
2136 * Split of every thing from simgrid v3.5 into a separate package.
2138 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2140 SimGrid (3.6) stable; urgency=medium
2142 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2145 * Bindings now constitute their own package, separated from the main one.
2146 Rationale: reduce our maintenance nightmare by reducing the module coupling
2147 They will soon be released on their own on gforge.
2149 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2150 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2152 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2153 * It's not quite deprecated for now because we have no replacement,
2154 but it may soon become the case.
2157 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2158 * Fortran: New user-level cache variable to store the rank of the running
2159 process. This improves performance by an order of magnitude.
2160 * C: New coccinelle script to automatically locate and modify global and
2161 local static variables.
2162 * Improved SMPI network model with a sender-side gap to account for multiple
2166 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2167 only say if a communication is finished, no matter whether it succeeded or
2168 failed. You can call MSG_comm_get_status() to know the status of a finished
2170 * New function MSG_task_dsend() to send a task and detach it. When a
2171 communication is detached, you are never notified of its success or failure
2172 and the memory is released automatically once it is finished. This function
2173 is useful when you don't care about the end nor the success of a
2175 * Change the prototypes of action replay. Sorry for inconvenience,
2176 but this is really more efficient this way (and to adapt your code,
2177 you just have to fix the initialization, that shouldn't be too long)
2178 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2179 ever managed to do anything useful with it.
2180 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2181 takes a dynar of communications. It returns immediately and gives the
2182 index of a finished communication (if any).
2183 * New example: a basic implementation of the Chord P2P algorithm.
2186 * New model for multi-core CPUs. You can now use the core attribute to
2187 precise the number of cores of a host. This is a basic model. Every
2188 process running on the host receives at most the power provided in
2189 the DTD (throughput<=power). Total throughput of process cannot exceed
2191 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2192 router linked by an up-link and a down-link (possibly asymmetrical).
2193 This kind of pattern allows to easily build last-mile model style platforms.
2194 Aggregating such patterns in a rule-based AS is thus the technique of
2195 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2196 * New model for Vivaldi routing. We transformed the Vivaldi network model
2197 into a Vivaldi routing model (based on the rule-based model). This allows to
2198 combine Vivaldi based latencies with last-mile platforms.
2201 * Added a check for NaN of IEEE754 infinite in the double entries of
2203 * Introduce a new context factory "raw", highly inspirated from the
2204 ucontext factory, but using manually crafted functions in assembly to
2205 do the work in an efficient manner.
2206 * Allow to change the used context factory at run time, not only at
2207 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2208 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2209 contexts at runtime (only with raw contexts or ucontexts).
2210 * Completely rewrote this module to allow parallel execution of user
2211 processes. Use --cfg=contexts/nthreads:N to execute user processes
2212 with N parallel threads (the default is 1, meaning no parallelism).
2213 * Allow to decide dynamically between sequential and parallel modes.
2214 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2215 processes in parallel only when their number is greater than or equal to P
2217 * Added a check for NaN of IEEE754 infinite in the double entries of
2221 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2222 won't output the state of processes when interrupted with Ctrl-C
2223 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2224 NULL-terminated array. This may solve backward compatibility issues
2225 due to the change to return type of SD_simulate. See also:
2226 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2227 * Add new macros with variable number of arguments.
2228 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2229 - in xbt/asserts.h: xbt_assert
2230 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2231 - in xbt/ex.h: THROWF and RETHROWF.
2232 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2234 * Change xbt_die() to accept a format string with arguments, just like printf.
2235 * New data structure: xbt_lib_t, like a dict but more general and with better
2239 * New configuration options
2240 Options triva/categorized and triva/uncategorized can be used to generate
2241 graph configuration files for Triva visualization tool.
2242 * Configuration option tracing/platform is renamed to tracing/categorized
2243 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2244 * New TRACE_user_link_variable interface:
2245 User provides the name of the link and the tracing variable to attach to it
2246 * the declaration of tracing categories must be done after the environment
2248 * simpler tracing interface, just one way to declare categories
2249 TRACE_category or TRACE_category_with_color, it is up to you
2250 * links in the trace file are again identified by their names
2251 * trace contains the full platform hierarchy exactly as declared using the ASes
2252 * Options tracing/msg/[task|process]:1 groups the process by hosts
2253 for both cases, tasks and processes must have names that are unique during the simulation
2254 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2255 * The experimental option tracing/msg/volume is deprecated
2256 its functionality may be reincorporated if needed
2258 The tracing generates a trace file with unordered timestamped events,
2259 because of the way the core simulator (surf) works. A script available
2260 at the tools directory (fix-paje-trace.sh) can be used to put the events
2261 in order. We have changed the tracing so it can generate ordered timestamped
2262 events in the final trace, but depending on the simulator (and how much time
2263 is simulated) that can lead to a huge memory utilization. It is deactivated
2264 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2266 Build Infrastructure
2267 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2268 - We are trying hard to keep the API stable, but it may happen that
2269 some things change (we're a research project after all, not a
2270 nuclear plant operating system). If such things should happen, you
2271 could rely on that macro to adapt.
2272 - current value: 30600 for 3.06.00, aka 3.6
2273 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2274 a number that can be compared with SIMGRID_VERSION.
2275 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2276 assertions and verbose/debug logging events are disabled at compile time.
2278 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2280 SimGrid (3.5) stable; urgency=medium
2283 * New feature to any SimGrid-based simulator: Model-Checking
2284 Check SIN#1 for more details.
2287 * New Model SMPI (three-interval linear regression for correction factors)
2288 See RR-7426, available at http://hal.inria.fr/inria-00527150
2289 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2290 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2291 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2292 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2293 * See also src/smpi/README
2298 - Tracing API changes: TRACE_start and TRACE_end should not be called
2299 by user-code. They are automatically called by simulators created
2300 with SimDAG, MSG and SMPI if the toolkit is compiled with
2301 tracing_enabled=ON. Categories declaration and utilization remain the
2302 same for MSG and SimDag.
2303 - A function was added to the tracing API to declare categories with
2305 - TRACE_category_with_color (char *category, char *color)
2306 where color must be in the following format
2307 "%f %f %f", red, green, blue
2308 and red, green, blue are float values in the interval [0, 1]
2309 - User can specify NULL as color parameter, or continue calling
2310 TRACE_category (cat)
2311 On that case, the tracing system will define random colors
2312 - The following command-line options are supported:
2313 --cfg=tracing/filename:msg.trace
2314 --cfg=tracing:1 (activate tracing, needed to use others)
2315 --cfg=tracing/platform:1 (categorized resource use)
2316 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2317 --cfg=tracing/msg/task:1 (task creation)
2318 --cfg=tracing/msg/process:1 (process creation, migration)
2319 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2320 --cfg=tracing/smpi:1 (SMPI interface tracing)
2321 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2322 - examples of examples/msg/tracing updated
2324 - DAXLoader and DOTLoader functions can generate tasks with categories
2325 - A new function to attribute a category to SD tasks:
2326 TRACE_sd_set_task_category (SD_task_t task, char *category)
2327 * Tracing the MPI interface implemented by SMPI
2328 - Collective operations are traced with states
2329 - Point-to-Point operations are traced with states/links
2330 - Tracing activated by a parameter "-trace filename" passed
2331 to smpirun during execution (considering that simgrid
2332 is compiled with tracing enabled)
2333 - To run the simulation with gdb, the simulator
2334 accepts --cfg=tracing/smpi:1 to trace SMPI
2335 - tesh files to check if smpi tracing is ok
2336 - See examples/smpi/NAS/DT-trace
2337 * GTNetS tracing re-worked
2338 - adaptation to the tracing system of GTNets to cope
2339 with modifications regarding the fullduplex mode
2340 - new tesh files to check if gtnets tracing is ok
2343 * Asynchronous communications through the functions:
2344 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2345 * New function: MSG_load_platform_script()
2346 to make possible using a lua script instead of XML files to set up platforms
2347 * New function: MSG_set_function
2348 to associate functions to processes, used when bypassing the parser
2349 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2351 Platforms: Add some more examples in examples/platforms
2352 * Grid'5000: see www.grid5000.fr
2353 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2356 * Change the XML format. This is a very important modification. SimGrid
2357 3.5 introduces a new hierarchical format based on the notion of
2358 Autonomous Systems. Compatibility with old format is ensured
2359 through the perl script provided in the install bin directory
2360 bin/simgrid_update_xml.
2361 It is now possible to build platforms with specific routing
2362 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2363 connect several platforms together. We will try to provide soon
2364 set of realistic platforms exploiting these properties (have a
2365 look at examples/platforms/ for the moment).
2366 * Take the opportunity of the XML format change to be a good XML citizen:
2367 rename link:ctn to link_ctn and similar changes (also dealt with by
2369 * Add a new routing scheme (rule-based) using regular expressions. It
2370 enables to have an extremely low memory footprint when the
2371 underlying routing is simple and can be compactly described. You
2372 need to have libpcre4-dev (perl regular expressions) installed if
2373 you want to use this routing scheme.
2374 * Revive the cluster TAG and allow to easily and efficiently (both in
2375 term of memory and speed) connect clusters together. Have a look
2376 at teshsuite/simdag/platforms/ to see how this can be done. With
2377 this tag, you can create clusters with thousands of tasks at no
2378 cost (have a look at examples/platforms/).
2379 Note: clusters are implemented as ASes, so there is no need for an
2380 enclosing AS tag if you have only one cluster in your platform.
2381 * Add new generic functions in the public interface that allows the user
2382 to call SURF 'create_resource' methods from your code (same
2383 functionality as the XML bypass mechanism but with a much lighter
2385 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2386 that uses a piecewise linear approximation to produce better
2387 results when exchanging small messages.
2388 * Add a new parameter to handle correctly full duplex link and account
2389 for interferences between uplink and downlink communications
2390 (activate with --cfg=fullduplex:1).
2393 * Rename the SD_READY (all dependencies are satisfied and task is
2394 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2395 dependencies are satisfied) state.
2396 This prevents a confusion between the notion of "ready to schedule"
2397 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2398 simulated" (SD_RUNNABLE) used by the simulation kernel.
2399 * Change the way a task is considered as ready. Instead of removing
2400 dependencies when a task is done, a counter is decreased. This way,
2401 it is always possible to reach ancestors thanks to the
2402 SD_taks_get_parents function (even after the end of the simulation.)
2403 * Change the return type of SD_Simulate from (SD_task_t*) into
2404 xbt_dynar_t. This function was in handling a dynar internally and
2405 converted it into a NULL terminated array for historical reasons.
2406 * New function SD_dotload(char*) to load a DAG described in dot
2407 format. This loader and the corresponding examples require the
2408 installation of the graphviz library.
2409 * Fix a bug in the management of tasks of size 0 in the surf network
2410 models. This problem was only visible with SIMDAG and you should
2411 thus disregard results produced with earlier versions if you
2412 relied on this feature (some tasks were blocked because of this).
2413 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2414 even though your applications were doing only point-to-point
2415 communications and sequential computations. Now you can really use any
2416 model you want (of course, if you create real parallel tasks, which are
2417 not implemented in most models beside ptaskL07, this will abort).
2418 * Add an example that schedules a DAX on an heterogeneous platform
2419 using a Min-Min strategy.
2420 * New function SD_workstation_get_current_task() that returns the kind
2421 of task currently running on a workstation in the sequential access
2423 * Raise some warnings when unexecuted tasks remains at the end of the
2424 simulation. This is usually caused by cycles in the DAG.
2427 * New function: SIMIX_process_set_function() called by MSG_set_function
2428 * Change the underlying waiting queue in semaphores so that a process
2429 can wait on several of them simultaneously (as in waitany).
2430 * Fix the way to handle tokens in semaphores so that all access patterns
2431 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2432 * kill the dirty pimple SIMIX_message_sizes_output()
2433 Please use (proper) visualization instead
2436 * New data container: setset (set of sets of elements)
2437 * New module: mmalloc (mapped malloc, allowing to have several
2438 independent segments of malloc)
2439 * New function: xbt_dict_cursor_set_data()
2440 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2441 * New function: xbt_dynar_is_empty()
2442 * New function: xbt_fifo_get_last_item()
2443 * Fix xbt_dynar_shrink(): use the right element size.
2444 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2445 uninitialized areas during expand.
2446 * Fix semaphores: previous implementation was severely broken.
2447 * Use library init/fini functions for our initialization.
2448 - you can use logs and other feature as soon as you want in your
2449 code (even before the xbt_init / MSG_init)
2450 - xbt_exit is now a no-op and produce a warning when used.
2453 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2454 benefit from the latest improvement to the simulation kernel.
2455 * Kill measurement sockets for now. If you rely on them, sorry. This
2456 release is not for you. This feature will be reintroduced in the
2457 future, but we cannot delay the release any further.
2458 * New function: gras_msgtype_get_name().
2459 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2460 * Fix (at last) the pmm example: it should not randomly fail anymore.
2462 Build chain: bug fixes and overall polishing
2463 * Cmake is now stable enough. Hence, we killed the autotools.
2464 * Port to windows ( TM :)
2465 * Fix the 'make install' target.
2466 No need to use 'make install-simgrid' anymore
2467 * Introduce a 'make dist' target compiling a *source* archive
2468 'make package' compiles a binary archive
2469 * Compile java files only on need
2470 * Add --cd and --setenv command line options to tesh
2471 * Out of source builds are not fully supported yet, but we are close
2472 * Enable supernovae and optimization flags by default for our users
2475 * Add layer to set up environment directly from lua, without XML.
2476 * The effect of gras_stub_generator can be achieved through
2477 lua too (check examples/gras/console/ping_generator.lua)
2479 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2481 SimGrid (3.4.1) stable; urgency=low
2483 The "Polishing easter eggs is probably a good idea" release.
2484 This is a bug fixes release only.
2488 * Fix a bug preventing the tasks from begin garbage collected.
2491 * Fix a bug occurring when a host involved in a communication fails.
2492 This was not detected properly by the other peer involved in the
2493 communication. Now, it's reported as a network error.
2496 * Warn the user about loop dependencies in data flow of DAX files
2497 * Obey the control-flow dependencies of DAX files
2500 * Add option "enable_smpi" allowing to not compile SMPI.
2501 Probably useful for the (Mac) users experiencing a build error here
2502 * Improve the detection of lua5.1 and ruby1.8
2504 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2506 SimGrid (3.4) stable; urgency=low
2508 The "Easter in Cargese" release. Also known as (major changes):
2510 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2511 ~> bindings were greatly improved
2512 ~> new tracing infrastructure for better visualization introduced
2514 * the "Welcome to configury modernity" release.
2515 ~> we switched from autotools to cmake, and improved our cdash
2518 A more detailled list of changes follow (full detail in svn log).
2520 Java Bindings: Various Cleanups
2521 * (install java-gcj-compat-dev on debian-like to use them)
2522 * Remove put/get: no need to export deprecated interface in Java
2523 Use send/receive instead.
2524 * Cleanup the examples and add a README per directory
2525 * Remove example autoDestination (that's the only way to go now)
2526 * Remove example explicitDestination (was a plain copy of basic)
2527 * Make JniException a runtime exception, so that there is no need to
2528 declare the fact that you may encounter such a beast. I guess that
2529 nobody will ever want to survive such error.
2530 * Create specific errors for each MSG case of failure:
2531 host failure, transfer failure, timeout, task cancelled
2532 * Cleanup the exceptions that may get thrown by each function
2533 * Other internal cleanups in Java bindings. Performance still bad :/
2534 Ruby and Lua Bindings: create them
2535 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2536 * That's new and great, you should try them out.
2537 Same functionalities than Java bindings, only even less polished
2539 * Kill the useless "rate" argument of SD_task_get_execution_time()
2540 Everyone used to provide -1 as a value, it was not used, and the
2541 semantic of a possible use wasn't even clear.
2542 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2543 either as comm costs or compute costs to mean that there is no
2544 such thing for that specific task.
2545 * Add a SD_task_set_name() function
2546 * Fix SD_task_unschedule() on typed tasks
2547 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2548 * In DAX loader, accept useless 'level' attributes to <job> since
2549 LIGO DAGs have them (seem to be to ease graphical representation).
2551 * Add an example masterslave_mailbox.c using send/receive and not
2552 the deprecated put/get interface.
2553 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2554 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2556 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2557 (declare MSG_USE_DEPRECATED to still have the old name)
2558 * Add a MSG_task_set_data() function
2559 * About trace replay (see examples/msg/actions):
2561 - Allow to work with splitted trace files for each process
2562 Give the specific trace file as argument of each process,
2563 and call MSG_action_trace_run(NULL)
2564 You can still have one merged file for all processes.
2565 - Fix implementation of collective operations
2566 * Allow task_execute() on 0-sized tasks (closes #10063)
2568 * This is the first release of SimGrid where SMPI is not considered
2569 beta anymore (even if some corners should still be improved)
2570 * Port over the new SIMIX_network submodule (internal refactoring)
2571 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2572 * Implement more missing elements of the standard:
2574 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2576 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2577 MPI_Scan() MPI_Get_processor_name()
2578 - Added implementation of missing case for Alltoall (warning: it's
2579 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2580 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2581 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2582 SMPI_MPI_Allgatherv()
2583 * Bug fixes include:
2584 - MPI_Waitsome() was broken
2585 - Allow relative includes in smpicc
2586 - Command line cfg argument 'reference_speed' was ignored...
2587 - Some functions did not properly lead to auto-benching of user code
2588 - smpicc passes -O2 by default (just like openmpi one)
2590 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2591 * Bug fixes about timeouts during communications
2592 * add SIMIX_message_sizes_output() as a pimple to write to file the
2593 amount of messages per size. Use gnuplot to get histogram.
2594 Pimple because that's the only user-visible function of simix,
2595 defined directly in xbt.h (irk, sorry)
2597 - Add a SIMIX_sem_get_capacity() function
2598 - Fix interactions with processe resume/suspende
2599 - release_forever() was stupidly broken
2600 - Fix SIMIX_display_process_status() for processes in a semaphore
2601 - Make SIMIX_sem_block_onto() user-visible
2602 * Refactoring context stuff:
2603 - Use pseudo-OOP for better modularity
2604 - reimplement SIMIX_process_kill() without process_schedule() so
2605 that the latter can take as invariant that it is called from
2607 - Merge context_start into context_new for sake of simplicity
2609 * Add a Vivaldi network model, coded live during SUD'10 ;)
2610 * Rename configuration variables to start a hierarchy:
2611 o cpu_model -> cpu/model
2612 o network_model -> network/model
2613 o workstation_model -> workstation/model
2614 * New configuration variables:
2615 o network/bandwidth_factor: correction to bandwidth
2616 o network/latency_factor: correction to latency
2617 o netwotk/weight_S: correction to the weight of competing streams
2618 * Add a long description to the models, that users can see with such
2619 argument on the command line: --cfg=cpu/model:help
2620 * --help-models display the long description of all known models
2622 * config: add the ability to set a default value after registration
2623 Does not override any previously set value (e.g. from cmd line)
2624 * dict: allow to have integer key and data.
2625 When so, you need to use the following functions
2626 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2627 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2628 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2629 In contrary to regular dicts, the key is not malloced before copy.
2630 Mixing scalar and regular elements in the same dict is not tested
2632 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2633 Tracing for Visualization:
2634 * SimGrid is now instrumented in order to generate a trace file for
2635 visualization analysis: to use it, need to compile SimGrid with the
2636 "tracing" option enabled, and instrument the program using SimGrid with
2637 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2638 (among other functions).
2639 * The instrumentation only traces the platform utilization for now
2640 * Documentation to use the tracing functions and how to analyze the
2641 traces with the Triva tool is written.
2642 * More information about: SimGrid FAQ (in the section Tracing Simulations
2645 * We moved to cmake as default build system. Autotools support will
2646 be dropped soon. Check the FAQ for more info about how to use it.
2647 * Greatly improved our cdash/ctest interactions
2648 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2649 * Added memory checking tests with valgrind; lot of memleak fixing.
2650 This may be the first release of simgrid with so few memory issues
2651 * Added code coverage tests.
2652 Our coverage is still improvable, but at least we see it on cdash.
2654 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2656 SimGrid (3.3.4) stable; urgency=low
2658 The "Desktop Grid needs love too" release (also called Xmas release).
2660 Models improvements:
2661 * Major speedup in the maxmin system solving by using lazy evaluation
2662 Instead of solving completely the maxmin system at each iteration,
2663 only invalidate (and recompute) the modified parts.
2664 This new feature is enabled in default models but you can try to
2665 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2666 * Cas01 IMproved as default CPU model
2667 This CPU model is the same Cas01 model, but it uses the
2668 maxmin-selective-update flag and a heap structure to manage
2669 actions on SURF kernel.
2670 It reduces the complexity to find the next action to finish and,
2671 consequently, it's faster than the old Cas01.
2672 This is the new default CPU model (Cas01).
2673 * Rename the old Cas01 model to Cas01_fullupdate
2674 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2675 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2676 * CpuTI (CPU Trace Integration)
2677 A new CPU model whose objective is simulate faster when using
2678 availability trace files.
2679 Instead of using a full featured, over engineered maxmin system for
2680 CPU modeling, this model does the pre-integration of traces files
2681 to calculate the amount of CPU power available, and so, executes
2682 faster than the old CPU models.
2683 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2684 * Use LV08 as default network model since it gives better accuracy
2685 for small messages and shouldn't change things for big ones.
2686 Use --cfg=network_model:CM02 to get the previous behavior.
2689 ******************************************
2690 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2691 ******************************************
2692 * The new CPU model may changes simulations!
2693 The point is that events occurring at the exact same timestamp
2694 are not scheduled in the same order with the old and new
2695 version. This may be enough to completely change the execution
2696 of simulations in some cases.
2697 * The new network model will change simulations!
2698 This new model is more realistic than the previous one, so you
2699 should consider redoing your old experiments with this model.
2700 Sorry for the inconvenience.
2703 * Introduce the supernovae compilation mode
2704 When compiled that way, the whole SimGrid (or almost) is put in a
2705 single compilation unit and compiled in one shoot.
2706 This is to help gcc which has difficulties to inline stuff from one
2708 The speedup seem to be above 15%, althrough more tests are needed on
2709 amd64 to confirm that gain.
2712 * Port of MSG's mailbox on top of SIMIX network
2713 The put/get mechanism was greatly simplified on the way.
2716 * New SIMIX network module. Provides:
2717 - Mailbox: rendez-vous mechanism to find with who you want to speak
2718 - Synchronous send/recv: easier and hopefully faster since the
2719 logic is handled in the maestro process directly now
2720 - Asynchronous send/recv: you dreamt of it? It's here now
2721 Too bad that nobody cared enough to propagate the change to MSG.
2722 * Add semaphores as SIMIX synchronization mechanism.
2725 * new function SD_daxload(char*) to load a DAX file
2726 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2727 * Introduce typed tasks. Specify its kind and cost at creation.
2728 At scheduling, just give where it should be placed, and the cost
2729 for each involved resource is automatically computed.
2730 Existing constructors so far (more to come of course):
2731 - SD_task_create_comm_e2e() for end-to-end communication
2732 - SD_task_create_comp_seq() for sequential computation
2733 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2734 * new function SD_task_dump() for debugging display
2735 * new function SD_task_dotty(task,FILE*) writing to file the info
2736 about the task in dotty format
2737 * SD_task_dependency_exists() can now cope with having one of its
2738 arguments NULL. If so, it tests whether the other argument has any
2740 * Add getters on list of preceding/following tasks:
2741 SD_task_get_parents(task) and SD_task_get_children(task)
2742 * Add getters on amount of workstations and list:
2743 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2744 * Add getter on task kind: SD_task_get_kind(task)
2745 * Update the start_time and finish_time of tasks on completion/failure
2746 * Bugfix: Remove task from state swags when destroyed
2749 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2752 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2753 dynar with one separate thread per value of the dynar.
2754 * Change the prototype of xbt_thread_create(), sorry.
2755 Added a boolean parameter indicating whether we want to join this
2756 thread (used in SG only for now)
2757 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2760 * GTNetS wrappers should now be usable again (and betterly tested too)
2761 * Fix a major regression from 3.2 where the timeout provided to
2762 MSG_task_put_with_timeout() was used as absolute time before which
2763 the comm should be done.
2764 * Start to fix the <cluster> tag.
2765 - Internal links should be good now (beside of the loopback, which
2766 use the private link instead)
2767 - paths to the external world is still rather broken
2768 - the <route:multi> tag is just broken. Actually that's brain-dead.
2769 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2770 to make it less stupid
2771 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2772 * Fix a source-level compatibility glitch from 3.2: after defining
2773 MSG_USE_DEPRECATED, you can use the old name
2774 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2775 * Allow to compile from the SVN with automake 1.11
2776 * Fix some problems when using the "start_time" tag in deployment XMLs.
2777 * Fix #8569: XBT/synchro.h has redundant declarations
2778 * Fix #8563: MSG return values and exceptions
2779 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2780 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2781 Thanks to Robson Peixoto.
2782 * Fix a vicious bug in dictionaries inducing that some elements were
2783 not freed on xbt_dict_free()
2785 Portability report of this version:
2786 * Main portability targets:
2787 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2788 on (amd64/i386/ia64)
2789 - mac leopard on i386
2790 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2791 but nothing critical.
2792 * Other platforms: windows, AIX and others were not tested for this release
2794 Timing report of this version:
2795 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2796 scenario parameters). From 8h to a few seconds in desktop grid settings.
2797 * Supernovae brings about 25% speedup on i386.
2799 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2801 SimGrid (3.3.3) stable; urgency=low
2803 The "Need for Speed" release.
2805 The timings done to validate the 3.3.2 were faulty.
2806 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2808 The problem was a conversion from a manually handled vector to
2809 xbt_dynar_t on the critical path.
2810 xbt_dynar_foreach calls functions, inducing stack management crap.
2812 We inlined these functions and xbt_dynar_foreach is now breath taking.
2813 We also inlined xbt_swag_belong on the way.
2815 Here are some approximate speedup measurements (on master/slaves
2816 simulations lasting between 10s and 20s each):
2817 3.3.1 -> 3.3.2: about same performance
2818 3.3.2 -> 3.3.3: 40% speedup
2819 3.3.1 -> 3.3.3: 40% speedup
2820 3.3.1 with inline patch -> 3.3.3: 30% speedup
2822 Our reading is that the refactoring which occurred in 3.3.2 made us
2823 suffer much more from the xbt_dynar_foreach low performance, but
2824 once we solved this, this refactoring proved to be very performance
2825 effective. From the 40% speedup, somehow, 10% are due to the
2826 inlining and 30% to the refactoring.
2828 That's a pitty that gcc cannot inline functions placed in other files
2829 alone. We have to choose between:
2830 - break the encapsulation (by putting private data structures and
2831 accessors in headers files to help gcc)
2832 - live with low performance
2833 - switch to a decent compiler such as icc (not quite possible).
2835 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2837 SimGrid (3.3.2) stable; urgency=low
2839 The "Simplicity does not preceed complexity, but follows it" release.
2841 The main contributors of this release were (lexical order):
2842 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2845 * Extract the routing logic into its own object.
2846 (was dupplicated in network.c and workstation_LV07.c;
2847 Allows to implement other ways of storing that info)
2848 => kill now useless network_card concept
2849 - Use dynar to represent routes (instead of void** + int*)
2850 - kill link_set (use surf_network_model->resource_set instead)
2851 - Add a command-line option to choose the routing schema to use
2852 - Add three new models:
2853 * Floyd (shortest path computed at initialization)
2854 * Dijikstra (shortest path recomputed all the time)
2855 * Cached Dijikstra (shortest path computed on need)
2856 All these models where contributed by Silas De Munck, and are
2857 described in his ICCS09 paper.
2859 * Simplify model declaration
2860 (less redirections, less function to write when defining a model)
2861 - Factorize stuff between models:
2864 surf_model_resource_set(model)
2865 surf_model_resource_by_name(model, name)
2866 - Unify the types of models in s_surf_model_t (using an union)
2867 - Embeed fields of common_public directly into s_surf_model_t
2868 - Rename model methods:
2869 action_free ~> action_unref
2870 action_change_state ~> action_state_set
2871 action_get_state ~> action_state_get
2872 - Change model methods into functions:
2873 (model)->common_public->action_use ~> surf_action_ref
2875 * Implement a generic resource; use it as ancestor to specific ones
2876 (allows to kill duplicated code in models)
2877 Drawback: timer command don't need no name nor properties;
2878 workstation_CLM03 don't need no properties
2879 (but I guess we can live with those few bytes wasted)
2881 * Improve the action object model
2882 - implement a constructor avoiding dupplicated code about field
2883 initialization in generic_action part.
2885 * Kill the SDP model: it has an external dependency, is deprecated
2886 in flavor of modern lmm models, and didn't compile since a while
2889 * Relocation of the context module from XBT to SIMIX.
2890 (the context were decoupled from the simix processes, duplicating a lot of code)
2891 => a lot of code was factorized
2892 - less overhead is introduced during scheduling
2893 - simpler API for the context factory
2894 - the logic for process creation,destruction and manipulation was simplified
2895 * Simplification of the s_smx_process_t data structure.
2896 => accesing the simix level data associated to a process is faster now,
2897 and the code is a lot more readable.
2900 * Implement some more MPI primitives:
2901 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2902 -implementation: Bcast: flat or 2-ary tree (default),
2903 Barrier: 4-ary tree,
2905 Allreduce: Reduce then Bcast
2906 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2907 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2908 Alltoallv: flat tree, like ompi
2910 * Add support for optimized collectives (Bcast is now binomial by default)
2911 * Port smpirun and smpicc to OS X
2914 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2915 Shout out if you used it.
2918 * Display the list of still queued messages in SG mode when existing
2922 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2923 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2924 * Add xbt_str_from_file(FILE*)
2925 * Add xbt_dict_get_key achieving a linear reverse search
2926 * Remove the context module
2928 Portability report of this version:
2929 * Main portability targets:
2930 - Linux(debian)/x86/context
2931 - Linux(debian)/x86/pthreads
2932 - Linux(debian)/amd64/context
2933 - Linux(debian)/amd64/pthreads
2934 On these, we still have the eratic breakages of gras/pmm and
2935 amok/saturate_sg reported in previous version. We still think
2936 that the tests are the cause of the fault, not the tested code.
2938 - Mac OSX Leopard/x86/context
2939 Still false negative in tesh autotesting.
2940 Smpi still fails, but this time because readlink does not accept -f
2941 Everything seems to work properly beside of that.
2944 - AIX version 5.3 (only tested contexts this time)
2945 Smpi still fails there because mktemp is not installed.
2946 Everything seems to work properly beside of that.
2948 I managed to compile it for the first time, but several breakages.
2949 Won't delay the release for this exotic platform.
2951 * Windows: it's still lagging behind. If you want to help, please
2954 Timing report of this version:
2955 This version seem to be more than 5% faster than 3.3.1 (on linux
2956 64bits with contextes). The gain is less than expected, we are
2957 investigating this for next release.
2959 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2961 SimGrid (3.3.1) stable; urgency=low
2964 * Implement a --cfg-help to show existing configuration variables
2965 * Build chain do not require doxygen in maintainer mode
2968 * fix a bug on struct sizeof computation, which prevented the
2969 exchange of arrays of structs in some conditions
2970 - added a regression test about this in datadesc_usage
2971 * Allow the exchange of 0-long dynamic vectors.
2972 - for that, use -1 as indicator of dynamic size instead of 0
2973 - This implied to change any size from unsigned long to long,
2974 reducing a bit communication abilities, but I guess that with
2975 64bits being quite common, this is more than enough.
2976 - This also induce a protocol change, thus bumping network protocol
2977 version from 0 to 1 (if we have external users, we have to get
2978 clean on that point too ;)
2979 - added two regression tests about this in datadesc_usage
2980 * Be more verbose when propagating local exceptions
2981 This helps debugging.
2982 * Display the status of simulated processes when receiving SIGINT in
2986 * Allow to control the simulation from a trace file.
2987 New functions MSG_action_register() and MSG_action_trace_run()
2988 The first one allows to associate a function execution to each
2989 kind of action while the second one parses a trace file and
2990 triggers the corresponding actions within the system.
2991 For now, only a toy example is provided in examples/msg/actions
2992 * Add an example of process migration in examples/msg/migration
2993 * Fix a bug in task exchange which broke MSG_task_get_sender()
2994 Add a teshsuite regression test for that.
2995 [Bug: if MSG_task_get_sender() is called after sender exit,
2997 * Fix a bug which prevented suspend/resume to work properly
2998 * Display the status of simulated processes when receiving SIGINT
2999 This fixes a regression of v3.3. due to the introduction of SIMIX
3000 * Bug fixing in failure management:
3001 - trace could not start by a failure at time 0
3002 - failure during communications were not working
3005 * Add SIMIX_process_set_name() to change the name of the current
3006 process in the log messages.
3007 * Store smx_hosts in a dict since we only retrieve them by name
3008 * Move the configuration infrastructure to surf
3011 * Move the configuration infrastructure to surf
3014 * Massive internal cleanups:
3015 - Store internal structures on processes instead of hosts (allows
3016 to have more than one process per host, in addition of being more
3018 - Cleanup the initialization/finalization process
3019 - Kill a whole bunch of unneeded synchronization:
3020 processes run in exclusive manner within the simulator
3021 - Move queues from global tables to process data fields
3023 - now accept -platform and -hostfile arguments
3024 - Pass the right rank value to processes according to the hostfile
3025 * Compile the examples by default, and use them as regression tests
3026 * Implement MPI_Wtime()
3027 * Change the reference speed to a command line option
3030 * TCP_gamma can now be specified as command line option using
3031 --cfg=TCP_gamma:10000000.0
3032 * Change the --surf-path cmd line option into --cfg=path:
3035 * Also include strbuff from xbt.h public header
3036 * xbt_ex_display(): do not free the exception after displaying
3037 This allows to do more with the given exception afterward.
3038 Users should call xbt_ex_free() themselves.
3042 Portability report of this version:
3043 * Main portability targets:
3044 - Linux(debian)/x86/context
3045 - Linux(debian)/x86/pthreads
3046 - Linux(debian)/amd64/context
3047 - Linux(debian)/amd64/pthreads
3048 These targets fail about 1/10 of times on gras/pmm, but we believe
3049 that this is because of the test, not because of simgrid.
3050 amok/saturate_sg fails even more rarely, and the test may not be
3053 - Mac OSX Leopard/x86/context
3054 The test suite still spits tons of errors because some obscure
3055 force prevents us from removing the temporary directories
3056 arguing that they still contain some metadata I've never heard of.
3057 Smpi fails because seq is not installed.
3058 Everything seems to work properly beside of that.
3061 - AIX version 5.3 (both contexts and pthread)
3062 Smpi still fails there because mktemp is not installed.
3063 XML inclusions seems rosty on AIX.
3065 * Windows: it's still lagging behind. If you want to help, please
3068 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3070 SimGrid (3.3) stable; urgency=high
3074 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3075 [Malek Cherier & Mt]
3077 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3080 Until now, GRAS were using MSG as an interface to SURF. It was
3081 quite difficult because both interface have several differences
3082 (MSG channels vs GRAS sockets were the most notable point).
3084 This also opens the gate to SMPI (which should occur soon) and speed
3085 up simulations by to 40% (even if it were not the main goal).
3087 **************************************
3088 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3089 **************************************
3090 The point is that events occurring at the exact same timestamp are
3091 not scheduled in the same order with the old and new version. This
3092 may be enough to completely change the execution of simulations in
3093 some cases. Sorry for the inconvenience.
3095 * Cleanup and upgrade the XML format to push further scalability
3096 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3098 * Improve the testing infrastructure with tesh. Now a very large part of
3099 the code is tested not only by being run but also by checking that the
3100 output match an expected output [Mt].
3102 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3103 is really less memory-demanding, which should allow you to use
3104 larger files in SimGrid [AL].
3106 * Inform valgrind about our contextes, so that it becomes usable
3107 with the default (and more effecient) version of SimGrid
3108 [contributed by Sékou Diakite, many thanks]
3111 * Introduce a listener thread in charge of receiving incoming
3112 messages from the network. It allows to overlap communication and
3113 computation but most notably, it removes some stupid deadlocks due
3114 to the fact that so far, a process could not send and receive at
3115 the same time. This made most non trivial communication schema
3117 * Convert the PIDs from long int to int to match the MSG ones (and
3118 linux ones too) [Mt]
3119 * New function: gras_agent_spawn() to launch a new process on
3120 current host. Only working in simulation for now. [Mt]
3121 * New function: gras_os_hostport() returning a constant form (ie,
3122 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3125 * Make the backtrace of exceptions more human readable [Mt]
3126 * New module: xbt/str [Mt]
3127 a ton of string utility functions (split, join, printf to a newly
3128 allocated buffer, trim, etc)
3129 * New module: xbt/hash [Mt]
3130 SHA1 hashing algorithm (more to come if needed)
3131 * New module: xbt/synchro [Mt]
3132 synchronization tools (mutex and conditions) working the same way
3133 in simulation and in real life (mainly useful for GRAS, but not
3135 * New module: xbt/queue [Mt]
3136 classical producer/consumer synchronization scheme
3137 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3138 (using the classical functions will get serialized) [Mt]
3139 * Make dictionary internal table dynamic. No need to specify its size
3140 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3142 * Make sure the log channels are organized as a tree under windows
3143 (because of ANSI C compatibility issue, any channel were child of
3144 root directly) [Mt].
3147 * Cleaned many thing in surf and fixed a few bugs [AL].
3148 * Add a nice command line configuration mechanism to compose models [AL].
3149 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3150 the previous one (KCCFLN05). It relies on something that looks like
3151 a max-min sharing mechanism but cannot be written as such. A new solver
3152 was thus designed [AL].
3153 * Add a new solver to lmm. Based on Lagrange optimization and
3154 gradient-based descent, it enables to efficiently maximise systems s.a
3156 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3158 This solver enables to propose two new network models for TCP Reno and
3159 TCP Vegas based on Low's work. These models still need to be fully
3160 tested though [Pedro Velho].
3163 * Bug fix in SD_simulate. Now the time bound given as argument is
3165 * Use the new parallel task model (ptask_L07) as default.
3166 * Use the SURF command line configuration mechanism.
3167 * 0-size tasks (for synchronization) should now work.
3169 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3171 SimGrid (3.2) stable; urgency=high
3175 We still experience issues on this platform, but we believe that at
3176 least MSG is usable.
3178 GRAS API BREAKAGE (for simplification purpose, sorry):
3179 * the gras_msgtype_by_name is not used anymore. Instead of
3180 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3181 you can write (and must)
3182 gras_msg_send(toserver, "request", &request);
3183 - If you still want to pass a gras_msgtype_t to the function (to cache
3184 the type and avoid the lookup time), use the gras_msg_send_() variant.
3185 - Impacted functions:
3186 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3187 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3188 * The callbacks are now expected to return 0 when everything went well
3189 (just like the main() function)
3191 GRAS new features and improvements:
3192 * New module mechanism where user code can use per process globals [Mt]
3193 This is similar to gras_userdata_*() functions, but for libraries. It
3194 factorize some code developped over and over in the examples and AMOK.
3195 It has still to be documented and used (only amok/peermanagement is
3197 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3198 loss when they were small enough to fit into the buffer and sent quickly
3199 enough so that they can all get received in one shoot.
3200 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3201 if not found. Use the *_or_null() variant for the old semantic.
3202 * In gras_msg_handle, do not discard messages without callback.
3203 They are probably messages to be explicitly awaited later (ie, proofs of
3204 mis-synchronization in userland since they are sent before being awaited)
3205 No big deal usually.
3206 * gras_socket_meas_send/recv: semantic changed!
3207 The numerical arguments used to be (1) the total amount of data to send
3208 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3209 messages. This was need for the fool willing to send more than MAXINT
3210 bytes on quite fat pipes.
3213 * Do really rename the hostmanagement module to peermanagement. [Mt]
3214 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3215 but this is rather new and this was documented in the module
3216 documentation (poor excuses, I admit)
3217 * Bandwidth measurement semantic changed! This follows the changes to
3218 gras_socket_meas_send/recv explained above.
3221 * A sequential mode has been added to the workstations. When a workstation
3222 is in sequential mode, it can execute only one task, and the other tasks
3223 are waiting in a FIFO. [Christophe Thiery]
3226 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3227 model for SIMDAG. [Christophe Thiery]
3228 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3229 equal to 0, which caused some bad behaviors in
3230 saturated_constraint_set_update. I now use a threshold mechanism like in
3234 * When running manually src/testall, you select specific units [Mt]
3235 testall is the result of our cunit mechanism, and should replace all
3236 the scripty thingy around since bash don't run easily on billware.
3238 * A mallocator system has been added. [Christophe Thiery]
3239 Mallocators allow you to recycle your unused objects instead of freeing them
3240 and allocating new ones.
3242 Documentation update:
3243 * FAQ reworking + New FAQs:
3244 - "Valgrind spits tons of errors!" [Mt]
3245 - "How to repport bugs" [Mt]
3246 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3247 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3248 - Communication time measurement within MSG [AL]
3249 - I experience weird communication times when I change the latency [AL]
3250 * GRAS tutorial [Mt]
3252 - an introduction to the framework and to the used communication model
3253 - an initiatic tour introducing the most proheminent features:
3255 . Lesson 0: Installing GRAS
3256 . Lesson 1: Setting up your own project
3257 o Part 2: Message passing
3258 . Lesson 2: Exchanging simple messages
3259 . Lesson 3: Passing arguments to the processes (in SG)
3260 . Lesson 4: Attaching callbacks to messages
3261 . Lesson 5: Using globals in processes
3262 . Lesson 6: Logging information properly
3263 . Lesson 7: Using internal timers
3264 . Lesson 8: Handling errors through exceptions
3265 . Lesson 9: Exchanging simple data
3266 . Lesson 10: Remote Procedure Calling (RPC)
3267 . Lesson 11: Explicitly waiting for messages
3268 . Recapping of message passing features in GRAS
3269 - A HOWTO section containing:
3270 o HOWTO design a GRAS application
3271 More are due, of course. They will come latter. In the meanwhile, you can
3272 check the examples which are still here.
3274 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3276 SimGrid (3.1) stable; urgency=high
3280 There was a stack corruption somewhere, visible only when optimizing
3281 with these versions. [Vince]
3284 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3285 module that enabled to play with Directed Acyclic Graphs. It is built
3286 directly on top of SURF and provides an API rather close to the old
3287 SG. Some old codes using SG are currently under rewrite to check that
3288 all needful functions are provided. [Christophe Thiery]
3291 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3292 extension of the classical CLM03 model that gracefully handles
3293 failures. This is now the default model for MSG and GRAS. It doesn't
3294 handle parallel tasks yet though. [AL]
3295 * Bug fix: Weights were not correctly set in the network part.
3296 WARNING: This may have resulted in incorrect results with simulations
3297 where there are more than one flow on a given link. [AL]
3300 * After a (long ?) discussion on simgrid-devel, we have decided that the
3301 convention we had on units was stupid. That is why it has been decided
3302 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3303 WARNING: This means that all previous platform files will not work as
3304 such with this version! A warning is issued to ask users to update
3306 A conversion script can be found in the contrib module of the CVS, under
3307 the name contrib/platform_generation/surfxml_update.pl [MQ]
3310 * Bug fix: Processes were started in reverse order, wrt deployment file.
3311 WARNING: if your code relies on this bug, please fix it. [AL]
3312 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3313 being executed on two different locations. [AL]
3314 * Bug fix: Failures are now better supported thanks to Derrick's tests
3315 (there was many failure situations I hadn't thought of and that weren't
3316 correctly handled). [AL]
3317 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3321 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3322 Exception occurring on server-side are propagated back to client (!).
3324 API CHANGE: the callback changed their prototype. Change:
3325 int my_handler(gras_socket_t expeditor, void *payload_data) {
3327 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3328 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3330 * New! function: gras_msg_handleall to deal with all messages arriving
3331 within a given period.
3332 * New! function: gras_socket_server_range to get a server socket in a
3333 range of port numbers (ease to avoid port number conflicts) [MQ]
3334 * New! gras processes display their backtrace when they get a SIGUSR1
3335 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3336 Sweet to debug RL processes [MQ]
3340 - Do not force experiment sizes to be expressed in kb, or it becomes
3341 impossible to measure the latency this way (needs one byte-long tests)
3342 WARNING: this changes the amok_bw_* function semantic. [MQ]
3343 - Implements the link saturation stuff. [MQ]
3344 * Peer management module:
3345 New! module factorizing code that we wrote over and over [MQ].
3348 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3349 - Test units are placed directly into the library code, they get extracted
3350 automatically and placed into the src/testall binary.
3351 - Convert most of the XBT tests to this system.
3352 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3353 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3354 live in the same C code. This required to change a little bit the API
3355 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3356 * New module: sparse graph structure with basic algorithms (this is work
3357 in progress and the API is not considered to be frozen yet). [AL]
3358 * Display more information on backtraces: source line & function names are
3359 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3360 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3361 * Reworked a little bit some #include statements to load only required
3362 headers. Some user code that relied on SimGrid to include stdlib or
3363 stdio may need to include it by themselves. [AL]
3364 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3366 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3367 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3368 but I guess that almost nobody used those parts. [MQ]
3370 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3372 SimGrid (3.0.1) stable; urgency=low
3375 * Unfortunately, I had missed 5 misnamed functions:
3376 xbt_fifo_item_t xbt_fifo_newitem(void);
3377 void xbt_fifo_freeitem(xbt_fifo_item_t);
3378 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3379 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3380 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3381 They're now deprecated. Please use their new versions:
3382 xbt_fifo_item_t xbt_fifo_new_item(void);
3383 void xbt_fifo_free_item(xbt_fifo_item_t);
3384 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3385 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3386 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3388 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3389 * Documentation: xbt_log module unmercifully reworked [MQ]
3390 * Bugfix: there was a problem with the ending of contexts with
3391 the pthread backend. It caused some weird deadlock or behavior
3392 depending on the pthread implementation. [AL]
3393 * Bugfix: get the exceptions raised in the simulator repport where
3394 and why they come from when they are not caught in time [AL, MQ]
3397 * Bugfix: Do repport the error when two non-connected hosts try to
3398 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3401 * Add additionnal checkings on communications. Assert that two
3402 communicating hosts are connected by a set of links... [AL]
3405 * Add additionnal checkings on channel values in communication [AL]
3406 * New: MSG_task_get_source to see on which host a task was generated [HC]
3407 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3408 the number of tasks waiting to be received on channel and sent
3410 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3411 waits for the first task coming from a given host.. [AL]
3413 GRAS new functionnalities: [MQ]
3414 * Enhance the parsing macro to allow the size of multidimentional objects
3415 to be given thru annotations.
3416 * New example (and documentation): Matrix Multiplication a la RPC
3417 (as when I was young!) and fix a bunch of bugs found on the way.
3419 GRAS performance improvements: [MQ]
3421 * Reduce the amount of cbps creation/destruction by making it static to
3422 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3425 * Change libdata to a set so that we can search for stuff by ID (and thus
3426 reduce the insane amount of dict lookups)
3429 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3430 RL communication on top of SG since it's so uneffective.
3431 It may also be used for inter-thread communication in RL, one day.
3432 * Use gras_datadesc_copy() to exchange messages on top of SG
3434 - improve message exchange performance on top of SG
3435 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3436 * Don't exchange on the network the size of the used part of buffer,
3437 instead, specify the possible buffer size to read().
3439 - reduces the amount of read/write calls (one pair per exchange)
3440 - reduces the amount of exchanged data (the size)
3441 - allows to retrieve all arrived data on receiver side, if we don't need
3442 it right now (subsequent read will peek the buffer)
3443 - allows the receiver to proceed with the beginning of the stream before
3444 everything is arrived
3445 - make it possible to build an iov transport (using readv/writev)
3447 - take care of the data with non-stable storage (like stacked data),
3449 * If possible, TCP send uses vector I/O (when writev() is here)
3450 - Don't use it for receive since we send data sizes and data on the
3451 same stream, so we wouldn't be able to chain large amount of chunks
3452 before having to flush the stuff to read the size.
3453 * Rework the transport plugin mechanism to simplify it and reduce the
3454 amount of pointer dereferencement when searching for the right function
3457 * I guess that now, we do almost as few system calls as possible while
3458 doing as few data copy as possible.
3460 To improve it further, we could try to send all the sizes first and then
3461 all the data (to use iov on receiving size), but it's only a partial
3462 solution: when you have 2 dimensional data, the sizes of the second
3463 dimension is data of the first dimension, so you need 3 streams.
3465 I'm not sure the potential performance gains justify the coding burden.
3467 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3469 SimGrid (3.00) stable; urgency=high
3472 * New! Give the possibility to hijack the surf parser and thus bypass
3473 MSG_create_environment and MSG_launch_application. Have a look at
3474 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3476 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3478 SimGrid (2.96) unstable; urgency=low
3483 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3484 This deprecates the xbt_error_t mechanisms.
3485 It modifies (simplifies) all XBT and GRAS API.
3486 MSG API keeps unchanged (exceptions raised by XBT are caught from
3487 within MSG and masked with existing error handling facilities)
3490 * New! Add a FATPIPE model. [AL]
3491 * New! Add a parallel task model. [AL]
3492 * New! Add automatically a loopback interface (in the default
3493 network model) if none was precised.
3496 * Bugfix: MSG_process_resume now works with the current running process.
3498 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3499 * Modification of MSG_task_get_compute_duration. Once a task has been
3500 processed, the value returned by this function is now equal to 0. [AL]
3501 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3502 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3503 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3504 of a m_task. For now, MSG_task_cancel only works with computation
3506 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3507 of the processor (in Mflop/s) regardless of the current load on the
3509 * API Change: use proper naming convention for MSG_getClock and
3510 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3512 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3513 This function changes the priority of a computation task. This priority
3514 doesn't affect the transfer rate. A priority of 2 will make a task
3515 receive two times more cpu power than the other ones. This function
3516 has been added to suit the needs of Nguyen The Loc and hasn't been that
3517 much tested yet. So if it fails, please report it and send me your code.
3519 * API Change: removed all functions and types that were marked "deprecated"
3520 since many months. Renamed MSG_global_init_args to MSG_global_init.
3522 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3524 SimGrid (2.95) unstable; urgency=low
3527 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3528 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3529 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3532 * Bugfix: Allow absolute paths for platform description files [MQ]
3533 * Bugfix: do free the variables after use. Leads to drastic performance
3535 * Implement max_duration (ie, timeouts) on resources [AL]
3538 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3540 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3541 the needed support (timeouts) [AL]
3544 * Implement measurement sockets. You can now get the bandwidth between two
3545 hosts thanks to AMOK (see below). [MQ]
3546 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3547 dynar over the network (yeah) [MQ]
3548 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3549 * simple/static token-ring example. [Alexandre Colucci and MQ]
3550 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3551 in place before [MQ]
3554 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3555 type parsing. "short" and "long" modifiers were also ignored; other
3556 modifier (such as reference level) are still ignored. [MQ]
3557 * Embeed the buffer size within the buffer itself on SG. [MQ]
3558 That way, send() are atomic and cannot get intermixed anymore (at least
3559 the ones which are less than 100k; bigger messages still have the issue)
3560 * Array size pushed by the field, not by the field type (or each
3561 and every long int will push stuff to the cbps) [MQ]
3562 * use select() to sleep since it allows to portably sleep less than one
3565 GRAS (minor cleanups)
3566 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3569 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3570 * type category 'ignored' killed as it was never used and were difficult
3572 * whether a type can cycle or not is now a flag, leaving room for more
3573 flags (as "ignored", if we feel the need one day ;)
3574 * Rename raw sockets to measurement sockets since "raw" has another
3575 meaning in networking community.
3578 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3579 and offers features not belonging to GRAS but that most applications
3580 need. One day, it may be a set of plugins loadable at runtime.
3581 * New module: bandwidth
3582 bandwidth measurement between arbitrary nodes running this module. [MQ]
3584 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3586 SimGrid (2.94) unstable; urgency=low
3588 The first beta release of SimGrid 3 !
3592 * Update the main page and the FAQ. Adding references to gforge.
3595 * Add a gras_os_getpid function.
3598 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3599 * Extend the logs so that they also print PID, hostname, date, ... if
3601 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3602 and kill the old version which were in testsuite/
3603 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3607 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3608 for users. Instead, each programming environment (GRAS, MSG,...) use it
3609 when needed to provide such a feature to users.
3610 Don't shortcut the mechanism or you will also shortcut the virtualization
3611 you need on the simulator.
3615 * Cleanups in configury with regard to compile optimization/warning flags.
3616 Also add -fno-loop-optimize to any powerpc since it's the optimization
3617 killing gcc (< 3.4.0).
3618 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3619 needed by MSG examples complications
3620 * Borrow configury beautifications from PHP
3623 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3624 warning (thanks loris for stumbling into this one).
3625 * Bugfix: stop loading private headers (gras_config.h) from the public
3629 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3630 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3631 RL would imply a DNS resolver, which is *hard* to do in a portable way
3632 (and therefore delayed).
3633 * Implement a real timer mechanism and use it in timing macros. This allows
3634 to avoid rounding errors and get a 0.000005 sec precision in timing
3635 macros. While I was at it, various cleanups:
3636 - allow to declare more than one timed section per file (fix a stupid bug)
3637 - move some private declaration to the right place
3638 - merge conditional execution and timing macros into emulation module
3639 - document the module
3640 - make sure the module cleanups its mess on gras_exit
3641 * Documentation improvements:
3642 - (new) how to compile applications using GRAS
3643 - (new) emulation support (timing macros)
3645 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3647 SimGrid (2.93) unstable; urgency=low
3649 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3652 - Use Paje properly where used. Still to be sanitized properly.
3653 - Portability fix: Add an implementation of the contexts using pthread
3657 - Add xbt_procname(): returns the name of the current process.
3658 Use it to show the current process's name in all logging.
3660 - fix detection of older flex version and the reaction, since we do
3661 depend on modern ones (we use lex_destroy)
3662 - Better separation of SG and RL in the libs: remove all simulation code
3663 from libgras. As a result, this lib is now only 200k when stripped.
3664 Some of the xbt modules may also be duplicated (two sets and such) and
3665 should be cleaned/killed before SG3.
3666 - Insist on using xlC on AIX because of weird problems involving gcc there.
3667 - Cleanup the make remote stuff. This is now done by scripts
3668 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3669 mainly for our private use, but we're working on changing them to user
3672 - Bugfix: flush the socket on close only if there is some *output*.
3673 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3675 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3678 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3680 SimGrid (2.92) unstable; urgency=low
3682 Alpha 3 on the path to SimGrid 3
3686 - New! First try of benchmarking macros.
3687 - New! First try so that gras_stub_generator generate deployment and
3688 remote compilation helpers.
3690 - Bugfix: Initialization fix in msg_test.
3694 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3696 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3698 - API Change: gras_os_sleep to take the amount of seconds as a double.
3699 Accepting an int was error prone since it was the only location where
3700 seconds were coded as such. It leaded to damn rounding errors.
3701 - Bugfix: Hard to belive that timers ever worked before this.
3703 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3705 SimGrid (2.91) unstable; urgency=low
3707 Alpha 2 on the path to SimGrid 3
3711 - Bug fix in the lmm_solver.
3713 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3714 through the function MSG_paje_output.
3715 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3716 - It is possible to bound the rate of a communication in MSG with
3717 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3718 to write it in the changelog).
3719 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3724 - Various cleanups to the autotools stuff
3725 - Begin to move Gras examples to examples/gras/
3726 - Let make distcheck work again (yeah!)
3728 - documentation overhauled using doxygen.
3729 gtk-doc-tools is dead in SimGrid now.
3730 - Automatically extract all existing logging categories, and add the list
3731 to the documentation (long standing one, to say the less)
3733 - Cleanup the known architecture table. Reorder the entries to group what
3734 should be, and use a more consistent naming scheme.
3735 (some of the test dataset are still to be regenerated)
3736 - New! Allow library to register globals on each process just as userdata
3738 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3739 lookup time (for now).
3740 Use it in msg and trp.
3741 This cleans a lot the internals and helps enforcing privacy of the
3742 headers between the gras components.
3743 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3744 - Bugfix: gras_os_time was delirious in RL.
3745 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3747 - Reenable GRAS now that it works.
3749 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3751 SimGrid (2.90) unstable; urgency=low
3753 Alpha 1 on the path to SimGrid 3
3755 * It is a long time since the last release of SimGrid. I'm sorry about
3756 that but as I had told you, I was rewriting a lot of things. I apologize
3757 to those who had been reporting bugs to me and that I had not answered.
3758 If your bug is still in the new version, please tell me. Here is a
3759 summary of the main changes.
3761 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3762 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3763 portability, speed, and a lot more but you'll figure it out later.
3764 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3765 I wanted to release the new MSG as soon as possible and I have
3766 broken GRAS, which is the reason why, for now, only MSG is fully
3767 functional. A laconic description of these projects is available
3768 in the documentation.
3770 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3771 kernel optimized for our needs (called SURF but only the developers
3772 should use it). Hence, MSG is now roughly 30 times faster and I think
3773 that by rewriting a little bit MSG, I could event speed it up a little
3774 bit more. Beside the gain in speed, it is also much easier to encode a
3775 new platform model with SURF than it was with SG. More to come...
3777 * REVOLUTION 3: I have tried to change a little as possible the API of
3778 MSG but a few things really had to disappear. The main differences
3779 with the previous version are:
3780 1) no more m_links_t and the corresponding functions. Platforms are
3781 directly read from a XML description and cannot be hard-coded
3782 anymore. The same format is used for application deployment
3783 description. The new format is described in the documentation.
3784 Have a look in tools/platform_generation. There is a tiny script
3785 that converts from the old platform format to the new one. Concerning
3786 the application deployment format, parsing the old one is tricky.
3787 I think most of you should however be able to convert your files. If
3788 it is really an issue, I can write a C code that does the conversion.
3790 2) the toolbox tbx does not exist anymore. We now have a library
3791 with much more data-structures but without the hash-tables (we have
3792 dictionaries that are much faster).
3794 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3796 *****************************************************************************
3797 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3798 * don't feel like converting the oldies. (Mt) *
3799 *****************************************************************************
3802 Version 2.90: "the long awaited one"
3803 - Finished rewriting and debugging MSG. Rewrote the documentation.
3804 - disable GRAS for now since it needs to be ported to the newest SG
3807 - Finish the port to windows (using mingw32 for cross-compile)
3810 - Main loop and datastructures of SURF. A cpu resource object is
3811 functional. Surf can thus be used to create cpu's with variable
3812 performance on which you can execute some actions.
3814 2004-11-15 Martin Quinson
3815 - Port to ARM. Simply added the alignment and size descriptions. Should
3816 work, but the ARM machines are so slow that I didn't had the opportunity
3817 to 'make check' over there yet.
3819 2004-11-15 Arnaud Legrand
3820 - Trace manager now written. It uses a heap structure and is therefore
3821 expected to be efficient. It may however be speeded up (particularly
3822 when many events occur at the same date) by using red and black
3823 trees. One day maybe...
3824 - Max-min linear system solver written. It uses a sparse matrix
3825 structure taking advantage of its expected use. Most operations are
3826 O(1) and free/calloc are called as few as possible. The computation of
3827 the minimum could however be improved by using a red and black tree
3830 2004-11-03 Arnaud Legrand
3831 - Rename every gras_* function that was in xbt/ to its xbt_
3833 - Add a heap and a doubly-linked list to xbt
3834 - Added a dichotomy to the dictionaries. make check works as well before
3835 so I assume that the patch is correct. I do not know however if things
3836 run effectively faster than before now. :)
3838 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3839 SimGrid, and the version number is bumped to 2.x
3841 2004-10-29 Martin Quinson
3842 - Introduction of the remote errors.
3843 They are the result of a RMI/RPC on the remote machine.
3844 ErrCodes being scalar values, you can't get the host on which those
3845 errors did happen. Extending the error mechanism as in Gnome is possible.
3846 No idea yet whether it is a good idea.
3848 2004-10-28 Martin Quinson
3849 - Interface revolution: the Starred Structure Eradication.
3850 I used to do typedef struct {} toto_t; and then handle *toto_t.
3851 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3852 - ???_t is a valid type (builded with typedef)
3853 - s_toto_t is a structure (access to fields with .)
3854 - s_toto is a structure needing 'struct' keyword to be used
3855 - e_toto_t is an enum
3856 - toto_t is an 'object' (struct*)
3858 typedef struct s_toto {} s_toto_t, *toto_t;
3859 typedef enum {} e_toto_t;
3860 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3861 s_toto_t) is private.
3863 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3864 so that it can set the variable to NULL. It was so for dicts and sets,
3865 it changed for dynars.
3867 - Fix a bunch of memleaks in dict_remove
3868 - Fix a bug in sg/server_socket opening: it failed all the time.
3870 2004-10-07 Martin Quinson
3871 - Speed up dynar lookup operation a bit.
3873 gras_dynar_get is dead.
3875 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3876 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3877 (which gives you the address of the stored data).
3879 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3880 copy of the data using an affectation to do the job and not a memcpy.
3882 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3883 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3885 It does not really speedup the dynar test because they are
3886 setting elements all the time (and look them seldom). But the dict does
3887 far more lookup than setting.
3889 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3891 2004-10-05 Martin Quinson
3892 - Allow to (en/dis)able the cycle detection at run time.
3894 Whether we should check for cycle or not is now a property of each
3895 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3896 datadesc_cycle_unset allow to remove this property when previously set.
3898 Note that the cycle detection is off by default since it impacts the
3899 performance. Watch the data you feed GRAS with ;)
3901 This property is hereditary. Any element embedded in a structure having it
3902 set have it set for the time of this data exchange.
3904 You should set it both on sender and receiver side. If you don't set it on
3905 sender side, it will enter an endless loop. If you forget on receiver
3906 side, the cycles won't be recreated after communication.
3908 - Header reorganization.
3909 Kill gras_private.h, each submodule must load the headers it needs.
3911 2004-10-04 Martin Quinson
3912 - Interface revolution: do not try to survive to malloc failure.
3914 Now, gras_malloc and friends call gras_abort() on failure.
3915 As a conclusion, malloc_error is not a valid error anymore, and all
3916 functions for which it was the only gras_error_t return value are
3917 changed. They now return void, or there result directly.
3918 This simplify the API a lot.
3920 2004-09-29 Martin Quinson
3921 - Re-enable raw sockets.
3922 Created by gras_socket_{client,server}_ext;
3923 Used with gras_raw_{send,recv}
3926 It should allow to kill the last bits of gras first version soon.
3928 This is not completely satisfactory yet (duplicate code with
3929 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3932 - Simplify transport plugin (internal) interface by not passing any
3933 argument to _server and _client, but embedding them in the socket
3936 2004-09-28 Martin Quinson
3937 - Finish the port to AIX.
3938 autoconf was my problem (segfault within the malloc replacement
3939 function. No idea why)
3941 2004-09-16 Martin Quinson
3942 - Fix some size_t madness on 64bit architectures.
3944 2004-09-08 Martin Quinson
3945 - Reduce the number of system headers loaded, overload some more system
3946 calls (such as malloc to cast the result of the system one, and work
3948 - Fix and reintroduce the config support
3950 2004-09-07 Martin Quinson
3951 - Source code reorganization to allow Arnaud to surf all over there.
3952 - Allow to document the logging categories.
3953 - Remove all uppercase from logging categories and useless cleanup in names.
3955 2004-08-18 Martin Quinson
3956 Version 0.6.2 (protocol not changed; API changed)
3957 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3958 gras_error_t), and NULL when not found. Functions expecting a msgtype
3959 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3960 hopefully usefull message.
3961 - Portability to prehistoric sparcs again
3963 2004-08-17 Martin Quinson
3964 Version 0.6.1 (protocol not changed; ABI not changed)
3965 - prealloc some buffers to speed things up
3967 2004-08-11 Martin Quinson
3968 Version 0.6 (protocol not changed; ABI expended)
3969 - The parsing macro can deal with the references, provided that you add
3970 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3972 2004-08-09 Martin Quinson
3973 Version 0.5 (protocol not changed; ABI changed)
3974 - Allow to off turn the cycle detection code in data exchange at
3975 compilation time. It should be at run time, but I'm short of time (and
3976 the config stuff is still broken). That way, we keep dict out of the
3977 critical path, which is good because the performance is poor:
3978 - search not dichotomial yet
3979 - dynar give no way to access their content and memcpy everytime
3980 - In composed data description (struct, ref and so on), stop foolness of
3981 keeping the subtype's ID, but store the type itself. This keeps sets out
3982 of the critical path, which is good since they rely on dynar and
3983 dictionaries. The only loose of that is that we cannot detect the
3984 redeclaration of a structure/union with another content (but I'm not sure
3985 the code detected well this error before anyway). We still can detect
3986 the redefinition discrepancy for the other types.
3987 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3988 it breaks the code because of type-punning used all over the place).
3989 This breaks on all non-gcc architectures (for now).
3991 All those changes (plus the buffer of last time) allow me to gain 2 order
3992 of magnitude on cruel tests consisting of 800000 array of integers on two
3993 level of a hierarchical structure (200 secondes -> 4 secondes)
3996 - the selector of reference must now return the type it points to, not
3997 the ID of this type.
3999 2004-08-06 Martin Quinson
4000 Version 0.4 (protocol changed; ABI not changed)
4001 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4002 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4003 - Implement a buffer transport, and use it by default (it relies on tcp in
4004 real life and on sg in simulation).
4005 That's a bit hackish since I had a new field to the structure to store
4006 its data without interfering with the subtype ones. Inheritance
4007 is tricky in C. And that's a kind of reverse inheritance with one class
4008 derivating two classes. Or maybe a game with java interfaces. Anyway,
4009 that's damn hard in C (at least).
4010 Moreover, I got tired while trying to ensure plugin separation and
4011 genericity in SG mode. MSG wants me to do weird things, so let's go for
4012 cruel hacks (temporarily of course ;).
4013 See comment in transport_private.h:71
4014 - do not include all the _interface headers in private but in the files
4015 which really need them (to cut the compilation time when they are
4018 2004-07-26 Martin Quinson
4019 Version 0.3 (protocol not changed; ABI changed)
4020 - Major overhault of the datadesc interface to simplify it:
4021 - shorted the function names:
4022 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4023 - add a trivial way to push/pop integers into the cbps without malloc.
4024 This allows to make really generic sub_type description, which simply
4025 pop their size of the stack.
4026 - add a function gras_datadesc_ref_pop_arr() which does what users want
4027 most of the time: Declare a dynamic array (which pops its size of the
4028 stack) and declare a reference to it. Poor name, but anyway.
4029 - kill the post-send callback, add a post-receive one
4031 2004-07-23 Martin Quinson
4032 Version 0.2 (protocol changed; ABI changed)
4033 - add some testing for cpbs in the test cases, and fix some more bugs.
4034 This invalidate again the little64 data file, since I cannot regenerate
4036 - remove an awfull optimization in the logging stuff, allowing me to:
4037 - understand it again
4038 - learn gcc how to check that the argument match the provided format
4039 - fix all errors revealed by gcc after that
4040 - internal keys of dict are not \0 terminated. Deal with it properly in
4041 loggings instead of segfaulting when the user want to see the logs :-/
4043 2004-07-22 Martin Quinson
4044 - Fix some stupid bug preventing cbps (callback postit) from working
4046 2004-07-21 Martin Quinson
4047 - Some documentation cleanups
4048 - remove the useless last argument of msgtype_declare
4049 - rename the Virtu functions to fit into the 'os' namespace
4050 - move headers src/include -> src/include/gras/ and stop fooling with
4051 gras -> . symbolic link
4052 - make distcheck is now successful
4054 2004-07-19 Martin Quinson
4056 - Build shared library also
4057 - Install html doc to the right location
4058 - stop removing maintainer files in make clean
4059 - build tests only on make check
4061 2004-07-13 Martin Quinson
4063 - No major issue in previous version => change versioning schema
4064 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4065 corresponding dataset.
4067 2004-07-11 Martin Quinson
4069 - Get it working with any kind of structure (we can compute the padding
4070 bytes remotely for all the architectures I have access to)
4071 - Implement the structure parsing macro (still not quite robust/complete)
4072 - Improvement to the remote testing toysuite
4074 2004-07-10 Martin Quinson
4075 [autoconf mechanism]
4076 - get ride of a bunch of deprecated macros
4077 - actually run the test for two-compliment, not only compile it :-/
4078 - test whether the structures get packed (and bail out if yes. Damn.
4079 Alignment is a serious matter)
4080 - test whether the structures get compacted (but respecting the alignment
4081 constraints of each types)
4082 - test whether the array fields of structures can straddle alignment boundaries
4084 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4085 (took me 2 hours to find that bug, looking at the wrong place)
4086 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4087 before sending/receiving and are used to compute the offsets of fields
4088 - Given that padding size depend even on compiler options, keep track of
4089 alignment and aligned_size only for the current architecture. Not a big
4090 deal since we send structure fields one after the other (seems
4092 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4093 program, let it work on linux/gcc/little32. portability todo.
4095 2004-07-08 Martin Quinson
4096 - import and improve remote compilation support from FAST
4097 - make sure make check works on half a dozen of machines out there
4099 2004-07-07 Martin Quinson
4100 Let's say it's version 0.0.3 ;)
4101 - Implement conversions (yuhu!)
4102 - Let it work on solaris (beside conversion, of course)
4103 - Stupid me, using rand() to generate the conversion datatests in not wise.
4105 2004-07-06 Martin Quinson
4106 - Let make dist work, since I'm gonna need it to compile on remote hosts
4107 - Let Tests/datadesc_usage write the architecture on which the file was
4108 generated as first byte.
4109 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4110 architecture descriptions.
4111 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4112 on those architectures.
4113 - Optimization: send/recv array of scalar in one shoot
4115 2004-07-05 Martin Quinson
4116 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4118 - Plug a whole bunch of memleaks
4119 - each process now have to call gras_{init,exit}. One day, their log
4120 settings will be separated
4121 - Continue the code factorisation between SG, RL and common in Transport.
4123 2004-07-04 Martin Quinson
4125 - Redistribution between SG and RL.
4126 We wanna have to accept in SG, so move accepted related parts of RL in
4127 the common part. (more precisely, the dynar of all known sockets is no
4128 more a static in transport.c, but part of the process_data)
4130 [gras_stub_generator]
4131 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4133 2004-07-03 Martin Quinson
4134 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4135 them all up in one shot)
4137 - Fix the ugly case of reference to dynamic array.
4138 - New (semi-public) function gras_datadesc_size to allow the messaging
4139 layer to malloc the needed space for the buffer.
4141 - gras_socket_close now expect the socket to close (and not its address to
4142 put NULL in it after it). This is because the socket passed to handlers
4143 is one of their argument (=> not writable).
4145 - propagate the interface cleanup from last week in datadesc, ie remove a
4146 superfluous level of indirection. User pass adress of variable
4147 containing data (both when sending and receiving), and not of a variable
4148 being a pointer to the data. Let's say that I like it better ;)
4149 The price for that is constructs like "int msg=*(int*)payload" in
4150 handlers, but it's a fine price, IMHO.
4152 - Let it work in RL (yuhu)
4154 2004-06-21 Martin Quinson
4156 - porting SG plugin and SG select to new standards (works almost).
4157 - plug memleaks and fix bugs around.
4160 - cleanup the prototype of data recv and force users to specify when they
4161 want to handle references to objects. Test case working even for cycles.
4162 - plug memleaks. Valgrind is perfectly ok with this.
4164 2004-06-12 Martin Quinson
4166 - cleanup the separation between plugin and main code in plugin creation
4168 2004-06-11 Martin Quinson
4170 - Reput hook for raw sockets, needed for BW experiments
4171 - kill a few lines of dead code
4172 [Data description] Interface cleanup
4173 - gras_datadesc_by_name returns the searched type or NULL.
4174 That way, no variable is needed to use a type desc once, which makes
4176 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4177 two parameters were strings (field name, type name), leading to
4179 [Dicos] Interface cleanup
4180 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4181 This is consistant with the dynar API.
4183 2004-04-21 Martin Quinson
4185 - Porting to new standards.
4187 - interface cleanup.
4188 There is no bag anymore, no need to take extra provision to mask the
4189 pointers behind "ID".
4190 Better splitup of functions between files create/exchange/convert.
4191 This is still a bit artificial since convert and receive are so
4192 interleaved, but anyway.
4194 - add a queued message list to procdata (the ones not matching criteria
4196 - factorize some more code between SG and RL wrt procdata
4198 - use gras_exit in example to track memleaks
4199 - get rid of gs_example now that GS is properly integrated into gras
4200 - update run_test to integrate the latest tests (datadesc)
4202 - rename WARNINGn macros to WARNn since it prooved error-prone
4204 2004-04-19 Martin Quinson
4206 - register init/exit functions within gras module mechanism
4207 - send/receive function.
4208 Convertion is not implemented, but short-cutted if not needed.
4209 struct/array elements are sent one by one (instead of block-wise), but
4210 nobody really cares (yet). Get a prototype before optimizing.
4211 - tests (using a file socket) for DD send/receive on:
4212 - base types: int, float
4213 - array: fixed size, string (ie ref to dynamic string)
4214 - structure: homogeneous, heterogeneous
4215 - chained list, graph with cycle
4216 Believe it or not, valgrind is not too unhappy with the results. The
4217 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4221 - Bugfix when using a filename explicitly (instead of '-')
4223 2004-04-09 Martin Quinson
4225 - factorize more code between RL and SG in socket creation
4226 - Complete the implementation and tests of:
4228 o file (only in RL, and mainly for debugging)
4230 I lost 3 days to design a portable address resolver, and then decided
4231 that the prototype mainly have to run on my box.
4232 Addressing portability too early may be like optimizing too early :-/
4234 - use gras_init in the Tests instead of the crappy parse_log_opt
4235 (the latter function is removed)
4236 [Conditional execution]
4237 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4238 [Code reorganisation]
4239 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4240 Build examples against the RL library, since there is no way to disable
4241 its creation for now.
4243 For information, the beginning of coding on GRAS was back in june
4244 2003. I guess that every line has been rewritten at least twice since