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#52: Zero-seconds timeout : "That's in the past already"
67 - FG#54: How to suspend a comm?
68 - FG!22: Stochastic Profiles
69 - FG!24: Documentation and fix for xbt/random
70 - FG!35: Add a modeling hint for parallel links in doc
71 - FG!36: [xbt/random] Read/Write the state of the RNG
72 - FG!42: Link cumulated load plugin
73 - FG!44: Test condition variable wait_until+notify_all
74 - GF#18137: Allow different stack sizes?
75 - GH#128: Parallelization of simulation with --cfg=contexts/nthreads
76 - GH#139: Allow pthread creation in SMPI
77 - GH#336: Packet-level simulation using SMPI?
78 - GH#346: [SMPI] error while loading shared libraries: libsimgrid.so
79 - GH!337: Fix link_energy plugin for wifi platforms
80 - GH!339: Add Mailbox set_receiver method to python binding
81 - GH!344: Cast hugepages macros parameters to int64
83 ----------------------------------------------------------------------------
85 SimGrid (3.25) February 2. 2020 (aka 02 02 2020)
87 The Palindrom Day Release.
89 Important user-visible changes:
90 - Improve the Python usability (stability and documentation).
91 - A nasty synchronization bug was ironed out, see also below.
92 - Python's doc was organized as a separate tree, now integrated with C++.
93 - C bindings of S4U were not part of the doc.
94 - The C++ doc was also improved as methods are now split by theme.
95 - Further deprecate MSG: you now have to pass -Denable_msg=ON to cmake.
96 - This is mandatory to use the Java bindings.
97 - OFF by default; The plan is to completely remove MSG by 2020Q4 or 2021Q1.
98 - SimDAG++: Automatic dependencies on S4U activities (experimental)
99 - Some features are already implemented but not all of them
100 - Cannot block an activity until it's scheduled on a resource
101 - No heterogeneous wait_any() that would mix Exec/Comm/Io activities.
102 - See examples/s4u/{io,exec,comm}-dependent for what's already there.
105 - Upgrade documented Java dependency to Java 8 (earlier versions don't work).
106 - Drop support for Viva/Triva (old visualization tools), hence removed
107 TRACE_get_node_types() and TRACE_get_edge_types() functions.
110 - Require pybind11 version 2.4.
111 - Greatly improve locking during context switches with Python.
112 - Declaring Python callbacks to this_actor::on_exit is now working.
113 - No more tests are failing currently, even on win32.
116 - Actor: Merge signals on_migration_start/end into on_host_change
117 - Actor: Rename migrate() into set_host()
118 - Disk: Allow users to get the read and write nominal bandwidth values
119 - Exec: Implement wait_for(timeout)
120 - Io: Implement wait_for(timeout)
123 - Parse errors now raise a simgrid::ParseError that you may want to catch.
126 - In simgrid::kernel::resource::Model, the methods next_occuring_event*() have
127 been renamed to fix a spelling error. As usual, the previous definitions are
128 kept with a deprecation warning. However, to avoid runtime errors, it is now
129 forbidden to override these deprecated methods in a derived class. Please use
130 the new names immediately if you need to override them.
133 - Remove unused parameter 'free_ctn' for xbt_dict_set() and xbt_dict_set_ext().
134 - Drop unused functions for dicts and dynars: xbt_dict_cursor_get_elm,
135 xbt_dict_get, xbt_dict_get_elm, xbt_dict_get_ext, xbt_dict_get_key,
136 xbt_dict_remove, xbt_dict_reset, xbt_dynar_compare, xbt_dynar_dump,
137 xbt_dynar_foreach_ptr, xbt_dynar_free_data, xbt_dynar_free_voidp,
138 xbt_dynar_init, xbt_dynar_insert_at_as, xbt_dynar_merge,
139 xbt_dynar_remove_n_at, xbt_dynar_replace, xbt_dynar_search, xbt_dynar_set.
140 - New module: random, providing classical random numbers generators.
143 - New option : "smpi/auto-shared-malloc-thresh" which sets a value for
144 allocations' size above which they are considered "shared" by default (as if
145 they were performed through SMPI_SHARED_MALLOC macros).
146 Default = 0 = disabled feature.
147 Note : malloc, calloc and free are now overriden by smpicc/cxx by default.
148 This can cause some troubles if codes are already overriding these. If this
149 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
150 help, but will make this feature unusable.
151 - Input check is now performed much more consistently and outputs precise
152 warnings for each error found
153 - New supported calls : MPI_File_set_view, MPI_File_get_view
154 - MPI I/O now only supports disks, as storage is discontinued.
155 - MPI I/O allows opening files without specifying absolute path
156 - Note : SMPI configuration options are now initialized by smpi_init_options()
157 call, which should be called if SMPI is not being used through smpirun.
159 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
160 - FG#9: Python bindings crashing
161 - FG#39: Missing s4u::Comm::wait_any_for example
162 - FG#42: Add support for ThreadSanitizer (TSan)
163 - FG!19: Removing RngStream
164 - FG!20: A module for RNG calls
165 - FG!21: Choice between ad-hoc and standard distributions implementations
167 - FG!25: Fix link in Mutex doc
168 - FG!26: Fix links in SMPI interface doc
169 - GH#31: [MC] please provide an option to make MPI_Send asynchronous
170 - GH#305: Unscheduled tasks are still excuted
171 - GH#313: smpirun: manual outdated w.r.t. --help
172 - GH#321: [S4U] Get task remaining work ratio
173 - GH#323: Crash when an actor turn off his physical host
174 - GH#335: Missing links on dragonfly example svg
175 - https://lists.gforge.inria.fr/pipermail/simgrid-user/2019-November/004653.html:
176 MPI_Cart_sub was not working properly. Kudos to Jonathan Borne for the report.
178 ----------------------------------------------------------------------------
180 SimGrid (3.24) October 9. 2019
182 The Clean Disk Release.
185 - Introduce a s4u::Disk interface to manage the newly introduced <disk>
186 tag. s4u::Disk is called to supplant s4u::Storage in a near future. It
187 thus becomes the default resource to execute I/O operations. Examples
188 and the file system plugin have been modified to use disks instead of
189 storage elements. Storage can still be used but is doomed to disappear.
190 Users are thus adviced to update their simulators accordingly.
191 - Barrier::wait returns SG_BARRIER_SERIAL_THREAD for (only) one actor
192 for consistency with pthread_barrier_wait()
193 - Host::get_englobing_zone() returns the englobing netzone
194 - Actor::on_destruction is now called in the destructor
195 Actor::on_termination new signal called when the actor terminates
197 - Global signals are now part of the Engine:
198 - on_platform_creation: after config settings, before the XML parsing
199 - on_platform_created: right after the XML parsing
200 - on_time_advance: each time the clock advances
201 - on_simulation_end: after simulation, before cleanups
202 - on_deadlock: as the name implies.
204 - sg_{actor,host,link}_{data,data_set}() now all exist.
205 Use them to attach user data to the object and retrieve it.
208 - Introduce an experimental Wifi model. It sounds reasonable
209 according to the state of the art, but it still has to be properly
210 validated, at least against ns-3 if not against reality.
211 - Improved the usability of ns-3. Several bugs were ironed out.
212 - host_energy: Wattage was expressed as 'idle:oneCore:allCores'.
213 It is now expressed as 'idle:epsilon:allCores' to properly model the
214 consumption of non-whole tasks on mono-core hosts. More info in the doc.
217 - convert a new set of functions to the S4U C interface and move the old MSG
218 versions to legacy (MSG_process_self*, MSG_process_{un}ref, ...)
221 - Fortran bindings for DVFS have been removed.
222 - Add support for MPI_Irsend, MPI_Rsend, MPI_Rsend_init, MPI_Bsend,
223 MPI_Ibsend, MPI_Bsend_init, MPI_Buffer_attach, MPI_Buffer_detach
224 - SMPI can now be selected by cmake's find_module(MPI) with
225 MPI_C_COMPILER, MPI_CXX_COMPILER, MPI_Fortran_COMPILER variables.
226 - Add support for MPI Errhandlers in Comm, File or Win. Default errhandler is now
227 MPI_ERRORS_ARE_FATAL, so codes which were sending warnings may start failing.
228 - trace-call-location can be used with TI traces, and replayed, and is compatible
229 with smpi/comp-adjustment-file.
230 - sleep events are now correctly traced/replayed.
231 - Default for trace-call-location is now to use file names and not full paths.
232 To revert to previous behavior (in case of collision of filenames), option
233 "smpi/trace-call-use-absolute-path" can be set to yes.
236 - Use the included xxHash as an hash implem when C++14 is usable.
237 - Option model-checker/hash was removed. This is always activated now.
238 - New option smpi/buffering controls the MPI buffering in MC mode.
239 - MPI calls now MC_assert() that no MPI_ERR_* code is returned.
240 This is useful to check for MPI compliance.
243 - New section on plugins: how to define a new one, and existing ones.
246 - xbt_mutex_t and xbt_cond_t are now marked as deprecated, a new C interface
247 on S4U is already available to replace them by sg_mutex_t and sg_cond_t.
250 - Introduce the <disk> tag as a replacement of the <storage>, <storage_type>,
251 and <mount> tags. The rationale is finish to decouple what should be handle
252 by SimGrid kernel (i.e., resources and activities) from plugins built in
253 user space (e.g., the file system plugin). This new tag currently accepts
254 only three attributes (id, read_bw, and write_bw). All the other information
255 that was declared with the storage related tags now has to be expressed as
256 properties. An example of platform using this new tag is available at
257 examples/platforms/hosts_with_disks.xml
260 - 'expect signal' can now accept more than one potential signal.
262 Fixed bugs (FG#.. -> framagit bugs; FG!.. -> framagit merge requests):
263 - FG#28: add sg_actor_self (and other wrappers on this_actor methods)
264 - FG#29 and FG#33: provide a new C API to mutexes and condition variables
265 - FG#30: convert MSG_process_{un}ref to sg_actor_{un}ref
266 - FG#31: per-actor data
267 - FG#34: SG_BARRIER_SERIAL_THREAD?
268 - FG#35: model-checker does not like buster-produced binaries
269 - FG!13: MC: complete workaround in the error msg seen on modern systems
270 - FG!15: execute_flops now logs compute
271 - FG!16: Fix the ns-3 bindings when several flows are simultaneously finishing
272 - FG!17: ns-3: unblock the right number of communications + others issues
273 - FG!18: Improving the performance of the ns-3 bindings
274 - GH#189: Energy consumption of parallel tasks
275 - GH#219: Error in the throughput of TCP transfer
276 - GH!330: Minor polishing of English: Configuring_Simgrid
277 - GH!331: Fix a doc error about actors (Tutorial_algorithms)
278 - GH!332: Add '-O0' flag for: cmake -DCMAKE_BUILD_TYPE=Debug
279 - GH!334: Rework energy plugin, again
281 ----------------------------------------------------------------------------
283 SimGrid (3.23.2) July 8. 2019
286 - Nicer introduction page.
287 - Migrate the "Deploy your application" page to the new doc.
288 - Move Java as a subtree of MSG.
291 - Rename simgrid::TimeoutError to simgrid::TimeoutException.
294 - Drop xbt_dynar_sort_strings().
297 - Really fix FG#26: Turning off a link should raise NetworkFailureException
298 - FG#27: Wrong exception thrown to wait_any when link is turned off
299 - GH#328: Java: Canceling multiple tasks in a single vm/host
301 ----------------------------------------------------------------------------
303 SimGrid (3.23) June 25. 2019
305 The Exotic Solstice Release.
308 - SunOS and Haiku OS support. Because exotic platforms are fun.
309 - Stop setting random seed with srand48() at initialization.
310 - Use addr2line as a fallback for stacktraces when backtrace is not available.
311 - Build option -Denable_documentation is now OFF by default.
312 - Network model 'NS3' was renamed into 'ns-3'.
315 - Simgrid can now hopefully be installed with pip.
318 - wait_any can now be used for asynchronous executions too.
321 - New log appenders: stdout and stderr. Use stdout for xbt_help.
322 - Drop xbt_dict_dump.
325 - 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
326 - MPI/IO is now supported over the Storage API (no files are written or read, storage is simulated). Supported calls are all synchronous ones.
327 - MPI interface is now const correct for input parameters
328 - MPI_Ireduce, MPI_Iallreduce, MPI_Iscan, MPI_Iexscan, MPI_Ireduce_scatter, MPI_Ireduce_scatter_block support
329 - Fortran bindings for async collectives.
330 - MPI_Comm_get_name, MPI_Comm_set_name, MPI_Count support.
333 - Remove option 'model-check/record': Paths are recorded in any cases now.
334 - Remove option 'model-check/sparse-checkpoint': Checkpoints are now
335 always sparse. That's an immense gain in memory, and can even be
336 faster because of cache effects. So there is no need to clutter the
337 code to allow the user to go for the unefficient mode.
340 - Remove the lagrange-based models (Reno/Reno2/Vegas). The regular
341 models proved to be more accurate than these old experiments.
343 Fixed bugs (FG=FramaGit; GH=GitHub -- Please prefer framagit for new bugs)
344 - FG#1: Broken link in error messages
345 - FG#2: missing installation documentation
346 - FG#3: missing documentation in smpirun
347 - FG#6: Python bindings not available on PyPI
348 - FG#7: simple cmake call requires doxygen
349 - FG#8: make python bindings an optional dependency
350 - FG#10: Can not use MSG_process_set_data from SMPI any more
351 - FG#11: Auto-restart actors forget their on_exit behavior
352 - FG#12: -Denable_lto=OFF doesn't disable LTO
353 - FG#13: Installs unstripped file 'bin/graphicator'
354 - FG#14: Installs the empty directory 'doc/simgrid/html'
355 - FG#15: Setting -Denable_python=OFF doesn't disable the search for pybind11
356 - FG#17: Dead link in doc (pls_ns3)
357 - FG#20: 'tesh --help' should return 0
358 - FG#21: Documentation link on http://simgrid.org/ broken
359 - FG#22: Debian installation instruction are broken
360 - FG#26: Turning off a link should raise NetworkFailureException exceptions
361 - GH#133: Java: a process can run on a VM even if its host is off
362 - GH#320: Stacktrace: Avoid the backtrace variant of Boost.Stacktrace?
363 - GH#326: Valgrind-detected error for join() when energy plugin is activated
365 ----------------------------------------------------------------------------
367 SimGrid (3.22) April 1. 2019
369 The Easter Christmas Release (this one is somewhat late).
372 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
373 - Not complete yet: asynchronous activities (amongst others) are still missing
374 - Still ongoing: the interface may change in the future. We need more
375 testers! Please report any glitches.
376 - No new project using Java should start now. Please switch to Python.
379 - Some of the internal cleanups may lead to speed improvements:
380 - The hard limitation on the amount of simulated actors with
381 Java+Mac was removed. Now, the available memory is the only limit.
382 - Our refcounting was tidyied, leading to 10% speedups in some cases.
383 - We are still working on making our code robust to the actor kills
384 and hosts' churn. Things are improving, but it's not perfect yet.
385 - Replaced our own code to display a backtrace (that was forking addr2line)
386 with the Boost.Stacktrace library.
387 You won't see your backtraces without this optional dependency.
388 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
389 - Stop setting random seed with srand() at initialization.
392 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
393 XML file version remains 4.2 since old files are still compatible.
396 - Process termination which was broken at version 3.21 has been repaired.
397 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
398 - Hide the examples into examples/deprecated. New users should use Python.
401 - Drop MSG_process_create_from_stdfunc() from the API.
402 This C++ function was a pimple in the C API, made necessary at some
403 point by the Java bindings. This is fixed now.
404 - Hide the examples into examples/deprecated. New users should use S4U.
405 - MSG_process_create and MSG_process_attach now crash if the host used
406 to run this process is off.
407 - Fix the protype of MSG_process_on_exit()
408 Now use (int,void*) callbacks instead of (void*,void*) ones.
409 The implementation was ways too messy to actually work, I guess.
412 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
413 too often and provide more accurate timings.
414 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
415 - MPI_Alltoallw support
416 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
417 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
418 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
419 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
421 - Basic implementation of generalized requests (SMPI doesn't
422 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
425 - Drop sg_cmdline. Please use xbt_cmdline instead.
426 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
427 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
428 OsSemaphore is implemented in a portable way with C++11 threads.
429 This should allow much more threads to be created at the same time,
430 allowing Mac and Java users to simulate many more actors.
431 - Implement the 'thread' factory with std::thread instead of xbt ones.
432 It is not possible to set the stack size with threads anymore, but
433 -fsplit-stack is the way to go nowadays when using threads.
434 - Drop the xbt_os_thread_t module (now unused)
435 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
436 - Drop xbt_str_join_array().
437 - Drop cunit, use Catch2 instead.
440 - Many cleanups in the kernel::activity namespace. This was long
441 overdue, and shall open the path to many future endeavors.
444 - #132: Java: a process can not shut down its own host
445 - #220: S4U: detached send causes exception if sender terminates soon after sending
446 - #261: Document the parameters of parallel execution's constructor
447 - #300: [s4u] BarrierPtr is missing
448 - #314: SMPI args internal cleanup
449 - #316: Fix a bug related to the CPU utilization of multi-core VM
450 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
451 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
452 - #325: Turning off a host has different behavior on sleeping actors and computing actors
454 ----------------------------------------------------------------------------
456 SimGrid (3.21) October 3. 2018
458 The Restarting Documentation (TRD) Release.
461 - Convert the user manual to Sphinx for improved usability.
462 Unstable version now hosted on https://simgrid.frama.io/simgrid
463 Many glitches remain; stay tuned.
464 - Restrict the Doxygen reference API to the installed header files.
465 (documenting all internals this way is overkill + doxygen scales badly)
466 - New tutorial on S4U (highly inspired from the old MSG one, but with
467 a git repository to fork as a starting point, and a docker image)
468 - Started but not finished an SMPI tutorial inspired from the JLPC'14 one.
469 - The Developper manual is still to be converted (not compiled until then)
470 - Some parts are still missing in this conversion (such as the
471 platform and deployment sections) while others were blindly converted
472 and would need more love (such as the configuration flags).
473 Things will certainly further improve in the future releases.
476 - Fully reimplement auto_restart mechanism that was utterly broken
477 (fix #22, #131 and #281 that were all very related)
478 - Implement semaphores (fix #303 and #291)
479 - When creating an actor from a function and its parameters,
480 move-only parameters are not allowed anymore, as it would prevent
481 the actor to be restartable if its parameters are consumed on run.
482 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
483 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
484 IO operations that can be started, waited for, or canceled as a regular
486 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
490 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
491 - New host variable: 'core_count'
494 - Remove xbt_os_thread_specific features
495 - Remove portability wrapper to condition variables
496 - Remove xbt_os_thread_yield()
499 - MPICH collective selector now mimics MPICH 3.3b
500 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
501 - MPI_Init(NULL,NULL) is now allowed
502 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
503 - Command-line arguments should now be handled properly in Fortran simulations
504 - MPI Attributes and related callbacks should work in Fortran
505 - Apps using C + Fortran should now work
506 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
507 - Support MPI_Error_String
510 - Due to an internal bug, Msg.run() must now be your last line.
511 We hope to fix it in a future release, and we are sorry for the inconvenience.
514 - #22: Process autorestart seem to only work with CAS01 cpus
515 - #56: Feature request: dynamic SMPI replay
516 - #93: simgrid should not eat --help
517 - #111: How to change the loopback link characteristics?
518 - #116: Platform cluster radical
519 - #117: Platform cluster/host order
520 - #129: Replay_multiple on steroid
521 - #131: Java : autostart does not work
522 - #143: Setting a breakpoint at a given time
523 - #235: Network energy models should be integrated
524 - #264: Add ptask L07 resource tracing
525 - #271: Dynamic and manual replay of SMPI traces
526 - #279: Breakpoints option uses time deltas instead of absolute time
527 - #281: Daemonized actor no longer daemonized after an auto-restart
528 - #282: TIT convertor erase traces when using absolute path in the trace list file
529 - #285: segfault when a s4u actor kills itself with msg process tracing activated
530 - #286: Pajé traces are not exposing the number of cores
531 - #287: Command-line parsing should stop when encountering '--'
532 - #288: MPI_Init(NULL, NULL)
533 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
534 - #290: Method to check if a mailbox has a message ready to be consumed
535 - #291: [s4u] semaphores do not seem available yet
536 - #294: Alltoall collective implementation of mpich
537 - #295: Starting a Host as OFF is not supported
538 - #296: DTD too permissive
539 - #299: Add a s4u barrier example
541 ----------------------------------------------------------------------------
543 SimGrid (3.20) Released June 24. 2018
545 The proxy snake_case() release.
548 - The whole API is now using snake_case() instead of mixing it with camelCase.
549 Compatibility wrappers in place for one year.
550 - We took the opportunity to increase the consistency of the API.
551 Please report any remaining glitches.
552 - New function simgrid::s4u::Host::get_actor_count:
553 Returns the number of actors running on a specific host.
556 - SMPI is now tested with ~45 proxy apps from various sources, with none or
557 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
558 - Replay: The replay file has been re-written in C++.
559 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
560 supported. They are stored in the trace and used when replayed.
561 - Basic support of MPI_Cancel. Robustness not guaranteed.
562 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
563 (only for MPI_COMM_TYPE_SHARED).
564 - New option: smpi/privatize-libs, to add external shared libs to be privatized
565 by SMPI. They will be copied locally and loaded separately by each process.
566 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
567 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
568 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
569 with internal index optind with multiple processes. Only works if MPI_Init has
571 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
572 syntax without preprocessing tricks.
575 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
576 behavior tracing is no longer limited to MSG
577 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
578 this means that traces now need to be updated or re-obtained; both Irecv
579 and Isend lines in your traces have now in total 5 mandatory fields:
580 <rankid> <command> <to/from rankid> <tag> <size>
581 To update your traces, it suffices to add a 0 for the tag here.
582 - TIT now also supports waiting for a distinct request via MPI_Wait.
583 Wait/Test now wait for a specific request, not just the last one that was
584 issued. This unfortunately means another update, because we need to
585 identify which request you want to wait for. We do this via the
586 triplet (sender, receiver, tag), which needs to be added:
587 <rankid> <command> <sender> <receiver> <tag>
588 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
589 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
592 - The deprecation of MSG is ongoing (but this should not impact you).
593 Many MSG functions are now simple wrappers on the C API of S4U. If
594 you wish to convert your code to S4U, find the S4U counterparts of
595 your MSG calls in src/msg/msg_legacy.cpp.
596 - MSG can still be used, but won't evolve anymore.
599 - Allow to run the Link energy plugin from the command line with
600 --cfg=plugin:link_energy
601 - Rename Energy plugin into host_energy
602 - Rename Load plugin into host_load
605 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
606 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
607 - Improve performance when handling timeouts of simix synchros.
610 - Config: the C API is now deprecated (will be removed in 3.23), and
611 the C++ API has been slightly improved.
614 - Fix several build issues on OSX.
615 - Move simgrid_config.h to simgrid/config.h (old header still working)
618 - #143: Setting a breakpoint at a given time
619 - #150: Inconsistent event names in SMPI replay
620 - #258: daemonized actors hang after all non-daemonized actors have completed
621 - #267: Linker error on unit_tmgr
622 - #269: SMPI: tracing of MPI_Wait/all/any broken
623 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
625 ----------------------------------------------------------------------------
627 SimGrid (3.19.1) Released March 22 2018
629 The Fixed ABI Release.
631 As you may know, we are currently refactoring SimGrid in deep.
632 Upcoming SimGrid4 will be really different from SimGrid3: modular,
633 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
635 Our goal is to smooth this transition, with backward compatibility
636 and automatic update paths, while still progressing toward SimGrid4.
638 SimGrid remains open during works: The last pure SimGrid3 release was
639 v3.12 while all subsequent versions are usable alpha versions of
640 SimGrid4: Existing interfaces remain unchanged, but the new S4U
641 interface is budding and the internals are deeply reorganized.
643 Since 2015, we work hard to reduce the changes to public APIs. When
644 we need to rename a public library symbol in S4U, we let your
645 compiler issue an explicative warning when you use the deprecated
646 function. These messages remain for four releases, i.e. for one full
647 year, before turning into an error. Starting with v3.15, your can
648 also adapt to API changes with the SIMGRID_VERSION macro, that is
649 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
651 Starting with this v3.19.1, our commitment to reduce the changes to
652 the public interfaces is extended from the API to the ABI: a program
653 using only MSG or SimDag and compiled against a given version of
654 simgrid can probably be used with a later version of SimGrid without
655 recompilation. We will do our best... but don't expect too much of
656 it, that's a really difficult goal during such profund refactoring.
658 The difference between v3.19 and v3.19.1 is that the former was
659 accidentally breaking the ABI of MSG, while the later is restoring
662 S4U and kernel APIs will still evolve until SimGrid4, with one-year
663 deprecation warnings as currently. In fact, cleaning up these
664 interfaces and converting them to snake_case() is one release goal of
665 v3.20. But don't worry, we are working to smooth this upgrade path.
667 In summary, new projects should start with S4U to benefit of the
668 future, but old MSG projects should still be usable with no change.
670 ----------------------------------------------------------------------------
672 SimGrid (3.19) Released March 20 2018
674 The Moscovitly-cold Spring release.
676 SMPI user-visible improvements:
677 - Support of MPI_Type_create_subarray
678 - Unimplemented MPI calls now abort (instead of issuing a warning)
679 Some harmless calls around errhandlers still only issue a warning.
680 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
681 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
682 is now possible (mmap privatization was refactored to this end).
683 Adding new SMPI instances should also be technically doable now.
685 SMPI internal cleanups:
686 - Removed index notion from SMPI -- no more "getPid() - 1"!
687 In tracing files, actors are called rank-<PROCESS-ID> but these are
688 NOT the real ranks (tracing several communicators is not supported yet).
689 - Deprecate smpi_process_index() (will be removed in 3.22)
690 - Better testing on classical proxy apps, through an external project:
691 https://github.com/simgrid/SMPI-proxy-apps/
694 - Use a graphical TOC to make it easier to find the documentation you need
695 - Revamp the MSG tutorial
698 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
699 - Remove parameter of MSG_process_killall().
700 Resetting the PID was bogus anyway (several actors could have the same PID).
703 - Execution->setHost() can be called after start(), to migrate it.
704 - Comm::test_any() is now implemented.
705 - s4u::Actor now has onCreation() and onDestruction() signals.
706 - Install some kernel header files for the users' plugins and more.
707 Warning, their API is really not stable yet.
708 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
709 include/simgrid/kernel/routing/*Zone.hpp
712 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
714 Build System and other cleanups:
715 - Remove unused run-time parameter "tracing/onelink-only".
716 - SimGrid now works with intel compilers (MC still buggy)
717 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
718 - tesh kills subprocesses on timeout (but not on windows)
721 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
724 - #194: Feature request: simgrid::s4u::Comm::test_any()
725 - #245: migrating an actor does not migrate its execution
726 - #253: Feature Request: expose clusters as objects
727 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
728 - #255: Tesh broken on Windows
729 - #256: Modernize FindSimGrid.cmake
730 - #257: Fix (ab)use of CMake install
732 ----------------------------------------------------------------------------
734 SimGrid (3.18) Released December 24 2017
736 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
738 MSG IS NOW DEPRECATED. Please use S4U for new projects.
739 - Support for MSG should not be removed from SimGrid before 2020,
740 but future evolutions will be limited to the integration of
741 user-provided patches.
744 - s4u::onDeadlock() signal, run before stopping the simulation.
745 - s4u::Exec: asynchronous executions (abstraction of background
747 - s4u::Host->getLoad() returns the achieved speed in flops/s
750 - Link::name() is deprecated and will be removed in v3.21, use
751 Link::getCname() instead.
752 - Mailbox::getName() changed to return a std::string, use
753 Mailbox::getCname() instead to get a char*.
754 - Storage::getName() changed to return a std::string, use
755 Storage::getCname() instead to get a char*.
756 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
757 to have both allocation and deallocation of the map in user space.
760 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
761 The amount of flops remaining to do is a vector, not a scalar.
762 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
763 name implies on both sequential and parallel tasks.
764 - Both changes fix GitHub's #223 using PR #237 as a basis.
765 Thanks Michael Mercier.
766 - Most examples were converted to S4U and hidden elsewhere at tests.
767 Rationale: we still want MSG to work; we want newcomers to use S4U.
770 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
771 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
774 - Switch to the faster dlopen privatization mechanism by default
775 - Documentation improvments
778 - Live migration is getting moved to a plugin. Dirty page tracking is
779 the first part of this plugin. This imply that VM migration is now
780 only possible if one this function is called:
781 - C/MSG: MSG_vm_live_migration_plugin_init()
782 - C/C++: sg_vm_live_migration_plugin_init()
783 - Java: Msg.liveMigrationInit()
784 For C and C++, "simgrid/plugins/live_migration.h" also has to be
788 - Define class simgrid::xbt::Path to manage file names.
789 - Removed unused functions:
790 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
791 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
792 xbt_graph_export_graphviz()
793 - xbt/str.h: xbt_str_join()
794 - Remove unused datatypes:
795 - xbt/heap.h: use std::priority_queue or boost::heap instead
796 - xbt/swag.h: use boost::intrusive::list instead
799 - New link_energy plugin for the consumption of the links.
800 - All of the operations on files and storage contents have been
801 packaged into a plugin (src/plugins/file_system). The current
802 public interface can be found in
803 include/simgrid/plugins/file_system.h
804 To use these functions you now have to initialize the plugin by
805 calling MSG_storage_file_system_init() just after calling
806 MSG_init() or sg_storage_file_system_init() just after creating
810 - Remove the undocumented/untested tag <include>
813 - Remove viva specific tracing as the tool is no longer maintained
816 - #248: Kill of finished processes leads to segfault
817 - #240: xbt_cond_wait_timeout should gracefully return for C apps
818 - #239: please implement signal s4u::onDeadlock()
819 - #230: segfaults when exit() without run()
820 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
821 - #223: MSG_task_get_flops_amount() not working with parallel tasks
822 - #222: Actor::kill() doesn't really kill and segfaults
823 - #221: odd LMM warning when killing an actor
824 - #120: Memory leak when the processes are forcefully killed
826 ----------------------------------------------------------------------------
828 SimGrid (3.17) Released October 8 2017
830 The Drained Leaks release: (almost) no known leaks despite the tests.
832 Even very long simulations will preserve your memory: our testsuite
833 stresses SimGrid for over 45mn on fast machines for a coverage of
834 over 80%, and there is only one single known leak, of about 4kb.
838 - Comm.detach(): start and forget about asynchronous emission. A cleanup
839 handler may be given to free resources if the comm cannot be completed.
840 - this_actor::send(mailbox) is now mailbox->put()
841 - New: simgrid::s4u::Comm::wait_all()
842 - New: Host.execute() for remote executions.
845 - Unused option network/sender-gap is removed.
848 - Deprecate MSG_task_isend_with_matching(): This unused feature
849 really complicates our internals. Will be removed in v3.20.
852 - Improved context termination. It is now possible to free resources, even
853 when a process is forcibly killed.
854 - ContextBoost: add support for Boost versions above 1.61.
857 - Bring back run-time option --cfg=exception/cutpath to remove exception
859 - Removed unused functions:
860 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
861 xbt_str_rtrim(), xbt_str_trim().
862 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
865 - Removed header files obsolete since SimGrid 3.12:
866 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
868 - #3: SD_exit should be made optional
869 - #120: Memory leak when the processes are forcefully killed
870 - #159: Threading test regression in Actor refcounting
871 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
872 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
873 - #186: Actor::killAll() segfaults if some process is blocked on wait()
874 - #191: VM migration and pstate
875 - #192: Updating the energy consumptions of all hosts crashes with VMs
876 - #195: All actors have PID=0 in the logs
877 - #204: Sometimes segfault with thread contexts and mmap privatization
878 - #222: Actor::kill() doesn't really kill and segfaults
879 - #225: Actor::kill() doesn't really kill when victims are doing a join()
881 ----------------------------------------------------------------------------
883 SimGrid (3.16) Released June 22. 2017.
885 The Blooming Spring Release: developments are budding.
888 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
889 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
890 - Java JAR file should be finally fully working on Mac OSX too.
892 XML platforms: Switch to platform v4.1 format.
893 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
894 - <zone> can be used as a synonym for the now deprecated <as>
895 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
896 - <bypassZoneRoute> can be used as a synonym for the now deprecated <bypassAsRoute>
897 - <actor> can be used as a synonym for the now deprecated <process>
898 - state_file and avail_file periodicity is now easier to express
899 (check the documentation) the old behavior should still work.
900 * Storage: not backward compatible, but it seems that nobody ever used it
901 - Remove attribute 'content_type' of <storage_type>: was never used
902 - Make attribute 'model' of <storage_type> optional: for future usage
903 - Remove Bconnection model property: was never (in)validated. Replaced by
904 the maximum of the read and write bandwidth as a resource constraint.
907 - New and Backwards Compatibility break:
908 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
909 When one wants to get the list of tasks whose states have changed during a
910 simulation round, s/he has to allocate and free a dynar and use it as
911 argument to this function. The former SD_simulate (double how_long)
915 - Allow multicore VMs, with the correct sharing (unless you overcommit)
916 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
919 - New function to update the consumption of all hosts at once.
920 - Fix the model for multi-core hosts, linear in the amount of busy
921 cores with an abnormality for pIdle (see paper).
922 BUG: Still not satisfactory for ptask on mono-cores.
925 - The netzone are now available from the MSG API.
926 The old names still work, but are now deprecated.
929 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
930 - New option: smpi/keep-temps to not cleanup temp files
931 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
932 mallocs mode is used, allows to change the size of the fake file used
933 (default 1MB), to potentially limit the number of mappings for large runs.
934 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
935 - Fortran ifort and flang compilers support
936 - New RMA calls supported (experimental):
937 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
938 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
939 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
940 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
941 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
942 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
945 - New: this_actor::isMaestro() returns whether we are in kernel mode.
946 - Behind the scene, ActivityImpl are now automatically refcounted.
947 Removing this nasty bottleneck will greatly simplify our internals.
950 - Replay: New function xbt_replay_action_get():
951 Retrieve the function previously associated to an event type.
952 - DROPPED FUNCTION: xbt_str_varsubst()
953 - DROPPED MODULE: strbuff. We don't need it anymore.
954 - DROPPED MODULE: matrix. We don't need it anymore.
955 - DROPPED MODULE: lib. We don't need it anymore.
957 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
959 SimGrid (3.15) Released March 22. 2017
961 The Spring Release: continuous integration servers become green
962 We fixed even the transient bugs on all target architectures:
963 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
964 Windows, FreeBSD, NetBSD.
966 - Rename NetCards to NetPoints.
967 This was intended to help NS3 users, but that's not a netcard.
968 That's a point in the routing algorithm, let's avoid wrong simplifications.
971 - New: MSG_process_yield(). Stop and yield to other processes.
972 - New: MSG_process_daemon(). Daemon processes are automatically killed
973 when the last non-daemon process terminates
974 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
975 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
976 to make room for the upcoming network energy plugin.
977 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
980 - Ensure that an actor can kill itself with Process::exit()
981 - Kill the obscure NativeException. Nobody want to survive the issues
982 it denotes, so use JniException that is a RuntimeException (not to
983 be caught explicitly).
984 - Partial bug fix in initialization. SimGrid flags on command line were
985 consumed at C level but stayed in the original Java String[] args.
986 This could mess users' args[i] if SG flags were not put at the end of
988 The SimGrid flags are now removed from the Java arguments. However,
989 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
990 if args.length is greater than the number of YOUR OWN ARGUMENTS.
991 It might be if you have --log or --cfg flags in the command line.
992 - Fix numerous memleaks all around the place. In particular, around VMs.
996 - s4u::onPlatformCreated: right before the simulation starts
997 - s4u::onSimulationEnd: right after the main simulation loop
998 - s4u::onTimeAdvance: right after a clock change (time is discrete)
999 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
1000 event from the availability_file changes the avail speed.
1001 - Links are now usable from s4u
1002 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
1003 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
1004 - New examples: The conversion of MSG examples to S4U has begun
1005 - Actors: create, kill, migration, and suspend
1006 - Applications: master-worker and token-ring
1007 - Action replay: communications and storage
1008 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
1011 - Backwards Compatibility breaks
1012 - The SD_route_get_list and SD_route_get_size functions have been removed.
1013 They are replaced by the unique
1014 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
1015 The route from src to dst is built in the links dynar whose size can
1016 be obtained with xbt_dynar_length.
1017 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
1018 sg_host_route_bandwidth and sg_host_route_latency.
1019 Macros ensure the backwards compatibility, but you should fix your code
1022 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
1023 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
1024 - Bump our claim of support from MPI 1.1 to MPI 2.2.
1025 We don't support 100% of it, but it should be enough. Contact us if not.
1026 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
1027 naive and probably inaccurate)
1028 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
1029 mapped onto a unique small file using some system magic.
1030 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
1031 - Fortran: cleanups, fixes, support of user-added operations
1032 - RMA: MPI_Accumulate are applied in correct order.
1033 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
1034 - Better support for MPI_IN_PLACE
1035 - Support for MPI_Win attrs and keyvals.
1036 - Support MPI_Comm_group_create, MPI_Type_size_x
1037 - MPI ops should be properly applied to all allowed types
1038 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
1039 algorithm by a non failing one from mpich.
1042 - Kill the fifo data container: we don't use it anymore.
1045 - A <cluster> can now be created with different speed values to represent pstates.
1046 see examples/platforms/energy_cluster.xml for an example.
1048 SimGrid (3.14.159) Released December 28. 2016
1050 The previous release was only a crude approximation of a Pi release;
1051 this one should be a bit better.
1053 - Revert a last minute change that broke on Mac OSX.
1054 - Fix the changelog and NEWS files.
1055 - Improve the documentation, in particular of the routing module.
1056 - Integrate some patches flying around in Debian.
1058 SimGrid (3.14) Released December 24. 2016
1061 * The whole documentation was reworked and reorganized. There is
1062 still much room for improvement here, but we're on it.
1065 * We now need python3 for our testing infrastructure.
1066 * Model checking needs libevent
1069 * Now works on FreeBSD
1072 * Privatization now works on FreeBSD
1074 * Privatization is now activated by default in smpirun
1075 You can turn it off with -no-privatize if something goes wrong.
1077 * Call-location tracing for SMPI.
1078 You can add the exact location (filename / linenumber) of an MPI call to
1079 your trace files and slow down or speed up the simulation between two
1080 consecutive calls by using an adjustment file (see the documentation).
1082 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
1083 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
1084 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
1085 values were ignored in some cases. The timings of these functions can now
1086 be significantly different.
1088 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
1089 smpi/running-power is renamed to smpi/host-speed
1091 * smpi/grow-injected-times option to enable or disable multiplication of the
1092 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
1093 make simulation less precise (but also much faster).
1095 * smpirun script should be (much) faster for large deployments.
1097 * SMPI tracing: fixed issue with poor matching of send/receives.
1099 * Replay: Fix broken waitall
1101 New functions and features
1102 * MSG_parallel_task_execute_with_timeout, to timeout computations.
1104 Dropped / renamed functions and features
1105 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
1106 - MSG_mailbox_is_empty() -> Mailbox::empty()
1107 - MSG_mailbox_front() -> Mailbox::front()
1108 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
1109 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
1110 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
1111 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
1112 * Don't pass the free_f parameter to property related functions:
1113 - MSG_host_set_property_value()
1114 - MSG_as_router_set_property_value()
1115 - MSG_storage_set_property_value()
1116 * VM properties. Since msg_vm_t are msg_host_t, just use
1117 MSG_host_get_property_value() and friends
1118 * VM I/O related things:
1119 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
1120 - Unimplemented save/restore methods
1121 * MSG_as_router_get_property_value() was redundant with
1122 MSG_environment_as_get_property_value().
1123 - Removed MSG_as_router_*propert*() functions
1124 - Added MSG_environment_as_set_property_value() for consistency
1125 * xbt heterogeneous dictionaries (created with xbt_dict_new()).
1126 Well, they are still there for now, but deprecated with a warning.
1127 Please switch to xbt_dict_new_homogeneous() before this is removed
1129 * Task affinity. Its intended behavior (that was very badly tested
1130 and probably not really working) was deceiving what most users
1131 would have hoped here.
1132 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
1133 * Option network/coordinates is now useless and should be dropped.
1136 * Added option storage/max_file_descriptors to allow more than 1024 files opened
1139 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
1140 * Added Dragonfly topology support
1143 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
1144 EiBps, EBps, Eibps, Ebps for bandwidth.
1145 They may become useful to some lucky ones.
1148 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
1149 * ex/app_tokenring: new example, very similar to the MSG Token Ring
1150 * ex/async_waitAll: new example, on asynchronous communications
1153 * Memory usage should be decreased for simulations with a large number
1154 of processes. This also helps for SMPI.
1156 SimGrid (3.13) stable; urgency=low
1158 The Half Release, a.k.a. the Zealous Easter Trim.
1160 - We removed half of the lines, that were mostly experimental cruft.
1161 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
1162 (not counting blanks and comments -- according to openhub.net)
1163 - The internals are now compiled in C++ (and will soon be clean C++)
1164 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
1165 59 klines of C, etc. We added only 29 klines of C++ in replacement.
1167 * Backwards Compatibility breaks
1168 - Removed Lua simulation bindings (switch to C or Java for that).
1169 Lua can still be used to describe platforms
1170 - Removed Java kernel plug-ins.
1171 Will be reintroduced after the ongoing major internals reorg.
1173 - the following functions were removed.
1174 They were too specific and should be reimplemented in a generic
1175 way, with filter function.
1176 - MSG_task_listen_from_host
1177 - MSG_mailbox_get_count_host_waiting_tasks
1178 - MSG_mailbox_put_with_timeout was removed.
1179 Please use MSG_task_send_with_timeout instead.
1181 - the SD_application_reinit function was removed. It has been a noop for a while.
1182 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
1183 replaced by a more flexible mechanism.
1184 - The following functions thus do not exist anymore
1185 - SD_workstation_get_access_mode
1186 - SD_workstation_set_access_mode
1187 - SD_workstation_get_current_task
1188 - Basic estimation functions have been removed but can easily be replaced
1189 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
1190 - SD_workstation_get_computation_time => amount / sg_host_speed()
1192 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
1193 Use VM.getSpeed()*load/100 for the legacy behavior.
1195 - option enable_tracing was removed. It was not doing anything for a while.
1196 - In the ModelChecker:
1197 - the model-checker now ptraces the model-checked process which means
1198 you cannot use a debugger on the latter anymore (we might make this
1199 optional in the feature);
1200 - removed soft-dirty page tracking;
1201 - remove model-checked side snapshot management,
1202 MC_snapshot() and MC_compare_snapshot();
1203 - keep the MC_cut() function as a stub (it was not really working
1204 in the previous release).
1207 * All options are consistently in kebab-case. Old names are kept as alias.
1210 * Switch to platform v4 format.
1211 - Rename from 'power' to 'speed' the attributes describing the amount of
1212 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
1213 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
1214 - In <host> and <link>, attributes availability and state are gone.
1215 It was redundant with state and availability traces, and with peak values.
1216 - In <cluster>, attributes availability_file and state_file are gone.
1217 It was too complex and unused.
1218 - Kill <gpu>. Was not doing anything.
1219 - The DOCTYPE points to the right URL:
1220 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
1221 (the file at this address now documents the changelog since its v1)
1222 - A warning is emitted for unit-less values (they are still accepted).
1223 - speed. Default: 'f' or 'flops'. Also defined:
1224 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
1225 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
1226 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
1227 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
1228 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
1229 - latency. Default: 's' second. Also defined:
1230 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
1232 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
1233 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
1236 * s4u::Host is now the preferred public interface to the Host features.
1237 sg_host_* functions are C bindings to the exact same behavior
1238 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
1241 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
1242 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
1243 They were deprecated since a few years, and probably did not even compile anymore.
1246 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
1247 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
1248 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
1249 encouraged to update their codes. A list of the modified functions can be found at the end of
1250 include/simgrid/simdag.h
1253 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
1254 the lambda or closure passed as a parameter will run in kernel mode.
1255 All the callback functions should be rewritten to that interface at some point.
1258 * Reorganizing and cleaning the internals all around the place.
1261 * Remove old default barrier/bcast buggy algorithms (see #18407)
1262 * Various bug fixes to handle more codes
1263 * Remove the need for the --foreground option of smpirun (it is still
1264 accepted for backward compatibility).
1267 * Kill the setset data container: MC don't use it anymore.
1268 * Kill the queue data container: it made more sense with GRAS.
1269 * Kill the xbt_peer_t data type: it's useless without GRAS.
1270 * Kill rm_cb feature of config sets: it was never useful.
1271 * Kill graphxml parsing feature. It was not used.
1272 * Kill the deprecated code protected by XBT_USE_DEPRECATED
1274 - xbt_dynar_sort_strings(), when the content is char*
1275 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
1276 They throw exceptions on invalid input;
1277 * C++ support for declaring CLI flags (simgrid::config::Flag);
1278 * class for abstracting different signal backends (simgrid::xbt::signal).
1279 with no external dependencies (we need very simple signals).
1282 * refactoring and cleanup of the code;
1283 * ongoing process to cleanly separate the model-checking algorithms
1284 from the code model-checking support.
1286 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1288 SimGrid (3.12) stable; urgency=low
1290 The Facelift Release.
1293 * Require g++ v4.7 at least to not speak prehistorical C++.
1294 * Require Boost 1.48 (for signal2 component).
1295 * Java must be version 7 at least when activated.
1296 * Builds on Windows again (including Java bindings).
1297 * Tracing is now always enabled (no way to turn it out)
1298 * Remove GTNetS. It was not working anyway.
1299 * Various cleanups in the cmake scripts.
1300 * Move headers around to sort them out on installed systems:
1301 - instr/instr.h -> simgrid/instr.h
1302 - instr/jedule/* -> simgrid/jedule
1303 - simdag/datatypes.h was removed
1304 - simdag/simdag.h -> simgrid/simdag.h
1305 - msg/datatypes.h was removed
1306 - msg/msg.h -> simgrid/msg.h
1309 * Interface improvement:
1310 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
1311 - Sanitize the interface in MSG_task_ module:
1312 - Merge two functions that were close enough but misleading:
1313 set_compute_duration(t) -> set_flops_amount(t)
1314 get_remaining_computation(t) -> get_flops_amount(t)
1315 - set_data_size(t) -> set_bytes_amount(t)
1316 get_data_size(t) -> get_bytes_amount(t)
1317 - Massive cleanups in the functions related to the energy
1318 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
1319 - New: MSG_host_get_pstate()
1321 - msg/energy/onoff: switching hosts on and off
1323 * Interface improvement:
1324 - Rename Host.isAvail() to Host.isOn()
1325 - Rename Process.currentProcess() to Process.getCurrentProcess()
1326 - Rename Task.setDataSize() to Task.setBytesAmount()
1327 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
1329 - #18874: Actually allows the GC to reclaim tasks
1333 - SIMIX_process_throw: raises an exception in a remote process
1334 * Refactoring: Separate sub-modules
1335 - libsmx: the public interface, as libc in a real system
1336 - popping: the strange dance that converts a user request into a kernel handling
1337 - smx_context_*: the virtualization mechanisms that embed the user code
1338 - smx_*: the handling of each simcalls
1339 * Interface cleanup:
1340 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
1341 * Rename smx_action_t into smx_synchro_t, making explicit that these
1342 things are used to synchronize processes with their environment.
1343 For example, a communication is a sort of synchronization involving
1344 the communicating processes (that may block until the exchange) and
1345 the platform. The same can be said from computations, etc.
1347 - #18888: segfault when a process ends before its kill_time
1351 - 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.
1352 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
1353 - MPI_Type_set_name, MPI_Type_get_name
1354 - MPI_*_c2f and MPI_*_f2c functions
1355 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
1356 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
1357 - Activate a lot of new tests from the mpich 3 testsuite
1359 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
1360 - InfiniBand network model added: Based on the works of Jerome Vienne
1361 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
1362 - When smpi/display_timing is set, also display global simulation time and application times
1363 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
1364 * Collective communications
1365 - 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
1366 - 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/).
1367 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1369 - Replay now uses algorithms from wanted collective selector
1370 - Replay can be used with SMP-aware algorithms
1371 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1372 - 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.
1374 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1375 - Correctly use loopback on fat-tree clusters
1376 - Asynchronous small messages shouldn't trigger deadlocks anymore
1377 * Energy/DVFS cleanup and improvement
1378 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1379 - new: smpi_get_host_pstate()
1382 - "Full" network optimization flag was broken since Surf++
1383 - Better handling of precision flags in maxmin
1384 - Fix bug causing sometimes "Impossible" errors
1385 - Properly pass cluster properties to included hosts
1386 * Improvement of the Energy plugin.
1387 - Always update the consumption before returning that value
1388 - New property: watt_off to denote the dissipation when the host is off
1389 - New functions getWattMinAt and getWattMaxAt to retrieve the
1390 dissipation of pstates that we are not currently at.
1391 * Java: class NetworkLink renamed to Link
1392 * New function: simcall_process_get_kill_time()
1393 * Massive rename s/workstation/host/
1394 - That's intrusive, but that's good for the project consistency. Sorry.
1395 - Change config option "workstation/model" into "host/model"
1398 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1399 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1400 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1401 - xbt_dirname and xbt_basename for non-POSIX systems
1403 * The model checker now runs as a separate process.
1404 * The model checker runs must now be launched with the new simgrid-mc program.
1405 * Record/Replay: the MC can display a textual representation of a path in the
1406 execution graph. It can then be replayed outside of the model checker.
1408 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1410 SimGrid (3.11) stable; urgency=low
1415 * Normalizing pointers addresses tool for better diff between logs
1418 * Add cloud examples using new VMs
1419 - examples/msg/cloud/two_tasks_vm.tesh
1420 - examples/msg/cloud/simple_vm.tesh
1421 - examples/java/cloud/cloud.tesh
1422 - examples/java/cloud/migration/migration.tesh
1423 * Add java surf examples:
1424 - examples/java/surfPlugin/surf_plugin.tesh
1425 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1426 - examples/java/surfCpuModel/surf_cpu_model.tesh
1427 * Add SMPI+MSG example:
1428 - examples/smpi/smpi_msg_masterslave/
1433 - msg task destroy cancel test
1434 - msg_host on/off test
1435 * Move all tests in testsuite to teshsuite (adding tesh files)
1436 * Restructure teshsuites
1437 - one folder for each kind of test
1438 * Restructure AddTests.cmake
1440 - structure the order of tests (with sections)
1443 * Add virtual machine
1444 - creation of a VM on a PM
1445 - migration of a VM from a PM to another PM
1447 - MSG_process_join(msg_process_t process, double timeout)
1448 - msg_bar_t MSG_barrier_init(unsigned int count)
1449 - int MSG_barrier_wait(msg_bar_t barrier)
1450 - void MSG_barrier_destroy(msg_bar_t barrier)
1451 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1452 * New option "msg/debug_multiple_use" to help debugging when a task is used
1455 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1457 - Increase code coverage in test suites
1461 * Protect context stacks against stack overflow. The number of protected memory
1462 pages allocated on the top of each stack (1 by default) can be configured
1463 with the new command line option --cfg=contexts/guard_size:<value>.
1464 * Simcalls are now generated by a python script that
1465 - generates files included by SimGrid
1466 - checks that all the functions exist, and proposes prototypes
1468 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1469 host_set_data, host_get_data
1471 - simcall_process_join(smx_process_t process, double timeout)
1472 * Fix bug where sleeping processing could not be suspended.
1475 * Translate surf models from C to C++
1476 - Generic classes for all models: Model, Resource, Action
1477 - A generic interface for each kind of model (CPU, Network, Storage
1478 Workstation, WorkstationVM)
1480 * Translate surf routings from C to C++
1481 * Add callbacks using sigc++ or boost::signals2
1482 - Add callback functions for resource creation/destruction
1483 - Add callback functions for action state change
1484 - Handle Energy as a plugin
1485 * Replace swag by boost::intrusive
1486 * Add new routing models for clusters. For documentation, see
1487 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1488 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1489 parameters for cluster tag
1490 - Fat trees, with topology="FAT_TREE" and
1491 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1492 - see examples/platforms/torus_cluster.xml and
1493 examples/platforms/fat_tree_cluster.xml
1494 * More documentation
1497 * Hostfiles support host:nb_processes construct to deploy several processes on
1499 * Collective communication algorithms should not crash if used with
1500 improper number of nodes and report the error.
1501 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1502 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1503 MPI_Dims_create, MPI_Cart_sub are supported.
1504 * New interface to use SMPI programmatically (still depends on MSG for
1505 some parts, see examples/smpi/smpi_msg_masterslave):
1506 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1510 * Global variables privatization in MPI executables is now performed at runtime
1511 with the option smpi/privatize_global_variables (default:no).
1512 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1513 dynamic libraries loaded by the application are not privatized (static
1514 linking with these libraries is advised in this case)
1517 * Options defined in XML work correctly now.
1520 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1521 copied into simgrid.jar or not (ON by default). Use this option if you want
1522 to reduce the size of the installed simgrid.jar, *and* the native libraries
1523 are kept installed elsewhere.
1524 * Surf binding with SWIG (code generated in maintainer mode only):
1525 - plugin to handle callbacks
1526 - CPU model only for the moment
1529 * Supernovae build mode is definitively removed. It was used to improve
1530 inlining and inter-module optimizations. It is nowadays superseded by
1531 link-time optimizations commonly available in compilers.
1532 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1533 * Add boost dependency for surf++
1534 * Add new macro for tests
1535 - ADD_TESH(name <tesh_args>)
1536 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1540 - xbt_bar_t XBT_barrier_init(unsigned int count)
1541 - int XBT_barrier_wait(xbt_bar_t barrier)
1542 - void XBT_barrier_destroy(xbt_bar_t barrier)
1543 * Make the xbt_os_time module public
1545 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1547 SimGrid (3.10) stable; urgency=low
1549 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1552 * Reintegrate Java to the main archive as desynchronizing these
1553 package is not acceptable anymore (Java is now considered stable)
1554 * Add explicit synchronization facilities through semaphores
1555 * Bug fix: Task.setDataSize() only changed the C world, not the value
1556 cached in the Java world
1559 * Dramatically change the way files are handled. API and internals changed, but
1560 this part of MSG was not considered as production grade either.
1561 * Add explicit synchronization facilities through semaphores
1562 * Add a new function MSG_host_get_process_list()
1563 * Preliminary DVFS support (see examples/msg/energy/ for details)
1566 * SMPI is now included directly in the libsimgrid as the windows
1567 linker doesn't force us on splitting it anymore.
1568 * Improvements of the SMPI replay tool:
1569 - Most of the collective communications are now rooted in the same process as
1570 in the original application.
1571 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1572 used until now). Multiple data types can now be used in a trace.
1573 - The replay tool now supports traces produce either by TAU or a modified
1575 - Bug Fix: the compute part of the reduce action is now taken into account.
1576 - Gatherv collective is now supported
1577 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1578 of smpirun outputs time independent traces for the current run. One file
1579 is created per process. If too many processes are simulated, this behavior
1580 can be changed to one file for all processes by using the
1581 tracing/smpi/format/ti_one_file flag
1582 * smpirun generates the host file if needed (with given host count and platform)
1583 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1584 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1585 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1586 --cfg=smpi/coll_selector:(mpich/ompi)
1587 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1589 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1590 happening inside a collective SMPI call.
1591 * Fix the behavior of complex data types handling.
1592 * Make MPI_Wtime another synchronization point to take computations into
1594 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1595 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1596 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1597 (no privatization of global variables yet)
1598 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1601 * Verification of liveness properties is now available for SMPI applications
1602 (in addition to MSG applications)
1603 * Bugged examples using SMPI in examples/smpi/mc/
1604 * Add --cfg=model-check/visited option. Allows the verification of infinite
1605 programs. Detection of loops in the execution thanks to the system state
1606 comparison and reduction of the state space to explore. Can be combined with
1607 DPOR for safety properties.
1610 * Allow to change SimGrid configuration (see --help) within the code
1611 thanks to SD_config() as it can be done in MSG.
1612 * Add a new function SD_task_set_amount() upon user request.
1615 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1616 * Remove rule based routing (no more PCRE dependency)
1617 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1618 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1619 * Add a loopback_bw and loopback_lat options to cluster tag.
1620 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1621 routing. Now peers are encapsulated in an AS and have their own private
1622 router but this is transparent.
1625 * Our own implementation of getline is renamed xbt_getline, and gets
1626 used even if the OS provide a getline(). This should reduce the
1627 configuration complexity by using the same code on all platforms.
1628 * New type: xbt_cfg_elm_boolean.
1629 * Allow to use yes/no for boolean configuration options in the command line.
1630 * Allow to disable SimGrid cleanups at exit from command line option.
1631 There are situations where one may want a simulation to end with an exit.
1632 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1633 annoying when scripting around the simulator. Adding a
1634 --cfg=clean_atexit:no allows to circumvent this issue.
1637 * Lots of memory leaks were corrected in this release.
1638 There is no known memory leaks anymore, in all of our 600+ tests.
1639 * New command line option --version, to get SimGrid version information.
1640 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1642 * Supernovae builds are deprecated, and expected to be removed in the next
1645 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1647 SimGrid (3.9) stable; urgency=low
1649 The Grasgory release: GRAS is really dead now.
1651 * Complete overhaul of the internal host structures scheme.
1654 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1655 was considered as experimental and badly maintained since then.
1656 * Keeping it was thus a trap to our potential users, that could take
1657 it instead of MSG or SMPI by mistake despite is pity state.
1658 * GRAS seems to have very few users (if any), and no one volunteered
1659 to maintain it further. It also induces a lot of XBT code (for
1660 portability sake), that must be maintained too.
1661 * For all these reasons, we killed GRAS. If someone wants to revive it
1662 in the future, don't cry, our git history still remembers of GRAS.
1665 * Major overhaul. Merge our documentation again as time proved that
1666 splitting it was really not helping our users.
1667 * Further improve the developer documentation to help newcomers
1668 hacking on SimGrid itself. The user documentation (and in
1669 particular, the beginner documentation) is still in a sorry state.
1672 * Now works on Windows too!
1673 * Much more extensive test suite, from MPICH
1676 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1677 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1678 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1679 * Bug fix: let task be scheduled when the last dependency to be solved is
1680 a control dependency.
1681 * Remove SD_load_environment_script function.
1682 Use the C sg_platf function if you want to declare a platform
1686 * New function: MSG_process_get_number()
1687 * Old function documented: MSG_config()
1688 * Remove MSG_load_platform_script function
1689 Use the C sg_platf function if you want to declare a platform
1693 * Change the default value of the TCP_gamma constant (maximal size of TCP
1694 congestion window) to a more realistic 4MiB value. If you notice changes in
1695 your simulation results, you can fall back to the previous 20k tiny window
1696 by adding --cfg=network/TCP_gamma:20000 on command line.
1697 * (Hopefully) fix a bug wrt periodic availability and state traces
1698 * Bug fix: use default values at start when first event in availability/state
1699 trace is not at time 0.
1702 * remove the "new_" part of function name sg_platf_new_trace_connect
1703 (resulting in sg_platf_trace_connect), since it does not create
1707 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1708 use case where it's really mandatory, and maintaining it was a pain in
1710 * New: xbt_fifo_search(), search an item with a user-provided
1711 comparison function instead of dumb pointer comparison.
1714 * Fix the lua deployment:
1715 Use `simgrid.init_application()` before deployment instead of
1716 `simgrid.msg_register_application()` after.
1719 * Transfer the tracing files into the corresponding modules.
1721 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1723 SimGrid (3.8.1) stable; urgency=low
1725 The "we are told that some people want to also *install* the simgrid
1728 * Add missing file "tesh.1" to the archive.
1730 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1732 SimGrid (3.8) stable; urgency=low
1734 The Psssshiiiit release: SimGrid jumps into the Cloud.
1737 * Add an experimental interface to manipulate VMs. They are mainly
1738 process groups with very few intrinsic semantic, but they should
1739 allow you to build the semantic you want easily.
1740 * New function: MSG_host_set_property_value()
1741 * New function: MSG_process_on_exit(). To clean memory in all cases.
1742 * Bug fixes that made the host (and link) failures unusable.
1743 * Add a way to auto-restart process when the host in which they are
1744 executing comes back (ON_FAILURE="RESTART" on deployment file,
1745 MSG_process_auto_restart_set).
1746 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1747 please stop using the old ones, they are DEPRECATED.
1749 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1750 Please use MSG_init() instead. (reducing the amount of entry
1751 points in the library helps us).
1752 * Make it impossible to link against the wrong version of the lib
1753 * Deprecate MSG_clean(). No need to call it anymore.
1754 * Function MSG_get_host_number() is not deprecated anymore.
1757 * Split the doc into a user guide and a reference guide.
1758 * Start a developper guide to help people hacking on SimGrid.
1761 * Enable tracing by default. This modules rocks you should use it.
1762 * Remove option custom_flags. Now use environment variables CFLAGS
1764 * Use default cmake things to detect lua instead of home grown ones.
1765 * New option "enable_mallocators" to disable mallocators, for debugging
1766 purpose ("on" by default).
1769 * Bug fixes around the resource failures: don't let the processes
1770 survive the host they are running onto.
1771 * Add an interface to auto-restart processes when the host in which they are
1772 executing comes back.
1773 * Ensures that SIMIX_clean is called automatically. It's not part of
1774 the public interface anymore (bindings should be updated).
1777 * Bug fix for when SD_Simulate is called with a positive value: be careful
1778 when comparing doubles. Sometimes they are different for non significant
1780 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1781 parallel task whose initial work is distributed among host according
1782 to the Amdahl's law. Such tasks are created with a parameter alpha
1783 that corresponds to the non-parallelizable part of the computation.
1784 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1785 between two sets of workstations assuming a 1D block distribution (each
1786 workstation owns a similar share of data) on both sides.
1788 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1789 Data redistribution will be automatically scheduled once parent and child
1790 are both scheduled. The filling of computation_amount and
1791 communication_amount structures is now done seamlessly thanks to the chosen
1793 * New function SD_workstation_dump to display various information
1794 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1795 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1796 bandwidth on the route onto which the task is scheduled and the amount of
1798 To divide the nominal bandwidth by 2, the rate then has to be:
1799 rate = bandwidth/(2*amount)
1800 * Compute tasks that have failed can now be rescheduled and executed again
1801 (from their beginning)
1802 * Increasing source code coverage (src/simdag is now covered at 95.8%
1806 * Re-implement time-independent trace replay using SMPI (at the
1807 smpi_smp_* level) instead of MSG. This should replace
1808 examples/msg/actions/actions.c
1809 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1810 hindexed and structs)
1811 * Implement the exchange of non-contiguous data.
1812 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1813 * Correct behavior of smpi/sender_gap and set its default value to 0
1814 * Add option to asynchronously send small messages to allow better
1815 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1816 specifies the size in bytes under which messages will be asynchronously sent.
1817 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1818 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1821 * New C interface to define a platform: XML is now optional.
1822 For more info, please check include/simgrid/platf.h.
1823 * New interface to define random platforms from the C:
1824 For more info, please check include/simgrid/platf_generator.h and
1825 examples/msg/masterslave/masterslave_platfgen.c
1826 * Export a sg_cmdline dynar containing all the arguments we got from
1830 * Two new tracing options for adding comments to trace file so you
1831 can track your experiments (see --help-tracing for details).
1832 * New option to generate an impoverished trace file (--cfg=tracing/basic:1)
1833 * Adding the SimGrid version that generated the trace file as a comment.
1834 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1835 * Fix to avoid key clashes on Paje links
1836 * Other minor fixes related to the Paje specification
1839 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1840 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1841 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1842 * Greatly improve the robustness of mmalloc to user errors (such as
1843 using an area after freeing it, or freeing it twice)
1845 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1847 SimGrid-java (3.8.1) stable; urgency=low
1849 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1850 Warning: all methods are visible, but only some of them are
1851 implemented so far. Check the source (src/jtrace.c)
1852 for further information.
1853 * New module: org.simgrid.msg.File (SimGrid File management functions)
1854 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1855 * Change the meaning of Process.restart: now restart the process from
1856 the beginning, like MSG_process_restart in C.
1857 * Add Process.setAutoRestart: handling of process restart when failed
1859 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1860 you to retrieve the properties of the processes/hosts
1861 * Deprecate Msg.clean(): you can just forget about it now.
1862 * New function Process.getCount(), that only works when compiling
1863 with the not yet released version 3.9 of the C library.
1865 * New context factory based on Coroutines. It mandates a modified JVM
1866 but then, the simulations run about five times faster, and there is
1867 no limit to the amount of processes (beside of the available memory).
1869 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1871 SimGrid (3.7.1) stable; urgency=low
1874 * Restore the prototype of MSG_process_create_with_environment() to
1875 the pre-3.7 situation by removing the kill_time argument.
1876 * Add a MSG_process_set_kill_time() function instead.
1879 * Fix weird behaviors when dealing with parallel tasks.
1882 * Simgrid is now built as a dll.
1883 * Simgrid-java now works on Windows.
1884 * Simgrid-Java is now included into Windows package.
1887 * First pre-build package for MacOSX.
1890 * Fix compilation when using MSG_USE_DEPRECATED.
1891 * Fix some compilation issues on Macs and Windows.
1892 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1893 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1895 We discovered that the Lua console is broken, but we are missing the
1896 manpower to fix it right now. The problem existed in 3.7 too, so we
1897 are not blocking the release for that. Sorry if you depended on this
1898 feature, any help would be really welcome.
1900 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1903 SimGrid-java (3.7.1) stable; urgency=low
1905 The "Java aint got to be bloated and slow" release
1908 * Various internal cleanups and performance improvement
1909 Simulations are expected to run up to twice faster or so
1910 * Make Process.kill(process) an instance method, not a static one
1911 * User processes are not java.lang.Thread subclasses.
1912 This breaks the compatibility (sorry), but previous API was
1913 brain-dead, making it impossible to have non-trivial
1914 initializations in the process constructor.
1915 * Require a full constructor per Process sub-class.
1916 Kinda breaks the compatibility (sorry), but this allows a much more
1917 efficient way to launch the processes at simulation startup.
1918 * Do not embeed our version of semaphores, java 1.5 can be considered
1919 as sufficiently prevalent for us to not dupplicate its features.
1923 * Add examples for almost every part of the API
1924 We spotted and fixed a lot of bugs in the process
1926 * New module: asynchronous communication API
1927 * New function: Process.sleep()
1928 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1929 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1930 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1932 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1934 SimGrid (3.7) stable; urgency=low
1936 The "spring cleanups (before the next Big Project kicks in)" release.
1939 * We can specify the SMPI latency/bandwidth factor with command line
1940 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1941 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1942 You can also use the "config tag" from platform file by adding this line
1943 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1944 (see "example/platforms/tag_config.xml" to use "config tag").
1945 Note that the command line supersedes the platform file configuration.
1946 * Change the correction factors used in LMM model, according to
1947 the latest experiments described in INRIA RR-7821.
1948 Accuracy should be improved this way.
1949 * Use the partial invalidation optimization by default for the
1950 network too. Should produce the exact same results, only faster.
1951 * Major cleanup in surf to merge models and split some optimization
1952 mechanisms from the core of the models. As a result you can now
1953 specify which model to use (e.g., --cfg=network/model:LV08
1954 --cfg=cpu/model:Cas01) and which optimization mode to use
1955 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1956 Incompatible combinations should err at initialization. See
1957 --help-models for the list of all models and optimization modes.
1958 * The CLM03 workstation model was dropped for simplicity because it
1959 used the deprecated CM02 network model. Use default instead.
1960 * Rename the TCP_gamma configuration option to network/TCP_gamma
1961 * Rename the coordinates configuration option to
1962 network/coordinates, and document it
1963 * Use now crosstraffic keyword instead of the terribly misleading
1964 fullduplex keyword. It is activated by default now in the current
1965 default model, use --cfg=network/crosstraffic:0 to turn it off.
1966 * Ongoing refactoring the model parsing to make XML files optional
1967 See include/simgrid/platf.h for details (still to be completed)
1970 * Major overhaul of the documentation. Almost instructive now :/
1971 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1972 functions and friends. This interface was considered as
1973 deprecated since over 2 years, it's time to inform our users that it is.
1974 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1975 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1976 use these (crappy) functions in your code.
1977 These functions will be removed soon. Stop using them now.
1978 * Deprecate MSG_get_host_{table,number}
1979 Implement MSG_hosts_as_dynar() instead.
1980 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1981 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1984 * Stabilize the parallel execution mode of user contexts
1985 * Introduce configuration variables to control parallel execution:
1986 - contexts/synchro: Synchronization mode to use when running
1987 contexts in parallel (either futex, posix or busy_wait)
1988 - contexts/parallel_threshold: Minimal number of user contexts
1989 that must be part of a scheduling round to switch to parallel
1990 execution mode (raw contexts only)
1991 * Fix bugs that prevented to use suspend/resume along with
1992 synchronization structures.
1993 * Fix bugs in process termination that lead to invalid memory access
1994 in very specific conditions.
1997 * Introduce a parallel mode for the models (controlled by surf/nthreads
1998 configuration item). In our tests, running the models in parallel
1999 never lead to any speedups because they are so fast that the gain
2000 of computing each model in parallel does not amortizes the
2001 synchronization costs, even when ultra fast futexes are used.
2002 This is released anyway because YMMV.
2005 * Performance boost by using a swag internally to compute the set of
2006 tasks that are finished and should constitute the return value of
2010 * Enable it by default now that it is considered rather stable.
2013 * Documentation of the tracing functions.
2014 * Performance gains when tracing categorized/uncategorized resource
2015 utilization by avoiding calls to get route when updating resource
2016 variables. LMM constraints are being used instead.
2017 * API changed to set task categories. Use MSG_task_set_category instead
2018 of TRACE_msg_set_task_category, and SD_task_set_category instead
2019 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
2020 * Bugfix for graphicator, routes not correctly obtained, memory leaks
2021 * Examples for link user variables added (see at examples/msg/tracing/)
2022 * Deprecated function TRACE_msg_set_process_category completely removed
2023 * Trace header updated according to the latest Paje file format
2024 * Tracing network lazy updates, no longer obligate users to use full updates
2025 * --cfg=tracing/platform:1 also registers power/bandwidth variables
2026 * Experimental: let user code declare/set/push/pop application states for hosts
2027 * API changed to allow the manual creation of graph configuration files
2028 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
2031 * Improve the API of Lua MSG bindings, using the Lua spirit.
2032 * Each simulated process now lives in its own Lua world (globals are
2033 automatically duplicated). It helps writing simulators. Will allow
2034 to run Splay programs within SimGrid in the future.
2035 * Add a Chord example in Lua, equivalent to the MSG one.
2038 * Start the implementation of a solution to express temporal
2039 properties, not only local assertions. This is still an
2040 experimental work in progress, stay clear from it to be safe.
2044 - Add new runtime parameters --help-logs and --help-log-categories
2045 to display information about supported logging parameters and
2047 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
2048 don't exists anymore.
2049 * Mallocators: allow value NULL for the reset function.
2051 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
2052 create homogeneous dictionaries, where all the elements share the
2053 same free function. Non homogeneous dictionaries will be
2054 deprecated in the next release.
2055 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
2056 - Multi-level dictionaries are deprecated.
2058 - new function xbt_dynar_search_or_negative() that is useful when
2059 you have less than 2 million elements in your dynar and don't
2060 want of the extra complexity of catching exceptions when the
2061 element is not found.
2063 - Make xbt_os_thread module (for thread portability) public.
2064 Documentation is still to come, sorry.
2066 - Cleanups and simplifications to make it maintainable again.
2067 - Exotic features (such as memalign and valloc) were removed.
2068 - The metadata were extended and improved so that the
2069 model-checker becomes able to explore and inspect the heaps.
2070 - This may induce a performance drop when enable_model-checking is
2071 ON in cmake (even if it's not used in the simulation), but it is
2072 necessary at this point to get MC working.
2074 Turn model-checking OFF if simulation performance matters to you.
2075 Not enabling it at runtime is not enough, disable it in cmake.
2077 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2080 SimGrid (3.6.2) stable; urgency=low
2082 The "Not coding new stuff allows to polish old things" release.
2085 * New bindings to the NS3 packet level simulator (experimental)
2086 * Use the raw (efficient) execution contexts instead of the sysv
2087 (portable) ones when possible.
2088 * libpcre is now mandatory in any cases since not using it led to
2089 severe performance loss and possibly other issues
2090 * Update the XML platforms:
2091 - G5K: include the latest machine in Nancy
2092 - GridPP and LCG: new platforms
2093 * Documentation was partially updated, at least (more to come)
2095 Bug fixes, cosmetics and small improvements
2096 * Free terminated processes before the end of the simulation to avoid
2097 exhausting the memory of users having very dynamic amount of
2099 * Bug fix and cosmetics about canceling non-running tasks
2100 * Bug fix about the dot loader's issues when using libcgraph
2103 * Create an installer for windows with nsis (amd64 and win32)
2104 - Add an hello world project to illustrate simgrid project creation.
2105 - Embed libpcre into the Simgrid installer to avoid
2106 its compilation burden
2107 * The raw execution contexts should work on Apple now
2108 * Port to Windows 64 bits
2109 - Sysv contexts now have an implementation for this arch
2110 - GRAS communication features now support this arch
2111 * Drop support for borland compiler on windows
2112 - this code was not maintained, and we kinda depend on gcc nowadays
2113 * Fix portability issues on kfreebsd/gnu: build error about semaphores
2114 * Fix portability issue on unstable ubuntu: linker became picky on
2117 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2119 SimGrid (3.6.1) stable; urgency=low
2121 The "Oops, we broke Macs too" release
2124 * Fixed contexts detection so that raw ones are used when possible
2125 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
2126 leads to a strange error, with user code segfaulting sometimes when
2127 the generated code is not perfectly aligned (which is not
2128 controllable from the user side, depends on the amount of code)
2131 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
2133 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2135 SimGrid-java (3.6) unstable; urgency=low
2138 * Split of every thing from simgrid v3.5 into a separate package.
2140 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2142 SimGrid (3.6) stable; urgency=medium
2144 The Summer Release, also known as the "OMG! They Killed Kenny!" version
2147 * Bindings now constitute their own package, separated from the main one.
2148 Rationale: reduce our maintenance nightmare by reducing the module coupling
2149 They will soon be released on their own on gforge.
2151 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
2152 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
2154 GRAS: It is not considered as stable anymore, but experimental. Sorry.
2155 * It's not quite deprecated for now because we have no replacement,
2156 but it may soon become the case.
2159 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
2160 * Fortran: New user-level cache variable to store the rank of the running
2161 process. This improves performance by an order of magnitude.
2162 * C: New coccinelle script to automatically locate and modify global and
2163 local static variables.
2164 * Improved SMPI network model with a sender-side gap to account for multiple
2168 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
2169 only say if a communication is finished, no matter whether it succeeded or
2170 failed. You can call MSG_comm_get_status() to know the status of a finished
2172 * New function MSG_task_dsend() to send a task and detach it. When a
2173 communication is detached, you are never notified of its success or failure
2174 and the memory is released automatically once it is finished. This function
2175 is useful when you don't care about the end nor the success of a
2177 * Change the prototypes of action replay. Sorry for inconvenience,
2178 but this is really more efficient this way (and to adapt your code,
2179 you just have to fix the initialization, that shouldn't be too long)
2180 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
2181 ever managed to do anything useful with it.
2182 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
2183 takes a dynar of communications. It returns immediately and gives the
2184 index of a finished communication (if any).
2185 * New example: a basic implementation of the Chord P2P algorithm.
2188 * New model for multi-core CPUs. You can now use the core attribute to
2189 precise the number of cores of a host. This is a basic model. Every
2190 process running on the host receives at most the power provided in
2191 the DTD (throughput<=power). Total throughput of process cannot exceed
2193 * New peer tag. This peer tag creates a tiny AS comprising a host and a
2194 router linked by an up-link and a down-link (possibly asymmetrical).
2195 This kind of pattern allows to easily build last-mile model style platforms.
2196 Aggregating such patterns in a rule-based AS is thus the technique of
2197 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
2198 * New model for Vivaldi routing. We transformed the Vivaldi network model
2199 into a Vivaldi routing model (based on the rule-based model). This allows to
2200 combine Vivaldi based latencies with last-mile platforms.
2203 * Added a check for NaN of IEEE754 infinite in the double entries of
2205 * Introduce a new context factory "raw", highly inspirated from the
2206 ucontext factory, but using manually crafted functions in assembly to
2207 do the work in an efficient manner.
2208 * Allow to change the used context factory at run time, not only at
2209 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
2210 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
2211 contexts at runtime (only with raw contexts or ucontexts).
2212 * Completely rewrote this module to allow parallel execution of user
2213 processes. Use --cfg=contexts/nthreads:N to execute user processes
2214 with N parallel threads (the default is 1, meaning no parallelism).
2215 * Allow to decide dynamically between sequential and parallel modes.
2216 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
2217 processes in parallel only when their number is greater than or equal to P
2219 * Added a check for NaN of IEEE754 infinite in the double entries of
2223 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
2224 won't output the state of processes when interrupted with Ctrl-C
2225 * Add a new function xbt_dynar_to_array that transforms a dynar into a
2226 NULL-terminated array. This may solve backward compatibility issues
2227 due to the change to return type of SD_simulate. See also:
2228 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
2229 * Add new macros with variable number of arguments.
2230 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
2231 - in xbt/asserts.h: xbt_assert
2232 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
2233 - in xbt/ex.h: THROWF and RETHROWF.
2234 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
2236 * Change xbt_die() to accept a format string with arguments, just like printf.
2237 * New data structure: xbt_lib_t, like a dict but more general and with better
2241 * New configuration options
2242 Options triva/categorized and triva/uncategorized can be used to generate
2243 graph configuration files for Triva visualization tool.
2244 * Configuration option tracing/platform is renamed to tracing/categorized
2245 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
2246 * New TRACE_user_link_variable interface:
2247 User provides the name of the link and the tracing variable to attach to it
2248 * the declaration of tracing categories must be done after the environment
2250 * simpler tracing interface, just one way to declare categories
2251 TRACE_category or TRACE_category_with_color, it is up to you
2252 * links in the trace file are again identified by their names
2253 * trace contains the full platform hierarchy exactly as declared using the ASes
2254 * Options tracing/msg/[task|process]:1 groups the process by hosts
2255 for both cases, tasks and processes must have names that are unique during the simulation
2256 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
2257 * The experimental option tracing/msg/volume is deprecated
2258 its functionality may be reincorporated if needed
2260 The tracing generates a trace file with unordered timestamped events,
2261 because of the way the core simulator (surf) works. A script available
2262 at the tools directory (fix-paje-trace.sh) can be used to put the events
2263 in order. We have changed the tracing so it can generate ordered timestamped
2264 events in the final trace, but depending on the simulator (and how much time
2265 is simulated) that can lead to a huge memory utilization. It is deactivated
2266 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
2268 Build Infrastructure
2269 * Define a SIMGRID_VERSION macro in simgrid_config.h.
2270 - We are trying hard to keep the API stable, but it may happen that
2271 some things change (we're a research project after all, not a
2272 nuclear plant operating system). If such things should happen, you
2273 could rely on that macro to adapt.
2274 - current value: 30600 for 3.06.00, aka 3.6
2275 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
2276 a number that can be compared with SIMGRID_VERSION.
2277 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
2278 assertions and verbose/debug logging events are disabled at compile time.
2280 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2282 SimGrid (3.5) stable; urgency=medium
2285 * New feature to any SimGrid-based simulator: Model-Checking
2286 Check SIN#1 for more details.
2289 * New Model SMPI (three-interval linear regression for correction factors)
2290 See RR-7426, available at http://hal.inria.fr/inria-00527150
2291 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
2292 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
2293 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
2294 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
2295 * See also src/smpi/README
2300 - Tracing API changes: TRACE_start and TRACE_end should not be called
2301 by user-code. They are automatically called by simulators created
2302 with SimDAG, MSG and SMPI if the toolkit is compiled with
2303 tracing_enabled=ON. Categories declaration and utilization remain the
2304 same for MSG and SimDag.
2305 - A function was added to the tracing API to declare categories with
2307 - TRACE_category_with_color (char *category, char *color)
2308 where color must be in the following format
2309 "%f %f %f", red, green, blue
2310 and red, green, blue are float values in the interval [0, 1]
2311 - User can specify NULL as color parameter, or continue calling
2312 TRACE_category (cat)
2313 On that case, the tracing system will define random colors
2314 - The following command-line options are supported:
2315 --cfg=tracing/filename:msg.trace
2316 --cfg=tracing:1 (activate tracing, needed to use others)
2317 --cfg=tracing/platform:1 (categorized resource use)
2318 --cfg=tracing/uncategorized:1 (uncategorized resource use)
2319 --cfg=tracing/msg/task:1 (task creation)
2320 --cfg=tracing/msg/process:1 (process creation, migration)
2321 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
2322 --cfg=tracing/smpi:1 (SMPI interface tracing)
2323 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
2324 - examples of examples/msg/tracing updated
2326 - DAXLoader and DOTLoader functions can generate tasks with categories
2327 - A new function to attribute a category to SD tasks:
2328 TRACE_sd_set_task_category (SD_task_t task, char *category)
2329 * Tracing the MPI interface implemented by SMPI
2330 - Collective operations are traced with states
2331 - Point-to-Point operations are traced with states/links
2332 - Tracing activated by a parameter "-trace filename" passed
2333 to smpirun during execution (considering that simgrid
2334 is compiled with tracing enabled)
2335 - To run the simulation with gdb, the simulator
2336 accepts --cfg=tracing/smpi:1 to trace SMPI
2337 - tesh files to check if smpi tracing is ok
2338 - See examples/smpi/NAS/DT-trace
2339 * GTNetS tracing re-worked
2340 - adaptation to the tracing system of GTNets to cope
2341 with modifications regarding the fullduplex mode
2342 - new tesh files to check if gtnets tracing is ok
2345 * Asynchronous communications through the functions:
2346 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
2347 * New function: MSG_load_platform_script()
2348 to make possible using a lua script instead of XML files to set up platforms
2349 * New function: MSG_set_function
2350 to associate functions to processes, used when bypassing the parser
2351 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
2353 Platforms: Add some more examples in examples/platforms
2354 * Grid'5000: see www.grid5000.fr
2355 * *_30000_hosts.xml: various huge files [mainly scalability testing]
2358 * Change the XML format. This is a very important modification. SimGrid
2359 3.5 introduces a new hierarchical format based on the notion of
2360 Autonomous Systems. Compatibility with old format is ensured
2361 through the perl script provided in the install bin directory
2362 bin/simgrid_update_xml.
2363 It is now possible to build platforms with specific routing
2364 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
2365 connect several platforms together. We will try to provide soon
2366 set of realistic platforms exploiting these properties (have a
2367 look at examples/platforms/ for the moment).
2368 * Take the opportunity of the XML format change to be a good XML citizen:
2369 rename link:ctn to link_ctn and similar changes (also dealt with by
2371 * Add a new routing scheme (rule-based) using regular expressions. It
2372 enables to have an extremely low memory footprint when the
2373 underlying routing is simple and can be compactly described. You
2374 need to have libpcre4-dev (perl regular expressions) installed if
2375 you want to use this routing scheme.
2376 * Revive the cluster TAG and allow to easily and efficiently (both in
2377 term of memory and speed) connect clusters together. Have a look
2378 at teshsuite/simdag/platforms/ to see how this can be done. With
2379 this tag, you can create clusters with thousands of tasks at no
2380 cost (have a look at examples/platforms/).
2381 Note: clusters are implemented as ASes, so there is no need for an
2382 enclosing AS tag if you have only one cluster in your platform.
2383 * Add new generic functions in the public interface that allows the user
2384 to call SURF 'create_resource' methods from your code (same
2385 functionality as the XML bypass mechanism but with a much lighter
2387 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2388 that uses a piecewise linear approximation to produce better
2389 results when exchanging small messages.
2390 * Add a new parameter to handle correctly full duplex link and account
2391 for interferences between uplink and downlink communications
2392 (activate with --cfg=fullduplex:1).
2395 * Rename the SD_READY (all dependencies are satisfied and task is
2396 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2397 dependencies are satisfied) state.
2398 This prevents a confusion between the notion of "ready to schedule"
2399 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2400 simulated" (SD_RUNNABLE) used by the simulation kernel.
2401 * Change the way a task is considered as ready. Instead of removing
2402 dependencies when a task is done, a counter is decreased. This way,
2403 it is always possible to reach ancestors thanks to the
2404 SD_taks_get_parents function (even after the end of the simulation.)
2405 * Change the return type of SD_Simulate from (SD_task_t*) into
2406 xbt_dynar_t. This function was in handling a dynar internally and
2407 converted it into a NULL terminated array for historical reasons.
2408 * New function SD_dotload(char*) to load a DAG described in dot
2409 format. This loader and the corresponding examples require the
2410 installation of the graphviz library.
2411 * Fix a bug in the management of tasks of size 0 in the surf network
2412 models. This problem was only visible with SIMDAG and you should
2413 thus disregard results produced with earlier versions if you
2414 relied on this feature (some tasks were blocked because of this).
2415 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2416 even though your applications were doing only point-to-point
2417 communications and sequential computations. Now you can really use any
2418 model you want (of course, if you create real parallel tasks, which are
2419 not implemented in most models beside ptaskL07, this will abort).
2420 * Add an example that schedules a DAX on an heterogeneous platform
2421 using a Min-Min strategy.
2422 * New function SD_workstation_get_current_task() that returns the kind
2423 of task currently running on a workstation in the sequential access
2425 * Raise some warnings when unexecuted tasks remains at the end of the
2426 simulation. This is usually caused by cycles in the DAG.
2429 * New function: SIMIX_process_set_function() called by MSG_set_function
2430 * Change the underlying waiting queue in semaphores so that a process
2431 can wait on several of them simultaneously (as in waitany).
2432 * Fix the way to handle tokens in semaphores so that all access patterns
2433 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2434 * kill the dirty pimple SIMIX_message_sizes_output()
2435 Please use (proper) visualization instead
2438 * New data container: setset (set of sets of elements)
2439 * New module: mmalloc (mapped malloc, allowing to have several
2440 independent segments of malloc)
2441 * New function: xbt_dict_cursor_set_data()
2442 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2443 * New function: xbt_dynar_is_empty()
2444 * New function: xbt_fifo_get_last_item()
2445 * Fix xbt_dynar_shrink(): use the right element size.
2446 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2447 uninitialized areas during expand.
2448 * Fix semaphores: previous implementation was severely broken.
2449 * Use library init/fini functions for our initialization.
2450 - you can use logs and other feature as soon as you want in your
2451 code (even before the xbt_init / MSG_init)
2452 - xbt_exit is now a no-op and produce a warning when used.
2455 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2456 benefit from the latest improvement to the simulation kernel.
2457 * Kill measurement sockets for now. If you rely on them, sorry. This
2458 release is not for you. This feature will be reintroduced in the
2459 future, but we cannot delay the release any further.
2460 * New function: gras_msgtype_get_name().
2461 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2462 * Fix (at last) the pmm example: it should not randomly fail anymore.
2464 Build chain: bug fixes and overall polishing
2465 * Cmake is now stable enough. Hence, we killed the autotools.
2466 * Port to windows ( TM :)
2467 * Fix the 'make install' target.
2468 No need to use 'make install-simgrid' anymore
2469 * Introduce a 'make dist' target compiling a *source* archive
2470 'make package' compiles a binary archive
2471 * Compile java files only on need
2472 * Add --cd and --setenv command line options to tesh
2473 * Out of source builds are not fully supported yet, but we are close
2474 * Enable supernovae and optimization flags by default for our users
2477 * Add layer to set up environment directly from lua, without XML.
2478 * The effect of gras_stub_generator can be achieved through
2479 lua too (check examples/gras/console/ping_generator.lua)
2481 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2483 SimGrid (3.4.1) stable; urgency=low
2485 The "Polishing easter eggs is probably a good idea" release.
2486 This is a bug fixes release only.
2490 * Fix a bug preventing the tasks from begin garbage collected.
2493 * Fix a bug occurring when a host involved in a communication fails.
2494 This was not detected properly by the other peer involved in the
2495 communication. Now, it's reported as a network error.
2498 * Warn the user about loop dependencies in data flow of DAX files
2499 * Obey the control-flow dependencies of DAX files
2502 * Add option "enable_smpi" allowing to not compile SMPI.
2503 Probably useful for the (Mac) users experiencing a build error here
2504 * Improve the detection of lua5.1 and ruby1.8
2506 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2508 SimGrid (3.4) stable; urgency=low
2510 The "Easter in Cargese" release. Also known as (major changes):
2512 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2513 ~> bindings were greatly improved
2514 ~> new tracing infrastructure for better visualization introduced
2516 * the "Welcome to configury modernity" release.
2517 ~> we switched from autotools to cmake, and improved our cdash
2520 A more detailled list of changes follow (full detail in svn log).
2522 Java Bindings: Various Cleanups
2523 * (install java-gcj-compat-dev on debian-like to use them)
2524 * Remove put/get: no need to export deprecated interface in Java
2525 Use send/receive instead.
2526 * Cleanup the examples and add a README per directory
2527 * Remove example autoDestination (that's the only way to go now)
2528 * Remove example explicitDestination (was a plain copy of basic)
2529 * Make JniException a runtime exception, so that there is no need to
2530 declare the fact that you may encounter such a beast. I guess that
2531 nobody will ever want to survive such error.
2532 * Create specific errors for each MSG case of failure:
2533 host failure, transfer failure, timeout, task cancelled
2534 * Cleanup the exceptions that may get thrown by each function
2535 * Other internal cleanups in Java bindings. Performance still bad :/
2536 Ruby and Lua Bindings: create them
2537 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2538 * That's new and great, you should try them out.
2539 Same functionalities than Java bindings, only even less polished
2541 * Kill the useless "rate" argument of SD_task_get_execution_time()
2542 Everyone used to provide -1 as a value, it was not used, and the
2543 semantic of a possible use wasn't even clear.
2544 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2545 either as comm costs or compute costs to mean that there is no
2546 such thing for that specific task.
2547 * Add a SD_task_set_name() function
2548 * Fix SD_task_unschedule() on typed tasks
2549 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2550 * In DAX loader, accept useless 'level' attributes to <job> since
2551 LIGO DAGs have them (seem to be to ease graphical representation).
2553 * Add an example masterslave_mailbox.c using send/receive and not
2554 the deprecated put/get interface.
2555 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2556 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2558 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2559 (declare MSG_USE_DEPRECATED to still have the old name)
2560 * Add a MSG_task_set_data() function
2561 * About trace replay (see examples/msg/actions):
2563 - Allow to work with splitted trace files for each process
2564 Give the specific trace file as argument of each process,
2565 and call MSG_action_trace_run(NULL)
2566 You can still have one merged file for all processes.
2567 - Fix implementation of collective operations
2568 * Allow task_execute() on 0-sized tasks (closes #10063)
2570 * This is the first release of SimGrid where SMPI is not considered
2571 beta anymore (even if some corners should still be improved)
2572 * Port over the new SIMIX_network submodule (internal refactoring)
2573 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2574 * Implement more missing elements of the standard:
2576 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2578 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2579 MPI_Scan() MPI_Get_processor_name()
2580 - Added implementation of missing case for Alltoall (warning: it's
2581 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2582 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2583 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2584 SMPI_MPI_Allgatherv()
2585 * Bug fixes include:
2586 - MPI_Waitsome() was broken
2587 - Allow relative includes in smpicc
2588 - Command line cfg argument 'reference_speed' was ignored...
2589 - Some functions did not properly lead to auto-benching of user code
2590 - smpicc passes -O2 by default (just like openmpi one)
2592 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2593 * Bug fixes about timeouts during communications
2594 * add SIMIX_message_sizes_output() as a pimple to write to file the
2595 amount of messages per size. Use gnuplot to get histogram.
2596 Pimple because that's the only user-visible function of simix,
2597 defined directly in xbt.h (irk, sorry)
2599 - Add a SIMIX_sem_get_capacity() function
2600 - Fix interactions with processe resume/suspende
2601 - release_forever() was stupidly broken
2602 - Fix SIMIX_display_process_status() for processes in a semaphore
2603 - Make SIMIX_sem_block_onto() user-visible
2604 * Refactoring context stuff:
2605 - Use pseudo-OOP for better modularity
2606 - reimplement SIMIX_process_kill() without process_schedule() so
2607 that the latter can take as invariant that it is called from
2609 - Merge context_start into context_new for sake of simplicity
2611 * Add a Vivaldi network model, coded live during SUD'10 ;)
2612 * Rename configuration variables to start a hierarchy:
2613 o cpu_model -> cpu/model
2614 o network_model -> network/model
2615 o workstation_model -> workstation/model
2616 * New configuration variables:
2617 o network/bandwidth_factor: correction to bandwidth
2618 o network/latency_factor: correction to latency
2619 o netwotk/weight_S: correction to the weight of competing streams
2620 * Add a long description to the models, that users can see with such
2621 argument on the command line: --cfg=cpu/model:help
2622 * --help-models display the long description of all known models
2624 * config: add the ability to set a default value after registration
2625 Does not override any previously set value (e.g. from cmd line)
2626 * dict: allow to have integer key and data.
2627 When so, you need to use the following functions
2628 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2629 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2630 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2631 In contrary to regular dicts, the key is not malloced before copy.
2632 Mixing scalar and regular elements in the same dict is not tested
2634 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2635 Tracing for Visualization:
2636 * SimGrid is now instrumented in order to generate a trace file for
2637 visualization analysis: to use it, need to compile SimGrid with the
2638 "tracing" option enabled, and instrument the program using SimGrid with
2639 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2640 (among other functions).
2641 * The instrumentation only traces the platform utilization for now
2642 * Documentation to use the tracing functions and how to analyze the
2643 traces with the Triva tool is written.
2644 * More information about: SimGrid FAQ (in the section Tracing Simulations
2647 * We moved to cmake as default build system. Autotools support will
2648 be dropped soon. Check the FAQ for more info about how to use it.
2649 * Greatly improved our cdash/ctest interactions
2650 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2651 * Added memory checking tests with valgrind; lot of memleak fixing.
2652 This may be the first release of simgrid with so few memory issues
2653 * Added code coverage tests.
2654 Our coverage is still improvable, but at least we see it on cdash.
2656 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2658 SimGrid (3.3.4) stable; urgency=low
2660 The "Desktop Grid needs love too" release (also called Xmas release).
2662 Models improvements:
2663 * Major speedup in the maxmin system solving by using lazy evaluation
2664 Instead of solving completely the maxmin system at each iteration,
2665 only invalidate (and recompute) the modified parts.
2666 This new feature is enabled in default models but you can try to
2667 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2668 * Cas01 IMproved as default CPU model
2669 This CPU model is the same Cas01 model, but it uses the
2670 maxmin-selective-update flag and a heap structure to manage
2671 actions on SURF kernel.
2672 It reduces the complexity to find the next action to finish and,
2673 consequently, it's faster than the old Cas01.
2674 This is the new default CPU model (Cas01).
2675 * Rename the old Cas01 model to Cas01_fullupdate
2676 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2677 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2678 * CpuTI (CPU Trace Integration)
2679 A new CPU model whose objective is simulate faster when using
2680 availability trace files.
2681 Instead of using a full featured, over engineered maxmin system for
2682 CPU modeling, this model does the pre-integration of traces files
2683 to calculate the amount of CPU power available, and so, executes
2684 faster than the old CPU models.
2685 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2686 * Use LV08 as default network model since it gives better accuracy
2687 for small messages and shouldn't change things for big ones.
2688 Use --cfg=network_model:CM02 to get the previous behavior.
2691 ******************************************
2692 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2693 ******************************************
2694 * The new CPU model may changes simulations!
2695 The point is that events occurring at the exact same timestamp
2696 are not scheduled in the same order with the old and new
2697 version. This may be enough to completely change the execution
2698 of simulations in some cases.
2699 * The new network model will change simulations!
2700 This new model is more realistic than the previous one, so you
2701 should consider redoing your old experiments with this model.
2702 Sorry for the inconvenience.
2705 * Introduce the supernovae compilation mode
2706 When compiled that way, the whole SimGrid (or almost) is put in a
2707 single compilation unit and compiled in one shoot.
2708 This is to help gcc which has difficulties to inline stuff from one
2710 The speedup seem to be above 15%, althrough more tests are needed on
2711 amd64 to confirm that gain.
2714 * Port of MSG's mailbox on top of SIMIX network
2715 The put/get mechanism was greatly simplified on the way.
2718 * New SIMIX network module. Provides:
2719 - Mailbox: rendez-vous mechanism to find with who you want to speak
2720 - Synchronous send/recv: easier and hopefully faster since the
2721 logic is handled in the maestro process directly now
2722 - Asynchronous send/recv: you dreamt of it? It's here now
2723 Too bad that nobody cared enough to propagate the change to MSG.
2724 * Add semaphores as SIMIX synchronization mechanism.
2727 * new function SD_daxload(char*) to load a DAX file
2728 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2729 * Introduce typed tasks. Specify its kind and cost at creation.
2730 At scheduling, just give where it should be placed, and the cost
2731 for each involved resource is automatically computed.
2732 Existing constructors so far (more to come of course):
2733 - SD_task_create_comm_e2e() for end-to-end communication
2734 - SD_task_create_comp_seq() for sequential computation
2735 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2736 * new function SD_task_dump() for debugging display
2737 * new function SD_task_dotty(task,FILE*) writing to file the info
2738 about the task in dotty format
2739 * SD_task_dependency_exists() can now cope with having one of its
2740 arguments NULL. If so, it tests whether the other argument has any
2742 * Add getters on list of preceding/following tasks:
2743 SD_task_get_parents(task) and SD_task_get_children(task)
2744 * Add getters on amount of workstations and list:
2745 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2746 * Add getter on task kind: SD_task_get_kind(task)
2747 * Update the start_time and finish_time of tasks on completion/failure
2748 * Bugfix: Remove task from state swags when destroyed
2751 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2754 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2755 dynar with one separate thread per value of the dynar.
2756 * Change the prototype of xbt_thread_create(), sorry.
2757 Added a boolean parameter indicating whether we want to join this
2758 thread (used in SG only for now)
2759 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2762 * GTNetS wrappers should now be usable again (and betterly tested too)
2763 * Fix a major regression from 3.2 where the timeout provided to
2764 MSG_task_put_with_timeout() was used as absolute time before which
2765 the comm should be done.
2766 * Start to fix the <cluster> tag.
2767 - Internal links should be good now (beside of the loopback, which
2768 use the private link instead)
2769 - paths to the external world is still rather broken
2770 - the <route:multi> tag is just broken. Actually that's brain-dead.
2771 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2772 to make it less stupid
2773 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2774 * Fix a source-level compatibility glitch from 3.2: after defining
2775 MSG_USE_DEPRECATED, you can use the old name
2776 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2777 * Allow to compile from the SVN with automake 1.11
2778 * Fix some problems when using the "start_time" tag in deployment XMLs.
2779 * Fix #8569: XBT/synchro.h has redundant declarations
2780 * Fix #8563: MSG return values and exceptions
2781 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2782 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2783 Thanks to Robson Peixoto.
2784 * Fix a vicious bug in dictionaries inducing that some elements were
2785 not freed on xbt_dict_free()
2787 Portability report of this version:
2788 * Main portability targets:
2789 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2790 on (amd64/i386/ia64)
2791 - mac leopard on i386
2792 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2793 but nothing critical.
2794 * Other platforms: windows, AIX and others were not tested for this release
2796 Timing report of this version:
2797 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2798 scenario parameters). From 8h to a few seconds in desktop grid settings.
2799 * Supernovae brings about 25% speedup on i386.
2801 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2803 SimGrid (3.3.3) stable; urgency=low
2805 The "Need for Speed" release.
2807 The timings done to validate the 3.3.2 were faulty.
2808 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2810 The problem was a conversion from a manually handled vector to
2811 xbt_dynar_t on the critical path.
2812 xbt_dynar_foreach calls functions, inducing stack management crap.
2814 We inlined these functions and xbt_dynar_foreach is now breath taking.
2815 We also inlined xbt_swag_belong on the way.
2817 Here are some approximate speedup measurements (on master/slaves
2818 simulations lasting between 10s and 20s each):
2819 3.3.1 -> 3.3.2: about same performance
2820 3.3.2 -> 3.3.3: 40% speedup
2821 3.3.1 -> 3.3.3: 40% speedup
2822 3.3.1 with inline patch -> 3.3.3: 30% speedup
2824 Our reading is that the refactoring which occurred in 3.3.2 made us
2825 suffer much more from the xbt_dynar_foreach low performance, but
2826 once we solved this, this refactoring proved to be very performance
2827 effective. From the 40% speedup, somehow, 10% are due to the
2828 inlining and 30% to the refactoring.
2830 That's a pitty that gcc cannot inline functions placed in other files
2831 alone. We have to choose between:
2832 - break the encapsulation (by putting private data structures and
2833 accessors in headers files to help gcc)
2834 - live with low performance
2835 - switch to a decent compiler such as icc (not quite possible).
2837 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2839 SimGrid (3.3.2) stable; urgency=low
2841 The "Simplicity does not preceed complexity, but follows it" release.
2843 The main contributors of this release were (lexical order):
2844 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2847 * Extract the routing logic into its own object.
2848 (was dupplicated in network.c and workstation_LV07.c;
2849 Allows to implement other ways of storing that info)
2850 => kill now useless network_card concept
2851 - Use dynar to represent routes (instead of void** + int*)
2852 - kill link_set (use surf_network_model->resource_set instead)
2853 - Add a command-line option to choose the routing schema to use
2854 - Add three new models:
2855 * Floyd (shortest path computed at initialization)
2856 * Dijikstra (shortest path recomputed all the time)
2857 * Cached Dijikstra (shortest path computed on need)
2858 All these models where contributed by Silas De Munck, and are
2859 described in his ICCS09 paper.
2861 * Simplify model declaration
2862 (less redirections, less function to write when defining a model)
2863 - Factorize stuff between models:
2866 surf_model_resource_set(model)
2867 surf_model_resource_by_name(model, name)
2868 - Unify the types of models in s_surf_model_t (using an union)
2869 - Embeed fields of common_public directly into s_surf_model_t
2870 - Rename model methods:
2871 action_free ~> action_unref
2872 action_change_state ~> action_state_set
2873 action_get_state ~> action_state_get
2874 - Change model methods into functions:
2875 (model)->common_public->action_use ~> surf_action_ref
2877 * Implement a generic resource; use it as ancestor to specific ones
2878 (allows to kill duplicated code in models)
2879 Drawback: timer command don't need no name nor properties;
2880 workstation_CLM03 don't need no properties
2881 (but I guess we can live with those few bytes wasted)
2883 * Improve the action object model
2884 - implement a constructor avoiding dupplicated code about field
2885 initialization in generic_action part.
2887 * Kill the SDP model: it has an external dependency, is deprecated
2888 in flavor of modern lmm models, and didn't compile since a while
2891 * Relocation of the context module from XBT to SIMIX.
2892 (the context were decoupled from the simix processes, duplicating a lot of code)
2893 => a lot of code was factorized
2894 - less overhead is introduced during scheduling
2895 - simpler API for the context factory
2896 - the logic for process creation,destruction and manipulation was simplified
2897 * Simplification of the s_smx_process_t data structure.
2898 => accesing the simix level data associated to a process is faster now,
2899 and the code is a lot more readable.
2902 * Implement some more MPI primitives:
2903 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2904 -implementation: Bcast: flat or 2-ary tree (default),
2905 Barrier: 4-ary tree,
2907 Allreduce: Reduce then Bcast
2908 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2909 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2910 Alltoallv: flat tree, like ompi
2912 * Add support for optimized collectives (Bcast is now binomial by default)
2913 * Port smpirun and smpicc to OS X
2916 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2917 Shout out if you used it.
2920 * Display the list of still queued messages in SG mode when existing
2924 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2925 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2926 * Add xbt_str_from_file(FILE*)
2927 * Add xbt_dict_get_key achieving a linear reverse search
2928 * Remove the context module
2930 Portability report of this version:
2931 * Main portability targets:
2932 - Linux(debian)/x86/context
2933 - Linux(debian)/x86/pthreads
2934 - Linux(debian)/amd64/context
2935 - Linux(debian)/amd64/pthreads
2936 On these, we still have the eratic breakages of gras/pmm and
2937 amok/saturate_sg reported in previous version. We still think
2938 that the tests are the cause of the fault, not the tested code.
2940 - Mac OSX Leopard/x86/context
2941 Still false negative in tesh autotesting.
2942 Smpi still fails, but this time because readlink does not accept -f
2943 Everything seems to work properly beside of that.
2946 - AIX version 5.3 (only tested contexts this time)
2947 Smpi still fails there because mktemp is not installed.
2948 Everything seems to work properly beside of that.
2950 I managed to compile it for the first time, but several breakages.
2951 Won't delay the release for this exotic platform.
2953 * Windows: it's still lagging behind. If you want to help, please
2956 Timing report of this version:
2957 This version seem to be more than 5% faster than 3.3.1 (on linux
2958 64bits with contextes). The gain is less than expected, we are
2959 investigating this for next release.
2961 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2963 SimGrid (3.3.1) stable; urgency=low
2966 * Implement a --cfg-help to show existing configuration variables
2967 * Build chain do not require doxygen in maintainer mode
2970 * fix a bug on struct sizeof computation, which prevented the
2971 exchange of arrays of structs in some conditions
2972 - added a regression test about this in datadesc_usage
2973 * Allow the exchange of 0-long dynamic vectors.
2974 - for that, use -1 as indicator of dynamic size instead of 0
2975 - This implied to change any size from unsigned long to long,
2976 reducing a bit communication abilities, but I guess that with
2977 64bits being quite common, this is more than enough.
2978 - This also induce a protocol change, thus bumping network protocol
2979 version from 0 to 1 (if we have external users, we have to get
2980 clean on that point too ;)
2981 - added two regression tests about this in datadesc_usage
2982 * Be more verbose when propagating local exceptions
2983 This helps debugging.
2984 * Display the status of simulated processes when receiving SIGINT in
2988 * Allow to control the simulation from a trace file.
2989 New functions MSG_action_register() and MSG_action_trace_run()
2990 The first one allows to associate a function execution to each
2991 kind of action while the second one parses a trace file and
2992 triggers the corresponding actions within the system.
2993 For now, only a toy example is provided in examples/msg/actions
2994 * Add an example of process migration in examples/msg/migration
2995 * Fix a bug in task exchange which broke MSG_task_get_sender()
2996 Add a teshsuite regression test for that.
2997 [Bug: if MSG_task_get_sender() is called after sender exit,
2999 * Fix a bug which prevented suspend/resume to work properly
3000 * Display the status of simulated processes when receiving SIGINT
3001 This fixes a regression of v3.3. due to the introduction of SIMIX
3002 * Bug fixing in failure management:
3003 - trace could not start by a failure at time 0
3004 - failure during communications were not working
3007 * Add SIMIX_process_set_name() to change the name of the current
3008 process in the log messages.
3009 * Store smx_hosts in a dict since we only retrieve them by name
3010 * Move the configuration infrastructure to surf
3013 * Move the configuration infrastructure to surf
3016 * Massive internal cleanups:
3017 - Store internal structures on processes instead of hosts (allows
3018 to have more than one process per host, in addition of being more
3020 - Cleanup the initialization/finalization process
3021 - Kill a whole bunch of unneeded synchronization:
3022 processes run in exclusive manner within the simulator
3023 - Move queues from global tables to process data fields
3025 - now accept -platform and -hostfile arguments
3026 - Pass the right rank value to processes according to the hostfile
3027 * Compile the examples by default, and use them as regression tests
3028 * Implement MPI_Wtime()
3029 * Change the reference speed to a command line option
3032 * TCP_gamma can now be specified as command line option using
3033 --cfg=TCP_gamma:10000000.0
3034 * Change the --surf-path cmd line option into --cfg=path:
3037 * Also include strbuff from xbt.h public header
3038 * xbt_ex_display(): do not free the exception after displaying
3039 This allows to do more with the given exception afterward.
3040 Users should call xbt_ex_free() themselves.
3044 Portability report of this version:
3045 * Main portability targets:
3046 - Linux(debian)/x86/context
3047 - Linux(debian)/x86/pthreads
3048 - Linux(debian)/amd64/context
3049 - Linux(debian)/amd64/pthreads
3050 These targets fail about 1/10 of times on gras/pmm, but we believe
3051 that this is because of the test, not because of simgrid.
3052 amok/saturate_sg fails even more rarely, and the test may not be
3055 - Mac OSX Leopard/x86/context
3056 The test suite still spits tons of errors because some obscure
3057 force prevents us from removing the temporary directories
3058 arguing that they still contain some metadata I've never heard of.
3059 Smpi fails because seq is not installed.
3060 Everything seems to work properly beside of that.
3063 - AIX version 5.3 (both contexts and pthread)
3064 Smpi still fails there because mktemp is not installed.
3065 XML inclusions seems rosty on AIX.
3067 * Windows: it's still lagging behind. If you want to help, please
3070 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
3072 SimGrid (3.3) stable; urgency=high
3076 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
3077 [Malek Cherier & Mt]
3079 * Introduce the SIMIX module: factorize code between MSG and GRAS.
3082 Until now, GRAS were using MSG as an interface to SURF. It was
3083 quite difficult because both interface have several differences
3084 (MSG channels vs GRAS sockets were the most notable point).
3086 This also opens the gate to SMPI (which should occur soon) and speed
3087 up simulations by to 40% (even if it were not the main goal).
3089 **************************************
3090 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
3091 **************************************
3092 The point is that events occurring at the exact same timestamp are
3093 not scheduled in the same order with the old and new version. This
3094 may be enough to completely change the execution of simulations in
3095 some cases. Sorry for the inconvenience.
3097 * Cleanup and upgrade the XML format to push further scalability
3098 issues (check http://hal.inria.fr/inria-00256883/ for more info)
3100 * Improve the testing infrastructure with tesh. Now a very large part of
3101 the code is tested not only by being run but also by checking that the
3102 output match an expected output [Mt].
3104 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
3105 is really less memory-demanding, which should allow you to use
3106 larger files in SimGrid [AL].
3108 * Inform valgrind about our contextes, so that it becomes usable
3109 with the default (and more effecient) version of SimGrid
3110 [contributed by Sékou Diakite, many thanks]
3113 * Introduce a listener thread in charge of receiving incoming
3114 messages from the network. It allows to overlap communication and
3115 computation but most notably, it removes some stupid deadlocks due
3116 to the fact that so far, a process could not send and receive at
3117 the same time. This made most non trivial communication schema
3119 * Convert the PIDs from long int to int to match the MSG ones (and
3120 linux ones too) [Mt]
3121 * New function: gras_agent_spawn() to launch a new process on
3122 current host. Only working in simulation for now. [Mt]
3123 * New function: gras_os_hostport() returning a constant form (ie,
3124 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
3127 * Make the backtrace of exceptions more human readable [Mt]
3128 * New module: xbt/str [Mt]
3129 a ton of string utility functions (split, join, printf to a newly
3130 allocated buffer, trim, etc)
3131 * New module: xbt/hash [Mt]
3132 SHA1 hashing algorithm (more to come if needed)
3133 * New module: xbt/synchro [Mt]
3134 synchronization tools (mutex and conditions) working the same way
3135 in simulation and in real life (mainly useful for GRAS, but not
3137 * New module: xbt/queue [Mt]
3138 classical producer/consumer synchronization scheme
3139 * xbt_dynar_new_sync() creates a synchronized dynar. All access
3140 (using the classical functions will get serialized) [Mt]
3141 * Make dictionary internal table dynamic. No need to specify its size
3142 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
3144 * Make sure the log channels are organized as a tree under windows
3145 (because of ANSI C compatibility issue, any channel were child of
3146 root directly) [Mt].
3149 * Cleaned many thing in surf and fixed a few bugs [AL].
3150 * Add a nice command line configuration mechanism to compose models [AL].
3151 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
3152 the previous one (KCCFLN05). It relies on something that looks like
3153 a max-min sharing mechanism but cannot be written as such. A new solver
3154 was thus designed [AL].
3155 * Add a new solver to lmm. Based on Lagrange optimization and
3156 gradient-based descent, it enables to efficiently maximise systems s.a
3158 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
3160 This solver enables to propose two new network models for TCP Reno and
3161 TCP Vegas based on Low's work. These models still need to be fully
3162 tested though [Pedro Velho].
3165 * Bug fix in SD_simulate. Now the time bound given as argument is
3167 * Use the new parallel task model (ptask_L07) as default.
3168 * Use the SURF command line configuration mechanism.
3169 * 0-size tasks (for synchronization) should now work.
3171 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
3173 SimGrid (3.2) stable; urgency=high
3177 We still experience issues on this platform, but we believe that at
3178 least MSG is usable.
3180 GRAS API BREAKAGE (for simplification purpose, sorry):
3181 * the gras_msgtype_by_name is not used anymore. Instead of
3182 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
3183 you can write (and must)
3184 gras_msg_send(toserver, "request", &request);
3185 - If you still want to pass a gras_msgtype_t to the function (to cache
3186 the type and avoid the lookup time), use the gras_msg_send_() variant.
3187 - Impacted functions:
3188 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
3189 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
3190 * The callbacks are now expected to return 0 when everything went well
3191 (just like the main() function)
3193 GRAS new features and improvements:
3194 * New module mechanism where user code can use per process globals [Mt]
3195 This is similar to gras_userdata_*() functions, but for libraries. It
3196 factorize some code developped over and over in the examples and AMOK.
3197 It has still to be documented and used (only amok/peermanagement is
3199 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
3200 loss when they were small enough to fit into the buffer and sent quickly
3201 enough so that they can all get received in one shoot.
3202 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
3203 if not found. Use the *_or_null() variant for the old semantic.
3204 * In gras_msg_handle, do not discard messages without callback.
3205 They are probably messages to be explicitly awaited later (ie, proofs of
3206 mis-synchronization in userland since they are sent before being awaited)
3207 No big deal usually.
3208 * gras_socket_meas_send/recv: semantic changed!
3209 The numerical arguments used to be (1) the total amount of data to send
3210 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
3211 messages. This was need for the fool willing to send more than MAXINT
3212 bytes on quite fat pipes.
3215 * Do really rename the hostmanagement module to peermanagement. [Mt]
3216 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
3217 but this is rather new and this was documented in the module
3218 documentation (poor excuses, I admit)
3219 * Bandwidth measurement semantic changed! This follows the changes to
3220 gras_socket_meas_send/recv explained above.
3223 * A sequential mode has been added to the workstations. When a workstation
3224 is in sequential mode, it can execute only one task, and the other tasks
3225 are waiting in a FIFO. [Christophe Thiery]
3228 * The KCCFLN05 workstation model now handles parallel tasks. It is the
3229 model for SIMDAG. [Christophe Thiery]
3230 * Bug fix in the maxmin solver: Some values were close to 0 instead of
3231 equal to 0, which caused some bad behaviors in
3232 saturated_constraint_set_update. I now use a threshold mechanism like in
3236 * When running manually src/testall, you select specific units [Mt]
3237 testall is the result of our cunit mechanism, and should replace all
3238 the scripty thingy around since bash don't run easily on billware.
3240 * A mallocator system has been added. [Christophe Thiery]
3241 Mallocators allow you to recycle your unused objects instead of freeing them
3242 and allocating new ones.
3244 Documentation update:
3245 * FAQ reworking + New FAQs:
3246 - "Valgrind spits tons of errors!" [Mt]
3247 - "How to repport bugs" [Mt]
3248 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
3249 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
3250 - Communication time measurement within MSG [AL]
3251 - I experience weird communication times when I change the latency [AL]
3252 * GRAS tutorial [Mt]
3254 - an introduction to the framework and to the used communication model
3255 - an initiatic tour introducing the most proheminent features:
3257 . Lesson 0: Installing GRAS
3258 . Lesson 1: Setting up your own project
3259 o Part 2: Message passing
3260 . Lesson 2: Exchanging simple messages
3261 . Lesson 3: Passing arguments to the processes (in SG)
3262 . Lesson 4: Attaching callbacks to messages
3263 . Lesson 5: Using globals in processes
3264 . Lesson 6: Logging information properly
3265 . Lesson 7: Using internal timers
3266 . Lesson 8: Handling errors through exceptions
3267 . Lesson 9: Exchanging simple data
3268 . Lesson 10: Remote Procedure Calling (RPC)
3269 . Lesson 11: Explicitly waiting for messages
3270 . Recapping of message passing features in GRAS
3271 - A HOWTO section containing:
3272 o HOWTO design a GRAS application
3273 More are due, of course. They will come latter. In the meanwhile, you can
3274 check the examples which are still here.
3276 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
3278 SimGrid (3.1) stable; urgency=high
3282 There was a stack corruption somewhere, visible only when optimizing
3283 with these versions. [Vince]
3286 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
3287 module that enabled to play with Directed Acyclic Graphs. It is built
3288 directly on top of SURF and provides an API rather close to the old
3289 SG. Some old codes using SG are currently under rewrite to check that
3290 all needful functions are provided. [Christophe Thiery]
3293 * Complete rewrite of the KCCFLN05 workstation model. It is now an
3294 extension of the classical CLM03 model that gracefully handles
3295 failures. This is now the default model for MSG and GRAS. It doesn't
3296 handle parallel tasks yet though. [AL]
3297 * Bug fix: Weights were not correctly set in the network part.
3298 WARNING: This may have resulted in incorrect results with simulations
3299 where there are more than one flow on a given link. [AL]
3302 * After a (long ?) discussion on simgrid-devel, we have decided that the
3303 convention we had on units was stupid. That is why it has been decided
3304 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
3305 WARNING: This means that all previous platform files will not work as
3306 such with this version! A warning is issued to ask users to update
3308 A conversion script can be found in the contrib module of the CVS, under
3309 the name contrib/platform_generation/surfxml_update.pl [MQ]
3312 * Bug fix: Processes were started in reverse order, wrt deployment file.
3313 WARNING: if your code relies on this bug, please fix it. [AL]
3314 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
3315 being executed on two different locations. [AL]
3316 * Bug fix: Failures are now better supported thanks to Derrick's tests
3317 (there was many failure situations I hadn't thought of and that weren't
3318 correctly handled). [AL]
3319 * New function: MSG_host_is_avail indicates you whether a given m_host_t
3323 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
3324 Exception occurring on server-side are propagated back to client (!).
3326 API CHANGE: the callback changed their prototype. Change:
3327 int my_handler(gras_socket_t expeditor, void *payload_data) {
3329 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
3330 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
3332 * New! function: gras_msg_handleall to deal with all messages arriving
3333 within a given period.
3334 * New! function: gras_socket_server_range to get a server socket in a
3335 range of port numbers (ease to avoid port number conflicts) [MQ]
3336 * New! gras processes display their backtrace when they get a SIGUSR1
3337 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
3338 Sweet to debug RL processes [MQ]
3342 - Do not force experiment sizes to be expressed in kb, or it becomes
3343 impossible to measure the latency this way (needs one byte-long tests)
3344 WARNING: this changes the amok_bw_* function semantic. [MQ]
3345 - Implements the link saturation stuff. [MQ]
3346 * Peer management module:
3347 New! module factorizing code that we wrote over and over [MQ].
3350 * New module: cunit (my jUnit implementation in ansi C) [MQ]
3351 - Test units are placed directly into the library code, they get extracted
3352 automatically and placed into the src/testall binary.
3353 - Convert most of the XBT tests to this system.
3354 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
3355 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
3356 live in the same C code. This required to change a little bit the API
3357 of surfxml parsing but shouldn't be an issue for end-users. [AL]
3358 * New module: sparse graph structure with basic algorithms (this is work
3359 in progress and the API is not considered to be frozen yet). [AL]
3360 * Display more information on backtraces: source line & function names are
3361 now displayed just like valgrind does (rely on addr2line tool) [MQ]
3362 * New function: xbt_backtrace_display(). Sweet while debugging [MQ]
3363 * Reworked a little bit some #include statements to load only required
3364 headers. Some user code that relied on SimGrid to include stdlib or
3365 stdio may need to include it by themselves. [AL]
3366 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3368 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3369 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3370 but I guess that almost nobody used those parts. [MQ]
3372 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3374 SimGrid (3.0.1) stable; urgency=low
3377 * Unfortunately, I had missed 5 misnamed functions:
3378 xbt_fifo_item_t xbt_fifo_newitem(void);
3379 void xbt_fifo_freeitem(xbt_fifo_item_t);
3380 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3381 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3382 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3383 They're now deprecated. Please use their new versions:
3384 xbt_fifo_item_t xbt_fifo_new_item(void);
3385 void xbt_fifo_free_item(xbt_fifo_item_t);
3386 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3387 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3388 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3390 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3391 * Documentation: xbt_log module unmercifully reworked [MQ]
3392 * Bugfix: there was a problem with the ending of contexts with
3393 the pthread backend. It caused some weird deadlock or behavior
3394 depending on the pthread implementation. [AL]
3395 * Bugfix: get the exceptions raised in the simulator repport where
3396 and why they come from when they are not caught in time [AL, MQ]
3399 * Bugfix: Do repport the error when two non-connected hosts try to
3400 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3403 * Add additionnal checkings on communications. Assert that two
3404 communicating hosts are connected by a set of links... [AL]
3407 * Add additionnal checkings on channel values in communication [AL]
3408 * New: MSG_task_get_source to see on which host a task was generated [HC]
3409 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3410 the number of tasks waiting to be received on channel and sent
3412 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3413 waits for the first task coming from a given host.. [AL]
3415 GRAS new functionnalities: [MQ]
3416 * Enhance the parsing macro to allow the size of multidimentional objects
3417 to be given thru annotations.
3418 * New example (and documentation): Matrix Multiplication a la RPC
3419 (as when I was young!) and fix a bunch of bugs found on the way.
3421 GRAS performance improvements: [MQ]
3423 * Reduce the amount of cbps creation/destruction by making it static to
3424 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3427 * Change libdata to a set so that we can search for stuff by ID (and thus
3428 reduce the insane amount of dict lookups)
3431 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3432 RL communication on top of SG since it's so uneffective.
3433 It may also be used for inter-thread communication in RL, one day.
3434 * Use gras_datadesc_copy() to exchange messages on top of SG
3436 - improve message exchange performance on top of SG
3437 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3438 * Don't exchange on the network the size of the used part of buffer,
3439 instead, specify the possible buffer size to read().
3441 - reduces the amount of read/write calls (one pair per exchange)
3442 - reduces the amount of exchanged data (the size)
3443 - allows to retrieve all arrived data on receiver side, if we don't need
3444 it right now (subsequent read will peek the buffer)
3445 - allows the receiver to proceed with the beginning of the stream before
3446 everything is arrived
3447 - make it possible to build an iov transport (using readv/writev)
3449 - take care of the data with non-stable storage (like stacked data),
3451 * If possible, TCP send uses vector I/O (when writev() is here)
3452 - Don't use it for receive since we send data sizes and data on the
3453 same stream, so we wouldn't be able to chain large amount of chunks
3454 before having to flush the stuff to read the size.
3455 * Rework the transport plugin mechanism to simplify it and reduce the
3456 amount of pointer dereferencement when searching for the right function
3459 * I guess that now, we do almost as few system calls as possible while
3460 doing as few data copy as possible.
3462 To improve it further, we could try to send all the sizes first and then
3463 all the data (to use iov on receiving size), but it's only a partial
3464 solution: when you have 2 dimensional data, the sizes of the second
3465 dimension is data of the first dimension, so you need 3 streams.
3467 I'm not sure the potential performance gains justify the coding burden.
3469 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3471 SimGrid (3.00) stable; urgency=high
3474 * New! Give the possibility to hijack the surf parser and thus bypass
3475 MSG_create_environment and MSG_launch_application. Have a look at
3476 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3478 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3480 SimGrid (2.96) unstable; urgency=low
3485 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3486 This deprecates the xbt_error_t mechanisms.
3487 It modifies (simplifies) all XBT and GRAS API.
3488 MSG API keeps unchanged (exceptions raised by XBT are caught from
3489 within MSG and masked with existing error handling facilities)
3492 * New! Add a FATPIPE model. [AL]
3493 * New! Add a parallel task model. [AL]
3494 * New! Add automatically a loopback interface (in the default
3495 network model) if none was precised.
3498 * Bugfix: MSG_process_resume now works with the current running process.
3500 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3501 * Modification of MSG_task_get_compute_duration. Once a task has been
3502 processed, the value returned by this function is now equal to 0. [AL]
3503 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3504 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3505 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3506 of a m_task. For now, MSG_task_cancel only works with computation
3508 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3509 of the processor (in Mflop/s) regardless of the current load on the
3511 * API Change: use proper naming convention for MSG_getClock and
3512 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3514 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3515 This function changes the priority of a computation task. This priority
3516 doesn't affect the transfer rate. A priority of 2 will make a task
3517 receive two times more cpu power than the other ones. This function
3518 has been added to suit the needs of Nguyen The Loc and hasn't been that
3519 much tested yet. So if it fails, please report it and send me your code.
3521 * API Change: removed all functions and types that were marked "deprecated"
3522 since many months. Renamed MSG_global_init_args to MSG_global_init.
3524 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3526 SimGrid (2.95) unstable; urgency=low
3529 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3530 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3531 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3534 * Bugfix: Allow absolute paths for platform description files [MQ]
3535 * Bugfix: do free the variables after use. Leads to drastic performance
3537 * Implement max_duration (ie, timeouts) on resources [AL]
3540 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3542 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3543 the needed support (timeouts) [AL]
3546 * Implement measurement sockets. You can now get the bandwidth between two
3547 hosts thanks to AMOK (see below). [MQ]
3548 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3549 dynar over the network (yeah) [MQ]
3550 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3551 * simple/static token-ring example. [Alexandre Colucci and MQ]
3552 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3553 in place before [MQ]
3556 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3557 type parsing. "short" and "long" modifiers were also ignored; other
3558 modifier (such as reference level) are still ignored. [MQ]
3559 * Embeed the buffer size within the buffer itself on SG. [MQ]
3560 That way, send() are atomic and cannot get intermixed anymore (at least
3561 the ones which are less than 100k; bigger messages still have the issue)
3562 * Array size pushed by the field, not by the field type (or each
3563 and every long int will push stuff to the cbps) [MQ]
3564 * use select() to sleep since it allows to portably sleep less than one
3567 GRAS (minor cleanups)
3568 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3571 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3572 * type category 'ignored' killed as it was never used and were difficult
3574 * whether a type can cycle or not is now a flag, leaving room for more
3575 flags (as "ignored", if we feel the need one day ;)
3576 * Rename raw sockets to measurement sockets since "raw" has another
3577 meaning in networking community.
3580 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3581 and offers features not belonging to GRAS but that most applications
3582 need. One day, it may be a set of plugins loadable at runtime.
3583 * New module: bandwidth
3584 bandwidth measurement between arbitrary nodes running this module. [MQ]
3586 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3588 SimGrid (2.94) unstable; urgency=low
3590 The first beta release of SimGrid 3 !
3594 * Update the main page and the FAQ. Adding references to gforge.
3597 * Add a gras_os_getpid function.
3600 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3601 * Extend the logs so that they also print PID, hostname, date, ... if
3603 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3604 and kill the old version which were in testsuite/
3605 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3609 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3610 for users. Instead, each programming environment (GRAS, MSG,...) use it
3611 when needed to provide such a feature to users.
3612 Don't shortcut the mechanism or you will also shortcut the virtualization
3613 you need on the simulator.
3617 * Cleanups in configury with regard to compile optimization/warning flags.
3618 Also add -fno-loop-optimize to any powerpc since it's the optimization
3619 killing gcc (< 3.4.0).
3620 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3621 needed by MSG examples complications
3622 * Borrow configury beautifications from PHP
3625 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3626 warning (thanks loris for stumbling into this one).
3627 * Bugfix: stop loading private headers (gras_config.h) from the public
3631 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3632 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3633 RL would imply a DNS resolver, which is *hard* to do in a portable way
3634 (and therefore delayed).
3635 * Implement a real timer mechanism and use it in timing macros. This allows
3636 to avoid rounding errors and get a 0.000005 sec precision in timing
3637 macros. While I was at it, various cleanups:
3638 - allow to declare more than one timed section per file (fix a stupid bug)
3639 - move some private declaration to the right place
3640 - merge conditional execution and timing macros into emulation module
3641 - document the module
3642 - make sure the module cleanups its mess on gras_exit
3643 * Documentation improvements:
3644 - (new) how to compile applications using GRAS
3645 - (new) emulation support (timing macros)
3647 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3649 SimGrid (2.93) unstable; urgency=low
3651 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3654 - Use Paje properly where used. Still to be sanitized properly.
3655 - Portability fix: Add an implementation of the contexts using pthread
3659 - Add xbt_procname(): returns the name of the current process.
3660 Use it to show the current process's name in all logging.
3662 - fix detection of older flex version and the reaction, since we do
3663 depend on modern ones (we use lex_destroy)
3664 - Better separation of SG and RL in the libs: remove all simulation code
3665 from libgras. As a result, this lib is now only 200k when stripped.
3666 Some of the xbt modules may also be duplicated (two sets and such) and
3667 should be cleaned/killed before SG3.
3668 - Insist on using xlC on AIX because of weird problems involving gcc there.
3669 - Cleanup the make remote stuff. This is now done by scripts
3670 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3671 mainly for our private use, but we're working on changing them to user
3674 - Bugfix: flush the socket on close only if there is some *output*.
3675 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3677 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3680 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3682 SimGrid (2.92) unstable; urgency=low
3684 Alpha 3 on the path to SimGrid 3
3688 - New! First try of benchmarking macros.
3689 - New! First try so that gras_stub_generator generate deployment and
3690 remote compilation helpers.
3692 - Bugfix: Initialization fix in msg_test.
3696 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3698 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3700 - API Change: gras_os_sleep to take the amount of seconds as a double.
3701 Accepting an int was error prone since it was the only location where
3702 seconds were coded as such. It leaded to damn rounding errors.
3703 - Bugfix: Hard to belive that timers ever worked before this.
3705 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3707 SimGrid (2.91) unstable; urgency=low
3709 Alpha 2 on the path to SimGrid 3
3713 - Bug fix in the lmm_solver.
3715 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3716 through the function MSG_paje_output.
3717 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3718 - It is possible to bound the rate of a communication in MSG with
3719 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3720 to write it in the changelog).
3721 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3726 - Various cleanups to the autotools stuff
3727 - Begin to move Gras examples to examples/gras/
3728 - Let make distcheck work again (yeah!)
3730 - documentation overhauled using doxygen.
3731 gtk-doc-tools is dead in SimGrid now.
3732 - Automatically extract all existing logging categories, and add the list
3733 to the documentation (long standing one, to say the less)
3735 - Cleanup the known architecture table. Reorder the entries to group what
3736 should be, and use a more consistent naming scheme.
3737 (some of the test dataset are still to be regenerated)
3738 - New! Allow library to register globals on each process just as userdata
3740 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3741 lookup time (for now).
3742 Use it in msg and trp.
3743 This cleans a lot the internals and helps enforcing privacy of the
3744 headers between the gras components.
3745 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3746 - Bugfix: gras_os_time was delirious in RL.
3747 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3749 - Reenable GRAS now that it works.
3751 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3753 SimGrid (2.90) unstable; urgency=low
3755 Alpha 1 on the path to SimGrid 3
3757 * It is a long time since the last release of SimGrid. I'm sorry about
3758 that but as I had told you, I was rewriting a lot of things. I apologize
3759 to those who had been reporting bugs to me and that I had not answered.
3760 If your bug is still in the new version, please tell me. Here is a
3761 summary of the main changes.
3763 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3764 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3765 portability, speed, and a lot more but you'll figure it out later.
3766 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3767 I wanted to release the new MSG as soon as possible and I have
3768 broken GRAS, which is the reason why, for now, only MSG is fully
3769 functional. A laconic description of these projects is available
3770 in the documentation.
3772 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3773 kernel optimized for our needs (called SURF but only the developers
3774 should use it). Hence, MSG is now roughly 30 times faster and I think
3775 that by rewriting a little bit MSG, I could event speed it up a little
3776 bit more. Beside the gain in speed, it is also much easier to encode a
3777 new platform model with SURF than it was with SG. More to come...
3779 * REVOLUTION 3: I have tried to change a little as possible the API of
3780 MSG but a few things really had to disappear. The main differences
3781 with the previous version are:
3782 1) no more m_links_t and the corresponding functions. Platforms are
3783 directly read from an XML description and cannot be hard-coded
3784 anymore. The same format is used for application deployment
3785 description. The new format is described in the documentation.
3786 Have a look in tools/platform_generation. There is a tiny script
3787 that converts from the old platform format to the new one. Concerning
3788 the application deployment format, parsing the old one is tricky.
3789 I think most of you should however be able to convert your files. If
3790 it is really an issue, I can write a C code that does the conversion.
3792 2) the toolbox tbx does not exist anymore. We now have a library
3793 with much more data-structures but without the hash-tables (we have
3794 dictionaries that are much faster).
3796 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3798 *****************************************************************************
3799 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3800 * don't feel like converting the oldies. (Mt) *
3801 *****************************************************************************
3804 Version 2.90: "the long awaited one"
3805 - Finished rewriting and debugging MSG. Rewrote the documentation.
3806 - disable GRAS for now since it needs to be ported to the newest SG
3809 - Finish the port to windows (using mingw32 for cross-compile)
3812 - Main loop and datastructures of SURF. A cpu resource object is
3813 functional. Surf can thus be used to create cpu's with variable
3814 performance on which you can execute some actions.
3816 2004-11-15 Martin Quinson
3817 - Port to ARM. Simply added the alignment and size descriptions. Should
3818 work, but the ARM machines are so slow that I didn't had the opportunity
3819 to 'make check' over there yet.
3821 2004-11-15 Arnaud Legrand
3822 - Trace manager now written. It uses a heap structure and is therefore
3823 expected to be efficient. It may however be speeded up (particularly
3824 when many events occur at the same date) by using red and black
3825 trees. One day maybe...
3826 - Max-min linear system solver written. It uses a sparse matrix
3827 structure taking advantage of its expected use. Most operations are
3828 O(1) and free/calloc are called as few as possible. The computation of
3829 the minimum could however be improved by using a red and black tree
3832 2004-11-03 Arnaud Legrand
3833 - Rename every gras_* function that was in xbt/ to its xbt_
3835 - Add a heap and a doubly-linked list to xbt
3836 - Added a dichotomy to the dictionaries. make check works as well before
3837 so I assume that the patch is correct. I do not know however if things
3838 run effectively faster than before now. :)
3840 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3841 SimGrid, and the version number is bumped to 2.x
3843 2004-10-29 Martin Quinson
3844 - Introduction of the remote errors.
3845 They are the result of a RMI/RPC on the remote machine.
3846 ErrCodes being scalar values, you can't get the host on which those
3847 errors did happen. Extending the error mechanism as in Gnome is possible.
3848 No idea yet whether it is a good idea.
3850 2004-10-28 Martin Quinson
3851 - Interface revolution: the Starred Structure Eradication.
3852 I used to do typedef struct {} toto_t; and then handle *toto_t.
3853 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3854 - ???_t is a valid type (builded with typedef)
3855 - s_toto_t is a structure (access to fields with .)
3856 - s_toto is a structure needing 'struct' keyword to be used
3857 - e_toto_t is an enum
3858 - toto_t is an 'object' (struct*)
3860 typedef struct s_toto {} s_toto_t, *toto_t;
3861 typedef enum {} e_toto_t;
3862 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3863 s_toto_t) is private.
3865 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3866 so that it can set the variable to NULL. It was so for dicts and sets,
3867 it changed for dynars.
3869 - Fix a bunch of memleaks in dict_remove
3870 - Fix a bug in sg/server_socket opening: it failed all the time.
3872 2004-10-07 Martin Quinson
3873 - Speed up dynar lookup operation a bit.
3875 gras_dynar_get is dead.
3877 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3878 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3879 (which gives you the address of the stored data).
3881 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3882 copy of the data using an affectation to do the job and not a memcpy.
3884 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3885 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3887 It does not really speedup the dynar test because they are
3888 setting elements all the time (and look them seldom). But the dict does
3889 far more lookup than setting.
3891 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3893 2004-10-05 Martin Quinson
3894 - Allow to (en/dis)able the cycle detection at run time.
3896 Whether we should check for cycle or not is now a property of each
3897 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3898 datadesc_cycle_unset allow to remove this property when previously set.
3900 Note that the cycle detection is off by default since it impacts the
3901 performance. Watch the data you feed GRAS with ;)
3903 This property is hereditary. Any element embedded in a structure having it
3904 set have it set for the time of this data exchange.
3906 You should set it both on sender and receiver side. If you don't set it on
3907 sender side, it will enter an endless loop. If you forget on receiver
3908 side, the cycles won't be recreated after communication.
3910 - Header reorganization.
3911 Kill gras_private.h, each submodule must load the headers it needs.
3913 2004-10-04 Martin Quinson
3914 - Interface revolution: do not try to survive to malloc failure.
3916 Now, gras_malloc and friends call gras_abort() on failure.
3917 As a conclusion, malloc_error is not a valid error anymore, and all
3918 functions for which it was the only gras_error_t return value are
3919 changed. They now return void, or there result directly.
3920 This simplify the API a lot.
3922 2004-09-29 Martin Quinson
3923 - Re-enable raw sockets.
3924 Created by gras_socket_{client,server}_ext;
3925 Used with gras_raw_{send,recv}
3928 It should allow to kill the last bits of gras first version soon.
3930 This is not completely satisfactory yet (duplicate code with
3931 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3934 - Simplify transport plugin (internal) interface by not passing any
3935 argument to _server and _client, but embedding them in the socket
3938 2004-09-28 Martin Quinson
3939 - Finish the port to AIX.
3940 autoconf was my problem (segfault within the malloc replacement
3941 function. No idea why)
3943 2004-09-16 Martin Quinson
3944 - Fix some size_t madness on 64bit architectures.
3946 2004-09-08 Martin Quinson
3947 - Reduce the number of system headers loaded, overload some more system
3948 calls (such as malloc to cast the result of the system one, and work
3950 - Fix and reintroduce the config support
3952 2004-09-07 Martin Quinson
3953 - Source code reorganization to allow Arnaud to surf all over there.
3954 - Allow to document the logging categories.
3955 - Remove all uppercase from logging categories and useless cleanup in names.
3957 2004-08-18 Martin Quinson
3958 Version 0.6.2 (protocol not changed; API changed)
3959 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3960 gras_error_t), and NULL when not found. Functions expecting a msgtype
3961 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3962 hopefully usefull message.
3963 - Portability to prehistoric sparcs again
3965 2004-08-17 Martin Quinson
3966 Version 0.6.1 (protocol not changed; ABI not changed)
3967 - prealloc some buffers to speed things up
3969 2004-08-11 Martin Quinson
3970 Version 0.6 (protocol not changed; ABI expended)
3971 - The parsing macro can deal with the references, provided that you add
3972 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3974 2004-08-09 Martin Quinson
3975 Version 0.5 (protocol not changed; ABI changed)
3976 - Allow to off turn the cycle detection code in data exchange at
3977 compilation time. It should be at run time, but I'm short of time (and
3978 the config stuff is still broken). That way, we keep dict out of the
3979 critical path, which is good because the performance is poor:
3980 - search not dichotomial yet
3981 - dynar give no way to access their content and memcpy everytime
3982 - In composed data description (struct, ref and so on), stop foolness of
3983 keeping the subtype's ID, but store the type itself. This keeps sets out
3984 of the critical path, which is good since they rely on dynar and
3985 dictionaries. The only loose of that is that we cannot detect the
3986 redeclaration of a structure/union with another content (but I'm not sure
3987 the code detected well this error before anyway). We still can detect
3988 the redefinition discrepancy for the other types.
3989 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3990 it breaks the code because of type-punning used all over the place).
3991 This breaks on all non-gcc architectures (for now).
3993 All those changes (plus the buffer of last time) allow me to gain 2 order
3994 of magnitude on cruel tests consisting of 800000 array of integers on two
3995 level of a hierarchical structure (200 secondes -> 4 secondes)
3998 - the selector of reference must now return the type it points to, not
3999 the ID of this type.
4001 2004-08-06 Martin Quinson
4002 Version 0.4 (protocol changed; ABI not changed)
4003 - Allow to pass --gras-log argument to processes in simulation mode. Really.
4004 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
4005 - Implement a buffer transport, and use it by default (it relies on tcp in
4006 real life and on sg in simulation).
4007 That's a bit hackish since I had a new field to the structure to store
4008 its data without interfering with the subtype ones. Inheritance
4009 is tricky in C. And that's a kind of reverse inheritance with one class
4010 derivating two classes. Or maybe a game with java interfaces. Anyway,
4011 that's damn hard in C (at least).
4012 Moreover, I got tired while trying to ensure plugin separation and
4013 genericity in SG mode. MSG wants me to do weird things, so let's go for
4014 cruel hacks (temporarily of course ;).
4015 See comment in transport_private.h:71
4016 - do not include all the _interface headers in private but in the files
4017 which really need them (to cut the compilation time when they are
4020 2004-07-26 Martin Quinson
4021 Version 0.3 (protocol not changed; ABI changed)
4022 - Major overhault of the datadesc interface to simplify it:
4023 - shorted the function names:
4024 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
4025 - add a trivial way to push/pop integers into the cbps without malloc.
4026 This allows to make really generic sub_type description, which simply
4027 pop their size of the stack.
4028 - add a function gras_datadesc_ref_pop_arr() which does what users want
4029 most of the time: Declare a dynamic array (which pops its size of the
4030 stack) and declare a reference to it. Poor name, but anyway.
4031 - kill the post-send callback, add a post-receive one
4033 2004-07-23 Martin Quinson
4034 Version 0.2 (protocol changed; ABI changed)
4035 - add some testing for cpbs in the test cases, and fix some more bugs.
4036 This invalidate again the little64 data file, since I cannot regenerate
4038 - remove an awfull optimization in the logging stuff, allowing me to:
4039 - understand it again
4040 - learn gcc how to check that the argument match the provided format
4041 - fix all errors revealed by gcc after that
4042 - internal keys of dict are not \0 terminated. Deal with it properly in
4043 loggings instead of segfaulting when the user want to see the logs :-/
4045 2004-07-22 Martin Quinson
4046 - Fix some stupid bug preventing cbps (callback postit) from working
4048 2004-07-21 Martin Quinson
4049 - Some documentation cleanups
4050 - remove the useless last argument of msgtype_declare
4051 - rename the Virtu functions to fit into the 'os' namespace
4052 - move headers src/include -> src/include/gras/ and stop fooling with
4053 gras -> . symbolic link
4054 - make distcheck is now successful
4056 2004-07-19 Martin Quinson
4058 - Build shared library also
4059 - Install html doc to the right location
4060 - stop removing maintainer files in make clean
4061 - build tests only on make check
4063 2004-07-13 Martin Quinson
4065 - No major issue in previous version => change versioning schema
4066 - Re-enable little64 convertion test now that Abdou kindly regenerated the
4067 corresponding dataset.
4069 2004-07-11 Martin Quinson
4071 - Get it working with any kind of structure (we can compute the padding
4072 bytes remotely for all the architectures I have access to)
4073 - Implement the structure parsing macro (still not quite robust/complete)
4074 - Improvement to the remote testing toysuite
4076 2004-07-10 Martin Quinson
4077 [autoconf mechanism]
4078 - get ride of a bunch of deprecated macros
4079 - actually run the test for two-compliment, not only compile it :-/
4080 - test whether the structures get packed (and bail out if yes. Damn.
4081 Alignment is a serious matter)
4082 - test whether the structures get compacted (but respecting the alignment
4083 constraints of each types)
4084 - test whether the array fields of structures can straddle alignment boundaries
4086 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
4087 (took me 2 hours to find that bug, looking at the wrong place)
4088 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
4089 before sending/receiving and are used to compute the offsets of fields
4090 - Given that padding size depend even on compiler options, keep track of
4091 alignment and aligned_size only for the current architecture. Not a big
4092 deal since we send structure fields one after the other (seems
4094 - Add the datastructure used for IEEE paper by the PBIO guys to the test
4095 program, let it work on linux/gcc/little32. portability todo.
4097 2004-07-08 Martin Quinson
4098 - import and improve remote compilation support from FAST
4099 - make sure make check works on half a dozen of machines out there
4101 2004-07-07 Martin Quinson
4102 Let's say it's version 0.0.3 ;)
4103 - Implement conversions (yuhu!)
4104 - Let it work on solaris (beside conversion, of course)
4105 - Stupid me, using rand() to generate the conversion datatests in not wise.
4107 2004-07-06 Martin Quinson
4108 - Let make dist work, since I'm gonna need it to compile on remote hosts
4109 - Let Tests/datadesc_usage write the architecture on which the file was
4110 generated as first byte.
4111 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
4112 architecture descriptions.
4113 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
4114 on those architectures.
4115 - Optimization: send/recv array of scalar in one shoot
4117 2004-07-05 Martin Quinson
4118 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
4120 - Plug a whole bunch of memleaks
4121 - each process now have to call gras_{init,exit}. One day, their log
4122 settings will be separated
4123 - Continue the code factorisation between SG, RL and common in Transport.
4125 2004-07-04 Martin Quinson
4127 - Redistribution between SG and RL.
4128 We wanna have to accept in SG, so move accepted related parts of RL in
4129 the common part. (more precisely, the dynar of all known sockets is no
4130 more a static in transport.c, but part of the process_data)
4132 [gras_stub_generator]
4133 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
4135 2004-07-03 Martin Quinson
4136 - Create a new log channel tbx containing dict, set, log, dynar (to shut
4137 them all up in one shot)
4139 - Fix the ugly case of reference to dynamic array.
4140 - New (semi-public) function gras_datadesc_size to allow the messaging
4141 layer to malloc the needed space for the buffer.
4143 - gras_socket_close now expect the socket to close (and not its address to
4144 put NULL in it after it). This is because the socket passed to handlers
4145 is one of their argument (=> not writable).
4147 - propagate the interface cleanup from last week in datadesc, ie remove a
4148 superfluous level of indirection. User pass adress of variable
4149 containing data (both when sending and receiving), and not of a variable
4150 being a pointer to the data. Let's say that I like it better ;)
4151 The price for that is constructs like "int msg=*(int*)payload" in
4152 handlers, but it's a fine price, IMHO.
4154 - Let it work in RL (yuhu)
4156 2004-06-21 Martin Quinson
4158 - porting SG plugin and SG select to new standards (works almost).
4159 - plug memleaks and fix bugs around.
4162 - cleanup the prototype of data recv and force users to specify when they
4163 want to handle references to objects. Test case working even for cycles.
4164 - plug memleaks. Valgrind is perfectly ok with this.
4166 2004-06-12 Martin Quinson
4168 - cleanup the separation between plugin and main code in plugin creation
4170 2004-06-11 Martin Quinson
4172 - Reput hook for raw sockets, needed for BW experiments
4173 - kill a few lines of dead code
4174 [Data description] Interface cleanup
4175 - gras_datadesc_by_name returns the searched type or NULL.
4176 That way, no variable is needed to use a type desc once, which makes
4178 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
4179 two parameters were strings (field name, type name), leading to
4181 [Dicos] Interface cleanup
4182 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
4183 This is consistant with the dynar API.
4185 2004-04-21 Martin Quinson
4187 - Porting to new standards.
4189 - interface cleanup.
4190 There is no bag anymore, no need to take extra provision to mask the
4191 pointers behind "ID".
4192 Better splitup of functions between files create/exchange/convert.
4193 This is still a bit artificial since convert and receive are so
4194 interleaved, but anyway.
4196 - add a queued message list to procdata (the ones not matching criteria
4198 - factorize some more code between SG and RL wrt procdata
4200 - use gras_exit in example to track memleaks
4201 - get rid of gs_example now that GS is properly integrated into gras
4202 - update run_test to integrate the latest tests (datadesc)
4204 - rename WARNINGn macros to WARNn since it prooved error-prone
4206 2004-04-19 Martin Quinson
4208 - register init/exit functions within gras module mechanism
4209 - send/receive function.
4210 Convertion is not implemented, but short-cutted if not needed.
4211 struct/array elements are sent one by one (instead of block-wise), but
4212 nobody really cares (yet). Get a prototype before optimizing.
4213 - tests (using a file socket) for DD send/receive on:
4214 - base types: int, float
4215 - array: fixed size, string (ie ref to dynamic string)
4216 - structure: homogeneous, heterogeneous
4217 - chained list, graph with cycle
4218 Believe it or not, valgrind is not too unhappy with the results. The
4219 cycle happily segfaults, but the others are ok. And I'm sick of pointers
4223 - Bugfix when using a filename explicitly (instead of '-')
4225 2004-04-09 Martin Quinson
4227 - factorize more code between RL and SG in socket creation
4228 - Complete the implementation and tests of:
4230 o file (only in RL, and mainly for debugging)
4232 I lost 3 days to design a portable address resolver, and then decided
4233 that the prototype mainly have to run on my box.
4234 Addressing portability too early may be like optimizing too early :-/
4236 - use gras_init in the Tests instead of the crappy parse_log_opt
4237 (the latter function is removed)
4238 [Conditional execution]
4239 - New functions: gras_if_RL/gras_if_SG (basic support for this)
4240 [Code reorganisation]
4241 - Get rid of libgrasutils.a since it makes more trouble than it solves.
4242 Build examples against the RL library, since there is no way to disable
4243 its creation for now.
4245 For information, the beginning of coding on GRAS was back in june
4246 2003. I guess that every line has been rewritten at least twice since