1 SimGrid (3.18) NOT RELEASED YET (target: December 24 2017)
4 - Link::name() is deprecated and will be removed in v3.21, use
5 Link::getCname() instead.
6 - Mailbox::getName() changed to return a std::string, use
7 Mailbox::getCname() instead to get a char*.
8 - Storage::getName() changed to return a std::string, use
9 Storage::getCname() instead to get a char*.
12 - Deprecate MSG_task_get_flops_amount(). Its semantic was weird:
13 before the start it returned the total amount of flops;
14 after it returned the ratio of remaining work. Split it:
15 MSG_task_get_initial_flops_amount() for first behavior
16 MSG_task_get_remaining_work_ratio() for the second one.
17 This fixes GitHub's #223 using PR #237. Thanks Michael Mercier.
20 - Define class simgrid::xbt::Path to manage file names.
21 - Removed unused functions:
22 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
23 - xbt/str.h: xbt_str_join()
24 - xbt/heap.h: use std::priority_queue or boost::heap instead
27 - New link_energy plugin for the consumption of the links.
28 - All of the operations on files and storage contents have been
29 packaged into a plugin (src/plugins/file_system). The current
30 public interface can be found in
31 include/simgrid/plugins/file_system.h
32 To use these functions you now have to initialize the plugin by
33 calling MSG_storage_file_system_init() just after calling
34 MSG_init() or sg_storage_file_system_init() just after creating
38 - Remove the undocumented/untested tag <include>
41 - Remove viva specific tracing as the tool is no longer maintained
44 - #248: Kill of finished MSG processes leads to segfault
46 SimGrid (3.17) Released October 8 2017
48 The Drained Leaks release: (almost) no known leaks despite the tests.
50 Even very long simulations will preserve your memory: our testsuite
51 stresses SimGrid for over 45mn on fast machines for a coverage of
52 over 80%, and there is only one single known leak, of about 4kb.
56 - Comm.detach(): start and forget about asynchronous emission. A cleanup
57 handler may be given to free resources if the comm cannot be completed.
58 - this_actor::send(mailbox) is now mailbox->put()
59 - New: simgrid::s4u::Comm::wait_all()
60 - New: Host.execute() for remote executions.
63 - Unused option network/sender-gap is removed.
66 - Deprecate MSG_task_isend_with_matching(): This unused feature
67 really complicates our internals. Will be removed in v3.20.
70 - Improved context termination. It is now possible to free resources, even
71 when a process is forcibly killed.
72 - ContextBoost: add support for Boost versions above 1.61.
75 - Bring back run-time option --cfg=exception/cutpath to remove exception
77 - Removed unused functions:
78 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
79 xbt_str_rtrim(), xbt_str_trim().
80 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
83 - Removed header files obsolete since SimGrid 3.12:
84 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
86 - #3: SD_exit should be made optional
87 - #120: Memory leak when the processes are forcefully killed
88 - #159: Threading test regression in Actor refcounting
89 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
90 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
91 - #186: Actor::killAll() segfaults if some process is blocked on wait()
92 - #191: VM migration and pstate
93 - #192: Updating the energy consumptions of all hosts crashes with VMs
94 - #195: All actors have PID=0 in the logs
95 - #204: Sometimes segfault with thread contexts and mmap privatization
96 - #222: Actor::kill() doesn't really kill and segfaults
97 - #225: Actor::kill() doesn't really kill when victims are doing a join()
99 ----------------------------------------------------------------------------
101 SimGrid (3.16) Released June 22. 2017.
103 The Blooming Spring Release: developments are budding.
106 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
107 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
108 - Java JAR file should be finally fully working on Mac OSX too.
110 XML platforms: Switch to platform v4.1 format.
111 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
112 - <zone> can be used as a synonym for the now deprecated <as>
113 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
114 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
115 - <actor> can be used as a synonym for the now deprecated <process>
116 - state_file and avail_file periodicity is now easier to express
117 (check the documentation) the old behavior should still work.
118 * Storage: not backward compatible, but it seems that nobody ever used it
119 - Remove attribute 'content_type' of <storage_type>: was never used
120 - Make attribute 'model' of <storage_type> optional: for future usage
121 - Remove Bconnection model property: was never (in)validated. Replaced by
122 the maximum of the read and write bandwidth as a resource constraint.
125 - New and Backwards Compatibility break:
126 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
127 When one wants to get the list of tasks whose states have changed during a
128 simulation round, s/he has to allocate and free a dynar and use it as
129 argument to this function. The former SD_simulate (double how_long)
133 - Allow multicore VMs, with the correct sharing (unless you overcommit)
134 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
137 - New function to update the consumption of all hosts at once.
138 - Fix the model for multi-core hosts, linear in the amount of busy
139 cores with an abnormality for pIdle (see paper).
140 BUG: Still not satisfactory for ptask on mono-cores.
143 - The netzone are now available from the MSG API.
144 The old names still work, but are now deprecated.
147 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
148 - New option: smpi/keep-temps to not cleanup temp files
149 - New option : smpi/shared-malloc-blocksize . Relevant only when global shared
150 mallocs mode is used, allows to change the size of the fake file used
151 (default 1MB), to potentially limit the number of mappings for large runs.
152 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
153 - Fortran ifort and flang compilers support
154 - New RMA calls supported (experimental) :
155 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
156 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
157 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
158 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
159 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
160 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
163 - New: this_actor::isMaestro() returns whether we are in kernel mode.
164 - Behind the scene, ActivityImpl are now automatically refcounted.
165 Removing this nasty bottleneck will greatly simplify our internals.
168 - Replay: New function xbt_replay_action_get():
169 Retrieve the function previously associated to an event type.
170 - DROPPED FUNCTION: xbt_str_varsubst()
171 - DROPPED MODULE: strbuff. We don't need it anymore.
172 - DROPPED MODULE: matrix. We don't need it anymore.
173 - DROPPED MODULE: lib. We don't need it anymore.
175 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
177 SimGrid (3.15) Released March 22. 2017
179 The Spring Release: continuous integration servers become green
180 We fixed even the transient bugs on all target architectures:
181 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
182 Windows, FreeBSD, NetBSD.
184 - Rename NetCards to NetPoints.
185 This was intended to help NS3 users, but that's not a netcard.
186 That's a point in the routing algorithm, let's avoid wrong simplifications.
189 - New: MSG_process_yield(). Stop and yield to other processes.
190 - New: MSG_process_daemon(). Daemon processes are automatically killed
191 when the last non-daemon process terminates
192 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
193 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
194 to make room for the upcoming network energy plugin.
195 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
198 - Ensure that an actor can kill itself with Process::exit()
199 - Kill the obscure NativeException. Nobody want to survive the issues
200 it denotes, so use JniException that is a RuntimeException (not to
201 be caught explicitly).
202 - Partial bug fix in initialization. SimGrid flags on command line were
203 consumed at C level but stayed in the original Java String[] args.
204 This could mess users' args[i] if SG flags were not put at the end of
206 The SimGrid flags are now removed from the Java arguments. However,
207 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
208 if args.length is greater than the number of YOUR OWN ARGUMENTS.
209 It might be if you have --log or --cfg flags in the command line.
210 - Fix numerous memleaks all around the place. In particular, around VMs.
214 - s4u::onPlatformCreated: right before the simulation starts
215 - s4u::onSimulationEnd: right after the main simulation loop
216 - s4u::onTimeAdvance: right after a clock change (time is discrete)
217 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
218 event from the availability_file changes the avail speed.
219 - Links are now usable from s4u
220 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
221 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
222 - New examples: The conversion of MSG examples to S4U has begun
223 - Actors: create, kill, migration, and suspend
224 - Applications: master-worker and token-ring
225 - Action replay: communications and storage
226 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
229 - Backwards Compatibility breaks
230 - The SD_route_get_list and SD_route_get_size functions have been removed.
231 They are replaced by the unique
232 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
233 The route from src to dst is built in the links dynar whose size can
234 be obtained with xbt_dynar_length.
235 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
236 sg_host_route_bandwidth and sg_host_route_latency.
237 Macros ensure the backwards compatibility, but you should fix your code
240 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
241 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
242 - Bump our claim of support from MPI 1.1 to MPI 2.2.
243 We don't support 100% of it, but it should be enough. Contact us if not.
244 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
245 naive and probably inaccurate)
246 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
247 mapped onto a unique small file using some system magic.
248 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
249 - Fortran: cleanups, fixes, support of user-added operations
250 - RMA: MPI_Accumulate are applied in correct order.
251 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
252 - Better support for MPI_IN_PLACE
253 - Support for MPI_Win attrs and keyvals.
254 - Support MPI_Comm_group_create, MPI_Type_size_x
255 - MPI ops should be properly applied to all allowed types
256 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
257 algorithm by a non failing one from mpich.
260 - Kill the fifo data container: we don't use it anymore.
263 - A <cluster> can now be created with different speed values to represent pstates.
264 see examples/platforms/energy_cluster.xml for an example.
266 SimGrid (3.14.159) Released December 28. 2016
268 The previous release was only a crude approximation of a Pi release;
269 this one should be a bit better.
271 - Revert a last minute change that broke on Mac OSX.
272 - Fix the changelog and NEWS files.
273 - Improve the documentation, in particular of the routing module.
274 - Integrate some patches flying around in Debian.
276 SimGrid (3.14) Released December 24. 2016
279 * The whole documentation was reworked and reorganized. There is
280 still much room for improvement here, but we're on it.
283 * We now need python3 for our testing infrastructure.
284 * Model checking needs libevent
287 * Now works on FreeBSD
290 * Privatization now works on FreeBSD
292 * Privatization is now activated by default in smpirun
293 You can turn it off with -no-privatize if something goes wrong.
295 * Call-location tracing for SMPI.
296 You can add the exact location (filename / linenumber) of an MPI call to
297 your trace files and slow down or speed up the simulation between two
298 consecutive calls by using an adjustment file (see the documentation).
300 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
301 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
302 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
303 values were ignored in some cases. The timings of these functions can now
304 be significantly different.
306 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
307 smpi/running-power is renamed to smpi/host-speed
309 * smpi/grow-injected-times option to enable or disable multiplication of the
310 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
311 make simulation less precise (but also much faster).
313 * smpirun script should be (much) faster for large deployments.
315 * SMPI tracing : fixed issue with poor matching of send/receives.
317 * Replay : Fix broken waitall
319 New functions and features
320 * MSG_parallel_task_execute_with_timeout, to timeout computations.
322 Dropped / renamed functions and features
323 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
324 - MSG_mailbox_is_empty() -> Mailbox::empty()
325 - MSG_mailbox_front() -> Mailbox::front()
326 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
327 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
328 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
329 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
330 * Don't pass the free_f parameter to property related functions:
331 - MSG_host_set_property_value()
332 - MSG_as_router_set_property_value()
333 - MSG_storage_set_property_value()
334 * VM properties. Since msg_vm_t are msg_host_t, just use
335 MSG_host_get_property_value() and friends
336 * VM I/O related things:
337 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
338 - Unimplemented save/restore methods
339 * MSG_as_router_get_property_value() was redundent with
340 MSG_environment_as_get_property_value().
341 - Removed MSG_as_router_*propert*() functions
342 - Added MSG_environment_as_set_property_value() for consistency
343 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
344 Well, they are still there for now, but deprecated with a warning.
345 Please switch to xbt_dict_new_homogeneous() before this is removed
347 * Task affinity. Its intended behavior (that was very badly tested
348 and probably not really working) was deceiving what most users
349 would have hoped here.
350 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
351 * Option network/coordinates is now useless and should be dropped.
354 * Added option storage/max_file_descriptors to allow more than 1024 files opened
357 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
358 * Added Dragonfly topology support
361 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
362 EiBps, EBps, Eibps, Ebps for bandwidth.
363 They may become useful to some lucky ones.
366 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
367 * ex/app_tokenring: new example, very similar to the MSG Token Ring
368 * ex/async_waitAll: new example, on asynchronous communications
371 * Memory usage should be decreased for simulations with a large number
372 of processes. This also helps for SMPI.
374 SimGrid (3.13) stable; urgency=low
376 The Half Release, a.k.a. the Zealous Easter Trim.
378 - We removed half of the lines, that were mostly experimental cruft.
379 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
380 (not counting blanks and comments -- according to openhub.net)
381 - The internals are now compiled in C++ (and will soon be clean C++)
382 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
383 59 klines of C, etc. We added only 29 klines of C++ in replacement.
385 * Backwards Compatibility breaks
386 - Removed Lua simulation bindings (switch to C or Java for that).
387 Lua can still be used to describe platforms
388 - Removed Java kernel plug-ins.
389 Will be reintroduced after the ongoing major internals reorg.
391 - the following functions were removed.
392 They were too specific and should be reimplemented in a generic
393 way, with filter function.
394 - MSG_task_listen_from_host
395 - MSG_mailbox_get_count_host_waiting_tasks
396 - MSG_mailbox_put_with_timeout was removed.
397 Please use MSG_task_send_with_timeout instead.
399 - the SD_application_reinit function was removed. It has been a noop for a while.
400 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
401 replaced by a more flexible mechanism.
402 - The following functions thus do not exist anymore
403 - SD_workstation_get_access_mode
404 - SD_workstation_set_access_mode
405 - SD_workstation_get_current_task
406 - Basic estimation functions have been removed but can easily be replaced
407 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
408 - SD_workstation_get_computation_time => amount / sg_host_speed()
410 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
411 Use VM.getSpeed()*load/100 for the legacy behavior.
413 - option enable_tracing was removed. It was not doing anything for a while.
414 - In the ModelChecker:
415 - the model-checker now ptraces the model-checked process which means
416 you cannot use a debugger on the latter anymore (we might make this
417 optional in the feature);
418 - removed soft-dirty page tracking;
419 - remove model-checked side snapshot management,
420 MC_snapshot() and MC_compare_snapshot();
421 - keep the MC_cut() function as a stub (it was not really working
422 in the previous release).
425 * All options are consistently in kebab-case. Old names are kept as alias.
428 * Switch to platform v4 format.
429 - Rename from 'power' to 'speed' the attributes describing the amount of
430 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
431 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
432 - In <host> and <link>, attributes availability and state are gone.
433 It was redundent with state and availability traces, and with peak values.
434 - In <cluster>, attributes availability_file and state_file are gone.
435 It was too complex and unused.
436 - Kill <gpu>. Was not doing anything.
437 - The DOCTYPE points to the right URL:
438 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
439 (the file at this address now documents the changelog since its v1)
440 - A warning is emitted for unit-less values (they are still accepted).
441 - speed. Default: 'f' or 'flops'. Also defined:
442 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
443 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
444 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
445 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
446 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
447 - latency. Default: 's' second. Also defined:
448 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
450 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
451 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
454 * s4u::Host is now the preferred public interface to the Host features.
455 sg_host_* functions are C bindings to the exact same behavior
456 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
459 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
460 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
461 They were deprecated since a few years, and probably did not even compile anymore.
464 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
465 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
466 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
467 encouraged to update their codes. A list of the modified functions can be found at the end of
468 include/simgrid/simdag.h
471 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
472 the lambda or closure passed as a parameter will run in kernel mode.
473 All the callback functions should be rewritten to that interface at some point.
476 * Reorganizing and cleaning the internals all around the place.
479 * Remove old default barrier/bcast buggy algorithms (see #18407)
480 * Various bug fixes to handle more codes
481 * Remove the need for the --foreground option of smpirun (it is still
482 accepted for backward compatibility).
485 * Kill the setset data container: MC don't use it anymore.
486 * Kill the queue data container: it made more sense with GRAS.
487 * Kill the xbt_peer_t data type: it's useless without GRAS.
488 * Kill rm_cb feature of config sets: it was never useful.
489 * Kill graphxml parsing feature. It was not used.
490 * Kill the deprecated code protected by XBT_USE_DEPRECATED
492 - xbt_dynar_sort_strings(), when the content is char*
493 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
494 They throw exceptions on invalid input;
495 * C++ support for declaring CLI flags (simgrid::config::Flag);
496 * class for abstracting different signal backends (simgrid::xbt::signal).
497 with no external dependencies (we need very simple signals).
500 * refactoring and cleanup of the code;
501 * ongoing process to cleanly separate the model-checking algorithms
502 from the code model-checking support.
504 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
506 SimGrid (3.12) stable; urgency=low
508 The Facelift Release.
511 * Require g++ v4.7 at least to not speak prehistorical C++.
512 * Require Boost 1.48 (for signal2 component).
513 * Java must be version 7 at least when activated.
514 * Builds on Windows again (including Java bindings).
515 * Tracing is now always enabled (no way to turn it out)
516 * Remove GTNetS. It was not working anyway.
517 * Various cleanups in the cmake scripts.
518 * Move headers around to sort them out on installed systems:
519 - instr/instr.h -> simgrid/instr.h
520 - instr/jedule/* -> simgrid/jedule
521 - simdag/datatypes.h was removed
522 - simdag/simdag.h -> simgrid/simdag.h
523 - msg/datatypes.h was removed
524 - msg/msg.h -> simgrid/msg.h
527 * Interface improvement:
528 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
529 - Sanitize the interface in MSG_task_ module:
530 - Merge two functions that were close enough but misleading:
531 set_compute_duration(t) -> set_flops_amount(t)
532 get_remaining_computation(t) -> get_flops_amount(t)
533 - set_data_size(t) -> set_bytes_amount(t)
534 get_data_size(t) -> get_bytes_amount(t)
535 - Massive cleanups in the functions related to the energy
536 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
537 - New: MSG_host_get_pstate()
539 - msg/energy/onoff: switching hosts on and off
541 * Interface improvement:
542 - Rename Host.isAvail() to Host.isOn()
543 - Rename Process.currentProcess() to Process.getCurrentProcess()
544 - Rename Task.setDataSize() to Task.setBytesAmount()
545 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
547 - #18874: Actually allows the GC to reclaim tasks
551 - SIMIX_process_throw: raises an exception in a remote process
552 * Refactoring: Separate sub-modules
553 - libsmx: the public interface, as libc in a real system
554 - popping: the strange dance that converts a user request into a kernel handling
555 - smx_context_*: the virtualization mechanisms that embed the user code
556 - smx_*: the handling of each simcalls
558 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
559 * Rename smx_action_t into smx_synchro_t, making explicit that these
560 things are used to synchronize processes with their environment.
561 For example, a communication is a sort of synchronization involving
562 the communicating processes (that may block until the exchange) and
563 the platform. The same can be said from computations, etc.
565 - #18888: segfault when a process ends before its kill_time
569 - 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.
570 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
571 - MPI_Type_set_name, MPI_Type_get_name
572 - MPI_*_c2f and MPI_*_f2c functions
573 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
574 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
575 - Activate a lot of new tests from the mpich 3 testsuite
577 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
578 - InfiniBand network model added : Based on the works of Jerome Vienne
579 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
580 - When smpi/display_timing is set, also display global simulation time and application times
581 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
582 * Collective communications
583 - 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
584 - 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/).
585 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
587 - Replay now uses algorithms from wanted collective selector
588 - Replay can be used with SMP-aware algorithms
589 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
590 - 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.
592 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
593 - Correctly use loopback on fat-tree clusters
594 - Asynchronous small messages shouldn't trigger deadlocks anymore
595 * Energy/DVFS cleanup and improvement
596 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
597 - new: smpi_get_host_pstate()
600 - "Full" network optimization flag was broken since Surf++
601 - Better handling of precision flags in maxmin
602 - Fix bug causing sometimes "Impossible" errors
603 - Properly pass cluster properties to included hosts
604 * Improvement of the Energy plugin.
605 - Always update the consumption before returning that value
606 - New property: watt_off to denote the dissipation when the host is off
607 - New functions getWattMinAt and getWattMaxAt to retrieve the
608 dissipation of pstates that we are not currently at.
609 * Java: class NetworkLink renamed to Link
610 * New function: simcall_process_get_kill_time()
611 * Massive rename s/workstation/host/
612 - That's intrusive, but that's good for the project consistency. Sorry.
613 - Change config option "workstation/model" into "host/model"
616 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
617 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
618 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
619 - xbt_dirname and xbt_basename for non-POSIX systems
621 * The model checker now runs as a separate process.
622 * The model checker runs must now be launched with the new simgrid-mc program.
623 * Record/Replay: the MC can display a textual representation of a path in the
624 execution graph. It can then be replayed outside of the model checker.
626 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
628 SimGrid (3.11) stable; urgency=low
633 * Normalizing pointers addresses tool for better diff between logs
636 * Add cloud examples using new VMs
637 - examples/msg/cloud/two_tasks_vm.tesh
638 - examples/msg/cloud/simple_vm.tesh
639 - examples/java/cloud/cloud.tesh
640 - examples/java/cloud/migration/migration.tesh
641 * Add java surf examples:
642 - examples/java/surfPlugin/surf_plugin.tesh
643 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
644 - examples/java/surfCpuModel/surf_cpu_model.tesh
645 * Add SMPI+MSG example:
646 - examples/smpi/smpi_msg_masterslave/
651 - msg task destroy cancel test
652 - msg_host on/off test
653 * Move all tests in testsuite to teshsuite (adding tesh files)
654 * Restructure teshsuites
655 - one folder for each kind of test
656 * Restructure AddTests.cmake
658 - structure the order of tests (with sections)
661 * Add virtual machine
662 - creation of a VM on a PM
663 - migration of a VM from a PM to another PM
665 - MSG_process_join(msg_process_t process, double timeout)
666 - msg_bar_t MSG_barrier_init(unsigned int count)
667 - int MSG_barrier_wait(msg_bar_t barrier)
668 - void MSG_barrier_destroy(msg_bar_t barrier)
669 - msg_as_t MSG_environment_as_get_by_name(const char * name)
670 * New option "msg/debug_multiple_use" to help debugging when a task is used
673 - Improvements and finalization of MSG_storage, MSG_file APIs and their
675 - Increase code coverage in test suites
679 * Protect context stacks against stack overflow. The number of protected memory
680 pages allocated on the top of each stack (1 by default) can be configured
681 with the new command line option --cfg=contexts/guard_size:<value>.
682 * Simcalls are now generated by a python script that
683 - generates files included by SimGrid
684 - checks that all the functions exist, and proposes prototypes
686 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
687 host_set_data, host_get_data
689 - simcall_process_join(smx_process_t process, double timeout)
690 * Fix bug where sleeping processing could not be suspended.
693 * Translate surf models from C to C++
694 - Generic classes for all models: Model, Resource, Action
695 - A generic interface for each kind of model (CPU, Network, Storage
696 Workstation, WorkstationVM)
698 * Translate surf routings from C to C++
699 * Add callbacks using sigc++ or boost::signals2
700 - Add callback functions for resource creation/destruction
701 - Add callback functions for action state change
702 - Handle Energy as a plugin
703 * Replace swag by boost::intrusive
704 * Add new routing models for clusters. For documentation, see
705 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
706 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
707 parameters for cluster tag
708 - Fat trees, with topology="FAT_TREE" and
709 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
710 - see examples/platforms/torus_cluster.xml and
711 examples/platforms/fat_tree_cluster.xml
715 * Hostfiles support host:nb_processes construct to deploy several processes on
717 * Collective communication algorithms should not crash if used with
718 improper number of nodes and report the error.
719 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
720 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
721 MPI_Dims_create, MPI_Cart_sub are supported.
722 * New interface to use SMPI programmatically (still depends on MSG for
723 some parts, see examples/smpi/smpi_msg_masterslave) :
724 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
728 * Global variables privatization in MPI executables is now performed at runtime
729 with the option smpi/privatize_global_variables (default:no).
730 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
731 dynamic libraries loaded by the application are not privatized (static
732 linking with these libraries is advised in this case)
735 * Options defined in XML work correctly now.
738 * New cmake option, enable_lib_in_jar, to control whether native libraries are
739 copied into simgrid.jar or not (ON by default). Use this option if you want
740 to reduce the size of the installed simgrid.jar, *and* the native libraries
741 are kept installed elsewhere.
742 * Surf binding with SWIG (code generated in maintainer mode only):
743 - plugin to handle callbacks
744 - CPU model only for the moment
747 * Supernovae build mode is definitively removed. It was used to improve
748 inlining and inter-module optimizations. It is nowadays superseded by
749 link-time optimizations commonly available in compilers.
750 * Update ns-3 find lib. Bindings for ns-3 should work again now.
751 * Add boost dependency for surf++
752 * Add new macro for tests
753 - ADD_TESH(name <tesh_args>)
754 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
758 - xbt_bar_t XBT_barrier_init(unsigned int count)
759 - int XBT_barrier_wait(xbt_bar_t barrier)
760 - void XBT_barrier_destroy(xbt_bar_t barrier)
761 * Make the xbt_os_time module public
763 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
765 SimGrid (3.10) stable; urgency=low
767 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
770 * Reintegrate Java to the main archive as desynchronizing these
771 package is not acceptable anymore (Java is now considered stable)
772 * Add explicit synchronization facilities through semaphores
773 * Bug fix: Task.setDataSize() only changed the C world, not the value
774 cached in the Java world
777 * Dramatically change the way files are handled. API and internals changed, but
778 this part of MSG was not considered as production grade either.
779 * Add explicit synchronization facilities through semaphores
780 * Add a new function MSG_host_get_process_list()
781 * Preliminary DVFS support (see examples/msg/energy/ for details)
784 * SMPI is now included directly in the libsimgrid as the windows
785 linker doesn't force us on splitting it anymore.
786 * Improvements of the SMPI replay tool:
787 - Most of the collective communications are now rooted in the same process as
788 in the original application.
789 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
790 used until now). Multiple data types can now be used in a trace.
791 - The replay tool now supports traces produce either by TAU or a modified
793 - Bug Fix: the compute part of the reduce action is now taken into account.
794 - Gatherv collective is now supported
795 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
796 of smpirun outputs time independent traces for the current run. One file
797 is created per process. If too many processes are simulated, this behavior
798 can be changed to one file for all processes by using the
799 tracing/smpi/format/ti_one_file flag
800 * smpirun generates the host file if needed (with given host count and platform)
801 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
802 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
803 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
804 --cfg=smpi/coll_selector:(mpich/ompi)
805 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
807 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
808 happening inside a collective SMPI call.
809 * Fix the behavior of complex data types handling.
810 * Make MPI_Wtime another synchronization point to take computations into
812 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
813 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
814 * Add all missing Fortran bindings, SMPI should work with Fortran 90
815 (no privatization of global variables yet)
816 * Preliminary DVFS support (see examples/smpi/energy/ for details)
819 * Verification of liveness properties is now available for SMPI applications
820 (in addition to MSG applications)
821 * Bugged examples using SMPI in examples/smpi/mc/
822 * Add --cfg=model-check/visited option. Allows the verification of infinite
823 programs. Detection of loops in the execution thanks to the system state
824 comparison and reduction of the state space to explore. Can be combined with
825 DPOR for safety properties.
828 * Allow to change SimGrid configuration (see --help) within the code
829 thanks to SD_config() as it can be done in MSG.
830 * Add a new function SD_task_set_amount() upon user request.
833 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
834 * Remove rule based routing (no more PCRE dependency)
835 * Add a limiter_link option to cluster tag, to specify a maximum reachable
836 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
837 * Add a loopback_bw and loopback_lat options to cluster tag.
838 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
839 routing. Now peers are encapsulated in an AS and have their own private
840 router but this is transparent.
843 * Our own implementation of getline is renamed xbt_getline, and gets
844 used even if the OS provide a getline(). This should reduce the
845 configuration complexity by using the same code on all platforms.
846 * New type: xbt_cfg_elm_boolean.
847 * Allow to use yes/no for boolean configuration options in the command line.
848 * Allow to disable SimGrid cleanups at exit from command line option.
849 There are situations where one may want a simulation to end with an exit.
850 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
851 annoying when scripting around the simulator. Adding a
852 --cfg=clean_atexit:no allows to circumvent this issue.
855 * Lots of memory leaks were corrected in this release.
856 There is no known memory leaks anymore, in all of our 600+ tests.
857 * New command line option --version, to get SimGrid version information.
858 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
860 * Supernovae builds are deprecated, and expected to be removed in the next
863 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
865 SimGrid (3.9) stable; urgency=low
867 The Grasgory release: GRAS is really dead now.
869 * Complete overhaul of the internal host structures scheme.
872 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
873 was considered as experimental and badly maintained since then.
874 * Keeping it was thus a trap to our potential users, that could take
875 it instead of MSG or SMPI by mistake despite is pity state.
876 * GRAS seems to have very few users (if any), and no one volunteered
877 to maintain it further. It also induces a lot of XBT code (for
878 portability sake), that must be maintained too.
879 * For all these reasons, we killed GRAS. If someone wants to revive it
880 in the future, don't cry, our git history still remembers of GRAS.
883 * Major overhaul. Merge our documentation again as time proved that
884 splitting it was really not helping our users.
885 * Further improve the developer documentation to help newcomers
886 hacking on SimGrid itself. The user documentation (and in
887 particular, the beginner documentation) is still in a sorry state.
890 * Now works on Windows too!
891 * Much more extensive test suite, from MPICH
894 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
895 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
896 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
897 * Bug fix: let task be scheduled when the last dependency to be solved is
898 a control dependency.
899 * Remove SD_load_environment_script function.
900 Use the C sg_platf function if you want to declare a platform
904 * New function: MSG_process_get_number()
905 * Old function documented: MSG_config()
906 * Remove MSG_load_platform_script function
907 Use the C sg_platf function if you want to declare a platform
911 * Change the default value of the TCP_gamma constant (maximal size of TCP
912 congestion window) to a more realistic 4MiB value. If you notice changes in
913 your simulation results, you can fall back to the previous 20k tiny window
914 by adding --cfg=network/TCP_gamma:20000 on command line.
915 * (Hopefully) fix a bug wrt periodic availability and state traces
916 * Bug fix: use default values at start when first event in availability/state
917 trace is not at time 0.
920 * remove the "new_" part of function name sg_platf_new_trace_connect
921 (resulting in sg_platf_trace_connect), since it does not create
925 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
926 use case where it's really mandatory, and maintaining it was a pain in
928 * New: xbt_fifo_search(), search an item with a user-provided
929 comparison function instead of dumb pointer comparison.
932 * Fix the lua deployment:
933 Use `simgrid.init_application()` before deployment instead of
934 `simgrid.msg_register_application()` after.
937 * Transfer the tracing files into the corresponding modules.
939 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
941 SimGrid (3.8.1) stable; urgency=low
943 The "we are told that some people want to also *install* the simgrid
946 * Add missing file "tesh.1" to the archive.
948 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
950 SimGrid (3.8) stable; urgency=low
952 The Psssshiiiit release: SimGrid jumps into the Cloud.
955 * Add an experimental interface to manipulate VMs. They are mainly
956 process groups with very few intrinsic semantic, but they should
957 allow you to build the semantic you want easily.
958 * New function: MSG_host_set_property_value()
959 * New function: MSG_process_on_exit(). To clean memory in all cases.
960 * Bug fixes that made the host (and link) failures unusable.
961 * Add a way to auto-restart process when the host in which they are
962 executing comes back (ON_FAILURE="RESTART" on deployment file,
963 MSG_process_auto_restart_set).
964 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
965 please stop using the old ones, they are DEPRECATED.
967 * Deprecate functions MSG_global_init() / MSG_global_init_args()
968 Please use MSG_init() instead. (reducing the amount of entry
969 points in the library helps us).
970 * Make it impossible to link against the wrong version of the lib
971 * Deprecate MSG_clean(). No need to call it anymore.
972 * Function MSG_get_host_number() is not deprecated anymore.
975 * Split the doc into a user guide and a reference guide.
976 * Start a developper guide to help people hacking on SimGrid.
979 * Enable tracing by default. This modules rocks you should use it.
980 * Remove option custom_flags. Now use environment variables CFLAGS
982 * Use default cmake things to detect lua instead of home grown ones.
983 * New option "enable_mallocators" to disable mallocators, for debugging
984 purpose ("on" by default).
987 * Bug fixes around the resource failures: don't let the processes
988 survive the host they are running onto.
989 * Add an interface to auto-restart processes when the host in which they are
990 executing comes back.
991 * Ensures that SIMIX_clean is called automatically. It's not part of
992 the public interface anymore (bindings should be updated).
995 * Bug fix for when SD_Simulate is called with a positive value: be careful
996 when comparing doubles. Sometimes they are different for non significant
998 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
999 parallel task whose initial work is distributed among host according
1000 to the Amdahl's law. Such tasks are created with a parameter alpha
1001 that corresponds to the non-parallelizable part of the computation.
1002 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1003 between two sets of workstations assuming a 1D block distribution (each
1004 workstation owns a similar share of data) on both sides.
1006 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1007 Data redistribution will be automatically scheduled once parent and child
1008 are both scheduled. The filling of computation_amount and
1009 communication_amount structures is now done seamlessly thanks to the chosen
1011 * New function SD_workstation_dump to display various information
1012 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1013 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1014 bandwidth on the route onto which the task is scheduled and the amount of
1016 To divide the nominal bandwidth by 2, the rate then has to be :
1017 rate = bandwidth/(2*amount)
1018 * Compute tasks that have failed can now be rescheduled and executed again
1019 (from their beginning)
1020 * Increasing source code coverage (src/simdag is now covered at 95.8%
1024 * Re-implement time-independent trace replay using SMPI (at the
1025 smpi_smp_* level) instead of MSG. This should replace
1026 examples/msg/actions/actions.c
1027 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1028 hindexed and structs)
1029 * Implement the exchange of non-contiguous data.
1030 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1031 * Correct behavior of smpi/sender_gap and set its default value to 0
1032 * Add option to asynchronously send small messages to allow better
1033 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1034 specifies the size in bytes under which messages will be asynchronously sent.
1035 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1036 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1039 * New C interface to define a platform: XML is now optional.
1040 For more info, please check include/simgrid/platf.h.
1041 * New interface to define random platforms from the C:
1042 For more info, please check include/simgrid/platf_generator.h and
1043 examples/msg/masterslave/masterslave_platfgen.c
1044 * Export a sg_cmdline dynar containing all the arguments we got from
1048 * Two new tracing options for adding comments to trace file so you
1049 can track your experiments (see --help-tracing for details).
1050 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1051 * Adding the SimGrid version that generated the trace file as a comment.
1052 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1053 * Fix to avoid key clashes on Paje links
1054 * Other minor fixes related to the Paje specification
1057 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1058 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1059 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1060 * Greatly improve the robustness of mmalloc to user errors (such as
1061 using an area after freeing it, or freeing it twice)
1063 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1065 SimGrid-java (3.8.1) stable; urgency=low
1067 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1068 Warning: all methods are visible, but only some of them are
1069 implemented so far. Check the source (src/jtrace.c)
1070 for further information.
1071 * New module: org.simgrid.msg.File (SimGrid File management functions)
1072 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1073 * Change the meaning of Process.restart: now restart the process from
1074 the begining, like MSG_process_restart in C.
1075 * Add Process.setAutoRestart: handling of process restart when failed
1077 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1078 you to retrieve the properties of the processes/hosts
1079 * Deprecate Msg.clean(): you can just forget about it now.
1080 * New function Process.getCount(), that only works when compiling
1081 with the not yet released version 3.9 of the C library.
1083 * New context factory based on Coroutines. It mandates a modified JVM
1084 but then, the simulations run about five times faster, and there is
1085 no limit to the amount of processes (beside of the available memory).
1087 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1089 SimGrid (3.7.1) stable; urgency=low
1092 * Restore the prototype of MSG_process_create_with_environment() to
1093 the pre-3.7 situation by removing the kill_time argument.
1094 * Add a MSG_process_set_kill_time() function instead.
1097 * Fix weird behaviors when dealing with parallel tasks.
1100 * Simgrid is now built as a dll.
1101 * Simgrid-java now works on Windows.
1102 * Simgrid-Java is now included into Windows package.
1105 * First pre-build package for MacOSX.
1108 * Fix compilation when using MSG_USE_DEPRECATED.
1109 * Fix some compilation issues on Macs and Windows.
1110 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1111 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1113 We discovered that the Lua console is broken, but we are missing the
1114 manpower to fix it right now. The problem existed in 3.7 too, so we
1115 are not blocking the release for that. Sorry if you depended on this
1116 feature, any help would be really welcome.
1118 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1121 SimGrid-java (3.7.1) stable; urgency=low
1123 The "Java aint got to be bloated and slow" release
1126 * Various internal cleanups and performance improvement
1127 Simulations are expected to run up to twice faster or so
1128 * Make Process.kill(process) an instance method, not a static one
1129 * User processes are not java.lang.Thread subclasses.
1130 This breaks the compatibility (sorry), but previous API was
1131 brain-dead, making it impossible to have non-trivial
1132 initializations in the process constructor.
1133 * Require a full constructor per Process sub-class.
1134 Kinda breaks the compatibility (sorry), but this allows a much more
1135 efficient way to launch the processes at simulation startup.
1136 * Do not embeed our version of semaphores, java 1.5 can be considered
1137 as sufficiently prevalent for us to not dupplicate its features.
1141 * Add examples for almost every part of the API
1142 We spotted and fixed a lot of bugs in the process
1144 * New module: asynchronous communication API
1145 * New function: Process.sleep()
1146 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1147 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1148 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1150 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1152 SimGrid (3.7) stable; urgency=low
1154 The "spring cleanups (before the next Big Project kicks in)" release.
1157 * We can specify the SMPI latency/bandwidth factor with command line
1158 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1159 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1160 You can also use the "config tag" from platform file by adding this line
1161 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1162 (see "example/platforms/tag_config.xml" to use "config tag").
1163 Note that the command line supersedes the platform file configuration.
1164 * Change the correction factors used in LMM model, according to
1165 the latest experiments described in INRIA RR-7821.
1166 Accuracy should be improved this way.
1167 * Use the partial invalidation optimization by default for the
1168 network too. Should produce the exact same results, only faster.
1169 * Major cleanup in surf to merge models and split some optimization
1170 mechanisms from the core of the models. As a result you can now
1171 specify which model to use (e.g., --cfg=network/model:LV08
1172 --cfg=cpu/model:Cas01) and which optimization mode to use
1173 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1174 Incompatible combinations should err at initialization. See
1175 --help-models for the list of all models and optimization modes.
1176 * The CLM03 workstation model was dropped for simplicity because it
1177 used the deprecated CM02 network model. Use default instead.
1178 * Rename the TCP_gamma configuration option to network/TCP_gamma
1179 * Rename the coordinates configuration option to
1180 network/coordinates, and document it
1181 * Use now crosstraffic keyword instead of the terribly misleading
1182 fullduplex keyword. It is activated by default now in the current
1183 default model, use --cfg=network/crosstraffic:0 to turn it off.
1184 * Ongoing refactoring the model parsing to make XML files optional
1185 See include/simgrid/platf.h for details (still to be completed)
1188 * Major overhaul of the documentation. Almost instructive now :/
1189 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1190 functions and friends. This interface was considered as
1191 deprecated since over 2 years, it's time to inform our users that it is.
1192 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1193 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1194 use these (crappy) functions in your code.
1195 These functions will be removed soon. Stop using them now.
1196 * Deprecate MSG_get_host_{table,number}
1197 Implement MSG_hosts_as_dynar() instead.
1198 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1199 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1202 * Stabilize the parallel execution mode of user contexts
1203 * Introduce configuration variables to control parallel execution:
1204 - contexts/synchro: Synchronization mode to use when running
1205 contexts in parallel (either futex, posix or busy_wait)
1206 - contexts/parallel_threshold: Minimal number of user contexts
1207 that must be part of a scheduling round to switch to parallel
1208 execution mode (raw contexts only)
1209 * Fix bugs that prevented to use suspend/resume along with
1210 synchronization structures.
1211 * Fix bugs in process termination that lead to invalid memory access
1212 in very specific conditions.
1215 * Introduce a parallel mode for the models (controlled by surf/nthreads
1216 configuration item). In our tests, running the models in parallel
1217 never lead to any speedups because they are so fast that the gain
1218 of computing each model in parallel does not amortizes the
1219 synchronization costs, even when ultra fast futexes are used.
1220 This is released anyway because YMMV.
1223 * Performance boost by using a swag internally to compute the set of
1224 tasks that are finished and should constitute the return value of
1228 * Enable it by default now that it is considered rather stable.
1231 * Documentation of the tracing functions.
1232 * Performance gains when tracing categorized/uncategorized resource
1233 utilization by avoiding calls to get route when updating resource
1234 variables. LMM constraints are being used instead.
1235 * API changed to set task categories. Use MSG_task_set_category instead
1236 of TRACE_msg_set_task_category, and SD_task_set_category instead
1237 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1238 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1239 * Examples for link user variables added (see at examples/msg/tracing/)
1240 * Deprecated function TRACE_msg_set_process_category completely removed
1241 * Trace header updated according to the latest Paje file format
1242 * Tracing network lazy updates, no longer obligate users to use full updates
1243 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1244 * Experimental: let user code declare/set/push/pop application states for hosts
1245 * API changed to allow the manual creation of graph configuration files
1246 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1249 * Improve the API of Lua MSG bindings, using the Lua spirit.
1250 * Each simulated process now lives in its own Lua world (globals are
1251 automatically duplicated). It helps writing simulators. Will allow
1252 to run Splay programs within SimGrid in the future.
1253 * Add a Chord example in Lua, equivalent to the MSG one.
1256 * Start the implementation of a solution to express temporal
1257 properties, not only local assertions. This is still an
1258 experimental work in progress, stay clear from it to be safe.
1262 - Add new runtime parameters --help-logs and --help-log-categories
1263 to display information about supported logging parameters and
1265 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1266 don't exists anymore.
1267 * Mallocators: allow value NULL for the reset function.
1269 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1270 create homogeneous dictionaries, where all the elements share the
1271 same free function. Non homogeneous dictionaries will be
1272 deprecated in the next release.
1273 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1274 - Multi-level dictionaries are deprecated.
1276 - new function xbt_dynar_search_or_negative() that is useful when
1277 you have less than 2 million elements in your dynar and don't
1278 want of the extra complexity of catching exceptions when the
1279 element is not found.
1281 - Make xbt_os_thread module (for thread portability) public.
1282 Documentation is still to come, sorry.
1284 - Cleanups and simplifications to make it maintainable again.
1285 - Exotic features (such as memalign and valloc) were removed.
1286 - The metadata were extended and improved so that the
1287 model-checker becomes able to explore and inspect the heaps.
1288 - This may induce a performance drop when enable_model-checking is
1289 ON in cmake (even if it's not used in the simulation), but it is
1290 necessary at this point to get MC working.
1292 Turn model-checking OFF if simulation performance matters to you.
1293 Not enabling it at runtime is not enough, disable it in cmake.
1295 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1298 SimGrid (3.6.2) stable; urgency=low
1300 The "Not coding new stuff allows to polish old things" release.
1303 * New bindings to the NS3 packet level simulator (experimental)
1304 * Use the raw (efficient) execution contexts instead of the sysv
1305 (portable) ones when possible.
1306 * libpcre is now mandatory in any cases since not using it led to
1307 severe performance loss and possibly other issues
1308 * Update the XML platforms:
1309 - G5K: include the latest machine in Nancy
1310 - GridPP and LCG: new platforms
1311 * Documentation was partially updated, at least (more to come)
1313 Bug fixes, cosmetics and small improvements
1314 * Free terminated processes before the end of the simulation to avoid
1315 exhausting the memory of users having very dynamic amount of
1317 * Bug fix and cosmetics about canceling non-running tasks
1318 * Bug fix about the dot loader's issues when using libcgraph
1321 * Create an installer for windows with nsis (amd64 and win32)
1322 - Add an hello world project to illustrate simgrid project creation.
1323 - Embed libpcre into the Simgrid installer to avoid
1324 its compilation burden
1325 * The raw execution contexts should work on Apple now
1326 * Port to Windows 64 bits
1327 - Sysv contexts now have an implementation for this arch
1328 - GRAS communication features now support this arch
1329 * Drop support for borland compiler on windows
1330 - this code was not maintained, and we kinda depend on gcc nowadays
1331 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1332 * Fix portability issue on unstable ubuntu: linker became picky on
1335 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1337 SimGrid (3.6.1) stable; urgency=low
1339 The "Oops, we broke Macs too" release
1342 * Fixed contexts detection so that raw ones are used when possible
1343 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1344 leads to a strange error, with user code segfaulting sometimes when
1345 the generated code is not perfectly aligned (which is not
1346 controllable from the user side, depends on the amount of code)
1349 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1351 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1353 SimGrid-java (3.6) unstable; urgency=low
1356 * Split of every thing from simgrid v3.5 into a separate package.
1358 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1360 SimGrid (3.6) stable; urgency=medium
1362 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1365 * Bindings now constitute their own package, separated from the main one.
1366 Rationale: reduce our maintenance nightmare by reducing the module coupling
1367 They will soon be released on their own on gforge.
1369 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1370 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1372 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1373 * It's not quite deprecated for now because we have no replacement,
1374 but it may soon become the case.
1377 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1378 * Fortran: New user-level cache variable to store the rank of the running
1379 process. This improves performance by an order of magnitude.
1380 * C: New coccinelle script to automatically locate and modify global and
1381 local static variables.
1382 * Improved SMPI network model with a sender-side gap to account for multiple
1386 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1387 only say if a communication is finished, no matter whether it succeeded or
1388 failed. You can call MSG_comm_get_status() to know the status of a finished
1390 * New function MSG_task_dsend() to send a task and detach it. When a
1391 communication is detached, you are never notified of its success or failure
1392 and the memory is released automatically once it is finished. This function
1393 is useful when you don't care about the end nor the success of a
1395 * Change the prototypes of action replay. Sorry for inconvenience,
1396 but this is really more efficient this way (and to adapt your code,
1397 you just have to fix the initialization, that shouldn't be too long)
1398 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1399 ever managed to do anything useful with it.
1400 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1401 takes a dynar of communications. It returns immediately and gives the
1402 index of a finished communication (if any).
1403 * New example: a basic implementation of the Chord P2P algorithm.
1406 * New model for multi-core CPUs. You can now use the core attribute to
1407 precise the number of cores of a host. This is a basic model. Every
1408 process running on the host receives at most the power provided in
1409 the DTD (throughput<=power). Total throughput of process cannot exceed
1411 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1412 router linked by an up-link and a down-link (possibly asymmetrical).
1413 This kind of pattern allows to easily build last-mile model style platforms.
1414 Aggregating such patterns in a rule-based AS is thus the technique of
1415 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1416 * New model for Vivaldi routing. We transformed the Vivaldi network model
1417 into a Vivaldi routing model (based on the rule-based model). This allows to
1418 combine Vivaldi based latencies with last-mile platforms.
1421 * Added a check for NaN of IEEE754 infinite in the double entries of
1423 * Introduce a new context factory "raw", highly inspirated from the
1424 ucontext factory, but using manually crafted functions in assembly to
1425 do the work in an efficient manner.
1426 * Allow to change the used context factory at run time, not only at
1427 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1428 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1429 contexts at runtime (only with raw contexts or ucontexts).
1430 * Completely rewrote this module to allow parallel execution of user
1431 processes. Use --cfg=contexts/nthreads:N to execute user processes
1432 with N parallel threads (the default is 1, meaning no parallelism).
1433 * Allow to decide dynamically between sequential and parallel modes.
1434 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1435 processes in parallel only when their number is greater than or equal to P
1437 * Added a check for NaN of IEEE754 infinite in the double entries of
1441 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1442 won't output the state of processes when interrupted with Ctrl-C
1443 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1444 NULL-terminated array. This may solve backward compatibility issues
1445 due to the change to return type of SD_simulate. See also:
1446 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1447 * Add new macros with variable number of arguments.
1448 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1449 - in xbt/asserts.h: xbt_assert
1450 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1451 - in xbt/ex.h: THROWF and RETHROWF.
1452 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1454 * Change xbt_die() to accept a format string with arguments, just like printf.
1455 * New data structure: xbt_lib_t, like a dict but more general and with better
1459 * New configuration options
1460 Options triva/categorized and triva/uncategorized can be used to generate
1461 graph configuration files for Triva visualization tool.
1462 * Configuration option tracing/platform is renamed to tracing/categorized
1463 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1464 * New TRACE_user_link_variable interface:
1465 User provides the name of the link and the tracing variable to attach to it
1466 * the declaration of tracing categories must be done after the environment
1468 * simpler tracing interface, just one way to declare categories
1469 TRACE_category or TRACE_category_with_color, it is up to you
1470 * links in the trace file are again identified by their names
1471 * trace contains the full platform hierarchy exactly as declared using the ASes
1472 * Options tracing/msg/[task|process]:1 groups the process by hosts
1473 for both cases, tasks and processes must have names that are unique during the simulation
1474 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1475 * The experimental option tracing/msg/volume is deprecated
1476 its functionality may be reincorporated if needed
1478 The tracing generates a trace file with unordered timestamped events,
1479 because of the way the core simulator (surf) works. A script available
1480 at the tools directory (fix-paje-trace.sh) can be used to put the events
1481 in order. We have changed the tracing so it can generate ordered timestamped
1482 events in the final trace, but depending on the simulator (and how much time
1483 is simulated) that can lead to a huge memory utilization. It is deactivated
1484 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1486 Build Infrastructure
1487 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1488 - We are trying hard to keep the API stable, but it may happen that
1489 some things change (we're a research project after all, not a
1490 nuclear plant operating system). If such things should happen, you
1491 could rely on that macro to adapt.
1492 - current value: 30600 for 3.06.00, aka 3.6
1493 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1494 a number that can be compared with SIMGRID_VERSION.
1495 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1496 assertions and verbose/debug logging events are disabled at compile time.
1498 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1500 SimGrid (3.5) stable; urgency=medium
1503 * New feature to any SimGrid-based simulator: Model-Checking
1504 Check SIN#1 for more details.
1507 * New Model SMPI (three-interval linear regression for correction factors)
1508 See RR-7426, available at http://hal.inria.fr/inria-00527150
1509 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1510 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1511 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1512 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1513 * See also src/smpi/README
1518 - Tracing API changes: TRACE_start and TRACE_end should not be called
1519 by user-code. They are automatically called by simulators created
1520 with SimDAG, MSG and SMPI if the toolkit is compiled with
1521 tracing_enabled=ON. Categories declaration and utilization remain the
1522 same for MSG and SimDag.
1523 - A function was added to the tracing API to declare categories with
1525 - TRACE_category_with_color (char *category, char *color)
1526 where color must be in the following format
1527 "%f %f %f", red, green, blue
1528 and red, green, blue are float values in the interval [0, 1]
1529 - User can specify NULL as color parameter, or continue calling
1530 TRACE_category (cat)
1531 On that case, the tracing system will define random colors
1532 - The following command-line options are supported:
1533 --cfg=tracing/filename:msg.trace
1534 --cfg=tracing:1 (activate tracing, needed to use others)
1535 --cfg=tracing/platform:1 (categorized resource use)
1536 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1537 --cfg=tracing/msg/task:1 (task creation)
1538 --cfg=tracing/msg/process:1 (process creation, migration)
1539 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1540 --cfg=tracing/smpi:1 (SMPI interface tracing)
1541 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1542 - examples of examples/msg/tracing updated
1544 - DAXLoader and DOTLoader functions can generate tasks with categories
1545 - A new function to attribute a category to SD tasks:
1546 TRACE_sd_set_task_category (SD_task_t task, char *category)
1547 * Tracing the MPI interface implemented by SMPI
1548 - Collective operations are traced with states
1549 - Point-to-Point operations are traced with states/links
1550 - Tracing activated by a parameter "-trace filename" passed
1551 to smpirun during execution (considering that simgrid
1552 is compiled with tracing enabled)
1553 - To run the simulation with gdb, the simulator
1554 accepts --cfg=tracing/smpi:1 to trace SMPI
1555 - tesh files to check if smpi tracing is ok
1556 - See examples/smpi/NAS/DT-trace
1557 * GTNetS tracing re-worked
1558 - adaptation to the tracing system of GTNets to cope
1559 with modifications regarding the fullduplex mode
1560 - new tesh files to check if gtnets tracing is ok
1563 * Asynchronous communications through the functions:
1564 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1565 * New function: MSG_load_platform_script()
1566 to make possible using a lua script instead of XML files to set up platforms
1567 * New function: MSG_set_function
1568 to associate functions to processes, used when bypassing the parser
1569 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1571 Platforms: Add some more examples in examples/platforms
1572 * Grid'5000: see www.grid5000.fr
1573 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1576 * Change the XML format. This is a very important modification. SimGrid
1577 3.5 introduces a new hierarchical format based on the notion of
1578 Autonomous Systems. Compatibility with old format is ensured
1579 through the perl script provided in the install bin directory
1580 bin/simgrid_update_xml.
1581 It is now possible to build platforms with specific routing
1582 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1583 connect several platforms together. We will try to provide soon
1584 set of realistic platforms exploiting these properties (have a
1585 look at examples/platforms/ for the moment).
1586 * Take the opportunity of the XML format change to be a good XML citizen:
1587 rename link:ctn to link_ctn and similar changes (also dealt with by
1589 * Add a new routing scheme (rule-based) using regular expressions. It
1590 enables to have an extremely low memory footprint when the
1591 underlying routing is simple and can be compactly described. You
1592 need to have libpcre4-dev (perl regular expressions) installed if
1593 you want to use this routing scheme.
1594 * Revive the cluster TAG and allow to easily and efficiently (both in
1595 term of memory and speed) connect clusters together. Have a look
1596 at teshsuite/simdag/platforms/ to see how this can be done. With
1597 this tag, you can create clusters with thousands of tasks at no
1598 cost (have a look at examples/platforms/).
1599 Note: clusters are implemented as ASes, so there is no need for an
1600 enclosing AS tag if you have only one cluster in your platform.
1601 * Add new generic functions in the public interface that allows the user
1602 to call SURF 'create_resource' methods from your code (same
1603 functionality as the XML bypass mechanism but with a much lighter
1605 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1606 that uses a piecewise linear approximation to produce better
1607 results when exchanging small messages.
1608 * Add a new parameter to handle correctly full duplex link and account
1609 for interferences between uplink and downlink communications
1610 (activate with --cfg=fullduplex:1).
1613 * Rename the SD_READY (all dependencies are satisfied and task is
1614 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1615 dependencies are satisfied) state.
1616 This prevents a confusion between the notion of "ready to schedule"
1617 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1618 simulated" (SD_RUNNABLE) used by the simulation kernel.
1619 * Change the way a task is considered as ready. Instead of removing
1620 dependencies when a task is done, a counter is decreased. This way,
1621 it is always possible to reach ancestors thanks to the
1622 SD_taks_get_parents function (even after the end of the simulation.)
1623 * Change the return type of SD_Simulate from (SD_task_t*) into
1624 xbt_dynar_t. This function was in handling a dynar internally and
1625 converted it into a NULL terminated array for historical reasons.
1626 * New function SD_dotload(char*) to load a DAG described in dot
1627 format. This loader and the corresponding examples require the
1628 installation of the graphviz library.
1629 * Fix a bug in the management of tasks of size 0 in the surf network
1630 models. This problem was only visible with SIMDAG and you should
1631 thus disregard results produced with earlier versions if you
1632 relied on this feature (some tasks were blocked because of this).
1633 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1634 even though your applications were doing only point-to-point
1635 communications and sequential computations. Now you can really use any
1636 model you want (of course, if you create real parallel tasks, which are
1637 not implemented in most models beside ptaskL07, this will abort).
1638 * Add an example that schedules a DAX on an heterogeneous platform
1639 using a Min-Min strategy.
1640 * New function SD_workstation_get_current_task() that returns the kind
1641 of task currently running on a workstation in the sequential access
1643 * Raise some warnings when unexecuted tasks remains at the end of the
1644 simulation. This is usually caused by cycles in the DAG.
1647 * New function: SIMIX_process_set_function() called by MSG_set_function
1648 * Change the underlying waiting queue in semaphores so that a process
1649 can wait on several of them simultaneously (as in waitany).
1650 * Fix the way to handle tokens in semaphores so that all access patterns
1651 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1652 * kill the dirty pimple SIMIX_message_sizes_output()
1653 Please use (proper) visualization instead
1656 * New data container: setset (set of sets of elements)
1657 * New module: mmalloc (mapped malloc, allowing to have several
1658 independent segments of malloc)
1659 * New function: xbt_dict_cursor_set_data()
1660 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1661 * New function: xbt_dynar_is_empty()
1662 * New function: xbt_fifo_get_last_item()
1663 * Fix xbt_dynar_shrink(): use the right element size.
1664 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1665 uninitialized areas during expand.
1666 * Fix semaphores: previous implementation was severely broken.
1667 * Use library init/fini functions for our initialization.
1668 - you can use logs and other feature as soon as you want in your
1669 code (even before the xbt_init / MSG_init)
1670 - xbt_exit is now a no-op and produce a warning when used.
1673 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1674 benefit from the latest improvement to the simulation kernel.
1675 * Kill measurement sockets for now. If you rely on them, sorry. This
1676 release is not for you. This feature will be reintroduced in the
1677 future, but we cannot delay the release any further.
1678 * New function: gras_msgtype_get_name().
1679 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1680 * Fix (at last) the pmm example: it should not randomly fail anymore.
1682 Build chain: bug fixes and overall polishing
1683 * Cmake is now stable enough. Hence, we killed the autotools.
1684 * Port to windows ( TM :)
1685 * Fix the 'make install' target.
1686 No need to use 'make install-simgrid' anymore
1687 * Introduce a 'make dist' target compiling a *source* archive
1688 'make package' compiles a binary archive
1689 * Compile java files only on need
1690 * Add --cd and --setenv command line options to tesh
1691 * Out of source builds are not fully supported yet, but we are close
1692 * Enable supernovae and optimization flags by default for our users
1695 * Add layer to set up environment directly from lua, without XML.
1696 * The effect of gras_stub_generator can be achieved through
1697 lua too (check examples/gras/console/ping_generator.lua)
1699 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1701 SimGrid (3.4.1) stable; urgency=low
1703 The "Polishing easter eggs is probably a good idea" release.
1704 This is a bug fixes release only.
1708 * Fix a bug preventing the tasks from begin garbage collected.
1711 * Fix a bug occuring when a host involved in a communication fails.
1712 This was not detected properly by the other peer involved in the
1713 communication. Now, it's reported as a network error.
1716 * Warn the user about loop dependencies in data flow of DAX files
1717 * Obey the control-flow dependencies of DAX files
1720 * Add option "enable_smpi" allowing to not compile SMPI.
1721 Probably useful for the (Mac) users experiencing a build error here
1722 * Improve the detection of lua5.1 and ruby1.8
1724 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1726 SimGrid (3.4) stable; urgency=low
1728 The "Easter in Cargese" release. Also known as (major changes):
1730 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1731 ~> bindings were greatly improved
1732 ~> new tracing infrastructure for better visualization introduced
1734 * the "Welcome to configury modernity" release.
1735 ~> we switched from autotools to cmake, and improved our cdash
1738 A more detailled list of changes follow (full detail in svn log).
1740 Java Bindings: Various Cleanups
1741 * (install java-gcj-compat-dev on debian-like to use them)
1742 * Remove put/get: no need to export deprecated interface in Java
1743 Use send/receive instead.
1744 * Cleanup the examples and add a README per directory
1745 * Remove example autoDestination (that's the only way to go now)
1746 * Remove example explicitDestination (was a plain copy of basic)
1747 * Make JniException a runtime exception, so that there is no need to
1748 declare the fact that you may encounter such a beast. I guess that
1749 nobody will ever want to survive such error.
1750 * Create specific errors for each MSG case of failure:
1751 host failure, transfer failure, timeout, task cancelled
1752 * Cleanup the exceptions that may get thrown by each function
1753 * Other internal cleanups in Java bindings. Performance still bad :/
1754 Ruby and Lua Bindings: create them
1755 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1756 * That's new and great, you should try them out.
1757 Same functionalities than Java bindings, only even less polished
1759 * Kill the useless "rate" argument of SD_task_get_execution_time()
1760 Everyone used to provide -1 as a value, it was not used, and the
1761 semantic of a possible use wasn't even clear.
1762 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1763 either as comm costs or compute costs to mean that there is no
1764 such thing for that specific task.
1765 * Add a SD_task_set_name() function
1766 * Fix SD_task_unschedule() on typed tasks
1767 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1768 * In DAX loader, accept useless 'level' attributes to <job> since
1769 LIGO DAGs have them (seem to be to ease graphical representation).
1771 * Add an example masterslave_mailbox.c using send/receive and not
1772 the deprecated put/get interface.
1773 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1774 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1776 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1777 (declare MSG_USE_DEPRECATED to still have the old name)
1778 * Add a MSG_task_set_data() function
1779 * About trace replay (see examples/msg/actions):
1781 - Allow to work with splitted trace files for each process
1782 Give the specific trace file as argument of each process,
1783 and call MSG_action_trace_run(NULL)
1784 You can still have one merged file for all processes.
1785 - Fix implementation of collective operations
1786 * Allow task_execute() on 0-sized tasks (closes #10063)
1788 * This is the first release of SimGrid where SMPI is not considered
1789 beta anymore (even if some corners should still be improved)
1790 * Port over the new SIMIX_network submodule (internal refactoring)
1791 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1792 * Implement more missing elements of the standard:
1794 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1796 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1797 MPI_Scan() MPI_Get_processor_name()
1798 - Added implementation of missing case for Alltoall (warning: it's
1799 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1800 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1801 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1802 SMPI_MPI_Allgatherv()
1803 * Bug fixes include:
1804 - MPI_Waitsome() was broken
1805 - Allow relative includes in smpicc
1806 - Command line cfg argument 'reference_speed' was ignored...
1807 - Some functions did not properly lead to auto-benching of user code
1808 - smpicc passes -O2 by default (just like openmpi one)
1810 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1811 * Bug fixes about timeouts during communications
1812 * add SIMIX_message_sizes_output() as a pimple to write to file the
1813 amount of messages per size. Use gnuplot to get histogram.
1814 Pimple because that's the only user-visible function of simix,
1815 defined directly in xbt.h (irk, sorry)
1817 - Add a SIMIX_sem_get_capacity() function
1818 - Fix interactions with processe resume/suspende
1819 - release_forever() was stupidly broken
1820 - Fix SIMIX_display_process_status() for processes in a semaphore
1821 - Make SIMIX_sem_block_onto() user-visible
1822 * Refactoring context stuff:
1823 - Use pseudo-OOP for better modularity
1824 - reimplement SIMIX_process_kill() without process_schedule() so
1825 that the latter can take as invariant that it is called from
1827 - Merge context_start into context_new for sake of simplicity
1829 * Add a Vivaldi network model, coded live during SUD'10 ;)
1830 * Rename configuration variables to start a hierarchy:
1831 o cpu_model -> cpu/model
1832 o network_model -> network/model
1833 o workstation_model -> workstation/model
1834 * New configuration variables:
1835 o network/bandwidth_factor: correction to bandwith
1836 o network/latency_factor: correction to latency
1837 o netwotk/weight_S: correction to the weight of competing streams
1838 * Add a long description to the models, that users can see with such
1839 argument on the command line: --cfg=cpu/model:help
1840 * --help-models display the long description of all known models
1842 * config: add the ability to set a default value after registration
1843 Does not override any previously set value (e.g. from cmd line)
1844 * dict: allow to have integer key and data.
1845 When so, you need to use the following functions
1846 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1847 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1848 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1849 In contrary to regular dicts, the key is not malloced before copy.
1850 Mixing scalar and regular elements in the same dict is not tested
1852 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1853 Tracing for Visualization:
1854 * SimGrid is now instrumented in order to generate a trace file for
1855 visualization analysis: to use it, need to compile SimGrid with the
1856 "tracing" option enabled, and instrument the program using SimGrid with
1857 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1858 (among other functions).
1859 * The instrumentation only traces the platform utilization for now
1860 * Documentation to use the tracing functions and how to analyze the
1861 traces with the Triva tool is written.
1862 * More information about: SimGrid FAQ (in the section Tracing Simulations
1865 * We moved to cmake as default build system. Autotools support will
1866 be dropped soon. Check the FAQ for more info about how to use it.
1867 * Greatly improved our cdash/ctest interactions
1868 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1869 * Added memory checking tests with valgrind; lot of memleak fixing.
1870 This may be the first release of simgrid with so few memory issues
1871 * Added code coverage tests.
1872 Our coverage is still improvable, but at least we see it on cdash.
1874 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1876 SimGrid (3.3.4) stable; urgency=low
1878 The "Desktop Grid needs love too" release (also called Xmas release).
1880 Models improvements:
1881 * Major speedup in the maxmin system solving by using lazy evaluation
1882 Instead of solving completely the maxmin system at each iteration,
1883 only invalidate (and recompute) the modified parts.
1884 This new feature is enabled in default models but you can try to
1885 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1886 * Cas01 IMproved as default CPU model
1887 This CPU model is the same Cas01 model, but it uses the
1888 maxmin-selective-update flag and a heap structure to manage
1889 actions on SURF kernel.
1890 It reduces the complexity to find the next action to finish and,
1891 consequently, it's faster than the old Cas01.
1892 This is the new default CPU model (Cas01).
1893 * Rename the old Cas01 model to Cas01_fullupdate
1894 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1895 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1896 * CpuTI (CPU Trace Integration)
1897 A new CPU model whose objective is simulate faster when using
1898 availability trace files.
1899 Instead of using a full featured, over engineered maxmin system for
1900 CPU modeling, this model does the pre-integration of traces files
1901 to calculate the amount of CPU power available, and so, executes
1902 faster than the old CPU models.
1903 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1904 * Use LV08 as default network model since it gives better accuracy
1905 for small messages and shouldn't change things for big ones.
1906 Use --cfg=network_model:CM02 to get the previous behavior.
1909 ******************************************
1910 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1911 ******************************************
1912 * The new CPU model may changes simulations!
1913 The point is that events occurring at the exact same timestamp
1914 are not scheduled in the same order with the old and new
1915 version. This may be enough to completely change the execution
1916 of simulations in some cases.
1917 * The new network model will change simulations!
1918 This new model is more realistic than the previous one, so you
1919 should consider redoing your old experiments with this model.
1920 Sorry for the inconvenience.
1923 * Introduce the supernovae compilation mode
1924 When compiled that way, the whole SimGrid (or almost) is put in a
1925 single compilation unit and compiled in one shoot.
1926 This is to help gcc which has difficulties to inline stuff from one
1928 The speedup seem to be above 15%, althrough more tests are needed on
1929 amd64 to confirm that gain.
1932 * Port of MSG's mailbox on top of SIMIX network
1933 The put/get mechanism was greatly simplified on the way.
1936 * New SIMIX network module. Provides:
1937 - Mailbox: rendez-vous mechanism to find with who you want to speak
1938 - Synchronous send/recv: easier and hopefully faster since the
1939 logic is handled in the maestro process directly now
1940 - Asynchronous send/recv: you dreamt of it? It's here now
1941 Too bad that nobody cared enough to propagate the change to MSG.
1942 * Add semaphores as SIMIX synchronization mechanism.
1945 * new function SD_daxload(char*) to load a DAX file
1946 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1947 * Introduce typed tasks. Specify its kind and cost at creation.
1948 At scheduling, just give where it should be placed, and the cost
1949 for each involved resource is automatically computed.
1950 Existing constructors so far (more to come of course):
1951 - SD_task_create_comm_e2e() for end-to-end communication
1952 - SD_task_create_comp_seq() for sequential computation
1953 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1954 * new function SD_task_dump() for debuging display
1955 * new function SD_task_dotty(task,FILE*) writing to file the info
1956 about the task in dotty format
1957 * SD_task_dependency_exists() can now cope with having one of its
1958 arguments NULL. If so, it tests whether the other argument has any
1960 * Add getters on list of preceding/following tasks:
1961 SD_task_get_parents(task) and SD_task_get_children(task)
1962 * Add getters on amount of workstations and list:
1963 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1964 * Add getter on task kind: SD_task_get_kind(task)
1965 * Update the start_time and finish_time of tasks on completion/failure
1966 * Bugfix: Remove task from state swags when destroyed
1969 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1972 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1973 dynar with one separate thread per value of the dynar.
1974 * Change the prototype of xbt_thread_create(), sorry.
1975 Added a boolean parameter indicating whether we want to join this
1976 thread (used in SG only for now)
1977 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1980 * GTNetS wrappers should now be usable again (and betterly tested too)
1981 * Fix a major regression from 3.2 where the timeout provided to
1982 MSG_task_put_with_timeout() was used as absolute time before which
1983 the comm should be done.
1984 * Start to fix the <cluster> tag.
1985 - Internal links should be good now (beside of the loopback, which
1986 use the private link instead)
1987 - paths to the external world is still rather broken
1988 - the <route:multi> tag is just broken. Actually that's brain-dead.
1989 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1990 to make it less stupid
1991 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1992 * Fix a source-level compatibility glitch from 3.2: after defining
1993 MSG_USE_DEPRECATED, you can use the old name
1994 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1995 * Allow to compile from the SVN with automake 1.11
1996 * Fix some problems when using the "start_time" tag in deployment XMLs.
1997 * Fix #8569: XBT/synchro.h has redundant declarations
1998 * Fix #8563: MSG return values and exceptions
1999 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2000 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2001 Thanks to Robson Peixoto.
2002 * Fix a vicious bug in dictionaries inducing that some elements were
2003 not freed on xbt_dict_free()
2005 Portability report of this version:
2006 * Main portability targets:
2007 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2008 on (amd64/i386/ia64)
2009 - mac leopard on i386
2010 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2011 but nothing critical.
2012 * Other platforms: windows, AIX and others were not tested for this release
2014 Timing report of this version:
2015 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2016 scenario parameters). From 8h to a few seconds in desktop grid settings.
2017 * Supernovae brings about 25% speedup on i386.
2019 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2021 SimGrid (3.3.3) stable; urgency=low
2023 The "Need for Speed" release.
2025 The timings done to validate the 3.3.2 were faulty.
2026 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2028 The problem was a conversion from a manually handled vector to
2029 xbt_dynar_t on the critical path.
2030 xbt_dynar_foreach calls functions, inducing stack management crap.
2032 We inlined these functions and xbt_dynar_foreach is now breath taking.
2033 We also inlined xbt_swag_belong on the way.
2035 Here are some approximate speedup measurements (on master/slaves
2036 simulations lasting between 10s and 20s each):
2037 3.3.1 -> 3.3.2: about same performance
2038 3.3.2 -> 3.3.3: 40% speedup
2039 3.3.1 -> 3.3.3: 40% speedup
2040 3.3.1 with inline patch -> 3.3.3: 30% speedup
2042 Our reading is that the refactoring which occurred in 3.3.2 made us
2043 suffer much more from the xbt_dynar_foreach low performance, but
2044 once we solved this, this refactoring proved to be very performance
2045 effective. From the 40% speedup, somehow, 10% are due to the
2046 inlining and 30% to the refactoring.
2048 That's a pitty that gcc cannot inline functions placed in other files
2049 alone. We have to choose between:
2050 - break the encapsulation (by putting private data structures and
2051 accessors in headers files to help gcc)
2052 - live with low performance
2053 - switch to a decent compiler such as icc (not quite possible).
2055 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2057 SimGrid (3.3.2) stable; urgency=low
2059 The "Simplicity does not preceed complexity, but follows it" release.
2061 The main contributors of this release were (lexical order):
2062 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2065 * Extract the routing logic into its own object.
2066 (was dupplicated in network.c and workstation_LV07.c;
2067 Allows to implement other ways of storing that info)
2068 => kill now useless network_card concept
2069 - Use dynar to represent routes (instead of void** + int*)
2070 - kill link_set (use surf_network_model->resource_set instead)
2071 - Add a command-line option to choose the routing schema to use
2072 - Add three new models:
2073 * Floyd (shortest path computed at initialization)
2074 * Dijikstra (shortest path recomputed all the time)
2075 * Cached Dijikstra (shortest path computed on need)
2076 All these models where contributed by Silas De Munck, and are
2077 described in his ICCS09 paper.
2079 * Simplify model declaration
2080 (less redirections, less function to write when defining a model)
2081 - Factorize stuff between models:
2084 surf_model_resource_set(model)
2085 surf_model_resource_by_name(model, name)
2086 - Unify the types of models in s_surf_model_t (using an union)
2087 - Embeed fields of common_public directly into s_surf_model_t
2088 - Rename model methods:
2089 action_free ~> action_unref
2090 action_change_state ~> action_state_set
2091 action_get_state ~> action_state_get
2092 - Change model methods into functions :
2093 (model)->common_public->action_use ~> surf_action_ref
2095 * Implement a generic resource; use it as ancestor to specific ones
2096 (allows to kill duplicated code in models)
2097 Drawback: timer command don't need no name nor properties;
2098 workstation_CLM03 don't need no properties
2099 (but I guess we can live with those few bytes wasted)
2101 * Improve the action object model
2102 - implement a constructor avoiding dupplicated code about field
2103 initialization in generic_action part.
2105 * Kill the SDP model: it has an external dependency, is deprecated
2106 in flavor of modern lmm models, and didn't compile since a while
2109 * Relocation of the context module from XBT to SIMIX.
2110 (the context were decoupled from the simix processes, duplicating a lot of code)
2111 => a lot of code was factorized
2112 - less overhead is introduced during scheduling
2113 - simpler API for the context factory
2114 - the logic for process creation,destruction and manipulation was simplified
2115 * Simplification of the s_smx_process_t data structure.
2116 => accesing the simix level data associated to a process is faster now,
2117 and the code is a lot more readable.
2120 * Implement some more MPI primitives:
2121 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2122 -implementation: Bcast: flat or 2-ary tree (default),
2123 Barrier: 4-ary tree,
2125 Allreduce: Reduce then Bcast
2126 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2127 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2128 Alltoallv: flat tree, like ompi
2130 * Add support for optimized collectives (Bcast is now binomial by default)
2131 * Port smpirun and smpicc to OS X
2134 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2135 Shout out if you used it.
2138 * Display the list of still queued messages in SG mode when existing
2142 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2143 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2144 * Add xbt_str_from_file(FILE*)
2145 * Add xbt_dict_get_key achieving a linear reverse search
2146 * Remove the context module
2148 Portability report of this version:
2149 * Main portability targets:
2150 - Linux(debian)/x86/context
2151 - Linux(debian)/x86/pthreads
2152 - Linux(debian)/amd64/context
2153 - Linux(debian)/amd64/pthreads
2154 On these, we still have the eratic breakages of gras/pmm and
2155 amok/saturate_sg reported in previous version. We still think
2156 that the tests are the cause of the fault, not the tested code.
2158 - Mac OSX Leopard/x86/context
2159 Still false negative in tesh autotesting.
2160 Smpi still fails, but this time because readlink does not accept -f
2161 Everything seems to work properly beside of that.
2164 - AIX version 5.3 (only tested contexts this time)
2165 Smpi still fails there because mktemp is not installed.
2166 Everything seems to work properly beside of that.
2168 I managed to compile it for the first time, but several breakages.
2169 Won't delay the release for this exotic platform.
2171 * Windows: it's still lagging behind. If you want to help, please
2174 Timing report of this version:
2175 This version seem to be more than 5% faster than 3.3.1 (on linux
2176 64bits with contextes). The gain is less than expected, we are
2177 investigating this for next release.
2179 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2181 SimGrid (3.3.1) stable; urgency=low
2184 * Implement a --cfg-help to show existing configuration variables
2185 * Build chain do not require doxygen in maintainer mode
2188 * fix a bug on struct sizeof computation, which prevented the
2189 exchange of arrays of structs in some conditions
2190 - added a regression test about this in datadesc_usage
2191 * Allow the exchange of 0-long dynamic vectors.
2192 - for that, use -1 as indicator of dynamic size instead of 0
2193 - This implied to change any size from unsigned long to long,
2194 reducing a bit communication abilities, but I guess that with
2195 64bits being quite common, this is more than enough.
2196 - This also induce a protocol change, thus bumping network protocol
2197 version from 0 to 1 (if we have external users, we have to get
2198 clean on that point too ;)
2199 - added two regression tests about this in datadesc_usage
2200 * Be more verbose when propagating local exceptions
2201 This helps debugging.
2202 * Display the status of simulated processes when receiving SIGINT in
2206 * Allow to control the simulation from a trace file.
2207 New functions MSG_action_register() and MSG_action_trace_run()
2208 The first one allows to associate a function execution to each
2209 kind of action while the second one parses a trace file and
2210 triggers the corresponding actions within the system.
2211 For now, only a toy example is provided in examples/msg/actions
2212 * Add an exemple of process migration in examples/msg/migration
2213 * Fix a bug in task exchange which broke MSG_task_get_sender()
2214 Add a teshsuite regression test for that.
2215 [Bug: if MSG_task_get_sender() is called after sender exit,
2217 * Fix a bug which prevented suspend/resume to work properly
2218 * Display the status of simulated processes when receiving SIGINT
2219 This fixes a regression of v3.3. due to the introduction of SIMIX
2220 * Bug fixing in failure management:
2221 - trace could not start by a failure at time 0
2222 - failure during communications were not working
2225 * Add SIMIX_process_set_name() to change the name of the current
2226 process in the log messages.
2227 * Store smx_hosts in a dict since we only retrieve them by name
2228 * Move the configuration infrastructure to surf
2231 * Move the configuration infrastructure to surf
2234 * Massive internal cleanups:
2235 - Store internal structures on processes instead of hosts (allows
2236 to have more than one process per host, in addition of being more
2238 - Cleanup the initialization/finalization process
2239 - Kill a whole bunch of unneeded synchronization:
2240 processes run in exclusive manner within the simulator
2241 - Move queues from global tables to process data fields
2243 - now accept -platform and -hostfile arguments
2244 - Pass the right rank value to processes according to the hostfile
2245 * Compile the examples by default, and use them as regression tests
2246 * Implement MPI_Wtime()
2247 * Change the reference speed to a command line option
2250 * TCP_gamma can now be specified as command line option using
2251 --cfg=TCP_gamma:10000000.0
2252 * Change the --surf-path cmd line option into --cfg=path:
2255 * Also include strbuff from xbt.h public header
2256 * xbt_ex_display(): do not free the exception after displaying
2257 This allows to do more with the given exception afterward.
2258 Users should call xbt_ex_free() themselves.
2262 Portability report of this version:
2263 * Main portability targets:
2264 - Linux(debian)/x86/context
2265 - Linux(debian)/x86/pthreads
2266 - Linux(debian)/amd64/context
2267 - Linux(debian)/amd64/pthreads
2268 These targets fail about 1/10 of times on gras/pmm, but we believe
2269 that this is because of the test, not because of simgrid.
2270 amok/saturate_sg fails even more rarely, and the test may not be
2273 - Mac OSX Leopard/x86/context
2274 The test suite still spits tons of errors because some obscure
2275 force prevents us from removing the temporary directories
2276 arguing that they still contain some metadata I've never heard of.
2277 Smpi fails because seq is not installed.
2278 Everything seems to work properly beside of that.
2281 - AIX version 5.3 (both contexts and pthread)
2282 Smpi still fails there because mktemp is not installed.
2283 XML inclusions seems rosty on AIX.
2285 * Windows: it's still lagging behind. If you want to help, please
2288 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2290 SimGrid (3.3) stable; urgency=high
2294 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2295 [Malek Cherier & Mt]
2297 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2300 Until now, GRAS were using MSG as an interface to SURF. It was
2301 quite difficult because both interface have several differences
2302 (MSG channels vs GRAS sockets were the most notable point).
2304 This also opens the gate to SMPI (which should occur soon) and speed
2305 up simulations by to 40% (even if it were not the main goal).
2307 **************************************
2308 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2309 **************************************
2310 The point is that events occuring at the exact same timestamp are
2311 not scheduled in the same order with the old and new version. This
2312 may be enough to completely change the execution of simulations in
2313 some cases. Sorry for the inconvenience.
2315 * Cleanup and upgrade the XML format to push further scalability
2316 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2318 * Improve the testing infrastructure with tesh. Now a very large part of
2319 the code is tested not only by being run but also by checking that the
2320 output match an expected output [Mt].
2322 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2323 is really less memory-demanding, which should allow you to use
2324 larger files in SimGrid [AL].
2326 * Inform valgrind about our contextes, so that it becomes usable
2327 with the default (and more effecient) version of SimGrid
2328 [contributed by Sékou Diakite, many thanks]
2331 * Introduce a listener thread in charge of receiving incoming
2332 messages from the network. It allows to overlap communication and
2333 computation but most notably, it removes some stupid deadlocks due
2334 to the fact that so far, a process could not send and receive at
2335 the same time. This made most non trivial communication schema
2337 * Convert the PIDs from long int to int to match the MSG ones (and
2338 linux ones too) [Mt]
2339 * New function: gras_agent_spawn() to launch a new process on
2340 current host. Only working in simulation for now. [Mt]
2341 * New function: gras_os_hostport() returning a constant form (ie,
2342 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2345 * Make the backtrace of exceptions more human readable [Mt]
2346 * New module: xbt/str [Mt]
2347 a ton of string utility functions (split, join, printf to a newly
2348 allocated buffer, trim, etc)
2349 * New module: xbt/hash [Mt]
2350 SHA1 hashing algorithm (more to come if needed)
2351 * New module: xbt/synchro [Mt]
2352 synchronization tools (mutex and conditions) working the same way
2353 in simulation and in real life (mainly useful for GRAS, but not
2355 * New module: xbt/queue [Mt]
2356 classical producer/consumer synchronization scheme
2357 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2358 (using the classical functions will get serialized) [Mt]
2359 * Make dictionary internal table dynamic. No need to specify its size
2360 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2362 * Make sure the log channels are organized as a tree under windows
2363 (because of ANSI C compatibility issue, any channel were child of
2364 root directly) [Mt].
2367 * Cleaned many thing in surf and fixed a few bugs [AL].
2368 * Add a nice command line configuration mechanism to compose models [AL].
2369 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2370 the previous one (KCCFLN05). It relies on something that looks like
2371 a max-min sharing mechanism but cannot be written as such. A new solver
2372 was thus designed [AL].
2373 * Add a new solver to lmm. Based on Lagrange optimization and
2374 gradient-based descent, it enables to efficiently maximise systems s.a
2376 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2378 This solver enables to propose two new network models for TCP Reno and
2379 TCP Vegas based on Low's work. These models still need to be fully
2380 tested though [Pedro Velho].
2383 * Bug fix in SD_simulate. Now the time bound given as argument is
2385 * Use the new parallel task model (ptask_L07) as default.
2386 * Use the SURF command line configuration mechanism.
2387 * 0-size tasks (for synchronization) should now work.
2389 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2391 SimGrid (3.2) stable; urgency=high
2395 We still experience issues on this platform, but we believe that at
2396 least MSG is usable.
2398 GRAS API BREAKAGE (for simplification purpose, sorry):
2399 * the gras_msgtype_by_name is not used anymore. Instead of
2400 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2401 you can write (and must)
2402 gras_msg_send(toserver, "request", &request);
2403 - If you still want to pass a gras_msgtype_t to the function (to cache
2404 the type and avoid the lookup time), use the gras_msg_send_() variant.
2405 - Impacted functions:
2406 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2407 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2408 * The callbacks are now expected to return 0 when everything went well
2409 (just like the main() function)
2411 GRAS new features and improvements:
2412 * New module mechanism where user code can use per process globals [Mt]
2413 This is similar to gras_userdata_*() functions, but for libraries. It
2414 factorize some code developped over and over in the examples and AMOK.
2415 It has still to be documented and used (only amok/peermanagement is
2417 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2418 loss when they were small enough to fit into the buffer and sent quickly
2419 enough so that they can all get received in one shoot.
2420 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2421 if not found. Use the *_or_null() variant for the old semantic.
2422 * In gras_msg_handle, do not discard messages without callback.
2423 They are probably messages to be explicitly awaited later (ie, proofs of
2424 mis-synchronization in userland since they are sent before being awaited)
2425 No big deal usually.
2426 * gras_socket_meas_send/recv: semantic changed!
2427 The numerical arguments used to be (1) the total amount of data to send
2428 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2429 messages. This was need for the fool willing to send more than MAXINT
2430 bytes on quite fat pipes.
2433 * Do really rename the hostmanagement module to peermanagement. [Mt]
2434 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2435 but this is rather new and this was documented in the module
2436 documentation (poor excuses, I admit)
2437 * Bandwidth measurement semantic changed! This follows the changes to
2438 gras_socket_meas_send/recv explained above.
2441 * A sequential mode has been added to the workstations. When a workstation
2442 is in sequential mode, it can execute only one task, and the other tasks
2443 are waiting in a FIFO. [Christophe Thiery]
2446 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2447 model for SIMDAG. [Christophe Thiery]
2448 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2449 equal to 0, which caused some bad behaviors in
2450 saturated_constraint_set_update. I now use a threshold mechanism like in
2454 * When running manually src/testall, you select specific units [Mt]
2455 testall is the result of our cunit mechanism, and should replace all
2456 the scripty thingy around since bash don't run easily on billware.
2458 * A mallocator system has been added. [Christophe Thiery]
2459 Mallocators allow you to recycle your unused objects instead of freeing them
2460 and allocating new ones.
2462 Documentation update:
2463 * FAQ reworking + New FAQs:
2464 - "Valgrind spits tons of errors!" [Mt]
2465 - "How to repport bugs" [Mt]
2466 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2467 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2468 - Communication time measurement within MSG [AL]
2469 - I experience weird communication times when I change the latency [AL]
2470 * GRAS tutorial [Mt]
2472 - an introduction to the framework and to the used communication model
2473 - an initiatic tour introducing the most proheminent features:
2475 . Lesson 0: Installing GRAS
2476 . Lesson 1: Setting up your own project
2477 o Part 2: Message passing
2478 . Lesson 2: Exchanging simple messages
2479 . Lesson 3: Passing arguments to the processes (in SG)
2480 . Lesson 4: Attaching callbacks to messages
2481 . Lesson 5: Using globals in processes
2482 . Lesson 6: Logging information properly
2483 . Lesson 7: Using internal timers
2484 . Lesson 8: Handling errors through exceptions
2485 . Lesson 9: Exchanging simple data
2486 . Lesson 10: Remote Procedure Calling (RPC)
2487 . Lesson 11: Explicitely waiting for messages
2488 . Recapping of message passing features in GRAS
2489 - A HOWTO section containing:
2490 o HOWTO design a GRAS application
2491 More are due, of course. They will come latter. In the meanwhile, you can
2492 check the examples which are still here.
2494 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2496 SimGrid (3.1) stable; urgency=high
2500 There was a stack corruption somewhere, visible only when optimizing
2501 with these versions. [Vince]
2504 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2505 module that enabled to play with Directed Acyclic Graphs. It is built
2506 directly on top of SURF and provides an API rather close to the old
2507 SG. Some old codes using SG are currently under rewrite to check that
2508 all needful functions are provided. [Christophe Thiery]
2511 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2512 extension of the classical CLM03 model that gracefully handles
2513 failures. This is now the default model for MSG and GRAS. It doesn't
2514 handle parallel tasks yet though. [AL]
2515 * Bug fix: Weights were not correctly set in the network part.
2516 WARNING: This may have resulted in incorrect results with simulations
2517 where there are more than one flow on a given link. [AL]
2520 * After a (long ?) discussion on simgrid-devel, we have decided that the
2521 convention we had on units was stupid. That is why it has been decided
2522 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2523 WARNING : This means that all previous platform files will not work as
2524 such with this version! A warning is issued to ask users to update
2526 A conversion script can be found in the contrib module of the CVS, under
2527 the name contrib/platform_generation/surfxml_update.pl [MQ]
2530 * Bug fix: Processes were started in reverse order, wrt deployment file.
2531 WARNING: if your code relies on this bug, please fix it. [AL]
2532 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2533 being executed on two different locations. [AL]
2534 * Bug fix: Failures are now better supported thanks to Derrick's tests
2535 (there was many failure situations I hadn't thought of and that weren't
2536 correctly handled). [AL]
2537 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2541 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2542 Exception occurring on server-side are propagated back to client (!).
2544 API CHANGE: the callback changed their prototype. Change:
2545 int my_handler(gras_socket_t expeditor, void *payload_data) {
2547 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2548 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2550 * New! function: gras_msg_handleall to deal with all messages arriving
2551 within a given period.
2552 * New! function: gras_socket_server_range to get a server socket in a
2553 range of port numbers (ease to avoid port number conflicts) [MQ]
2554 * New! gras processes display their backtrace when they get a SIGUSR1
2555 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2556 Sweet to debug RL processes [MQ]
2560 - Do not force experiment sizes to be expressed in kb, or it becomes
2561 impossible to measure the latency this way (needs one byte-long tests)
2562 WARNING: this changes the amok_bw_* function semantic. [MQ]
2563 - Implements the link saturation stuff. [MQ]
2564 * Peer management module:
2565 New! module factorizing code that we wrote over and over [MQ].
2568 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2569 - Test units are placed directly into the library code, they get extracted
2570 automatically and placed into the src/testall binary.
2571 - Convert most of the XBT tests to this system.
2572 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2573 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2574 live in the same C code. This required to change a little bit the API
2575 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2576 * New module: sparse graph structure with basic algorithms (this is work
2577 in progress and the API is not considered to be frozen yet). [AL]
2578 * Display more information on backtraces: source line & function names are
2579 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2580 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2581 * Reworked a little bit some #include statements to load only required
2582 headers. Some user code that relied on SimGrid to include stdlib or
2583 stdio may need to include it by themselves. [AL]
2584 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2586 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2587 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2588 but I guess that almost nobody used those parts. [MQ]
2590 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2592 SimGrid (3.0.1) stable; urgency=low
2595 * Unfortunately, I had missed 5 misnamed functions:
2596 xbt_fifo_item_t xbt_fifo_newitem(void);
2597 void xbt_fifo_freeitem(xbt_fifo_item_t);
2598 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2599 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2600 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2601 They're now deprecated. Please use their new versions:
2602 xbt_fifo_item_t xbt_fifo_new_item(void);
2603 void xbt_fifo_free_item(xbt_fifo_item_t);
2604 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2605 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2606 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2608 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2609 * Documentation: xbt_log module unmercifully reworked [MQ]
2610 * Bugfix: there was a problem with the ending of contexts with
2611 the pthread backend. It caused some weird deadlock or behavior
2612 depending on the pthread implementation. [AL]
2613 * Bugfix: get the exceptions raised in the simulator repport where
2614 and why they come from when they are not catched in time [AL, MQ]
2617 * Bugfix: Do repport the error when two non-connected hosts try to
2618 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2621 * Add additionnal checkings on communications. Assert that two
2622 communicating hosts are connected by a set of links... [AL]
2625 * Add additionnal checkings on channel values in communication [AL]
2626 * New: MSG_task_get_source to see on which host a task was generated [HC]
2627 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2628 the number of tasks waiting to be received on channel and sent
2630 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2631 waits for the first task coming from a given host.. [AL]
2633 GRAS new functionnalities: [MQ]
2634 * Enhance the parsing macro to allow the size of multidimentional objects
2635 to be given thru annotations.
2636 * New example (and documentation): Matrix Multiplication a la RPC
2637 (as when I was young!) and fix a bunch of bugs found on the way.
2639 GRAS performance improvements: [MQ]
2641 * Reduce the amount of cbps creation/destruction by making it static to
2642 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2645 * Change libdata to a set so that we can search for stuff by ID (and thus
2646 reduce the insane amount of dict lookups)
2649 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2650 RL communication on top of SG since it's so uneffective.
2651 It may also be used for inter-thread communication in RL, one day.
2652 * Use gras_datadesc_copy() to exchange messages on top of SG
2654 - improve message exchange performance on top of SG
2655 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2656 * Don't exchange on the network the size of the used part of buffer,
2657 instead, specify the possible buffer size to read().
2659 - reduces the amount of read/write calls (one pair per exchange)
2660 - reduces the amount of exchanged data (the size)
2661 - allows to retrieve all arrived data on receiver side, if we don't need
2662 it right now (subsequent read will peek the buffer)
2663 - allows the receiver to proceed with the begining of the stream before
2664 everything is arrived
2665 - make it possible to build an iov transport (using readv/writev)
2667 - take care of the data with non-stable storage (like stacked data),
2669 * If possible, TCP send uses vector I/O (when writev() is here)
2670 - Don't use it for receive since we send data sizes and data on the
2671 same stream, so we wouldn't be able to chain large amount of chunks
2672 before having to flush the stuff to read the size.
2673 * Rework the transport plugin mechanism to simplify it and reduce the
2674 amount of pointer dereferencement when searching for the right function
2677 * I guess that now, we do almost as few system calls as possible while
2678 doing as few data copy as possible.
2680 To improve it further, we could try to send all the sizes first and then
2681 all the data (to use iov on receiving size), but it's only a partial
2682 solution: when you have 2 dimensional data, the sizes of the second
2683 dimension is data of the first dimension, so you need 3 streams.
2685 I'm not sure the potential performance gains justify the coding burden.
2687 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2689 SimGrid (3.00) stable; urgency=high
2692 * New! Give the possibility to hijack the surf parser and thus bypass
2693 MSG_create_environment and MSG_launch_application. Have a look at
2694 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2696 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2698 SimGrid (2.96) unstable; urgency=low
2703 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2704 This deprecates the xbt_error_t mechanisms.
2705 It modifies (simplifies) all XBT and GRAS API.
2706 MSG API keeps unchanged (exceptions raised by XBT are catched from
2707 within MSG and masked with existing error handling facilities)
2710 * New! Add a FATPIPE model. [AL]
2711 * New! Add a parallel task model. [AL]
2712 * New! Add automatically a loopback interface (in the default
2713 network model) if none was precised.
2716 * Bugfix: MSG_process_resume now works with the current running process.
2718 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2719 * Modification of MSG_task_get_compute_duration. Once a task has been
2720 processed, the value returned by this function is now equal to 0. [AL]
2721 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2722 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2723 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2724 of a m_task. For now, MSG_task_cancel only works with computation
2726 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2727 of the processor (in Mflop/s) regardless of the current load on the
2729 * API Change: use proper naming convention for MSG_getClock and
2730 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2732 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2733 This function changes the priority of a computation task. This priority
2734 doesn't affect the transfer rate. A priority of 2 will make a task
2735 receive two times more cpu power than the other ones. This function
2736 has been added to suit the needs of Nguyen The Loc and hasn't been that
2737 much tested yet. So if it fails, please report it and send me your code.
2739 * API Change: removed all functions and types that were marked "deprecated"
2740 since many months. Renamed MSG_global_init_args to MSG_global_init.
2742 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2744 SimGrid (2.95) unstable; urgency=low
2747 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2748 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2749 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2752 * Bugfix: Allow absolute paths for platform description files [MQ]
2753 * Bugfix: do free the variables after use. Leads to drastic performance
2755 * Implement max_duration (ie, timeouts) on resources [AL]
2758 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2760 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2761 the needed support (timeouts) [AL]
2764 * Implement measurement sockets. You can now get the bandwidth between two
2765 hosts thanks to AMOK (see below). [MQ]
2766 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2767 dynar over the network (yeah) [MQ]
2768 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2769 * simple/static token-ring example. [Alexandre Colucci and MQ]
2770 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2771 in place before [MQ]
2774 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2775 type parsing. "short" and "long" modifiers were also ignored; other
2776 modifier (such as reference level) are still ignored. [MQ]
2777 * Embeed the buffer size within the buffer itself on SG. [MQ]
2778 That way, send() are atomic and cannot get intermixed anymore (at least
2779 the ones which are less than 100k; bigger messages still have the issue)
2780 * Array size pushed by the field, not by the field type (or each
2781 and every long int will push stuff to the cbps) [MQ]
2782 * use select() to sleep since it allows to portably sleep less than one
2785 GRAS (minor cleanups)
2786 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2789 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2790 * type category 'ignored' killed as it was never used and were difficult
2792 * whether a type can cycle or not is now a flag, leaving room for more
2793 flags (as "ignored", if we feel the need one day ;)
2794 * Rename raw sockets to measurement sockets since "raw" has another
2795 meaning in networking community.
2798 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2799 and offers features not belonging to GRAS but that most applications
2800 need. One day, it may be a set of plugins loadable at runtime.
2801 * New module: bandwidth
2802 bandwidth measurement between arbitrary nodes running this module. [MQ]
2804 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2806 SimGrid (2.94) unstable; urgency=low
2808 The first beta release of SimGrid 3 !
2812 * Update the main page and the FAQ. Adding references to gforge.
2815 * Add a gras_os_getpid function.
2818 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2819 * Extend the logs so that they also print PID, hostname, date, ... if
2821 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2822 and kill the old version which were in testsuite/
2823 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2827 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2828 for users. Instead, each programming environment (GRAS, MSG,...) use it
2829 when needed to provide such a feature to users.
2830 Don't shortcut the mechanism or you will also shortcut the virtualization
2831 you need on the simulator.
2835 * Cleanups in configury with regard to compile optimization/warning flags.
2836 Also add -fno-loop-optimize to any powerpc since it's the optimization
2837 killing gcc (< 3.4.0).
2838 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2839 needed by MSG examples complications
2840 * Borrow configury beautifications from PHP
2843 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2844 warning (thanks loris for stumbling into this one).
2845 * Bugfix: stop loading private headers (gras_config.h) from the public
2849 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2850 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2851 RL would imply a DNS resolver, which is *hard* to do in a portable way
2852 (and therefore delayed).
2853 * Implement a real timer mechanism and use it in timing macros. This allows
2854 to avoid rounding errors and get a 0.000005 sec precision in timing
2855 macros. While I was at it, various cleanups:
2856 - allow to declare more than one timed section per file (fix a stupid bug)
2857 - move some private declaration to the right place
2858 - merge conditional execution and timing macros into emulation module
2859 - document the module
2860 - make sure the module cleanups its mess on gras_exit
2861 * Documentation improvements:
2862 - (new) how to compile applications using GRAS
2863 - (new) emulation support (timing macros)
2865 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2867 SimGrid (2.93) unstable; urgency=low
2869 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2872 - Use Paje properly where used. Still to be sanitized properly.
2873 - Portability fix: Add an implementation of the contexts using pthread
2877 - Add xbt_procname(): returns the name of the current process.
2878 Use it to show the current process's name in all logging.
2880 - fix detection of older flex version and the reaction, since we do
2881 depend on modern ones (we use lex_destroy)
2882 - Better separation of SG and RL in the libs: remove all simulation code
2883 from libgras. As a result, this lib is now only 200k when stripped.
2884 Some of the xbt modules may also be duplicated (two sets and such) and
2885 should be cleaned/killed before SG3.
2886 - Insist on using xlC on AIX because of weird problems involving gcc there.
2887 - Cleanup the make remote stuff. This is now done by scripts
2888 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2889 mainly for our private use, but we're working on changing them to user
2892 - Bugfix: flush the socket on close only if there is some *output*.
2893 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2895 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2898 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2900 SimGrid (2.92) unstable; urgency=low
2902 Alpha 3 on the path to SimGrid 3
2906 - New! First try of benchmarking macros.
2907 - New! First try so that gras_stub_generator generate deployment and
2908 remote compilation helpers.
2910 - Bugfix: Initialization fix in msg_test.
2914 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2916 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2918 - API Change: gras_os_sleep to take the amount of seconds as a double.
2919 Accepting an int was error prone since it was the only location where
2920 seconds were coded as such. It leaded to damn rounding errors.
2921 - Bugfix: Hard to belive that timers ever worked before this.
2923 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2925 SimGrid (2.91) unstable; urgency=low
2927 Alpha 2 on the path to SimGrid 3
2931 - Bug fix in the lmm_solver.
2933 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2934 through the function MSG_paje_output.
2935 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2936 - It is possible to bound the rate of a communication in MSG with
2937 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2938 to write it in the changelog).
2939 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2944 - Various cleanups to the autotools stuff
2945 - Begin to move Gras examples to examples/gras/
2946 - Let make distcheck work again (yeah!)
2948 - documentation overhauled using doxygen.
2949 gtk-doc-tools is dead in SimGrid now.
2950 - Automatically extract all existing logging categories, and add the list
2951 to the documentation (long standing one, to say the less)
2953 - Cleanup the known architecture table. Reorder the entries to group what
2954 should be, and use a more consistent naming scheme.
2955 (some of the test dataset are still to be regenerated)
2956 - New! Allow library to register globals on each process just as userdata
2958 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2959 lookup time (for now).
2960 Use it in msg and trp.
2961 This cleans a lot the internals and helps enforcing privacy of the
2962 headers between the gras components.
2963 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2964 - Bugfix: gras_os_time was delirious in RL.
2965 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2967 - Reenable GRAS now that it works.
2969 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2971 SimGrid (2.90) unstable; urgency=low
2973 Alpha 1 on the path to SimGrid 3
2975 * It is a long time since the last release of SimGrid. I'm sorry about
2976 that but as I had told you, I was rewriting a lot of things. I apologize
2977 to those who had been reporting bugs to me and that I had not answered.
2978 If your bug is still in the new version, please tell me. Here is a
2979 summary of the main changes.
2981 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2982 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2983 portability, speed, and a lot more but you'll figure it out later.
2984 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2985 I wanted to release the new MSG as soon as possible and I have
2986 broken GRAS, which is the reason why, for now, only MSG is fully
2987 functional. A laconic description of these projects is available
2988 in the documentation.
2990 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2991 kernel optimized for our needs (called SURF but only the developers
2992 should use it). Hence, MSG is now roughly 30 times faster and I think
2993 that by rewriting a little bit MSG, I could event speed it up a little
2994 bit more. Beside the gain in speed, it is also much easier to encode a
2995 new platform model with SURF than it was with SG. More to come...
2997 * REVOLUTION 3: I have tried to change a little as possible the API of
2998 MSG but a few things really had to disappear. The main differences
2999 with the previous version are :
3000 1) no more m_links_t and the corresponding functions. Platforms are
3001 directly read from a XML description and cannot be hard-coded
3002 anymore. The same format is used for application deployment
3003 description. The new format is described in the documentation.
3004 Have a look in tools/platform_generation. There is a tiny script
3005 that converts from the old platform format to the new one. Concerning
3006 the application deployment format, parsing the old one is tricky.
3007 I think most of you should however be able to convert your files. If
3008 it is really an issue, I can write a C code that does the conversion.
3010 2) the toolbox tbx does not exist anymore. We now have a library
3011 with much more data-structures but without the hash-tables (we have
3012 dictionaries that are much faster).
3014 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3016 *****************************************************************************
3017 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3018 * don't feel like converting the oldies. (Mt) *
3019 *****************************************************************************
3022 Version 2.90: "the long awaited one"
3023 - Finished rewriting and debugging MSG. Rewrote the documentation.
3024 - disable GRAS for now since it needs to be ported to the newest SG
3027 - Finish the port to windows (using mingw32 for cross-compile)
3030 - Main loop and datastructures of SURF. A cpu resource object is
3031 functional. Surf can thus be used to create cpu's with variable
3032 performance on which you can execute some actions.
3034 2004-11-15 Martin Quinson
3035 - Port to ARM. Simply added the alignment and size descriptions. Should
3036 work, but the ARM machines are so slow that I didn't had the opportunity
3037 to 'make check' over there yet.
3039 2004-11-15 Arnaud Legrand
3040 - Trace manager now written. It uses a heap structure and is therefore
3041 expected to be efficient. It may however be speeded up (particularly
3042 when many events occur at the same date) by using red and black
3043 trees. One day maybe...
3044 - Max-min linear system solver written. It uses a sparse matrix
3045 structure taking advantage of its expected use. Most operations are
3046 O(1) and free/calloc are called as few as possible. The computation of
3047 the minimum could however be improved by using a red and black tree
3050 2004-11-03 Arnaud Legrand
3051 - Rename every gras_* function that was in xbt/ to its xbt_
3053 - Add a heap and a doubly-linked list to xbt
3054 - Added a dichotomy to the dictionaries. make check works as well before
3055 so I assume that the patch is correct. I do not know however if things
3056 run effectively faster than before now. :)
3058 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3059 SimGrid, and the version number is bumped to 2.x
3061 2004-10-29 Martin Quinson
3062 - Introduction of the remote errors.
3063 They are the result of a RMI/RPC on the remote machine.
3064 ErrCodes being scalar values, you can't get the host on which those
3065 errors did happen. Extending the error mechanism as in Gnome is possible.
3066 No idea yet whether it is a good idea.
3068 2004-10-28 Martin Quinson
3069 - Interface revolution: the Starred Structure Eradication.
3070 I used to do typedef struct {} toto_t; and then handle *toto_t.
3071 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3072 - ???_t is a valid type (builded with typedef)
3073 - s_toto_t is a structure (access to fields with .)
3074 - s_toto is a structure needing 'struct' keyword to be used
3075 - e_toto_t is an enum
3076 - toto_t is an 'object' (struct*)
3078 typedef struct s_toto {} s_toto_t, *toto_t;
3079 typedef enum {} e_toto_t;
3080 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3081 s_toto_t) is private.
3083 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3084 so that it can set the variable to NULL. It was so for dicts and sets,
3085 it changed for dynars.
3087 - Fix a bunch of memleaks in dict_remove
3088 - Fix a bug in sg/server_socket opening: it failed all the time.
3090 2004-10-07 Martin Quinson
3091 - Speed up dynar lookup operation a bit.
3093 gras_dynar_get is dead.
3095 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3096 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3097 (which gives you the address of the stored data).
3099 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3100 copy of the data using an affectation to do the job and not a memcpy.
3102 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3103 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3105 It does not really speedup the dynar test because they are
3106 setting elements all the time (and look them seldom). But the dict does
3107 far more lookup than setting.
3109 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3111 2004-10-05 Martin Quinson
3112 - Allow to (en/dis)able the cycle detection at run time.
3114 Whether we should check for cycle or not is now a property of each
3115 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3116 datadesc_cycle_unset allow to remove this property when previously set.
3118 Note that the cycle detection is off by default since it impacts the
3119 performance. Watch the data you feed GRAS with ;)
3121 This property is hereditary. Any element embedded in a structure having it
3122 set have it set for the time of this data exchange.
3124 You should set it both on sender and receiver side. If you don't set it on
3125 sender side, it will enter an endless loop. If you forget on receiver
3126 side, the cycles won't be recreated after communication.
3128 - Header reorganization.
3129 Kill gras_private.h, each submodule must load the headers it needs.
3131 2004-10-04 Martin Quinson
3132 - Interface revolution: do not try to survive to malloc failure.
3134 Now, gras_malloc and friends call gras_abort() on failure.
3135 As a conclusion, malloc_error is not a valid error anymore, and all
3136 functions for which it was the only gras_error_t return value are
3137 changed. They now return void, or there result directly.
3138 This simplify the API a lot.
3140 2004-09-29 Martin Quinson
3141 - Re-enable raw sockets.
3142 Created by gras_socket_{client,server}_ext;
3143 Used with gras_raw_{send,recv}
3146 It should allow to kill the last bits of gras first version soon.
3148 This is not completely satisfactory yet (duplicate code with
3149 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3152 - Simplify transport plugin (internal) interface by not passing any
3153 argument to _server and _client, but embedding them in the socket
3156 2004-09-28 Martin Quinson
3157 - Finish the port to AIX.
3158 autoconf was my problem (segfault within the malloc replacement
3159 function. No idea why)
3161 2004-09-16 Martin Quinson
3162 - Fix some size_t madness on 64bit architectures.
3164 2004-09-08 Martin Quinson
3165 - Reduce the number of system headers loaded, overload some more system
3166 calls (such as malloc to cast the result of the system one, and work
3168 - Fix and reintroduce the config support
3170 2004-09-07 Martin Quinson
3171 - Source code reorganization to allow Arnaud to surf all over there.
3172 - Allow to document the logging categories.
3173 - Remove all uppercase from logging categories and useless cleanup in names.
3175 2004-08-18 Martin Quinson
3176 Version 0.6.2 (protocol not changed; API changed)
3177 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3178 gras_error_t), and NULL when not found. Functions expecting a msgtype
3179 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3180 hopefully usefull message.
3181 - Portability to prehistoric sparcs again
3183 2004-08-17 Martin Quinson
3184 Version 0.6.1 (protocol not changed; ABI not changed)
3185 - prealloc some buffers to speed things up
3187 2004-08-11 Martin Quinson
3188 Version 0.6 (protocol not changed; ABI expended)
3189 - The parsing macro can deal with the references, provided that you add
3190 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3192 2004-08-09 Martin Quinson
3193 Version 0.5 (protocol not changed; ABI changed)
3194 - Allow to off turn the cycle detection code in data exchange at
3195 compilation time. It should be at run time, but I'm short of time (and
3196 the config stuff is still broken). That way, we keep dict out of the
3197 critical path, which is good because the performance is poor:
3198 - search not dichotomial yet
3199 - dynar give no way to access their content and memcpy everytime
3200 - In composed data description (struct, ref and so on), stop foolness of
3201 keeping the subtype's ID, but store the type itself. This keeps sets out
3202 of the critical path, which is good since they rely on dynar and
3203 dictionnaries. The only loose of that is that we cannot detect the
3204 redeclaration of a structure/union with another content (but I'm not sure
3205 the code detected well this error before anyway). We still can detect
3206 the redefinition discrepancy for the other types.
3207 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3208 it breaks the code because of type-punning used all over the place).
3209 This breaks on all non-gcc architectures (for now).
3211 All those changes (plus the buffer of last time) allow me to gain 2 order
3212 of magnitude on cruel tests consisting of 800000 array of integers on two
3213 level of a hierarchical structure (200 secondes -> 4 secondes)
3216 - the selector of reference must now return the type it points to, not
3217 the ID of this type.
3219 2004-08-06 Martin Quinson
3220 Version 0.4 (protocol changed; ABI not changed)
3221 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3222 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3223 - Implement a buffer transport, and use it by default (it relies on tcp in
3224 real life and on sg in simulation).
3225 That's a bit hackish since I had a new field to the structure to store
3226 its data without interfering with the subtype ones. Inheritance
3227 is tricky in C. And that's a kind of reverse inheritance with one class
3228 derivating two classes. Or maybe a game with java interfaces. Anyway,
3229 that's damn hard in C (at least).
3230 Moreover, I got tired while trying to ensure plugin separation and
3231 genericity in SG mode. MSG wants me to do weird things, so let's go for
3232 cruel hacks (temporarily of course ;).
3233 See comment in transport_private.h:71
3234 - do not include all the _interface headers in private but in the files
3235 which really need them (to cut the compilation time when they are
3238 2004-07-26 Martin Quinson
3239 Version 0.3 (protocol not changed; ABI changed)
3240 - Major overhault of the datadesc interface to simplify it:
3241 - shorted the function names:
3242 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3243 - add a trivial way to push/pop integers into the cbps without malloc.
3244 This allows to make really generic sub_type description, which simply
3245 pop their size of the stack.
3246 - add a function gras_datadesc_ref_pop_arr() which does what users want
3247 most of the time: Declare a dynamic array (which pops its size of the
3248 stack) and declare a reference to it. Poor name, but anyway.
3249 - kill the post-send callback, add a post-receive one
3251 2004-07-23 Martin Quinson
3252 Version 0.2 (protocol changed; ABI changed)
3253 - add some testing for cpbs in the test cases, and fix some more bugs.
3254 This invalidate again the little64 data file, since I cannot regenerate
3256 - remove an awfull optimization in the logging stuff, allowing me to:
3257 - understand it again
3258 - learn gcc how to check that the argument match the provided format
3259 - fix all errors revealed by gcc after that
3260 - internal keys of dict are not \0 terminated. Deal with it properly in
3261 loggings instead of segfaulting when the user want to see the logs :-/
3263 2004-07-22 Martin Quinson
3264 - Fix some stupid bug preventing cbps (callback postit) from working
3266 2004-07-21 Martin Quinson
3267 - Some documentation cleanups
3268 - remove the useless last argument of msgtype_declare
3269 - rename the Virtu functions to fit into the 'os' namespace
3270 - move headers src/include -> src/include/gras/ and stop fooling with
3271 gras -> . symbolic link
3272 - make distcheck is now successful
3274 2004-07-19 Martin Quinson
3276 - Build shared library also
3277 - Install html doc to the right location
3278 - stop removing maintainer files in make clean
3279 - build tests only on make check
3281 2004-07-13 Martin Quinson
3283 - No major issue in previous version => change versioning schema
3284 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3285 corresponding dataset.
3287 2004-07-11 Martin Quinson
3289 - Get it working with any kind of structure (we can compute the padding
3290 bytes remotely for all the architectures I have access to)
3291 - Implement the structure parsing macro (still not quite robust/complete)
3292 - Improvement to the remote testing toysuite
3294 2004-07-10 Martin Quinson
3295 [autoconf mechanism]
3296 - get ride of a bunch of deprecated macros
3297 - actually run the test for two-compliment, not only compile it :-/
3298 - test whether the structures get packed (and bail out if yes. Damn.
3299 Alignment is a serious matter)
3300 - test whether the structures get compacted (but respecting the alignment
3301 constraints of each types)
3302 - test whether the array fields of structures can straddle alignment boundaries
3304 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3305 (took me 2 hours to find that bug, looking at the wrong place)
3306 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3307 before sending/receiving and are used to compute the offsets of fields
3308 - Given that padding size depend even on compiler options, keep track of
3309 alignment and aligned_size only for the current architecture. Not a big
3310 deal since we send structure fields one after the other (seems
3312 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3313 program, let it work on linux/gcc/little32. portability todo.
3315 2004-07-08 Martin Quinson
3316 - import and improve remote compilation support from FAST
3317 - make sure make check works on half a dozen of machines out there
3319 2004-07-07 Martin Quinson
3320 Let's say it's version 0.0.3 ;)
3321 - Implement conversions (yuhu!)
3322 - Let it work on solaris (beside conversion, of course)
3323 - Stupid me, using rand() to generate the conversion datatests in not wise.
3325 2004-07-06 Martin Quinson
3326 - Let make dist work, since I'm gonna need it to compile on remote hosts
3327 - Let Tests/datadesc_usage write the architecture on which the file was
3328 generated as first byte.
3329 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3330 architecture descriptions.
3331 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3332 on those architectures.
3333 - Optimization: send/recv array of scalar in one shoot
3335 2004-07-05 Martin Quinson
3336 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3338 - Plug a whole bunch of memleaks
3339 - each process now have to call gras_{init,exit}. One day, their log
3340 settings will be separated
3341 - Continue the code factorisation between SG, RL and common in Transport.
3343 2004-07-04 Martin Quinson
3345 - Redistribution between SG and RL.
3346 We wanna have to accept in SG, so move accepted related parts of RL in
3347 the common part. (more precisely, the dynar of all known sockets is no
3348 more a static in transport.c, but part of the process_data)
3350 [gras_stub_generator]
3351 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3353 2004-07-03 Martin Quinson
3354 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3355 them all up in one shot)
3357 - Fix the ugly case of reference to dynamic array.
3358 - New (semi-public) function gras_datadesc_size to allow the messaging
3359 layer to malloc the needed space for the buffer.
3361 - gras_socket_close now expect the socket to close (and not its address to
3362 put NULL in it after it). This is because the socket passed to handlers
3363 is one of their argument (=> not writable).
3365 - propagate the interface cleanup from last week in datadesc, ie remove a
3366 superfluous level of indirection. User pass adress of variable
3367 containing data (both when sending and receiving), and not of a variable
3368 being a pointer to the data. Let's say that I like it better ;)
3369 The price for that is constructs like "int msg=*(int*)payload" in
3370 handlers, but it's a fine price, IMHO.
3372 - Let it work in RL (yuhu)
3374 2004-06-21 Martin Quinson
3376 - porting SG plugin and SG select to new standards (works almost).
3377 - plug memleaks and fix bugs around.
3380 - cleanup the prototype of data recv and force users to specify when they
3381 want to handle references to objects. Test case working even for cycles.
3382 - plug memleaks. Valgrind is perfectly ok with this.
3384 2004-06-12 Martin Quinson
3386 - cleanup the separation between plugin and main code in plugin creation
3388 2004-06-11 Martin Quinson
3390 - Reput hook for raw sockets, needed for BW experiments
3391 - kill a few lines of dead code
3392 [Data description] Interface cleanup
3393 - gras_datadesc_by_name returns the searched type or NULL.
3394 That way, no variable is needed to use a type desc once, which makes
3396 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3397 two parameters were strings (field name, type name), leading to
3399 [Dicos] Interface cleanup
3400 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3401 This is consistant with the dynar API.
3403 2004-04-21 Martin Quinson
3405 - Porting to new standards.
3407 - interface cleanup.
3408 There is no bag anymore, no need to take extra provision to mask the
3409 pointers behind "ID".
3410 Better splitup of functions between files create/exchange/convert.
3411 This is still a bit artificial since convert and receive are so
3412 interleaved, but anyway.
3414 - add a queued message list to procdata (the ones not matching criteria
3416 - factorize some more code between SG and RL wrt procdata
3418 - use gras_exit in example to track memleaks
3419 - get rid of gs_example now that GS is properly integrated into gras
3420 - update run_test to integrate the latest tests (datadesc)
3422 - rename WARNINGn macros to WARNn since it prooved error-prone
3424 2004-04-19 Martin Quinson
3426 - register init/exit functions within gras module mechanism
3427 - send/receive function.
3428 Convertion is not implemented, but short-cutted if not needed.
3429 struct/array elements are sent one by one (instead of block-wise), but
3430 nobody really cares (yet). Get a prototype before optimizing.
3431 - tests (using a file socket) for DD send/receive on:
3432 - base types: int, float
3433 - array: fixed size, string (ie ref to dynamic string)
3434 - structure: homogeneous, heterogeneous
3435 - chained list, graph with cycle
3436 Believe it or not, valgrind is not too unhappy with the results. The
3437 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3441 - Bugfix when using a filename explicitely (instead of '-')
3443 2004-04-09 Martin Quinson
3445 - factorize more code between RL and SG in socket creation
3446 - Complete the implementation and tests of:
3448 o file (only in RL, and mainly for debugging)
3450 I lost 3 days to design a portable address resolver, and then decided
3451 that the prototype mainly have to run on my box.
3452 Addressing portability too early may be like optimizing too early :-/
3454 - use gras_init in the Tests instead of the crappy parse_log_opt
3455 (the latter function is removed)
3456 [Conditional execution]
3457 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3458 [Code reorganisation]
3459 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3460 Build examples against the RL library, since there is no way to disable
3461 its creation for now.
3463 For information, the beginning of coding on GRAS was back in june
3464 2003. I guess that every line has been rewritten at least twice since