1 SimGrid (3.16) UNRELEASED
4 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
5 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
7 XML platforms: Switch to platform v4.1 format.
8 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
9 - <zone> can be used as a synonym for the now deprecated <as>
10 - <zoneZoute> an be used as a synonym for the now deprecated <asroute>
11 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
12 - <actor> can be used as a synonym for the now deprecated <process>
13 - state_file and avail_file periodicity is now easier to express
14 (check the documentation) the old behavior should still work.
15 * Storage: not backward compatible, but it seems that nobody ever used it
16 - Remove attribute 'content_type' of <storage_type>: was never used
17 - Make attribute 'model' of <storage_type> optional: for future usage
18 - Remove Bconnection model property: was never (in)validated. Replaced by
19 the maximum of the read and write bandwidth as a resource constraint.
22 - New and Backwards Compatibility break:
23 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
24 When one wants to get the list of tasks whose states have changed during a
25 simulation round, s/he has to allocate and free a dynar and use it as argument
26 to this function. The former SD_simulate (double how_long) now returns void.
29 - The netzone are now available from the MSG API.
30 The old names still work, but are now deprecated.
33 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
34 - New option: smpi/keep-temps to not cleanup temp files
35 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
38 - Replay: New function xbt_replay_action_get():
39 Retrieve the function previously associated to an event type.
40 - DROPPED FUNCTION: xbt_str_varsubst()
41 - DROPPED MODULE: strbuff. We don't need it anymore.
42 - DROPPED MODULE: matrix. We don't need it anymore.
44 -- Release target: June 21 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
46 SimGrid (3.15) stable; urgency=low
48 The Spring Release: continuous integration servers become green
49 We fixed even the transient bugs on all target architectures:
50 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
51 Windows, FreeBSD, NetBSD.
53 - Rename NetCards to NetPoints.
54 This was intended to help NS3 users, but that's not a netcard.
55 That's a point in the routing algorithm, let's avoid wrong simplifications.
58 - New: MSG_process_yield(). Stop and yield to other processes.
59 - New: MSG_process_daemon(). Daemon processes are automatically killed
60 when the last non-daemon process terminates
61 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
62 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
63 to make room for the upcoming network energy plugin.
64 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
67 - Ensure that an actor can kill itself with Process::exit()
68 - Kill the obscure NativeException. Nobody want to survive the issues
69 it denotes, so use JniException that is a RuntimeException (not to
70 be caught explicitly).
71 - Partial bug fix in initialization. SimGrid flags on command line were
72 consumed at C level but stayed in the original Java String[] args.
73 This could mess users' args[i] if SG flags were not put at the end of
75 The SimGrid flags are now removed from the Java arguments. However,
76 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
77 if args.length is greater than the number of YOUR OWN ARGUMENTS.
78 It might be if you have --log or --cfg flags in the command line.
79 - Fix numerous memleaks all around the place. In particular, around VMs.
83 - s4u::onPlatformCreated: right before the simulation starts
84 - s4u::onSimulationEnd: right after the main simulation loop
85 - s4u::onTimeAdvance: right after a clock change (time is discrete)
86 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
87 event from the availability_file changes the avail speed.
88 - Links are now usable from s4u
89 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
90 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
91 - New examples: The conversion of MSG examples to S4U has begun
92 - Actors: create, kill, migration, and suspend
93 - Applications: master-worker and token-ring
94 - Action replay: communications and storage
95 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
98 - Backwards Compatibility breaks
99 - The SD_route_get_list and SD_route_get_size functions have been removed.
100 They are replaced by the unique
101 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
102 The route from src to dst is built in the links dynar whose size can
103 be obtained with xbt_dynar_length.
104 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
105 sg_host_route_bandwidth and sg_host_route_latency.
106 Macros ensure the backwards compatibility, but you should fix your code
109 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
110 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
111 - Bump our claim of support from MPI 1.1 to MPI 2.2.
112 We don't support 100% of it, but it should be enough. Contact us if not.
113 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
114 naive and probably inaccurate)
115 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
116 mapped onto a unique small file using some system magic.
117 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
118 - Fortran: cleanups, fixes, support of user-added operations
119 - RMA: MPI_Accumulate are applied in correct order.
120 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
121 - Better support for MPI_IN_PLACE
122 - Support for MPI_Win attrs and keyvals.
123 - Support MPI_Comm_group_create, MPI_Type_size_x
124 - MPI ops should be properly applied to all allowed types
125 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
126 algorithm by a non failing one from mpich.
129 - Kill the fifo data container: we don't use it anymore.
132 - A <cluster> can now be created with different speed values to represent pstates.
133 see examples/platforms/energy_cluster.xml for an example.
135 -- Wed Mar 22 17:50:21 CET 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
137 SimGrid (3.14.159) stable; urgency=low
139 The previous release was only a crude approximation of a Pi release;
140 this one should be a bit better.
142 - Revert a last minute change that broke on Mac OSX.
143 - Fix the changelog and NEWS files.
144 - Improve the documentation, in particular of the routing module.
145 - Integrate some patches flying around in Debian.
147 -- Wed Dec 28 17:14:49 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
149 SimGrid (3.14) stable; urgency=low
152 * The whole documentation was reworked and reorganized. There is
153 still much room for improvement here, but we're on it.
156 * We now need python3 for our testing infrastructure.
157 * Model checking needs libevent
160 * Now works on FreeBSD
163 * Privatization now works on FreeBSD
165 * Privatization is now activated by default in smpirun
166 You can turn it off with -no-privatize if something goes wrong.
168 * Call-location tracing for SMPI.
169 You can add the exact location (filename / linenumber) of an MPI call to
170 your trace files and slow down or speed up the simulation between two
171 consecutive calls by using an adjustment file (see the documentation).
173 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
174 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
175 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
176 values were ignored in some cases. The timings of these functions can now
177 be significantly different.
179 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
180 smpi/running-power is renamed to smpi/host-speed
182 * smpi/grow-injected-times option to enable or disable multiplication of the
183 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
184 make simulation less precise (but also much faster).
186 * smpirun script should be (much) faster for large deployments.
188 * SMPI tracing : fixed issue with poor matching of send/receives.
190 * Replay : Fix broken waitall
192 New functions and features
193 * MSG_parallel_task_execute_with_timeout, to timeout computations.
195 Dropped / renamed functions and features
196 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
197 - MSG_mailbox_is_empty() -> Mailbox::empty()
198 - MSG_mailbox_front() -> Mailbox::front()
199 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
200 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
201 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
202 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
203 * Don't pass the free_f parameter to property related functions:
204 - MSG_host_set_property_value()
205 - MSG_as_router_set_property_value()
206 - MSG_storage_set_property_value()
207 * VM properties. Since msg_vm_t are msg_host_t, just use
208 MSG_host_get_property_value() and friends
209 * VM I/O related things:
210 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
211 - Unimplemented save/restore methods
212 * MSG_as_router_get_property_value() was redundent with
213 MSG_environment_as_get_property_value().
214 - Removed MSG_as_router_*propert*() functions
215 - Added MSG_environment_as_set_property_value() for consistency
216 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
217 Well, they are still there for now, but deprecated with a warning.
218 Please switch to xbt_dict_new_homogeneous() before this is removed
220 * Task affinity. Its intended behavior (that was very badly tested
221 and probably not really working) was deceiving what most users
222 would have hoped here.
223 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
224 * Option network/coordinates is now useless and should be dropped.
227 * Added option storage/max_file_descriptors to allow more than 1024 files opened
230 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
231 * Added Dragonfly topology support
234 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
235 EiBps, EBps, Eibps, Ebps for bandwidth.
236 They may become useful to some lucky ones.
239 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
240 * ex/app_tokenring: new example, very similar to the MSG Token Ring
241 * ex/async_waitAll: new example, on asynchronous communications
244 * Memory usage should be decreased for simulations with a large number
245 of processes. This also helps for SMPI.
247 -- Sat Dec 24 22:34:26 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
249 SimGrid (3.13) stable; urgency=low
251 The Half Release, a.k.a. the Zealous Easter Trim.
253 - We removed half of the lines, that were mostly experimental cruft.
254 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
255 (not counting blanks and comments -- according to openhub.net)
256 - The internals are now compiled in C++ (and will soon be clean C++)
257 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
258 59 klines of C, etc. We added only 29 klines of C++ in replacement.
260 * Backwards Compatibility breaks
261 - Removed Lua simulation bindings (switch to C or Java for that).
262 Lua can still be used to describe platforms
263 - Removed Java kernel plug-ins.
264 Will be reintroduced after the ongoing major internals reorg.
266 - the following functions were removed.
267 They were too specific and should be reimplemented in a generic
268 way, with filter function.
269 - MSG_task_listen_from_host
270 - MSG_mailbox_get_count_host_waiting_tasks
271 - MSG_mailbox_put_with_timeout was removed.
272 Please use MSG_task_send_with_timeout instead.
274 - the SD_application_reinit function was removed. It has been a noop for a while.
275 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
276 replaced by a more flexible mechanism.
277 - The following functions thus do not exist anymore
278 - SD_workstation_get_access_mode
279 - SD_workstation_set_access_mode
280 - SD_workstation_get_current_task
281 - Basic estimation functions have been removed but can easily be replaced
282 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
283 - SD_workstation_get_computation_time => amount / sg_host_speed()
285 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
286 Use VM.getSpeed()*load/100 for the legacy behavior.
288 - option enable_tracing was removed. It was not doing anything for a while.
289 - In the ModelChecker:
290 - the model-checker now ptraces the model-checked process which means
291 you cannot use a debugger on the latter anymore (we might make this
292 optional in the feature);
293 - removed soft-dirty page tracking;
294 - remove model-checked side snapshot management,
295 MC_snapshot() and MC_compare_snapshot();
296 - keep the MC_cut() function as a stub (it was not really working
297 in the previous release).
300 * All options are consistently in kebab-case. Old names are kept as alias.
303 * Switch to platform v4 format.
304 - Rename from 'power' to 'speed' the attributes describing the amount of
305 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
306 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
307 - In <host> and <link>, attributes availability and state are gone.
308 It was redundent with state and availability traces, and with peak values.
309 - In <cluster>, attributes availability_file and state_file are gone.
310 It was too complex and unused.
311 - Kill <gpu>. Was not doing anything.
312 - The DOCTYPE points to the right URL:
313 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
314 (the file at this address now documents the changelog since its v1)
315 - A warning is emitted for unit-less values (they are still accepted).
316 - speed. Default: 'f' or 'flops'. Also defined:
317 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
318 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
319 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
320 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
321 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
322 - latency. Default: 's' second. Also defined:
323 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
325 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
326 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
329 * s4u::Host is now the preferred public interface to the Host features.
330 sg_host_* functions are C bindings to the exact same behavior
331 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
334 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
335 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
336 They were deprecated since a few years, and probably did not even compile anymore.
339 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
340 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
341 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
342 encouraged to update their codes. A list of the modified functions can be found at the end of
343 include/simgrid/simdag.h
346 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
347 the lambda or closure passed as a parameter will run in kernel mode.
348 All the callback functions should be rewritten to that interface at some point.
351 * Reorganizing and cleaning the internals all around the place.
354 * Remove old default barrier/bcast buggy algorithms (see #18407)
355 * Various bug fixes to handle more codes
356 * Remove the need for the --foreground option of smpirun (it is still
357 accepted for backward compatibility).
360 * Kill the setset data container: MC don't use it anymore.
361 * Kill the queue data container: it made more sense with GRAS.
362 * Kill the xbt_peer_t data type: it's useless without GRAS.
363 * Kill rm_cb feature of config sets: it was never useful.
364 * Kill graphxml parsing feature. It was not used.
365 * Kill the deprecated code protected by XBT_USE_DEPRECATED
367 - xbt_dynar_sort_strings(), when the content is char*
368 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
369 They throw exceptions on invalid input;
370 * C++ support for declaring CLI flags (simgrid::config::Flag);
371 * class for abstracting different signal backends (simgrid::xbt::signal).
372 with no external dependencies (we need very simple signals).
375 * refactoring and cleanup of the code;
376 * ongoing process to cleanly separate the model-checking algorithms
377 from the code model-checking support.
379 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
381 SimGrid (3.12) stable; urgency=low
383 The Facelift Release.
386 * Require g++ v4.7 at least to not speak prehistorical C++.
387 * Require Boost 1.48 (for signal2 component).
388 * Java must be version 7 at least when activated.
389 * Builds on Windows again (including Java bindings).
390 * Tracing is now always enabled (no way to turn it out)
391 * Remove GTNetS. It was not working anyway.
392 * Various cleanups in the cmake scripts.
393 * Move headers around to sort them out on installed systems:
394 - instr/instr.h -> simgrid/instr.h
395 - instr/jedule/* -> simgrid/jedule
396 - simdag/datatypes.h was removed
397 - simdag/simdag.h -> simgrid/simdag.h
398 - msg/datatypes.h was removed
399 - msg/msg.h -> simgrid/msg.h
402 * Interface improvement:
403 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
404 - Sanitize the interface in MSG_task_ module:
405 - Merge two functions that were close enough but misleading:
406 set_compute_duration(t) -> set_flops_amount(t)
407 get_remaining_computation(t) -> get_flops_amount(t)
408 - set_data_size(t) -> set_bytes_amount(t)
409 get_data_size(t) -> get_bytes_amount(t)
410 - Massive cleanups in the functions related to the energy
411 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
412 - New: MSG_host_get_pstate()
414 - msg/energy/onoff: switching hosts on and off
416 * Interface improvement:
417 - Rename Host.isAvail() to Host.isOn()
418 - Rename Process.currentProcess() to Process.getCurrentProcess()
419 - Rename Task.setDataSize() to Task.setBytesAmount()
420 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
422 - #18874: Actually allows the GC to reclaim tasks
426 - SIMIX_process_throw: raises an exception in a remote process
427 * Refactoring: Separate sub-modules
428 - libsmx: the public interface, as libc in a real system
429 - popping: the strange dance that converts a user request into a kernel handling
430 - smx_context_*: the virtualization mechanisms that embed the user code
431 - smx_*: the handling of each simcalls
433 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
434 * Rename smx_action_t into smx_synchro_t, making explicit that these
435 things are used to synchronize processes with their environment.
436 For example, a communication is a sort of synchronization involving
437 the communicating processes (that may block until the exchange) and
438 the platform. The same can be said from computations, etc.
440 - #18888: segfault when a process ends before its kill_time
444 - 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.
445 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
446 - MPI_Type_set_name, MPI_Type_get_name
447 - MPI_*_c2f and MPI_*_f2c functions
448 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
449 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
450 - Activate a lot of new tests from the mpich 3 testsuite
452 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
453 - InfiniBand network model added : Based on the works of Jerome Vienne
454 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
455 - When smpi/display_timing is set, also display global simulation time and application times
456 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
457 * Collective communications
458 - 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
459 - 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/).
460 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
462 - Replay now uses algorithms from wanted collective selector
463 - Replay can be used with SMP-aware algorithms
464 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
465 - 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.
467 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
468 - Correctly use loopback on fat-tree clusters
469 - Asynchronous small messages shouldn't trigger deadlocks anymore
470 * Energy/DVFS cleanup and improvement
471 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
472 - new: smpi_get_host_pstate()
475 - "Full" network optimization flag was broken since Surf++
476 - Better handling of precision flags in maxmin
477 - Fix bug causing sometimes "Impossible" errors
478 - Properly pass cluster properties to included hosts
479 * Improvement of the Energy plugin.
480 - Always update the consumption before returning that value
481 - New property: watt_off to denote the dissipation when the host is off
482 - New functions getWattMinAt and getWattMaxAt to retrieve the
483 dissipation of pstates that we are not currently at.
484 * Java: class NetworkLink renamed to Link
485 * New function: simcall_process_get_kill_time()
486 * Massive rename s/workstation/host/
487 - That's intrusive, but that's good for the project consistency. Sorry.
488 - Change config option "workstation/model" into "host/model"
491 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
492 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
493 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
494 - xbt_dirname and xbt_basename for non-POSIX systems
496 * The model checker now runs as a separate process.
497 * The model checker runs must now be launched with the new simgrid-mc program.
498 * Record/Replay: the MC can display a textual representation of a path in the
499 execution graph. It can then be replayed outside of the model checker.
501 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
503 SimGrid (3.11) stable; urgency=low
508 * Normalizing pointers addresses tool for better diff between logs
511 * Add cloud examples using new VMs
512 - examples/msg/cloud/two_tasks_vm.tesh
513 - examples/msg/cloud/simple_vm.tesh
514 - examples/java/cloud/cloud.tesh
515 - examples/java/cloud/migration/migration.tesh
516 * Add java surf examples:
517 - examples/java/surfPlugin/surf_plugin.tesh
518 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
519 - examples/java/surfCpuModel/surf_cpu_model.tesh
520 * Add SMPI+MSG example:
521 - examples/smpi/smpi_msg_masterslave/
526 - msg task destroy cancel test
527 - msg_host on/off test
528 * Move all tests in testsuite to teshsuite (adding tesh files)
529 * Restructure teshsuites
530 - one folder for each kind of test
531 * Restructure AddTests.cmake
533 - structure the order of tests (with sections)
536 * Add virtual machine
537 - creation of a VM on a PM
538 - migration of a VM from a PM to another PM
540 - MSG_process_join(msg_process_t process, double timeout)
541 - msg_bar_t MSG_barrier_init(unsigned int count)
542 - int MSG_barrier_wait(msg_bar_t barrier)
543 - void MSG_barrier_destroy(msg_bar_t barrier)
544 - msg_as_t MSG_environment_as_get_by_name(const char * name)
545 * New option "msg/debug_multiple_use" to help debugging when a task is used
548 - Improvements and finalization of MSG_storage, MSG_file APIs and their
550 - Increase code coverage in test suites
554 * Protect context stacks against stack overflow. The number of protected memory
555 pages allocated on the top of each stack (1 by default) can be configured
556 with the new command line option --cfg=contexts/guard_size:<value>.
557 * Simcalls are now generated by a python script that
558 - generates files included by SimGrid
559 - checks that all the functions exist, and proposes prototypes
561 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
562 host_set_data, host_get_data
564 - simcall_process_join(smx_process_t process, double timeout)
565 * Fix bug where sleeping processing could not be suspended.
568 * Translate surf models from C to C++
569 - Generic classes for all models: Model, Resource, Action
570 - A generic interface for each kind of model (CPU, Network, Storage
571 Workstation, WorkstationVM)
573 * Translate surf routings from C to C++
574 * Add callbacks using sigc++ or boost::signals2
575 - Add callback functions for resource creation/destruction
576 - Add callback functions for action state change
577 - Handle Energy as a plugin
578 * Replace swag by boost::intrusive
579 * Add new routing models for clusters. For documentation, see
580 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
581 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
582 parameters for cluster tag
583 - Fat trees, with topology="FAT_TREE" and
584 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
585 - see examples/platforms/torus_cluster.xml and
586 examples/platforms/fat_tree_cluster.xml
590 * Hostfiles support host:nb_processes construct to deploy several processes on
592 * Collective communication algorithms should not crash if used with
593 improper number of nodes and report the error.
594 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
595 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
596 MPI_Dims_create, MPI_Cart_sub are supported.
597 * New interface to use SMPI programmatically (still depends on MSG for
598 some parts, see examples/smpi/smpi_msg_masterslave) :
599 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
603 * Global variables privatization in MPI executables is now performed at runtime
604 with the option smpi/privatize_global_variables (default:no).
605 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
606 dynamic libraries loaded by the application are not privatized (static
607 linking with these libraries is advised in this case)
610 * Options defined in XML work correctly now.
613 * New cmake option, enable_lib_in_jar, to control whether native libraries are
614 copied into simgrid.jar or not (ON by default). Use this option if you want
615 to reduce the size of the installed simgrid.jar, *and* the native libraries
616 are kept installed elsewhere.
617 * Surf binding with SWIG (code generated in maintainer mode only):
618 - plugin to handle callbacks
619 - CPU model only for the moment
622 * Supernovae build mode is definitively removed. It was used to improve
623 inlining and inter-module optimizations. It is nowadays superseded by
624 link-time optimizations commonly available in compilers.
625 * Update ns-3 find lib. Bindings for ns-3 should work again now.
626 * Add boost dependency for surf++
627 * Add new macro for tests
628 - ADD_TESH(name <tesh_args>)
629 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
633 - xbt_bar_t XBT_barrier_init(unsigned int count)
634 - int XBT_barrier_wait(xbt_bar_t barrier)
635 - void XBT_barrier_destroy(xbt_bar_t barrier)
636 * Make the xbt_os_time module public
638 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
640 SimGrid (3.10) stable; urgency=low
642 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
645 * Reintegrate Java to the main archive as desynchronizing these
646 package is not acceptable anymore (Java is now considered stable)
647 * Add explicit synchronization facilities through semaphores
648 * Bug fix: Task.setDataSize() only changed the C world, not the value
649 cached in the Java world
652 * Dramatically change the way files are handled. API and internals changed, but
653 this part of MSG was not considered as production grade either.
654 * Add explicit synchronization facilities through semaphores
655 * Add a new function MSG_host_get_process_list()
656 * Preliminary DVFS support (see examples/msg/energy/ for details)
659 * SMPI is now included directly in the libsimgrid as the windows
660 linker doesn't force us on splitting it anymore.
661 * Improvements of the SMPI replay tool:
662 - Most of the collective communications are now rooted in the same process as
663 in the original application.
664 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
665 used until now). Multiple data types can now be used in a trace.
666 - The replay tool now supports traces produce either by TAU or a modified
668 - Bug Fix: the compute part of the reduce action is now taken into account.
669 - Gatherv collective is now supported
670 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
671 of smpirun outputs time independent traces for the current run. One file
672 is created per process. If too many processes are simulated, this behavior
673 can be changed to one file for all processes by using the
674 tracing/smpi/format/ti_one_file flag
675 * smpirun generates the host file if needed (with given host count and platform)
676 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
677 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
678 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
679 --cfg=smpi/coll_selector:(mpich/ompi)
680 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
682 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
683 happening inside a collective SMPI call.
684 * Fix the behavior of complex data types handling.
685 * Make MPI_Wtime another synchronization point to take computations into
687 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
688 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
689 * Add all missing Fortran bindings, SMPI should work with Fortran 90
690 (no privatization of global variables yet)
691 * Preliminary DVFS support (see examples/smpi/energy/ for details)
694 * Verification of liveness properties is now available for SMPI applications
695 (in addition to MSG applications)
696 * Bugged examples using SMPI in examples/smpi/mc/
697 * Add --cfg=model-check/visited option. Allows the verification of infinite
698 programs. Detection of loops in the execution thanks to the system state
699 comparison and reduction of the state space to explore. Can be combined with
700 DPOR for safety properties.
703 * Allow to change SimGrid configuration (see --help) within the code
704 thanks to SD_config() as it can be done in MSG.
705 * Add a new function SD_task_set_amount() upon user request.
708 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
709 * Remove rule based routing (no more PCRE dependency)
710 * Add a limiter_link option to cluster tag, to specify a maximum reachable
711 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
712 * Add a loopback_bw and loopback_lat options to cluster tag.
713 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
714 routing. Now peers are encapsulated in an AS and have their own private
715 router but this is transparent.
718 * Our own implementation of getline is renamed xbt_getline, and gets
719 used even if the OS provide a getline(). This should reduce the
720 configuration complexity by using the same code on all platforms.
721 * New type: xbt_cfg_elm_boolean.
722 * Allow to use yes/no for boolean configuration options in the command line.
723 * Allow to disable SimGrid cleanups at exit from command line option.
724 There are situations where one may want a simulation to end with an exit.
725 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
726 annoying when scripting around the simulator. Adding a
727 --cfg=clean_atexit:no allows to circumvent this issue.
730 * Lots of memory leaks were corrected in this release.
731 There is no known memory leaks anymore, in all of our 600+ tests.
732 * New command line option --version, to get SimGrid version information.
733 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
735 * Supernovae builds are deprecated, and expected to be removed in the next
738 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
740 SimGrid (3.9) stable; urgency=low
742 The Grasgory release: GRAS is really dead now.
744 * Complete overhaul of the internal host structures scheme.
747 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
748 was considered as experimental and badly maintained since then.
749 * Keeping it was thus a trap to our potential users, that could take
750 it instead of MSG or SMPI by mistake despite is pity state.
751 * GRAS seems to have very few users (if any), and no one volunteered
752 to maintain it further. It also induces a lot of XBT code (for
753 portability sake), that must be maintained too.
754 * For all these reasons, we killed GRAS. If someone wants to revive it
755 in the future, don't cry, our git history still remembers of GRAS.
758 * Major overhaul. Merge our documentation again as time proved that
759 splitting it was really not helping our users.
760 * Further improve the developer documentation to help newcomers
761 hacking on SimGrid itself. The user documentation (and in
762 particular, the beginner documentation) is still in a sorry state.
765 * Now works on Windows too!
766 * Much more extensive test suite, from MPICH
769 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
770 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
771 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
772 * Bug fix: let task be scheduled when the last dependency to be solved is
773 a control dependency.
774 * Remove SD_load_environment_script function.
775 Use the C sg_platf function if you want to declare a platform
779 * New function: MSG_process_get_number()
780 * Old function documented: MSG_config()
781 * Remove MSG_load_platform_script function
782 Use the C sg_platf function if you want to declare a platform
786 * Change the default value of the TCP_gamma constant (maximal size of TCP
787 congestion window) to a more realistic 4MiB value. If you notice changes in
788 your simulation results, you can fall back to the previous 20k tiny window
789 by adding --cfg=network/TCP_gamma:20000 on command line.
790 * (Hopefully) fix a bug wrt periodic availability and state traces
791 * Bug fix: use default values at start when first event in availability/state
792 trace is not at time 0.
795 * remove the "new_" part of function name sg_platf_new_trace_connect
796 (resulting in sg_platf_trace_connect), since it does not create
800 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
801 use case where it's really mandatory, and maintaining it was a pain in
803 * New: xbt_fifo_search(), search an item with a user-provided
804 comparison function instead of dumb pointer comparison.
807 * Fix the lua deployment:
808 Use `simgrid.init_application()` before deployment instead of
809 `simgrid.msg_register_application()` after.
812 * Transfer the tracing files into the corresponding modules.
814 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
816 SimGrid (3.8.1) stable; urgency=low
818 The "we are told that some people want to also *install* the simgrid
821 * Add missing file "tesh.1" to the archive.
823 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
825 SimGrid (3.8) stable; urgency=low
827 The Psssshiiiit release: SimGrid jumps into the Cloud.
830 * Add an experimental interface to manipulate VMs. They are mainly
831 process groups with very few intrinsic semantic, but they should
832 allow you to build the semantic you want easily.
833 * New function: MSG_host_set_property_value()
834 * New function: MSG_process_on_exit(). To clean memory in all cases.
835 * Bug fixes that made the host (and link) failures unusable.
836 * Add a way to auto-restart process when the host in which they are
837 executing comes back (ON_FAILURE="RESTART" on deployment file,
838 MSG_process_auto_restart_set).
839 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
840 please stop using the old ones, they are DEPRECATED.
842 * Deprecate functions MSG_global_init() / MSG_global_init_args()
843 Please use MSG_init() instead. (reducing the amount of entry
844 points in the library helps us).
845 * Make it impossible to link against the wrong version of the lib
846 * Deprecate MSG_clean(). No need to call it anymore.
847 * Function MSG_get_host_number() is not deprecated anymore.
850 * Split the doc into a user guide and a reference guide.
851 * Start a developper guide to help people hacking on SimGrid.
854 * Enable tracing by default. This modules rocks you should use it.
855 * Remove option custom_flags. Now use environment variables CFLAGS
857 * Use default cmake things to detect lua instead of home grown ones.
858 * New option "enable_mallocators" to disable mallocators, for debugging
859 purpose ("on" by default).
862 * Bug fixes around the resource failures: don't let the processes
863 survive the host they are running onto.
864 * Add an interface to auto-restart processes when the host in which they are
865 executing comes back.
866 * Ensures that SIMIX_clean is called automatically. It's not part of
867 the public interface anymore (bindings should be updated).
870 * Bug fix for when SD_Simulate is called with a positive value: be careful
871 when comparing doubles. Sometimes they are different for non significant
873 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
874 parallel task whose initial work is distributed among host according
875 to the Amdahl's law. Such tasks are created with a parameter alpha
876 that corresponds to the non-parallelizable part of the computation.
877 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
878 between two sets of workstations assuming a 1D block distribution (each
879 workstation owns a similar share of data) on both sides.
881 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
882 Data redistribution will be automatically scheduled once parent and child
883 are both scheduled. The filling of computation_amount and
884 communication_amount structures is now done seamlessly thanks to the chosen
886 * New function SD_workstation_dump to display various information
887 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
888 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
889 bandwidth on the route onto which the task is scheduled and the amount of
891 To divide the nominal bandwidth by 2, the rate then has to be :
892 rate = bandwidth/(2*amount)
893 * Compute tasks that have failed can now be rescheduled and executed again
894 (from their beginning)
895 * Increasing source code coverage (src/simdag is now covered at 95.8%
899 * Re-implement time-independent trace replay using SMPI (at the
900 smpi_smp_* level) instead of MSG. This should replace
901 examples/msg/actions/actions.c
902 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
903 hindexed and structs)
904 * Implement the exchange of non-contiguous data.
905 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
906 * Correct behavior of smpi/sender_gap and set its default value to 0
907 * Add option to asynchronously send small messages to allow better
908 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
909 specifies the size in bytes under which messages will be asynchronously sent.
910 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
911 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
914 * New C interface to define a platform: XML is now optional.
915 For more info, please check include/simgrid/platf.h.
916 * New interface to define random platforms from the C:
917 For more info, please check include/simgrid/platf_generator.h and
918 examples/msg/masterslave/masterslave_platfgen.c
919 * Export a sg_cmdline dynar containing all the arguments we got from
923 * Two new tracing options for adding comments to trace file so you
924 can track your experiments (see --help-tracing for details).
925 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
926 * Adding the SimGrid version that generated the trace file as a comment.
927 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
928 * Fix to avoid key clashes on Paje links
929 * Other minor fixes related to the Paje specification
932 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
933 and renamed to xbt_str_hash() and xbt_str_hash_ext().
934 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
935 * Greatly improve the robustness of mmalloc to user errors (such as
936 using an area after freeing it, or freeing it twice)
938 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
940 SimGrid-java (3.8.1) stable; urgency=low
942 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
943 Warning: all methods are visible, but only some of them are
944 implemented so far. Check the source (src/jtrace.c)
945 for further information.
946 * New module: org.simgrid.msg.File (SimGrid File management functions)
947 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
948 * Change the meaning of Process.restart: now restart the process from
949 the begining, like MSG_process_restart in C.
950 * Add Process.setAutoRestart: handling of process restart when failed
952 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
953 you to retrieve the properties of the processes/hosts
954 * Deprecate Msg.clean(): you can just forget about it now.
955 * New function Process.getCount(), that only works when compiling
956 with the not yet released version 3.9 of the C library.
958 * New context factory based on Coroutines. It mandates a modified JVM
959 but then, the simulations run about five times faster, and there is
960 no limit to the amount of processes (beside of the available memory).
962 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
964 SimGrid (3.7.1) stable; urgency=low
967 * Restore the prototype of MSG_process_create_with_environment() to
968 the pre-3.7 situation by removing the kill_time argument.
969 * Add a MSG_process_set_kill_time() function instead.
972 * Fix weird behaviors when dealing with parallel tasks.
975 * Simgrid is now built as a dll.
976 * Simgrid-java now works on Windows.
977 * Simgrid-Java is now included into Windows package.
980 * First pre-build package for MacOSX.
983 * Fix compilation when using MSG_USE_DEPRECATED.
984 * Fix some compilation issues on Macs and Windows.
985 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
986 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
988 We discovered that the Lua console is broken, but we are missing the
989 manpower to fix it right now. The problem existed in 3.7 too, so we
990 are not blocking the release for that. Sorry if you depended on this
991 feature, any help would be really welcome.
993 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
996 SimGrid-java (3.7.1) stable; urgency=low
998 The "Java aint got to be bloated and slow" release
1001 * Various internal cleanups and performance improvement
1002 Simulations are expected to run up to twice faster or so
1003 * Make Process.kill(process) an instance method, not a static one
1004 * User processes are not java.lang.Thread subclasses.
1005 This breaks the compatibility (sorry), but previous API was
1006 brain-dead, making it impossible to have non-trivial
1007 initializations in the process constructor.
1008 * Require a full constructor per Process sub-class.
1009 Kinda breaks the compatibility (sorry), but this allows a much more
1010 efficient way to launch the processes at simulation startup.
1011 * Do not embeed our version of semaphores, java 1.5 can be considered
1012 as sufficiently prevalent for us to not dupplicate its features.
1016 * Add examples for almost every part of the API
1017 We spotted and fixed a lot of bugs in the process
1019 * New module: asynchronous communication API
1020 * New function: Process.sleep()
1021 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1022 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1023 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1025 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1027 SimGrid (3.7) stable; urgency=low
1029 The "spring cleanups (before the next Big Project kicks in)" release.
1032 * We can specify the SMPI latency/bandwidth factor with command line
1033 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1034 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1035 You can also use the "config tag" from platform file by adding this line
1036 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1037 (see "example/platforms/tag_config.xml" to use "config tag").
1038 Note that the command line supersedes the platform file configuration.
1039 * Change the correction factors used in LMM model, according to
1040 the latest experiments described in INRIA RR-7821.
1041 Accuracy should be improved this way.
1042 * Use the partial invalidation optimization by default for the
1043 network too. Should produce the exact same results, only faster.
1044 * Major cleanup in surf to merge models and split some optimization
1045 mechanisms from the core of the models. As a result you can now
1046 specify which model to use (e.g., --cfg=network/model:LV08
1047 --cfg=cpu/model:Cas01) and which optimization mode to use
1048 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1049 Incompatible combinations should err at initialization. See
1050 --help-models for the list of all models and optimization modes.
1051 * The CLM03 workstation model was dropped for simplicity because it
1052 used the deprecated CM02 network model. Use default instead.
1053 * Rename the TCP_gamma configuration option to network/TCP_gamma
1054 * Rename the coordinates configuration option to
1055 network/coordinates, and document it
1056 * Use now crosstraffic keyword instead of the terribly misleading
1057 fullduplex keyword. It is activated by default now in the current
1058 default model, use --cfg=network/crosstraffic:0 to turn it off.
1059 * Ongoing refactoring the model parsing to make XML files optional
1060 See include/simgrid/platf.h for details (still to be completed)
1063 * Major overhaul of the documentation. Almost instructive now :/
1064 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1065 functions and friends. This interface was considered as
1066 deprecated since over 2 years, it's time to inform our users that it is.
1067 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1068 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1069 use these (crappy) functions in your code.
1070 These functions will be removed soon. Stop using them now.
1071 * Deprecate MSG_get_host_{table,number}
1072 Implement MSG_hosts_as_dynar() instead.
1073 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1074 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1077 * Stabilize the parallel execution mode of user contexts
1078 * Introduce configuration variables to control parallel execution:
1079 - contexts/synchro: Synchronization mode to use when running
1080 contexts in parallel (either futex, posix or busy_wait)
1081 - contexts/parallel_threshold: Minimal number of user contexts
1082 that must be part of a scheduling round to switch to parallel
1083 execution mode (raw contexts only)
1084 * Fix bugs that prevented to use suspend/resume along with
1085 synchronization structures.
1086 * Fix bugs in process termination that lead to invalid memory access
1087 in very specific conditions.
1090 * Introduce a parallel mode for the models (controlled by surf/nthreads
1091 configuration item). In our tests, running the models in parallel
1092 never lead to any speedups because they are so fast that the gain
1093 of computing each model in parallel does not amortizes the
1094 synchronization costs, even when ultra fast futexes are used.
1095 This is released anyway because YMMV.
1098 * Performance boost by using a swag internally to compute the set of
1099 tasks that are finished and should constitute the return value of
1103 * Enable it by default now that it is considered rather stable.
1106 * Documentation of the tracing functions.
1107 * Performance gains when tracing categorized/uncategorized resource
1108 utilization by avoiding calls to get route when updating resource
1109 variables. LMM constraints are being used instead.
1110 * API changed to set task categories. Use MSG_task_set_category instead
1111 of TRACE_msg_set_task_category, and SD_task_set_category instead
1112 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1113 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1114 * Examples for link user variables added (see at examples/msg/tracing/)
1115 * Deprecated function TRACE_msg_set_process_category completely removed
1116 * Trace header updated according to the latest Paje file format
1117 * Tracing network lazy updates, no longer obligate users to use full updates
1118 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1119 * Experimental: let user code declare/set/push/pop application states for hosts
1120 * API changed to allow the manual creation of graph configuration files
1121 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1124 * Improve the API of Lua MSG bindings, using the Lua spirit.
1125 * Each simulated process now lives in its own Lua world (globals are
1126 automatically duplicated). It helps writing simulators. Will allow
1127 to run Splay programs within SimGrid in the future.
1128 * Add a Chord example in Lua, equivalent to the MSG one.
1131 * Start the implementation of a solution to express temporal
1132 properties, not only local assertions. This is still an
1133 experimental work in progress, stay clear from it to be safe.
1137 - Add new runtime parameters --help-logs and --help-log-categories
1138 to display information about supported logging parameters and
1140 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1141 don't exists anymore.
1142 * Mallocators: allow value NULL for the reset function.
1144 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1145 create homogeneous dictionaries, where all the elements share the
1146 same free function. Non homogeneous dictionaries will be
1147 deprecated in the next release.
1148 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1149 - Multi-level dictionaries are deprecated.
1151 - new function xbt_dynar_search_or_negative() that is useful when
1152 you have less than 2 million elements in your dynar and don't
1153 want of the extra complexity of catching exceptions when the
1154 element is not found.
1156 - Make xbt_os_thread module (for thread portability) public.
1157 Documentation is still to come, sorry.
1159 - Cleanups and simplifications to make it maintainable again.
1160 - Exotic features (such as memalign and valloc) were removed.
1161 - The metadata were extended and improved so that the
1162 model-checker becomes able to explore and inspect the heaps.
1163 - This may induce a performance drop when enable_model-checking is
1164 ON in cmake (even if it's not used in the simulation), but it is
1165 necessary at this point to get MC working.
1167 Turn model-checking OFF if simulation performance matters to you.
1168 Not enabling it at runtime is not enough, disable it in cmake.
1170 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1173 SimGrid (3.6.2) stable; urgency=low
1175 The "Not coding new stuff allows to polish old things" release.
1178 * New bindings to the NS3 packet level simulator (experimental)
1179 * Use the raw (efficient) execution contexts instead of the sysv
1180 (portable) ones when possible.
1181 * libpcre is now mandatory in any cases since not using it led to
1182 severe performance loss and possibly other issues
1183 * Update the XML platforms:
1184 - G5K: include the latest machine in Nancy
1185 - GridPP and LCG: new platforms
1186 * Documentation was partially updated, at least (more to come)
1188 Bug fixes, cosmetics and small improvements
1189 * Free terminated processes before the end of the simulation to avoid
1190 exhausting the memory of users having very dynamic amount of
1192 * Bug fix and cosmetics about canceling non-running tasks
1193 * Bug fix about the dot loader's issues when using libcgraph
1196 * Create an installer for windows with nsis (amd64 and win32)
1197 - Add an hello world project to illustrate simgrid project creation.
1198 - Embed libpcre into the Simgrid installer to avoid
1199 its compilation burden
1200 * The raw execution contexts should work on Apple now
1201 * Port to Windows 64 bits
1202 - Sysv contexts now have an implementation for this arch
1203 - GRAS communication features now support this arch
1204 * Drop support for borland compiler on windows
1205 - this code was not maintained, and we kinda depend on gcc nowadays
1206 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1207 * Fix portability issue on unstable ubuntu: linker became picky on
1210 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1212 SimGrid (3.6.1) stable; urgency=low
1214 The "Oops, we broke Macs too" release
1217 * Fixed contexts detection so that raw ones are used when possible
1218 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1219 leads to a strange error, with user code segfaulting sometimes when
1220 the generated code is not perfectly aligned (which is not
1221 controllable from the user side, depends on the amount of code)
1224 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1226 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1228 SimGrid-java (3.6) unstable; urgency=low
1231 * Split of every thing from simgrid v3.5 into a separate package.
1233 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1235 SimGrid (3.6) stable; urgency=medium
1237 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1240 * Bindings now constitute their own package, separated from the main one.
1241 Rationale: reduce our maintenance nightmare by reducing the module coupling
1242 They will soon be released on their own on gforge.
1244 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1245 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1247 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1248 * It's not quite deprecated for now because we have no replacement,
1249 but it may soon become the case.
1252 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1253 * Fortran: New user-level cache variable to store the rank of the running
1254 process. This improves performance by an order of magnitude.
1255 * C: New coccinelle script to automatically locate and modify global and
1256 local static variables.
1257 * Improved SMPI network model with a sender-side gap to account for multiple
1261 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1262 only say if a communication is finished, no matter whether it succeeded or
1263 failed. You can call MSG_comm_get_status() to know the status of a finished
1265 * New function MSG_task_dsend() to send a task and detach it. When a
1266 communication is detached, you are never notified of its success or failure
1267 and the memory is released automatically once it is finished. This function
1268 is useful when you don't care about the end nor the success of a
1270 * Change the prototypes of action replay. Sorry for inconvenience,
1271 but this is really more efficient this way (and to adapt your code,
1272 you just have to fix the initialization, that shouldn't be too long)
1273 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1274 ever managed to do anything useful with it.
1275 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1276 takes a dynar of communications. It returns immediately and gives the
1277 index of a finished communication (if any).
1278 * New example: a basic implementation of the Chord P2P algorithm.
1281 * New model for multi-core CPUs. You can now use the core attribute to
1282 precise the number of cores of a host. This is a basic model. Every
1283 process running on the host receives at most the power provided in
1284 the DTD (throughput<=power). Total throughput of process cannot exceed
1286 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1287 router linked by an up-link and a down-link (possibly asymmetrical).
1288 This kind of pattern allows to easily build last-mile model style platforms.
1289 Aggregating such patterns in a rule-based AS is thus the technique of
1290 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1291 * New model for Vivaldi routing. We transformed the Vivaldi network model
1292 into a Vivaldi routing model (based on the rule-based model). This allows to
1293 combine Vivaldi based latencies with last-mile platforms.
1296 * Added a check for NaN of IEEE754 infinite in the double entries of
1298 * Introduce a new context factory "raw", highly inspirated from the
1299 ucontext factory, but using manually crafted functions in assembly to
1300 do the work in an efficient manner.
1301 * Allow to change the used context factory at run time, not only at
1302 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1303 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1304 contexts at runtime (only with raw contexts or ucontexts).
1305 * Completely rewrote this module to allow parallel execution of user
1306 processes. Use --cfg=contexts/nthreads:N to execute user processes
1307 with N parallel threads (the default is 1, meaning no parallelism).
1308 * Allow to decide dynamically between sequential and parallel modes.
1309 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1310 processes in parallel only when their number is greater than or equal to P
1312 * Added a check for NaN of IEEE754 infinite in the double entries of
1316 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1317 won't output the state of processes when interrupted with Ctrl-C
1318 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1319 NULL-terminated array. This may solve backward compatibility issues
1320 due to the change to return type of SD_simulate. See also:
1321 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1322 * Add new macros with variable number of arguments.
1323 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1324 - in xbt/asserts.h: xbt_assert
1325 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1326 - in xbt/ex.h: THROWF and RETHROWF.
1327 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1329 * Change xbt_die() to accept a format string with arguments, just like printf.
1330 * New data structure: xbt_lib_t, like a dict but more general and with better
1334 * New configuration options
1335 Options triva/categorized and triva/uncategorized can be used to generate
1336 graph configuration files for Triva visualization tool.
1337 * Configuration option tracing/platform is renamed to tracing/categorized
1338 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1339 * New TRACE_user_link_variable interface:
1340 User provides the name of the link and the tracing variable to attach to it
1341 * the declaration of tracing categories must be done after the environment
1343 * simpler tracing interface, just one way to declare categories
1344 TRACE_category or TRACE_category_with_color, it is up to you
1345 * links in the trace file are again identified by their names
1346 * trace contains the full platform hierarchy exactly as declared using the ASes
1347 * Options tracing/msg/[task|process]:1 groups the process by hosts
1348 for both cases, tasks and processes must have names that are unique during the simulation
1349 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1350 * The experimental option tracing/msg/volume is deprecated
1351 its functionality may be reincorporated if needed
1353 The tracing generates a trace file with unordered timestamped events,
1354 because of the way the core simulator (surf) works. A script available
1355 at the tools directory (fix-paje-trace.sh) can be used to put the events
1356 in order. We have changed the tracing so it can generate ordered timestamped
1357 events in the final trace, but depending on the simulator (and how much time
1358 is simulated) that can lead to a huge memory utilization. It is deactivated
1359 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1361 Build Infrastructure
1362 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1363 - We are trying hard to keep the API stable, but it may happen that
1364 some things change (we're a research project after all, not a
1365 nuclear plant operating system). If such things should happen, you
1366 could rely on that macro to adapt.
1367 - current value: 30600 for 3.06.00, aka 3.6
1368 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1369 a number that can be compared with SIMGRID_VERSION.
1370 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1371 assertions and verbose/debug logging events are disabled at compile time.
1373 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1375 SimGrid (3.5) stable; urgency=medium
1378 * New feature to any SimGrid-based simulator: Model-Checking
1379 Check SIN#1 for more details.
1382 * New Model SMPI (three-interval linear regression for correction factors)
1383 See RR-7426, available at http://hal.inria.fr/inria-00527150
1384 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1385 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1386 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1387 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1388 * See also src/smpi/README
1393 - Tracing API changes: TRACE_start and TRACE_end should not be called
1394 by user-code. They are automatically called by simulators created
1395 with SimDAG, MSG and SMPI if the toolkit is compiled with
1396 tracing_enabled=ON. Categories declaration and utilization remain the
1397 same for MSG and SimDag.
1398 - A function was added to the tracing API to declare categories with
1400 - TRACE_category_with_color (char *category, char *color)
1401 where color must be in the following format
1402 "%f %f %f", red, green, blue
1403 and red, green, blue are float values in the interval [0, 1]
1404 - User can specify NULL as color parameter, or continue calling
1405 TRACE_category (cat)
1406 On that case, the tracing system will define random colors
1407 - The following command-line options are supported:
1408 --cfg=tracing/filename:msg.trace
1409 --cfg=tracing:1 (activate tracing, needed to use others)
1410 --cfg=tracing/platform:1 (categorized resource use)
1411 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1412 --cfg=tracing/msg/task:1 (task creation)
1413 --cfg=tracing/msg/process:1 (process creation, migration)
1414 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1415 --cfg=tracing/smpi:1 (SMPI interface tracing)
1416 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1417 - examples of examples/msg/tracing updated
1419 - DAXLoader and DOTLoader functions can generate tasks with categories
1420 - A new function to attribute a category to SD tasks:
1421 TRACE_sd_set_task_category (SD_task_t task, char *category)
1422 * Tracing the MPI interface implemented by SMPI
1423 - Collective operations are traced with states
1424 - Point-to-Point operations are traced with states/links
1425 - Tracing activated by a parameter "-trace filename" passed
1426 to smpirun during execution (considering that simgrid
1427 is compiled with tracing enabled)
1428 - To run the simulation with gdb, the simulator
1429 accepts --cfg=tracing/smpi:1 to trace SMPI
1430 - tesh files to check if smpi tracing is ok
1431 - See examples/smpi/NAS/DT-trace
1432 * GTNetS tracing re-worked
1433 - adaptation to the tracing system of GTNets to cope
1434 with modifications regarding the fullduplex mode
1435 - new tesh files to check if gtnets tracing is ok
1438 * Asynchronous communications through the functions:
1439 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1440 * New function: MSG_load_platform_script()
1441 to make possible using a lua script instead of XML files to set up platforms
1442 * New function: MSG_set_function
1443 to associate functions to processes, used when bypassing the parser
1444 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1446 Platforms: Add some more examples in examples/platforms
1447 * Grid'5000: see www.grid5000.fr
1448 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1451 * Change the XML format. This is a very important modification. SimGrid
1452 3.5 introduces a new hierarchical format based on the notion of
1453 Autonomous Systems. Compatibility with old format is ensured
1454 through the perl script provided in the install bin directory
1455 bin/simgrid_update_xml.
1456 It is now possible to build platforms with specific routing
1457 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1458 connect several platforms together. We will try to provide soon
1459 set of realistic platforms exploiting these properties (have a
1460 look at examples/platforms/ for the moment).
1461 * Take the opportunity of the XML format change to be a good XML citizen:
1462 rename link:ctn to link_ctn and similar changes (also dealt with by
1464 * Add a new routing scheme (rule-based) using regular expressions. It
1465 enables to have an extremely low memory footprint when the
1466 underlying routing is simple and can be compactly described. You
1467 need to have libpcre4-dev (perl regular expressions) installed if
1468 you want to use this routing scheme.
1469 * Revive the cluster TAG and allow to easily and efficiently (both in
1470 term of memory and speed) connect clusters together. Have a look
1471 at teshsuite/simdag/platforms/ to see how this can be done. With
1472 this tag, you can create clusters with thousands of tasks at no
1473 cost (have a look at examples/platforms/).
1474 Note: clusters are implemented as ASes, so there is no need for an
1475 enclosing AS tag if you have only one cluster in your platform.
1476 * Add new generic functions in the public interface that allows the user
1477 to call SURF 'create_resource' methods from your code (same
1478 functionality as the XML bypass mechanism but with a much lighter
1480 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1481 that uses a piecewise linear approximation to produce better
1482 results when exchanging small messages.
1483 * Add a new parameter to handle correctly full duplex link and account
1484 for interferences between uplink and downlink communications
1485 (activate with --cfg=fullduplex:1).
1488 * Rename the SD_READY (all dependencies are satisfied and task is
1489 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1490 dependencies are satisfied) state.
1491 This prevents a confusion between the notion of "ready to schedule"
1492 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1493 simulated" (SD_RUNNABLE) used by the simulation kernel.
1494 * Change the way a task is considered as ready. Instead of removing
1495 dependencies when a task is done, a counter is decreased. This way,
1496 it is always possible to reach ancestors thanks to the
1497 SD_taks_get_parents function (even after the end of the simulation.)
1498 * Change the return type of SD_Simulate from (SD_task_t*) into
1499 xbt_dynar_t. This function was in handling a dynar internally and
1500 converted it into a NULL terminated array for historical reasons.
1501 * New function SD_dotload(char*) to load a DAG described in dot
1502 format. This loader and the corresponding examples require the
1503 installation of the graphviz library.
1504 * Fix a bug in the management of tasks of size 0 in the surf network
1505 models. This problem was only visible with SIMDAG and you should
1506 thus disregard results produced with earlier versions if you
1507 relied on this feature (some tasks were blocked because of this).
1508 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1509 even though your applications were doing only point-to-point
1510 communications and sequential computations. Now you can really use any
1511 model you want (of course, if you create real parallel tasks, which are
1512 not implemented in most models beside ptaskL07, this will abort).
1513 * Add an example that schedules a DAX on an heterogeneous platform
1514 using a Min-Min strategy.
1515 * New function SD_workstation_get_current_task() that returns the kind
1516 of task currently running on a workstation in the sequential access
1518 * Raise some warnings when unexecuted tasks remains at the end of the
1519 simulation. This is usually caused by cycles in the DAG.
1522 * New function: SIMIX_process_set_function() called by MSG_set_function
1523 * Change the underlying waiting queue in semaphores so that a process
1524 can wait on several of them simultaneously (as in waitany).
1525 * Fix the way to handle tokens in semaphores so that all access patterns
1526 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1527 * kill the dirty pimple SIMIX_message_sizes_output()
1528 Please use (proper) visualization instead
1531 * New data container: setset (set of sets of elements)
1532 * New module: mmalloc (mapped malloc, allowing to have several
1533 independent segments of malloc)
1534 * New function: xbt_dict_cursor_set_data()
1535 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1536 * New function: xbt_dynar_is_empty()
1537 * New function: xbt_fifo_get_last_item()
1538 * Fix xbt_dynar_shrink(): use the right element size.
1539 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1540 uninitialized areas during expand.
1541 * Fix semaphores: previous implementation was severely broken.
1542 * Use library init/fini functions for our initialization.
1543 - you can use logs and other feature as soon as you want in your
1544 code (even before the xbt_init / MSG_init)
1545 - xbt_exit is now a no-op and produce a warning when used.
1548 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1549 benefit from the latest improvement to the simulation kernel.
1550 * Kill measurement sockets for now. If you rely on them, sorry. This
1551 release is not for you. This feature will be reintroduced in the
1552 future, but we cannot delay the release any further.
1553 * New function: gras_msgtype_get_name().
1554 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1555 * Fix (at last) the pmm example: it should not randomly fail anymore.
1557 Build chain: bug fixes and overall polishing
1558 * Cmake is now stable enough. Hence, we killed the autotools.
1559 * Port to windows ( TM :)
1560 * Fix the 'make install' target.
1561 No need to use 'make install-simgrid' anymore
1562 * Introduce a 'make dist' target compiling a *source* archive
1563 'make package' compiles a binary archive
1564 * Compile java files only on need
1565 * Add --cd and --setenv command line options to tesh
1566 * Out of source builds are not fully supported yet, but we are close
1567 * Enable supernovae and optimization flags by default for our users
1570 * Add layer to set up environment directly from lua, without XML.
1571 * The effect of gras_stub_generator can be achieved through
1572 lua too (check examples/gras/console/ping_generator.lua)
1574 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1576 SimGrid (3.4.1) stable; urgency=low
1578 The "Polishing easter eggs is probably a good idea" release.
1579 This is a bug fixes release only.
1583 * Fix a bug preventing the tasks from begin garbage collected.
1586 * Fix a bug occuring when a host involved in a communication fails.
1587 This was not detected properly by the other peer involved in the
1588 communication. Now, it's reported as a network error.
1591 * Warn the user about loop dependencies in data flow of DAX files
1592 * Obey the control-flow dependencies of DAX files
1595 * Add option "enable_smpi" allowing to not compile SMPI.
1596 Probably useful for the (Mac) users experiencing a build error here
1597 * Improve the detection of lua5.1 and ruby1.8
1599 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1601 SimGrid (3.4) stable; urgency=low
1603 The "Easter in Cargese" release. Also known as (major changes):
1605 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1606 ~> bindings were greatly improved
1607 ~> new tracing infrastructure for better visualization introduced
1609 * the "Welcome to configury modernity" release.
1610 ~> we switched from autotools to cmake, and improved our cdash
1613 A more detailled list of changes follow (full detail in svn log).
1615 Java Bindings: Various Cleanups
1616 * (install java-gcj-compat-dev on debian-like to use them)
1617 * Remove put/get: no need to export deprecated interface in Java
1618 Use send/receive instead.
1619 * Cleanup the examples and add a README per directory
1620 * Remove example autoDestination (that's the only way to go now)
1621 * Remove example explicitDestination (was a plain copy of basic)
1622 * Make JniException a runtime exception, so that there is no need to
1623 declare the fact that you may encounter such a beast. I guess that
1624 nobody will ever want to survive such error.
1625 * Create specific errors for each MSG case of failure:
1626 host failure, transfer failure, timeout, task cancelled
1627 * Cleanup the exceptions that may get thrown by each function
1628 * Other internal cleanups in Java bindings. Performance still bad :/
1629 Ruby and Lua Bindings: create them
1630 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1631 * That's new and great, you should try them out.
1632 Same functionalities than Java bindings, only even less polished
1634 * Kill the useless "rate" argument of SD_task_get_execution_time()
1635 Everyone used to provide -1 as a value, it was not used, and the
1636 semantic of a possible use wasn't even clear.
1637 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1638 either as comm costs or compute costs to mean that there is no
1639 such thing for that specific task.
1640 * Add a SD_task_set_name() function
1641 * Fix SD_task_unschedule() on typed tasks
1642 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1643 * In DAX loader, accept useless 'level' attributes to <job> since
1644 LIGO DAGs have them (seem to be to ease graphical representation).
1646 * Add an example masterslave_mailbox.c using send/receive and not
1647 the deprecated put/get interface.
1648 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1649 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1651 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1652 (declare MSG_USE_DEPRECATED to still have the old name)
1653 * Add a MSG_task_set_data() function
1654 * About trace replay (see examples/msg/actions):
1656 - Allow to work with splitted trace files for each process
1657 Give the specific trace file as argument of each process,
1658 and call MSG_action_trace_run(NULL)
1659 You can still have one merged file for all processes.
1660 - Fix implementation of collective operations
1661 * Allow task_execute() on 0-sized tasks (closes #10063)
1663 * This is the first release of SimGrid where SMPI is not considered
1664 beta anymore (even if some corners should still be improved)
1665 * Port over the new SIMIX_network submodule (internal refactoring)
1666 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1667 * Implement more missing elements of the standard:
1669 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1671 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1672 MPI_Scan() MPI_Get_processor_name()
1673 - Added implementation of missing case for Alltoall (warning: it's
1674 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1675 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1676 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1677 SMPI_MPI_Allgatherv()
1678 * Bug fixes include:
1679 - MPI_Waitsome() was broken
1680 - Allow relative includes in smpicc
1681 - Command line cfg argument 'reference_speed' was ignored...
1682 - Some functions did not properly lead to auto-benching of user code
1683 - smpicc passes -O2 by default (just like openmpi one)
1685 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1686 * Bug fixes about timeouts during communications
1687 * add SIMIX_message_sizes_output() as a pimple to write to file the
1688 amount of messages per size. Use gnuplot to get histogram.
1689 Pimple because that's the only user-visible function of simix,
1690 defined directly in xbt.h (irk, sorry)
1692 - Add a SIMIX_sem_get_capacity() function
1693 - Fix interactions with processe resume/suspende
1694 - release_forever() was stupidly broken
1695 - Fix SIMIX_display_process_status() for processes in a semaphore
1696 - Make SIMIX_sem_block_onto() user-visible
1697 * Refactoring context stuff:
1698 - Use pseudo-OOP for better modularity
1699 - reimplement SIMIX_process_kill() without process_schedule() so
1700 that the latter can take as invariant that it is called from
1702 - Merge context_start into context_new for sake of simplicity
1704 * Add a Vivaldi network model, coded live during SUD'10 ;)
1705 * Rename configuration variables to start a hierarchy:
1706 o cpu_model -> cpu/model
1707 o network_model -> network/model
1708 o workstation_model -> workstation/model
1709 * New configuration variables:
1710 o network/bandwidth_factor: correction to bandwith
1711 o network/latency_factor: correction to latency
1712 o netwotk/weight_S: correction to the weight of competing streams
1713 * Add a long description to the models, that users can see with such
1714 argument on the command line: --cfg=cpu/model:help
1715 * --help-models display the long description of all known models
1717 * config: add the ability to set a default value after registration
1718 Does not override any previously set value (e.g. from cmd line)
1719 * dict: allow to have integer key and data.
1720 When so, you need to use the following functions
1721 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1722 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1723 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1724 In contrary to regular dicts, the key is not malloced before copy.
1725 Mixing scalar and regular elements in the same dict is not tested
1727 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1728 Tracing for Visualization:
1729 * SimGrid is now instrumented in order to generate a trace file for
1730 visualization analysis: to use it, need to compile SimGrid with the
1731 "tracing" option enabled, and instrument the program using SimGrid with
1732 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1733 (among other functions).
1734 * The instrumentation only traces the platform utilization for now
1735 * Documentation to use the tracing functions and how to analyze the
1736 traces with the Triva tool is written.
1737 * More information about: SimGrid FAQ (in the section Tracing Simulations
1740 * We moved to cmake as default build system. Autotools support will
1741 be dropped soon. Check the FAQ for more info about how to use it.
1742 * Greatly improved our cdash/ctest interactions
1743 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1744 * Added memory checking tests with valgrind; lot of memleak fixing.
1745 This may be the first release of simgrid with so few memory issues
1746 * Added code coverage tests.
1747 Our coverage is still improvable, but at least we see it on cdash.
1749 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1751 SimGrid (3.3.4) stable; urgency=low
1753 The "Desktop Grid needs love too" release (also called Xmas release).
1755 Models improvements:
1756 * Major speedup in the maxmin system solving by using lazy evaluation
1757 Instead of solving completely the maxmin system at each iteration,
1758 only invalidate (and recompute) the modified parts.
1759 This new feature is enabled in default models but you can try to
1760 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1761 * Cas01 IMproved as default CPU model
1762 This CPU model is the same Cas01 model, but it uses the
1763 maxmin-selective-update flag and a heap structure to manage
1764 actions on SURF kernel.
1765 It reduces the complexity to find the next action to finish and,
1766 consequently, it's faster than the old Cas01.
1767 This is the new default CPU model (Cas01).
1768 * Rename the old Cas01 model to Cas01_fullupdate
1769 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1770 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1771 * CpuTI (CPU Trace Integration)
1772 A new CPU model whose objective is simulate faster when using
1773 availability trace files.
1774 Instead of using a full featured, over engineered maxmin system for
1775 CPU modeling, this model does the pre-integration of traces files
1776 to calculate the amount of CPU power available, and so, executes
1777 faster than the old CPU models.
1778 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1779 * Use LV08 as default network model since it gives better accuracy
1780 for small messages and shouldn't change things for big ones.
1781 Use --cfg=network_model:CM02 to get the previous behavior.
1784 ******************************************
1785 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1786 ******************************************
1787 * The new CPU model may changes simulations!
1788 The point is that events occurring at the exact same timestamp
1789 are not scheduled in the same order with the old and new
1790 version. This may be enough to completely change the execution
1791 of simulations in some cases.
1792 * The new network model will change simulations!
1793 This new model is more realistic than the previous one, so you
1794 should consider redoing your old experiments with this model.
1795 Sorry for the inconvenience.
1798 * Introduce the supernovae compilation mode
1799 When compiled that way, the whole SimGrid (or almost) is put in a
1800 single compilation unit and compiled in one shoot.
1801 This is to help gcc which has difficulties to inline stuff from one
1803 The speedup seem to be above 15%, althrough more tests are needed on
1804 amd64 to confirm that gain.
1807 * Port of MSG's mailbox on top of SIMIX network
1808 The put/get mechanism was greatly simplified on the way.
1811 * New SIMIX network module. Provides:
1812 - Mailbox: rendez-vous mechanism to find with who you want to speak
1813 - Synchronous send/recv: easier and hopefully faster since the
1814 logic is handled in the maestro process directly now
1815 - Asynchronous send/recv: you dreamt of it? It's here now
1816 Too bad that nobody cared enough to propagate the change to MSG.
1817 * Add semaphores as SIMIX synchronization mechanism.
1820 * new function SD_daxload(char*) to load a DAX file
1821 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1822 * Introduce typed tasks. Specify its kind and cost at creation.
1823 At scheduling, just give where it should be placed, and the cost
1824 for each involved resource is automatically computed.
1825 Existing constructors so far (more to come of course):
1826 - SD_task_create_comm_e2e() for end-to-end communication
1827 - SD_task_create_comp_seq() for sequential computation
1828 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1829 * new function SD_task_dump() for debuging display
1830 * new function SD_task_dotty(task,FILE*) writing to file the info
1831 about the task in dotty format
1832 * SD_task_dependency_exists() can now cope with having one of its
1833 arguments NULL. If so, it tests whether the other argument has any
1835 * Add getters on list of preceding/following tasks:
1836 SD_task_get_parents(task) and SD_task_get_children(task)
1837 * Add getters on amount of workstations and list:
1838 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1839 * Add getter on task kind: SD_task_get_kind(task)
1840 * Update the start_time and finish_time of tasks on completion/failure
1841 * Bugfix: Remove task from state swags when destroyed
1844 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1847 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1848 dynar with one separate thread per value of the dynar.
1849 * Change the prototype of xbt_thread_create(), sorry.
1850 Added a boolean parameter indicating whether we want to join this
1851 thread (used in SG only for now)
1852 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1855 * GTNetS wrappers should now be usable again (and betterly tested too)
1856 * Fix a major regression from 3.2 where the timeout provided to
1857 MSG_task_put_with_timeout() was used as absolute time before which
1858 the comm should be done.
1859 * Start to fix the <cluster> tag.
1860 - Internal links should be good now (beside of the loopback, which
1861 use the private link instead)
1862 - paths to the external world is still rather broken
1863 - the <route:multi> tag is just broken. Actually that's brain-dead.
1864 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1865 to make it less stupid
1866 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1867 * Fix a source-level compatibility glitch from 3.2: after defining
1868 MSG_USE_DEPRECATED, you can use the old name
1869 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1870 * Allow to compile from the SVN with automake 1.11
1871 * Fix some problems when using the "start_time" tag in deployment XMLs.
1872 * Fix #8569: XBT/synchro.h has redundant declarations
1873 * Fix #8563: MSG return values and exceptions
1874 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1875 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1876 Thanks to Robson Peixoto.
1877 * Fix a vicious bug in dictionaries inducing that some elements were
1878 not freed on xbt_dict_free()
1880 Portability report of this version:
1881 * Main portability targets:
1882 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1883 on (amd64/i386/ia64)
1884 - mac leopard on i386
1885 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1886 but nothing critical.
1887 * Other platforms: windows, AIX and others were not tested for this release
1889 Timing report of this version:
1890 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1891 scenario parameters). From 8h to a few seconds in desktop grid settings.
1892 * Supernovae brings about 25% speedup on i386.
1894 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1896 SimGrid (3.3.3) stable; urgency=low
1898 The "Need for Speed" release.
1900 The timings done to validate the 3.3.2 were faulty.
1901 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1903 The problem was a conversion from a manually handled vector to
1904 xbt_dynar_t on the critical path.
1905 xbt_dynar_foreach calls functions, inducing stack management crap.
1907 We inlined these functions and xbt_dynar_foreach is now breath taking.
1908 We also inlined xbt_swag_belong on the way.
1910 Here are some approximate speedup measurements (on master/slaves
1911 simulations lasting between 10s and 20s each):
1912 3.3.1 -> 3.3.2: about same performance
1913 3.3.2 -> 3.3.3: 40% speedup
1914 3.3.1 -> 3.3.3: 40% speedup
1915 3.3.1 with inline patch -> 3.3.3: 30% speedup
1917 Our reading is that the refactoring which occurred in 3.3.2 made us
1918 suffer much more from the xbt_dynar_foreach low performance, but
1919 once we solved this, this refactoring proved to be very performance
1920 effective. From the 40% speedup, somehow, 10% are due to the
1921 inlining and 30% to the refactoring.
1923 That's a pitty that gcc cannot inline functions placed in other files
1924 alone. We have to choose between:
1925 - break the encapsulation (by putting private data structures and
1926 accessors in headers files to help gcc)
1927 - live with low performance
1928 - switch to a decent compiler such as icc (not quite possible).
1930 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1932 SimGrid (3.3.2) stable; urgency=low
1934 The "Simplicity does not preceed complexity, but follows it" release.
1936 The main contributors of this release were (lexical order):
1937 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1940 * Extract the routing logic into its own object.
1941 (was dupplicated in network.c and workstation_LV07.c;
1942 Allows to implement other ways of storing that info)
1943 => kill now useless network_card concept
1944 - Use dynar to represent routes (instead of void** + int*)
1945 - kill link_set (use surf_network_model->resource_set instead)
1946 - Add a command-line option to choose the routing schema to use
1947 - Add three new models:
1948 * Floyd (shortest path computed at initialization)
1949 * Dijikstra (shortest path recomputed all the time)
1950 * Cached Dijikstra (shortest path computed on need)
1951 All these models where contributed by Silas De Munck, and are
1952 described in his ICCS09 paper.
1954 * Simplify model declaration
1955 (less redirections, less function to write when defining a model)
1956 - Factorize stuff between models:
1959 surf_model_resource_set(model)
1960 surf_model_resource_by_name(model, name)
1961 - Unify the types of models in s_surf_model_t (using an union)
1962 - Embeed fields of common_public directly into s_surf_model_t
1963 - Rename model methods:
1964 action_free ~> action_unref
1965 action_change_state ~> action_state_set
1966 action_get_state ~> action_state_get
1967 - Change model methods into functions :
1968 (model)->common_public->action_use ~> surf_action_ref
1970 * Implement a generic resource; use it as ancestor to specific ones
1971 (allows to kill duplicated code in models)
1972 Drawback: timer command don't need no name nor properties;
1973 workstation_CLM03 don't need no properties
1974 (but I guess we can live with those few bytes wasted)
1976 * Improve the action object model
1977 - implement a constructor avoiding dupplicated code about field
1978 initialization in generic_action part.
1980 * Kill the SDP model: it has an external dependency, is deprecated
1981 in flavor of modern lmm models, and didn't compile since a while
1984 * Relocation of the context module from XBT to SIMIX.
1985 (the context were decoupled from the simix processes, duplicating a lot of code)
1986 => a lot of code was factorized
1987 - less overhead is introduced during scheduling
1988 - simpler API for the context factory
1989 - the logic for process creation,destruction and manipulation was simplified
1990 * Simplification of the s_smx_process_t data structure.
1991 => accesing the simix level data associated to a process is faster now,
1992 and the code is a lot more readable.
1995 * Implement some more MPI primitives:
1996 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1997 -implementation: Bcast: flat or 2-ary tree (default),
1998 Barrier: 4-ary tree,
2000 Allreduce: Reduce then Bcast
2001 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2002 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2003 Alltoallv: flat tree, like ompi
2005 * Add support for optimized collectives (Bcast is now binomial by default)
2006 * Port smpirun and smpicc to OS X
2009 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2010 Shout out if you used it.
2013 * Display the list of still queued messages in SG mode when existing
2017 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2018 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2019 * Add xbt_str_from_file(FILE*)
2020 * Add xbt_dict_get_key achieving a linear reverse search
2021 * Remove the context module
2023 Portability report of this version:
2024 * Main portability targets:
2025 - Linux(debian)/x86/context
2026 - Linux(debian)/x86/pthreads
2027 - Linux(debian)/amd64/context
2028 - Linux(debian)/amd64/pthreads
2029 On these, we still have the eratic breakages of gras/pmm and
2030 amok/saturate_sg reported in previous version. We still think
2031 that the tests are the cause of the fault, not the tested code.
2033 - Mac OSX Leopard/x86/context
2034 Still false negative in tesh autotesting.
2035 Smpi still fails, but this time because readlink does not accept -f
2036 Everything seems to work properly beside of that.
2039 - AIX version 5.3 (only tested contexts this time)
2040 Smpi still fails there because mktemp is not installed.
2041 Everything seems to work properly beside of that.
2043 I managed to compile it for the first time, but several breakages.
2044 Won't delay the release for this exotic platform.
2046 * Windows: it's still lagging behind. If you want to help, please
2049 Timing report of this version:
2050 This version seem to be more than 5% faster than 3.3.1 (on linux
2051 64bits with contextes). The gain is less than expected, we are
2052 investigating this for next release.
2054 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2056 SimGrid (3.3.1) stable; urgency=low
2059 * Implement a --cfg-help to show existing configuration variables
2060 * Build chain do not require doxygen in maintainer mode
2063 * fix a bug on struct sizeof computation, which prevented the
2064 exchange of arrays of structs in some conditions
2065 - added a regression test about this in datadesc_usage
2066 * Allow the exchange of 0-long dynamic vectors.
2067 - for that, use -1 as indicator of dynamic size instead of 0
2068 - This implied to change any size from unsigned long to long,
2069 reducing a bit communication abilities, but I guess that with
2070 64bits being quite common, this is more than enough.
2071 - This also induce a protocol change, thus bumping network protocol
2072 version from 0 to 1 (if we have external users, we have to get
2073 clean on that point too ;)
2074 - added two regression tests about this in datadesc_usage
2075 * Be more verbose when propagating local exceptions
2076 This helps debugging.
2077 * Display the status of simulated processes when receiving SIGINT in
2081 * Allow to control the simulation from a trace file.
2082 New functions MSG_action_register() and MSG_action_trace_run()
2083 The first one allows to associate a function execution to each
2084 kind of action while the second one parses a trace file and
2085 triggers the corresponding actions within the system.
2086 For now, only a toy example is provided in examples/msg/actions
2087 * Add an exemple of process migration in examples/msg/migration
2088 * Fix a bug in task exchange which broke MSG_task_get_sender()
2089 Add a teshsuite regression test for that.
2090 [Bug: if MSG_task_get_sender() is called after sender exit,
2092 * Fix a bug which prevented suspend/resume to work properly
2093 * Display the status of simulated processes when receiving SIGINT
2094 This fixes a regression of v3.3. due to the introduction of SIMIX
2095 * Bug fixing in failure management:
2096 - trace could not start by a failure at time 0
2097 - failure during communications were not working
2100 * Add SIMIX_process_set_name() to change the name of the current
2101 process in the log messages.
2102 * Store smx_hosts in a dict since we only retrieve them by name
2103 * Move the configuration infrastructure to surf
2106 * Move the configuration infrastructure to surf
2109 * Massive internal cleanups:
2110 - Store internal structures on processes instead of hosts (allows
2111 to have more than one process per host, in addition of being more
2113 - Cleanup the initialization/finalization process
2114 - Kill a whole bunch of unneeded synchronization:
2115 processes run in exclusive manner within the simulator
2116 - Move queues from global tables to process data fields
2118 - now accept -platform and -hostfile arguments
2119 - Pass the right rank value to processes according to the hostfile
2120 * Compile the examples by default, and use them as regression tests
2121 * Implement MPI_Wtime()
2122 * Change the reference speed to a command line option
2125 * TCP_gamma can now be specified as command line option using
2126 --cfg=TCP_gamma:10000000.0
2127 * Change the --surf-path cmd line option into --cfg=path:
2130 * Also include strbuff from xbt.h public header
2131 * xbt_ex_display(): do not free the exception after displaying
2132 This allows to do more with the given exception afterward.
2133 Users should call xbt_ex_free() themselves.
2137 Portability report of this version:
2138 * Main portability targets:
2139 - Linux(debian)/x86/context
2140 - Linux(debian)/x86/pthreads
2141 - Linux(debian)/amd64/context
2142 - Linux(debian)/amd64/pthreads
2143 These targets fail about 1/10 of times on gras/pmm, but we believe
2144 that this is because of the test, not because of simgrid.
2145 amok/saturate_sg fails even more rarely, and the test may not be
2148 - Mac OSX Leopard/x86/context
2149 The test suite still spits tons of errors because some obscure
2150 force prevents us from removing the temporary directories
2151 arguing that they still contain some metadata I've never heard of.
2152 Smpi fails because seq is not installed.
2153 Everything seems to work properly beside of that.
2156 - AIX version 5.3 (both contexts and pthread)
2157 Smpi still fails there because mktemp is not installed.
2158 XML inclusions seems rosty on AIX.
2160 * Windows: it's still lagging behind. If you want to help, please
2163 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2165 SimGrid (3.3) stable; urgency=high
2169 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2170 [Malek Cherier & Mt]
2172 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2175 Until now, GRAS were using MSG as an interface to SURF. It was
2176 quite difficult because both interface have several differences
2177 (MSG channels vs GRAS sockets were the most notable point).
2179 This also opens the gate to SMPI (which should occur soon) and speed
2180 up simulations by to 40% (even if it were not the main goal).
2182 **************************************
2183 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2184 **************************************
2185 The point is that events occuring at the exact same timestamp are
2186 not scheduled in the same order with the old and new version. This
2187 may be enough to completely change the execution of simulations in
2188 some cases. Sorry for the inconvenience.
2190 * Cleanup and upgrade the XML format to push further scalability
2191 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2193 * Improve the testing infrastructure with tesh. Now a very large part of
2194 the code is tested not only by being run but also by checking that the
2195 output match an expected output [Mt].
2197 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2198 is really less memory-demanding, which should allow you to use
2199 larger files in SimGrid [AL].
2201 * Inform valgrind about our contextes, so that it becomes usable
2202 with the default (and more effecient) version of SimGrid
2203 [contributed by Sékou Diakite, many thanks]
2206 * Introduce a listener thread in charge of receiving incoming
2207 messages from the network. It allows to overlap communication and
2208 computation but most notably, it removes some stupid deadlocks due
2209 to the fact that so far, a process could not send and receive at
2210 the same time. This made most non trivial communication schema
2212 * Convert the PIDs from long int to int to match the MSG ones (and
2213 linux ones too) [Mt]
2214 * New function: gras_agent_spawn() to launch a new process on
2215 current host. Only working in simulation for now. [Mt]
2216 * New function: gras_os_hostport() returning a constant form (ie,
2217 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2220 * Make the backtrace of exceptions more human readable [Mt]
2221 * New module: xbt/str [Mt]
2222 a ton of string utility functions (split, join, printf to a newly
2223 allocated buffer, trim, etc)
2224 * New module: xbt/hash [Mt]
2225 SHA1 hashing algorithm (more to come if needed)
2226 * New module: xbt/synchro [Mt]
2227 synchronization tools (mutex and conditions) working the same way
2228 in simulation and in real life (mainly useful for GRAS, but not
2230 * New module: xbt/queue [Mt]
2231 classical producer/consumer synchronization scheme
2232 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2233 (using the classical functions will get serialized) [Mt]
2234 * Make dictionary internal table dynamic. No need to specify its size
2235 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2237 * Make sure the log channels are organized as a tree under windows
2238 (because of ANSI C compatibility issue, any channel were child of
2239 root directly) [Mt].
2242 * Cleaned many thing in surf and fixed a few bugs [AL].
2243 * Add a nice command line configuration mechanism to compose models [AL].
2244 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2245 the previous one (KCCFLN05). It relies on something that looks like
2246 a max-min sharing mechanism but cannot be written as such. A new solver
2247 was thus designed [AL].
2248 * Add a new solver to lmm. Based on Lagrange optimization and
2249 gradient-based descent, it enables to efficiently maximise systems s.a
2251 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2253 This solver enables to propose two new network models for TCP Reno and
2254 TCP Vegas based on Low's work. These models still need to be fully
2255 tested though [Pedro Velho].
2258 * Bug fix in SD_simulate. Now the time bound given as argument is
2260 * Use the new parallel task model (ptask_L07) as default.
2261 * Use the SURF command line configuration mechanism.
2262 * 0-size tasks (for synchronization) should now work.
2264 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2266 SimGrid (3.2) stable; urgency=high
2270 We still experience issues on this platform, but we believe that at
2271 least MSG is usable.
2273 GRAS API BREAKAGE (for simplification purpose, sorry):
2274 * the gras_msgtype_by_name is not used anymore. Instead of
2275 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2276 you can write (and must)
2277 gras_msg_send(toserver, "request", &request);
2278 - If you still want to pass a gras_msgtype_t to the function (to cache
2279 the type and avoid the lookup time), use the gras_msg_send_() variant.
2280 - Impacted functions:
2281 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2282 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2283 * The callbacks are now expected to return 0 when everything went well
2284 (just like the main() function)
2286 GRAS new features and improvements:
2287 * New module mechanism where user code can use per process globals [Mt]
2288 This is similar to gras_userdata_*() functions, but for libraries. It
2289 factorize some code developped over and over in the examples and AMOK.
2290 It has still to be documented and used (only amok/peermanagement is
2292 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2293 loss when they were small enough to fit into the buffer and sent quickly
2294 enough so that they can all get received in one shoot.
2295 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2296 if not found. Use the *_or_null() variant for the old semantic.
2297 * In gras_msg_handle, do not discard messages without callback.
2298 They are probably messages to be explicitly awaited later (ie, proofs of
2299 mis-synchronization in userland since they are sent before being awaited)
2300 No big deal usually.
2301 * gras_socket_meas_send/recv: semantic changed!
2302 The numerical arguments used to be (1) the total amount of data to send
2303 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2304 messages. This was need for the fool willing to send more than MAXINT
2305 bytes on quite fat pipes.
2308 * Do really rename the hostmanagement module to peermanagement. [Mt]
2309 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2310 but this is rather new and this was documented in the module
2311 documentation (poor excuses, I admit)
2312 * Bandwidth measurement semantic changed! This follows the changes to
2313 gras_socket_meas_send/recv explained above.
2316 * A sequential mode has been added to the workstations. When a workstation
2317 is in sequential mode, it can execute only one task, and the other tasks
2318 are waiting in a FIFO. [Christophe Thiery]
2321 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2322 model for SIMDAG. [Christophe Thiery]
2323 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2324 equal to 0, which caused some bad behaviors in
2325 saturated_constraint_set_update. I now use a threshold mechanism like in
2329 * When running manually src/testall, you select specific units [Mt]
2330 testall is the result of our cunit mechanism, and should replace all
2331 the scripty thingy around since bash don't run easily on billware.
2333 * A mallocator system has been added. [Christophe Thiery]
2334 Mallocators allow you to recycle your unused objects instead of freeing them
2335 and allocating new ones.
2337 Documentation update:
2338 * FAQ reworking + New FAQs:
2339 - "Valgrind spits tons of errors!" [Mt]
2340 - "How to repport bugs" [Mt]
2341 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2342 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2343 - Communication time measurement within MSG [AL]
2344 - I experience weird communication times when I change the latency [AL]
2345 * GRAS tutorial [Mt]
2347 - an introduction to the framework and to the used communication model
2348 - an initiatic tour introducing the most proheminent features:
2350 . Lesson 0: Installing GRAS
2351 . Lesson 1: Setting up your own project
2352 o Part 2: Message passing
2353 . Lesson 2: Exchanging simple messages
2354 . Lesson 3: Passing arguments to the processes (in SG)
2355 . Lesson 4: Attaching callbacks to messages
2356 . Lesson 5: Using globals in processes
2357 . Lesson 6: Logging information properly
2358 . Lesson 7: Using internal timers
2359 . Lesson 8: Handling errors through exceptions
2360 . Lesson 9: Exchanging simple data
2361 . Lesson 10: Remote Procedure Calling (RPC)
2362 . Lesson 11: Explicitely waiting for messages
2363 . Recapping of message passing features in GRAS
2364 - A HOWTO section containing:
2365 o HOWTO design a GRAS application
2366 More are due, of course. They will come latter. In the meanwhile, you can
2367 check the examples which are still here.
2369 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2371 SimGrid (3.1) stable; urgency=high
2375 There was a stack corruption somewhere, visible only when optimizing
2376 with these versions. [Vince]
2379 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2380 module that enabled to play with Directed Acyclic Graphs. It is built
2381 directly on top of SURF and provides an API rather close to the old
2382 SG. Some old codes using SG are currently under rewrite to check that
2383 all needful functions are provided. [Christophe Thiery]
2386 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2387 extension of the classical CLM03 model that gracefully handles
2388 failures. This is now the default model for MSG and GRAS. It doesn't
2389 handle parallel tasks yet though. [AL]
2390 * Bug fix: Weights were not correctly set in the network part.
2391 WARNING: This may have resulted in incorrect results with simulations
2392 where there are more than one flow on a given link. [AL]
2395 * After a (long ?) discussion on simgrid-devel, we have decided that the
2396 convention we had on units was stupid. That is why it has been decided
2397 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2398 WARNING : This means that all previous platform files will not work as
2399 such with this version! A warning is issued to ask users to update
2401 A conversion script can be found in the contrib module of the CVS, under
2402 the name contrib/platform_generation/surfxml_update.pl [MQ]
2405 * Bug fix: Processes were started in reverse order, wrt deployment file.
2406 WARNING: if your code relies on this bug, please fix it. [AL]
2407 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2408 being executed on two different locations. [AL]
2409 * Bug fix: Failures are now better supported thanks to Derrick's tests
2410 (there was many failure situations I hadn't thought of and that weren't
2411 correctly handled). [AL]
2412 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2416 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2417 Exception occurring on server-side are propagated back to client (!).
2419 API CHANGE: the callback changed their prototype. Change:
2420 int my_handler(gras_socket_t expeditor, void *payload_data) {
2422 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2423 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2425 * New! function: gras_msg_handleall to deal with all messages arriving
2426 within a given period.
2427 * New! function: gras_socket_server_range to get a server socket in a
2428 range of port numbers (ease to avoid port number conflicts) [MQ]
2429 * New! gras processes display their backtrace when they get a SIGUSR1
2430 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2431 Sweet to debug RL processes [MQ]
2435 - Do not force experiment sizes to be expressed in kb, or it becomes
2436 impossible to measure the latency this way (needs one byte-long tests)
2437 WARNING: this changes the amok_bw_* function semantic. [MQ]
2438 - Implements the link saturation stuff. [MQ]
2439 * Peer management module:
2440 New! module factorizing code that we wrote over and over [MQ].
2443 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2444 - Test units are placed directly into the library code, they get extracted
2445 automatically and placed into the src/testall binary.
2446 - Convert most of the XBT tests to this system.
2447 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2448 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2449 live in the same C code. This required to change a little bit the API
2450 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2451 * New module: sparse graph structure with basic algorithms (this is work
2452 in progress and the API is not considered to be frozen yet). [AL]
2453 * Display more information on backtraces: source line & function names are
2454 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2455 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2456 * Reworked a little bit some #include statements to load only required
2457 headers. Some user code that relied on SimGrid to include stdlib or
2458 stdio may need to include it by themselves. [AL]
2459 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2461 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2462 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2463 but I guess that almost nobody used those parts. [MQ]
2465 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2467 SimGrid (3.0.1) stable; urgency=low
2470 * Unfortunately, I had missed 5 misnamed functions:
2471 xbt_fifo_item_t xbt_fifo_newitem(void);
2472 void xbt_fifo_freeitem(xbt_fifo_item_t);
2473 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2474 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2475 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2476 They're now deprecated. Please use their new versions:
2477 xbt_fifo_item_t xbt_fifo_new_item(void);
2478 void xbt_fifo_free_item(xbt_fifo_item_t);
2479 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2480 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2481 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2483 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2484 * Documentation: xbt_log module unmercifully reworked [MQ]
2485 * Bugfix: there was a problem with the ending of contexts with
2486 the pthread backend. It caused some weird deadlock or behavior
2487 depending on the pthread implementation. [AL]
2488 * Bugfix: get the exceptions raised in the simulator repport where
2489 and why they come from when they are not catched in time [AL, MQ]
2492 * Bugfix: Do repport the error when two non-connected hosts try to
2493 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2496 * Add additionnal checkings on communications. Assert that two
2497 communicating hosts are connected by a set of links... [AL]
2500 * Add additionnal checkings on channel values in communication [AL]
2501 * New: MSG_task_get_source to see on which host a task was generated [HC]
2502 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2503 the number of tasks waiting to be received on channel and sent
2505 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2506 waits for the first task coming from a given host.. [AL]
2508 GRAS new functionnalities: [MQ]
2509 * Enhance the parsing macro to allow the size of multidimentional objects
2510 to be given thru annotations.
2511 * New example (and documentation): Matrix Multiplication a la RPC
2512 (as when I was young!) and fix a bunch of bugs found on the way.
2514 GRAS performance improvements: [MQ]
2516 * Reduce the amount of cbps creation/destruction by making it static to
2517 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2520 * Change libdata to a set so that we can search for stuff by ID (and thus
2521 reduce the insane amount of dict lookups)
2524 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2525 RL communication on top of SG since it's so uneffective.
2526 It may also be used for inter-thread communication in RL, one day.
2527 * Use gras_datadesc_copy() to exchange messages on top of SG
2529 - improve message exchange performance on top of SG
2530 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2531 * Don't exchange on the network the size of the used part of buffer,
2532 instead, specify the possible buffer size to read().
2534 - reduces the amount of read/write calls (one pair per exchange)
2535 - reduces the amount of exchanged data (the size)
2536 - allows to retrieve all arrived data on receiver side, if we don't need
2537 it right now (subsequent read will peek the buffer)
2538 - allows the receiver to proceed with the begining of the stream before
2539 everything is arrived
2540 - make it possible to build an iov transport (using readv/writev)
2542 - take care of the data with non-stable storage (like stacked data),
2544 * If possible, TCP send uses vector I/O (when writev() is here)
2545 - Don't use it for receive since we send data sizes and data on the
2546 same stream, so we wouldn't be able to chain large amount of chunks
2547 before having to flush the stuff to read the size.
2548 * Rework the transport plugin mechanism to simplify it and reduce the
2549 amount of pointer dereferencement when searching for the right function
2552 * I guess that now, we do almost as few system calls as possible while
2553 doing as few data copy as possible.
2555 To improve it further, we could try to send all the sizes first and then
2556 all the data (to use iov on receiving size), but it's only a partial
2557 solution: when you have 2 dimensional data, the sizes of the second
2558 dimension is data of the first dimension, so you need 3 streams.
2560 I'm not sure the potential performance gains justify the coding burden.
2562 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2564 SimGrid (3.00) stable; urgency=high
2567 * New! Give the possibility to hijack the surf parser and thus bypass
2568 MSG_create_environment and MSG_launch_application. Have a look at
2569 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2571 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2573 SimGrid (2.96) unstable; urgency=low
2578 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2579 This deprecates the xbt_error_t mechanisms.
2580 It modifies (simplifies) all XBT and GRAS API.
2581 MSG API keeps unchanged (exceptions raised by XBT are catched from
2582 within MSG and masked with existing error handling facilities)
2585 * New! Add a FATPIPE model. [AL]
2586 * New! Add a parallel task model. [AL]
2587 * New! Add automatically a loopback interface (in the default
2588 network model) if none was precised.
2591 * Bugfix: MSG_process_resume now works with the current running process.
2593 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2594 * Modification of MSG_task_get_compute_duration. Once a task has been
2595 processed, the value returned by this function is now equal to 0. [AL]
2596 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2597 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2598 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2599 of a m_task. For now, MSG_task_cancel only works with computation
2601 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2602 of the processor (in Mflop/s) regardless of the current load on the
2604 * API Change: use proper naming convention for MSG_getClock and
2605 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2607 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2608 This function changes the priority of a computation task. This priority
2609 doesn't affect the transfer rate. A priority of 2 will make a task
2610 receive two times more cpu power than the other ones. This function
2611 has been added to suit the needs of Nguyen The Loc and hasn't been that
2612 much tested yet. So if it fails, please report it and send me your code.
2614 * API Change: removed all functions and types that were marked "deprecated"
2615 since many months. Renamed MSG_global_init_args to MSG_global_init.
2617 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2619 SimGrid (2.95) unstable; urgency=low
2622 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2623 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2624 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2627 * Bugfix: Allow absolute paths for platform description files [MQ]
2628 * Bugfix: do free the variables after use. Leads to drastic performance
2630 * Implement max_duration (ie, timeouts) on resources [AL]
2633 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2635 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2636 the needed support (timeouts) [AL]
2639 * Implement measurement sockets. You can now get the bandwidth between two
2640 hosts thanks to AMOK (see below). [MQ]
2641 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2642 dynar over the network (yeah) [MQ]
2643 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2644 * simple/static token-ring example. [Alexandre Colucci and MQ]
2645 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2646 in place before [MQ]
2649 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2650 type parsing. "short" and "long" modifiers were also ignored; other
2651 modifier (such as reference level) are still ignored. [MQ]
2652 * Embeed the buffer size within the buffer itself on SG. [MQ]
2653 That way, send() are atomic and cannot get intermixed anymore (at least
2654 the ones which are less than 100k; bigger messages still have the issue)
2655 * Array size pushed by the field, not by the field type (or each
2656 and every long int will push stuff to the cbps) [MQ]
2657 * use select() to sleep since it allows to portably sleep less than one
2660 GRAS (minor cleanups)
2661 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2664 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2665 * type category 'ignored' killed as it was never used and were difficult
2667 * whether a type can cycle or not is now a flag, leaving room for more
2668 flags (as "ignored", if we feel the need one day ;)
2669 * Rename raw sockets to measurement sockets since "raw" has another
2670 meaning in networking community.
2673 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2674 and offers features not belonging to GRAS but that most applications
2675 need. One day, it may be a set of plugins loadable at runtime.
2676 * New module: bandwidth
2677 bandwidth measurement between arbitrary nodes running this module. [MQ]
2679 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2681 SimGrid (2.94) unstable; urgency=low
2683 The first beta release of SimGrid 3 !
2687 * Update the main page and the FAQ. Adding references to gforge.
2690 * Add a gras_os_getpid function.
2693 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2694 * Extend the logs so that they also print PID, hostname, date, ... if
2696 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2697 and kill the old version which were in testsuite/
2698 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2702 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2703 for users. Instead, each programming environment (GRAS, MSG,...) use it
2704 when needed to provide such a feature to users.
2705 Don't shortcut the mechanism or you will also shortcut the virtualization
2706 you need on the simulator.
2710 * Cleanups in configury with regard to compile optimization/warning flags.
2711 Also add -fno-loop-optimize to any powerpc since it's the optimization
2712 killing gcc (< 3.4.0).
2713 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2714 needed by MSG examples complications
2715 * Borrow configury beautifications from PHP
2718 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2719 warning (thanks loris for stumbling into this one).
2720 * Bugfix: stop loading private headers (gras_config.h) from the public
2724 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2725 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2726 RL would imply a DNS resolver, which is *hard* to do in a portable way
2727 (and therefore delayed).
2728 * Implement a real timer mechanism and use it in timing macros. This allows
2729 to avoid rounding errors and get a 0.000005 sec precision in timing
2730 macros. While I was at it, various cleanups:
2731 - allow to declare more than one timed section per file (fix a stupid bug)
2732 - move some private declaration to the right place
2733 - merge conditional execution and timing macros into emulation module
2734 - document the module
2735 - make sure the module cleanups its mess on gras_exit
2736 * Documentation improvements:
2737 - (new) how to compile applications using GRAS
2738 - (new) emulation support (timing macros)
2740 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2742 SimGrid (2.93) unstable; urgency=low
2744 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2747 - Use Paje properly where used. Still to be sanitized properly.
2748 - Portability fix: Add an implementation of the contexts using pthread
2752 - Add xbt_procname(): returns the name of the current process.
2753 Use it to show the current process's name in all logging.
2755 - fix detection of older flex version and the reaction, since we do
2756 depend on modern ones (we use lex_destroy)
2757 - Better separation of SG and RL in the libs: remove all simulation code
2758 from libgras. As a result, this lib is now only 200k when stripped.
2759 Some of the xbt modules may also be duplicated (two sets and such) and
2760 should be cleaned/killed before SG3.
2761 - Insist on using xlC on AIX because of weird problems involving gcc there.
2762 - Cleanup the make remote stuff. This is now done by scripts
2763 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2764 mainly for our private use, but we're working on changing them to user
2767 - Bugfix: flush the socket on close only if there is some *output*.
2768 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2770 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2773 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2775 SimGrid (2.92) unstable; urgency=low
2777 Alpha 3 on the path to SimGrid 3
2781 - New! First try of benchmarking macros.
2782 - New! First try so that gras_stub_generator generate deployment and
2783 remote compilation helpers.
2785 - Bugfix: Initialization fix in msg_test.
2789 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2791 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2793 - API Change: gras_os_sleep to take the amount of seconds as a double.
2794 Accepting an int was error prone since it was the only location where
2795 seconds were coded as such. It leaded to damn rounding errors.
2796 - Bugfix: Hard to belive that timers ever worked before this.
2798 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2800 SimGrid (2.91) unstable; urgency=low
2802 Alpha 2 on the path to SimGrid 3
2806 - Bug fix in the lmm_solver.
2808 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2809 through the function MSG_paje_output.
2810 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2811 - It is possible to bound the rate of a communication in MSG with
2812 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2813 to write it in the changelog).
2814 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2819 - Various cleanups to the autotools stuff
2820 - Begin to move Gras examples to examples/gras/
2821 - Let make distcheck work again (yeah!)
2823 - documentation overhauled using doxygen.
2824 gtk-doc-tools is dead in SimGrid now.
2825 - Automatically extract all existing logging categories, and add the list
2826 to the documentation (long standing one, to say the less)
2828 - Cleanup the known architecture table. Reorder the entries to group what
2829 should be, and use a more consistent naming scheme.
2830 (some of the test dataset are still to be regenerated)
2831 - New! Allow library to register globals on each process just as userdata
2833 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2834 lookup time (for now).
2835 Use it in msg and trp.
2836 This cleans a lot the internals and helps enforcing privacy of the
2837 headers between the gras components.
2838 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2839 - Bugfix: gras_os_time was delirious in RL.
2840 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2842 - Reenable GRAS now that it works.
2844 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2846 SimGrid (2.90) unstable; urgency=low
2848 Alpha 1 on the path to SimGrid 3
2850 * It is a long time since the last release of SimGrid. I'm sorry about
2851 that but as I had told you, I was rewriting a lot of things. I apologize
2852 to those who had been reporting bugs to me and that I had not answered.
2853 If your bug is still in the new version, please tell me. Here is a
2854 summary of the main changes.
2856 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2857 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2858 portability, speed, and a lot more but you'll figure it out later.
2859 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2860 I wanted to release the new MSG as soon as possible and I have
2861 broken GRAS, which is the reason why, for now, only MSG is fully
2862 functional. A laconic description of these projects is available
2863 in the documentation.
2865 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2866 kernel optimized for our needs (called SURF but only the developers
2867 should use it). Hence, MSG is now roughly 30 times faster and I think
2868 that by rewriting a little bit MSG, I could event speed it up a little
2869 bit more. Beside the gain in speed, it is also much easier to encode a
2870 new platform model with SURF than it was with SG. More to come...
2872 * REVOLUTION 3: I have tried to change a little as possible the API of
2873 MSG but a few things really had to disappear. The main differences
2874 with the previous version are :
2875 1) no more m_links_t and the corresponding functions. Platforms are
2876 directly read from a XML description and cannot be hard-coded
2877 anymore. The same format is used for application deployment
2878 description. The new format is described in the documentation.
2879 Have a look in tools/platform_generation. There is a tiny script
2880 that converts from the old platform format to the new one. Concerning
2881 the application deployment format, parsing the old one is tricky.
2882 I think most of you should however be able to convert your files. If
2883 it is really an issue, I can write a C code that does the conversion.
2885 2) the toolbox tbx does not exist anymore. We now have a library
2886 with much more data-structures but without the hash-tables (we have
2887 dictionaries that are much faster).
2889 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2891 *****************************************************************************
2892 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2893 * don't feel like converting the oldies. (Mt) *
2894 *****************************************************************************
2897 Version 2.90: "the long awaited one"
2898 - Finished rewriting and debugging MSG. Rewrote the documentation.
2899 - disable GRAS for now since it needs to be ported to the newest SG
2902 - Finish the port to windows (using mingw32 for cross-compile)
2905 - Main loop and datastructures of SURF. A cpu resource object is
2906 functional. Surf can thus be used to create cpu's with variable
2907 performance on which you can execute some actions.
2909 2004-11-15 Martin Quinson
2910 - Port to ARM. Simply added the alignment and size descriptions. Should
2911 work, but the ARM machines are so slow that I didn't had the opportunity
2912 to 'make check' over there yet.
2914 2004-11-15 Arnaud Legrand
2915 - Trace manager now written. It uses a heap structure and is therefore
2916 expected to be efficient. It may however be speeded up (particularly
2917 when many events occur at the same date) by using red and black
2918 trees. One day maybe...
2919 - Max-min linear system solver written. It uses a sparse matrix
2920 structure taking advantage of its expected use. Most operations are
2921 O(1) and free/calloc are called as few as possible. The computation of
2922 the minimum could however be improved by using a red and black tree
2925 2004-11-03 Arnaud Legrand
2926 - Rename every gras_* function that was in xbt/ to its xbt_
2928 - Add a heap and a doubly-linked list to xbt
2929 - Added a dichotomy to the dictionaries. make check works as well before
2930 so I assume that the patch is correct. I do not know however if things
2931 run effectively faster than before now. :)
2933 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2934 SimGrid, and the version number is bumped to 2.x
2936 2004-10-29 Martin Quinson
2937 - Introduction of the remote errors.
2938 They are the result of a RMI/RPC on the remote machine.
2939 ErrCodes being scalar values, you can't get the host on which those
2940 errors did happen. Extending the error mechanism as in Gnome is possible.
2941 No idea yet whether it is a good idea.
2943 2004-10-28 Martin Quinson
2944 - Interface revolution: the Starred Structure Eradication.
2945 I used to do typedef struct {} toto_t; and then handle *toto_t.
2946 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2947 - ???_t is a valid type (builded with typedef)
2948 - s_toto_t is a structure (access to fields with .)
2949 - s_toto is a structure needing 'struct' keyword to be used
2950 - e_toto_t is an enum
2951 - toto_t is an 'object' (struct*)
2953 typedef struct s_toto {} s_toto_t, *toto_t;
2954 typedef enum {} e_toto_t;
2955 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2956 s_toto_t) is private.
2958 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2959 so that it can set the variable to NULL. It was so for dicts and sets,
2960 it changed for dynars.
2962 - Fix a bunch of memleaks in dict_remove
2963 - Fix a bug in sg/server_socket opening: it failed all the time.
2965 2004-10-07 Martin Quinson
2966 - Speed up dynar lookup operation a bit.
2968 gras_dynar_get is dead.
2970 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2971 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2972 (which gives you the address of the stored data).
2974 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2975 copy of the data using an affectation to do the job and not a memcpy.
2977 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2978 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2980 It does not really speedup the dynar test because they are
2981 setting elements all the time (and look them seldom). But the dict does
2982 far more lookup than setting.
2984 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2986 2004-10-05 Martin Quinson
2987 - Allow to (en/dis)able the cycle detection at run time.
2989 Whether we should check for cycle or not is now a property of each
2990 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2991 datadesc_cycle_unset allow to remove this property when previously set.
2993 Note that the cycle detection is off by default since it impacts the
2994 performance. Watch the data you feed GRAS with ;)
2996 This property is hereditary. Any element embedded in a structure having it
2997 set have it set for the time of this data exchange.
2999 You should set it both on sender and receiver side. If you don't set it on
3000 sender side, it will enter an endless loop. If you forget on receiver
3001 side, the cycles won't be recreated after communication.
3003 - Header reorganization.
3004 Kill gras_private.h, each submodule must load the headers it needs.
3006 2004-10-04 Martin Quinson
3007 - Interface revolution: do not try to survive to malloc failure.
3009 Now, gras_malloc and friends call gras_abort() on failure.
3010 As a conclusion, malloc_error is not a valid error anymore, and all
3011 functions for which it was the only gras_error_t return value are
3012 changed. They now return void, or there result directly.
3013 This simplify the API a lot.
3015 2004-09-29 Martin Quinson
3016 - Re-enable raw sockets.
3017 Created by gras_socket_{client,server}_ext;
3018 Used with gras_raw_{send,recv}
3021 It should allow to kill the last bits of gras first version soon.
3023 This is not completely satisfactory yet (duplicate code with
3024 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3027 - Simplify transport plugin (internal) interface by not passing any
3028 argument to _server and _client, but embedding them in the socket
3031 2004-09-28 Martin Quinson
3032 - Finish the port to AIX.
3033 autoconf was my problem (segfault within the malloc replacement
3034 function. No idea why)
3036 2004-09-16 Martin Quinson
3037 - Fix some size_t madness on 64bit architectures.
3039 2004-09-08 Martin Quinson
3040 - Reduce the number of system headers loaded, overload some more system
3041 calls (such as malloc to cast the result of the system one, and work
3043 - Fix and reintroduce the config support
3045 2004-09-07 Martin Quinson
3046 - Source code reorganization to allow Arnaud to surf all over there.
3047 - Allow to document the logging categories.
3048 - Remove all uppercase from logging categories and useless cleanup in names.
3050 2004-08-18 Martin Quinson
3051 Version 0.6.2 (protocol not changed; API changed)
3052 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3053 gras_error_t), and NULL when not found. Functions expecting a msgtype
3054 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3055 hopefully usefull message.
3056 - Portability to prehistoric sparcs again
3058 2004-08-17 Martin Quinson
3059 Version 0.6.1 (protocol not changed; ABI not changed)
3060 - prealloc some buffers to speed things up
3062 2004-08-11 Martin Quinson
3063 Version 0.6 (protocol not changed; ABI expended)
3064 - The parsing macro can deal with the references, provided that you add
3065 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3067 2004-08-09 Martin Quinson
3068 Version 0.5 (protocol not changed; ABI changed)
3069 - Allow to off turn the cycle detection code in data exchange at
3070 compilation time. It should be at run time, but I'm short of time (and
3071 the config stuff is still broken). That way, we keep dict out of the
3072 critical path, which is good because the performance is poor:
3073 - search not dichotomial yet
3074 - dynar give no way to access their content and memcpy everytime
3075 - In composed data description (struct, ref and so on), stop foolness of
3076 keeping the subtype's ID, but store the type itself. This keeps sets out
3077 of the critical path, which is good since they rely on dynar and
3078 dictionnaries. The only loose of that is that we cannot detect the
3079 redeclaration of a structure/union with another content (but I'm not sure
3080 the code detected well this error before anyway). We still can detect
3081 the redefinition discrepancy for the other types.
3082 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3083 it breaks the code because of type-punning used all over the place).
3084 This breaks on all non-gcc architectures (for now).
3086 All those changes (plus the buffer of last time) allow me to gain 2 order
3087 of magnitude on cruel tests consisting of 800000 array of integers on two
3088 level of a hierarchical structure (200 secondes -> 4 secondes)
3091 - the selector of reference must now return the type it points to, not
3092 the ID of this type.
3094 2004-08-06 Martin Quinson
3095 Version 0.4 (protocol changed; ABI not changed)
3096 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3097 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3098 - Implement a buffer transport, and use it by default (it relies on tcp in
3099 real life and on sg in simulation).
3100 That's a bit hackish since I had a new field to the structure to store
3101 its data without interfering with the subtype ones. Inheritance
3102 is tricky in C. And that's a kind of reverse inheritance with one class
3103 derivating two classes. Or maybe a game with java interfaces. Anyway,
3104 that's damn hard in C (at least).
3105 Moreover, I got tired while trying to ensure plugin separation and
3106 genericity in SG mode. MSG wants me to do weird things, so let's go for
3107 cruel hacks (temporarily of course ;).
3108 See comment in transport_private.h:71
3109 - do not include all the _interface headers in private but in the files
3110 which really need them (to cut the compilation time when they are
3113 2004-07-26 Martin Quinson
3114 Version 0.3 (protocol not changed; ABI changed)
3115 - Major overhault of the datadesc interface to simplify it:
3116 - shorted the function names:
3117 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3118 - add a trivial way to push/pop integers into the cbps without malloc.
3119 This allows to make really generic sub_type description, which simply
3120 pop their size of the stack.
3121 - add a function gras_datadesc_ref_pop_arr() which does what users want
3122 most of the time: Declare a dynamic array (which pops its size of the
3123 stack) and declare a reference to it. Poor name, but anyway.
3124 - kill the post-send callback, add a post-receive one
3126 2004-07-23 Martin Quinson
3127 Version 0.2 (protocol changed; ABI changed)
3128 - add some testing for cpbs in the test cases, and fix some more bugs.
3129 This invalidate again the little64 data file, since I cannot regenerate
3131 - remove an awfull optimization in the logging stuff, allowing me to:
3132 - understand it again
3133 - learn gcc how to check that the argument match the provided format
3134 - fix all errors revealed by gcc after that
3135 - internal keys of dict are not \0 terminated. Deal with it properly in
3136 loggings instead of segfaulting when the user want to see the logs :-/
3138 2004-07-22 Martin Quinson
3139 - Fix some stupid bug preventing cbps (callback postit) from working
3141 2004-07-21 Martin Quinson
3142 - Some documentation cleanups
3143 - remove the useless last argument of msgtype_declare
3144 - rename the Virtu functions to fit into the 'os' namespace
3145 - move headers src/include -> src/include/gras/ and stop fooling with
3146 gras -> . symbolic link
3147 - make distcheck is now successful
3149 2004-07-19 Martin Quinson
3151 - Build shared library also
3152 - Install html doc to the right location
3153 - stop removing maintainer files in make clean
3154 - build tests only on make check
3156 2004-07-13 Martin Quinson
3158 - No major issue in previous version => change versioning schema
3159 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3160 corresponding dataset.
3162 2004-07-11 Martin Quinson
3164 - Get it working with any kind of structure (we can compute the padding
3165 bytes remotely for all the architectures I have access to)
3166 - Implement the structure parsing macro (still not quite robust/complete)
3167 - Improvement to the remote testing toysuite
3169 2004-07-10 Martin Quinson
3170 [autoconf mechanism]
3171 - get ride of a bunch of deprecated macros
3172 - actually run the test for two-compliment, not only compile it :-/
3173 - test whether the structures get packed (and bail out if yes. Damn.
3174 Alignment is a serious matter)
3175 - test whether the structures get compacted (but respecting the alignment
3176 constraints of each types)
3177 - test whether the array fields of structures can straddle alignment boundaries
3179 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3180 (took me 2 hours to find that bug, looking at the wrong place)
3181 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3182 before sending/receiving and are used to compute the offsets of fields
3183 - Given that padding size depend even on compiler options, keep track of
3184 alignment and aligned_size only for the current architecture. Not a big
3185 deal since we send structure fields one after the other (seems
3187 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3188 program, let it work on linux/gcc/little32. portability todo.
3190 2004-07-08 Martin Quinson
3191 - import and improve remote compilation support from FAST
3192 - make sure make check works on half a dozen of machines out there
3194 2004-07-07 Martin Quinson
3195 Let's say it's version 0.0.3 ;)
3196 - Implement conversions (yuhu!)
3197 - Let it work on solaris (beside conversion, of course)
3198 - Stupid me, using rand() to generate the conversion datatests in not wise.
3200 2004-07-06 Martin Quinson
3201 - Let make dist work, since I'm gonna need it to compile on remote hosts
3202 - Let Tests/datadesc_usage write the architecture on which the file was
3203 generated as first byte.
3204 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3205 architecture descriptions.
3206 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3207 on those architectures.
3208 - Optimization: send/recv array of scalar in one shoot
3210 2004-07-05 Martin Quinson
3211 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3213 - Plug a whole bunch of memleaks
3214 - each process now have to call gras_{init,exit}. One day, their log
3215 settings will be separated
3216 - Continue the code factorisation between SG, RL and common in Transport.
3218 2004-07-04 Martin Quinson
3220 - Redistribution between SG and RL.
3221 We wanna have to accept in SG, so move accepted related parts of RL in
3222 the common part. (more precisely, the dynar of all known sockets is no
3223 more a static in transport.c, but part of the process_data)
3225 [gras_stub_generator]
3226 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3228 2004-07-03 Martin Quinson
3229 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3230 them all up in one shot)
3232 - Fix the ugly case of reference to dynamic array.
3233 - New (semi-public) function gras_datadesc_size to allow the messaging
3234 layer to malloc the needed space for the buffer.
3236 - gras_socket_close now expect the socket to close (and not its address to
3237 put NULL in it after it). This is because the socket passed to handlers
3238 is one of their argument (=> not writable).
3240 - propagate the interface cleanup from last week in datadesc, ie remove a
3241 superfluous level of indirection. User pass adress of variable
3242 containing data (both when sending and receiving), and not of a variable
3243 being a pointer to the data. Let's say that I like it better ;)
3244 The price for that is constructs like "int msg=*(int*)payload" in
3245 handlers, but it's a fine price, IMHO.
3247 - Let it work in RL (yuhu)
3249 2004-06-21 Martin Quinson
3251 - porting SG plugin and SG select to new standards (works almost).
3252 - plug memleaks and fix bugs around.
3255 - cleanup the prototype of data recv and force users to specify when they
3256 want to handle references to objects. Test case working even for cycles.
3257 - plug memleaks. Valgrind is perfectly ok with this.
3259 2004-06-12 Martin Quinson
3261 - cleanup the separation between plugin and main code in plugin creation
3263 2004-06-11 Martin Quinson
3265 - Reput hook for raw sockets, needed for BW experiments
3266 - kill a few lines of dead code
3267 [Data description] Interface cleanup
3268 - gras_datadesc_by_name returns the searched type or NULL.
3269 That way, no variable is needed to use a type desc once, which makes
3271 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3272 two parameters were strings (field name, type name), leading to
3274 [Dicos] Interface cleanup
3275 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3276 This is consistant with the dynar API.
3278 2004-04-21 Martin Quinson
3280 - Porting to new standards.
3282 - interface cleanup.
3283 There is no bag anymore, no need to take extra provision to mask the
3284 pointers behind "ID".
3285 Better splitup of functions between files create/exchange/convert.
3286 This is still a bit artificial since convert and receive are so
3287 interleaved, but anyway.
3289 - add a queued message list to procdata (the ones not matching criteria
3291 - factorize some more code between SG and RL wrt procdata
3293 - use gras_exit in example to track memleaks
3294 - get rid of gs_example now that GS is properly integrated into gras
3295 - update run_test to integrate the latest tests (datadesc)
3297 - rename WARNINGn macros to WARNn since it prooved error-prone
3299 2004-04-19 Martin Quinson
3301 - register init/exit functions within gras module mechanism
3302 - send/receive function.
3303 Convertion is not implemented, but short-cutted if not needed.
3304 struct/array elements are sent one by one (instead of block-wise), but
3305 nobody really cares (yet). Get a prototype before optimizing.
3306 - tests (using a file socket) for DD send/receive on:
3307 - base types: int, float
3308 - array: fixed size, string (ie ref to dynamic string)
3309 - structure: homogeneous, heterogeneous
3310 - chained list, graph with cycle
3311 Believe it or not, valgrind is not too unhappy with the results. The
3312 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3316 - Bugfix when using a filename explicitely (instead of '-')
3318 2004-04-09 Martin Quinson
3320 - factorize more code between RL and SG in socket creation
3321 - Complete the implementation and tests of:
3323 o file (only in RL, and mainly for debugging)
3325 I lost 3 days to design a portable address resolver, and then decided
3326 that the prototype mainly have to run on my box.
3327 Addressing portability too early may be like optimizing too early :-/
3329 - use gras_init in the Tests instead of the crappy parse_log_opt
3330 (the latter function is removed)
3331 [Conditional execution]
3332 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3333 [Code reorganisation]
3334 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3335 Build examples against the RL library, since there is no way to disable
3336 its creation for now.
3338 For information, the beginning of coding on GRAS was back in june
3339 2003. I guess that every line has been rewritten at least twice since