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
19 - New and Backwards Compatibility break:
20 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
21 When one wants to get the list of tasks whose states have changed during a
22 simulation round, s/he has to allocate and free a dynar and use it as argument
23 to this function. The former SD_simulate (double how_long) now returns void.
26 - The netzone are now available from the MSG API.
27 The old names still work, but are now deprecated.
30 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
31 - New option: smpi/keep-temps to not cleanup temp files
32 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
35 - Replay: New function xbt_replay_action_get():
36 Retrieve the function previously associated to an event type.
37 - DROPPED FUNCTION: xbt_str_varsubst()
38 - DROPPED MODULE: strbuff. We don't need it anymore.
40 -- Release target: June 21 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
42 SimGrid (3.15) stable; urgency=low
44 The Spring Release: continuous integration servers become green
45 We fixed even the transient bugs on all target architectures:
46 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
47 Windows, FreeBSD, NetBSD.
49 - Rename NetCards to NetPoints.
50 This was intended to help NS3 users, but that's not a netcard.
51 That's a point in the routing algorithm, let's avoid wrong simplifications.
54 - New: MSG_process_yield(). Stop and yield to other processes.
55 - New: MSG_process_daemon(). Daemon processes are automatically killed
56 when the last non-daemon process terminates
57 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
58 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
59 to make room for the upcoming network energy plugin.
60 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
63 - Ensure that an actor can kill itself with Process::exit()
64 - Kill the obscure NativeException. Nobody want to survive the issues
65 it denotes, so use JniException that is a RuntimeException (not to
66 be caught explicitly).
67 - Partial bug fix in initialization. SimGrid flags on command line were
68 consumed at C level but stayed in the original Java String[] args.
69 This could mess users' args[i] if SG flags were not put at the end of
71 The SimGrid flags are now removed from the Java arguments. However,
72 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
73 if args.length is greater than the number of YOUR OWN ARGUMENTS.
74 It might be if you have --log or --cfg flags in the command line.
75 - Fix numerous memleaks all around the place. In particular, around VMs.
79 - s4u::onPlatformCreated: right before the simulation starts
80 - s4u::onSimulationEnd: right after the main simulation loop
81 - s4u::onTimeAdvance: right after a clock change (time is discrete)
82 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
83 event from the availability_file changes the avail speed.
84 - Links are now usable from s4u
85 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
86 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
87 - New examples: The conversion of MSG examples to S4U has begun
88 - Actors: create, kill, migration, and suspend
89 - Applications: master-worker and token-ring
90 - Action replay: communications and storage
91 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
94 - Backwards Compatibility breaks
95 - The SD_route_get_list and SD_route_get_size functions have been removed.
96 They are replaced by the unique
97 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
98 The route from src to dst is built in the links dynar whose size can
99 be obtained with xbt_dynar_length.
100 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
101 sg_host_route_bandwidth and sg_host_route_latency.
102 Macros ensure the backwards compatibility, but you should fix your code
105 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
106 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
107 - Bump our claim of support from MPI 1.1 to MPI 2.2.
108 We don't support 100% of it, but it should be enough. Contact us if not.
109 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
110 naive and probably inaccurate)
111 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
112 mapped onto a unique small file using some system magic.
113 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
114 - Fortran: cleanups, fixes, support of user-added operations
115 - RMA: MPI_Accumulate are applied in correct order.
116 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
117 - Better support for MPI_IN_PLACE
118 - Support for MPI_Win attrs and keyvals.
119 - Support MPI_Comm_group_create, MPI_Type_size_x
120 - MPI ops should be properly applied to all allowed types
121 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
122 algorithm by a non failing one from mpich.
125 - Kill the fifo data container: we don't use it anymore.
128 - A <cluster> can now be created with different speed values to represent pstates.
129 see examples/platforms/energy_cluster.xml for an example.
131 -- Wed Mar 22 17:50:21 CET 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
133 SimGrid (3.14.159) stable; urgency=low
135 The previous release was only a crude approximation of a Pi release;
136 this one should be a bit better.
138 - Revert a last minute change that broke on Mac OSX.
139 - Fix the changelog and NEWS files.
140 - Improve the documentation, in particular of the routing module.
141 - Integrate some patches flying around in Debian.
143 -- Wed Dec 28 17:14:49 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
145 SimGrid (3.14) stable; urgency=low
148 * The whole documentation was reworked and reorganized. There is
149 still much room for improvement here, but we're on it.
152 * We now need python3 for our testing infrastructure.
153 * Model checking needs libevent
156 * Now works on FreeBSD
159 * Privatization now works on FreeBSD
161 * Privatization is now activated by default in smpirun
162 You can turn it off with -no-privatize if something goes wrong.
164 * Call-location tracing for SMPI.
165 You can add the exact location (filename / linenumber) of an MPI call to
166 your trace files and slow down or speed up the simulation between two
167 consecutive calls by using an adjustment file (see the documentation).
169 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
170 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
171 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
172 values were ignored in some cases. The timings of these functions can now
173 be significantly different.
175 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
176 smpi/running-power is renamed to smpi/host-speed
178 * smpi/grow-injected-times option to enable or disable multiplication of the
179 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
180 make simulation less precise (but also much faster).
182 * smpirun script should be (much) faster for large deployments.
184 * SMPI tracing : fixed issue with poor matching of send/receives.
186 * Replay : Fix broken waitall
188 New functions and features
189 * MSG_parallel_task_execute_with_timeout, to timeout computations.
191 Dropped / renamed functions and features
192 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
193 - MSG_mailbox_is_empty() -> Mailbox::empty()
194 - MSG_mailbox_front() -> Mailbox::front()
195 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
196 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
197 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
198 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
199 * Don't pass the free_f parameter to property related functions:
200 - MSG_host_set_property_value()
201 - MSG_as_router_set_property_value()
202 - MSG_storage_set_property_value()
203 * VM properties. Since msg_vm_t are msg_host_t, just use
204 MSG_host_get_property_value() and friends
205 * VM I/O related things:
206 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
207 - Unimplemented save/restore methods
208 * MSG_as_router_get_property_value() was redundent with
209 MSG_environment_as_get_property_value().
210 - Removed MSG_as_router_*propert*() functions
211 - Added MSG_environment_as_set_property_value() for consistency
212 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
213 Well, they are still there for now, but deprecated with a warning.
214 Please switch to xbt_dict_new_homogeneous() before this is removed
216 * Task affinity. Its intended behavior (that was very badly tested
217 and probably not really working) was deceiving what most users
218 would have hoped here.
219 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
220 * Option network/coordinates is now useless and should be dropped.
223 * Added option storage/max_file_descriptors to allow more than 1024 files opened
226 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
227 * Added Dragonfly topology support
230 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
231 EiBps, EBps, Eibps, Ebps for bandwidth.
232 They may become useful to some lucky ones.
235 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
236 * ex/app_tokenring: new example, very similar to the MSG Token Ring
237 * ex/async_waitAll: new example, on asynchronous communications
240 * Memory usage should be decreased for simulations with a large number
241 of processes. This also helps for SMPI.
243 -- Sat Dec 24 22:34:26 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
245 SimGrid (3.13) stable; urgency=low
247 The Half Release, a.k.a. the Zealous Easter Trim.
249 - We removed half of the lines, that were mostly experimental cruft.
250 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
251 (not counting blanks and comments -- according to openhub.net)
252 - The internals are now compiled in C++ (and will soon be clean C++)
253 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
254 59 klines of C, etc. We added only 29 klines of C++ in replacement.
256 * Backwards Compatibility breaks
257 - Removed Lua simulation bindings (switch to C or Java for that).
258 Lua can still be used to describe platforms
259 - Removed Java kernel plug-ins.
260 Will be reintroduced after the ongoing major internals reorg.
262 - the following functions were removed.
263 They were too specific and should be reimplemented in a generic
264 way, with filter function.
265 - MSG_task_listen_from_host
266 - MSG_mailbox_get_count_host_waiting_tasks
267 - MSG_mailbox_put_with_timeout was removed.
268 Please use MSG_task_send_with_timeout instead.
270 - the SD_application_reinit function was removed. It has been a noop for a while.
271 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
272 replaced by a more flexible mechanism.
273 - The following functions thus do not exist anymore
274 - SD_workstation_get_access_mode
275 - SD_workstation_set_access_mode
276 - SD_workstation_get_current_task
277 - Basic estimation functions have been removed but can easily be replaced
278 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
279 - SD_workstation_get_computation_time => amount / sg_host_speed()
281 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
282 Use VM.getSpeed()*load/100 for the legacy behavior.
284 - option enable_tracing was removed. It was not doing anything for a while.
285 - In the ModelChecker:
286 - the model-checker now ptraces the model-checked process which means
287 you cannot use a debugger on the latter anymore (we might make this
288 optional in the feature);
289 - removed soft-dirty page tracking;
290 - remove model-checked side snapshot management,
291 MC_snapshot() and MC_compare_snapshot();
292 - keep the MC_cut() function as a stub (it was not really working
293 in the previous release).
296 * All options are consistently in kebab-case. Old names are kept as alias.
299 * Switch to platform v4 format.
300 - Rename from 'power' to 'speed' the attributes describing the amount of
301 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
302 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
303 - In <host> and <link>, attributes availability and state are gone.
304 It was redundent with state and availability traces, and with peak values.
305 - In <cluster>, attributes availability_file and state_file are gone.
306 It was too complex and unused.
307 - Kill <gpu>. Was not doing anything.
308 - The DOCTYPE points to the right URL:
309 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
310 (the file at this address now documents the changelog since its v1)
311 - A warning is emitted for unit-less values (they are still accepted).
312 - speed. Default: 'f' or 'flops'. Also defined:
313 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
314 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
315 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
316 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
317 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
318 - latency. Default: 's' second. Also defined:
319 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
321 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
322 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
325 * s4u::Host is now the preferred public interface to the Host features.
326 sg_host_* functions are C bindings to the exact same behavior
327 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
330 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
331 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
332 They were deprecated since a few years, and probably did not even compile anymore.
335 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
336 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
337 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
338 encouraged to update their codes. A list of the modified functions can be found at the end of
339 include/simgrid/simdag.h
342 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
343 the lambda or closure passed as a parameter will run in kernel mode.
344 All the callback functions should be rewritten to that interface at some point.
347 * Reorganizing and cleaning the internals all around the place.
350 * Remove old default barrier/bcast buggy algorithms (see #18407)
351 * Various bug fixes to handle more codes
352 * Remove the need for the --foreground option of smpirun (it is still
353 accepted for backward compatibility).
356 * Kill the setset data container: MC don't use it anymore.
357 * Kill the queue data container: it made more sense with GRAS.
358 * Kill the xbt_peer_t data type: it's useless without GRAS.
359 * Kill rm_cb feature of config sets: it was never useful.
360 * Kill graphxml parsing feature. It was not used.
361 * Kill the deprecated code protected by XBT_USE_DEPRECATED
363 - xbt_dynar_sort_strings(), when the content is char*
364 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
365 They throw exceptions on invalid input;
366 * C++ support for declaring CLI flags (simgrid::config::Flag);
367 * class for abstracting different signal backends (simgrid::xbt::signal).
368 with no external dependencies (we need very simple signals).
371 * refactoring and cleanup of the code;
372 * ongoing process to cleanly separate the model-checking algorithms
373 from the code model-checking support.
375 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
377 SimGrid (3.12) stable; urgency=low
379 The Facelift Release.
382 * Require g++ v4.7 at least to not speak prehistorical C++.
383 * Require Boost 1.48 (for signal2 component).
384 * Java must be version 7 at least when activated.
385 * Builds on Windows again (including Java bindings).
386 * Tracing is now always enabled (no way to turn it out)
387 * Remove GTNetS. It was not working anyway.
388 * Various cleanups in the cmake scripts.
389 * Move headers around to sort them out on installed systems:
390 - instr/instr.h -> simgrid/instr.h
391 - instr/jedule/* -> simgrid/jedule
392 - simdag/datatypes.h was removed
393 - simdag/simdag.h -> simgrid/simdag.h
394 - msg/datatypes.h was removed
395 - msg/msg.h -> simgrid/msg.h
398 * Interface improvement:
399 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
400 - Sanitize the interface in MSG_task_ module:
401 - Merge two functions that were close enough but misleading:
402 set_compute_duration(t) -> set_flops_amount(t)
403 get_remaining_computation(t) -> get_flops_amount(t)
404 - set_data_size(t) -> set_bytes_amount(t)
405 get_data_size(t) -> get_bytes_amount(t)
406 - Massive cleanups in the functions related to the energy
407 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
408 - New: MSG_host_get_pstate()
410 - msg/energy/onoff: switching hosts on and off
412 * Interface improvement:
413 - Rename Host.isAvail() to Host.isOn()
414 - Rename Process.currentProcess() to Process.getCurrentProcess()
415 - Rename Task.setDataSize() to Task.setBytesAmount()
416 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
418 - #18874: Actually allows the GC to reclaim tasks
422 - SIMIX_process_throw: raises an exception in a remote process
423 * Refactoring: Separate sub-modules
424 - libsmx: the public interface, as libc in a real system
425 - popping: the strange dance that converts a user request into a kernel handling
426 - smx_context_*: the virtualization mechanisms that embed the user code
427 - smx_*: the handling of each simcalls
429 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
430 * Rename smx_action_t into smx_synchro_t, making explicit that these
431 things are used to synchronize processes with their environment.
432 For example, a communication is a sort of synchronization involving
433 the communicating processes (that may block until the exchange) and
434 the platform. The same can be said from computations, etc.
436 - #18888: segfault when a process ends before its kill_time
440 - 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.
441 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
442 - MPI_Type_set_name, MPI_Type_get_name
443 - MPI_*_c2f and MPI_*_f2c functions
444 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
445 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
446 - Activate a lot of new tests from the mpich 3 testsuite
448 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
449 - InfiniBand network model added : Based on the works of Jerome Vienne
450 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
451 - When smpi/display_timing is set, also display global simulation time and application times
452 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
453 * Collective communications
454 - 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
455 - 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/).
456 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
458 - Replay now uses algorithms from wanted collective selector
459 - Replay can be used with SMP-aware algorithms
460 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
461 - 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.
463 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
464 - Correctly use loopback on fat-tree clusters
465 - Asynchronous small messages shouldn't trigger deadlocks anymore
466 * Energy/DVFS cleanup and improvement
467 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
468 - new: smpi_get_host_pstate()
471 - "Full" network optimization flag was broken since Surf++
472 - Better handling of precision flags in maxmin
473 - Fix bug causing sometimes "Impossible" errors
474 - Properly pass cluster properties to included hosts
475 * Improvement of the Energy plugin.
476 - Always update the consumption before returning that value
477 - New property: watt_off to denote the dissipation when the host is off
478 - New functions getWattMinAt and getWattMaxAt to retrieve the
479 dissipation of pstates that we are not currently at.
480 * Java: class NetworkLink renamed to Link
481 * New function: simcall_process_get_kill_time()
482 * Massive rename s/workstation/host/
483 - That's intrusive, but that's good for the project consistency. Sorry.
484 - Change config option "workstation/model" into "host/model"
487 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
488 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
489 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
490 - xbt_dirname and xbt_basename for non-POSIX systems
492 * The model checker now runs as a separate process.
493 * The model checker runs must now be launched with the new simgrid-mc program.
494 * Record/Replay: the MC can display a textual representation of a path in the
495 execution graph. It can then be replayed outside of the model checker.
497 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
499 SimGrid (3.11) stable; urgency=low
504 * Normalizing pointers addresses tool for better diff between logs
507 * Add cloud examples using new VMs
508 - examples/msg/cloud/two_tasks_vm.tesh
509 - examples/msg/cloud/simple_vm.tesh
510 - examples/java/cloud/cloud.tesh
511 - examples/java/cloud/migration/migration.tesh
512 * Add java surf examples:
513 - examples/java/surfPlugin/surf_plugin.tesh
514 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
515 - examples/java/surfCpuModel/surf_cpu_model.tesh
516 * Add SMPI+MSG example:
517 - examples/smpi/smpi_msg_masterslave/
522 - msg task destroy cancel test
523 - msg_host on/off test
524 * Move all tests in testsuite to teshsuite (adding tesh files)
525 * Restructure teshsuites
526 - one folder for each kind of test
527 * Restructure AddTests.cmake
529 - structure the order of tests (with sections)
532 * Add virtual machine
533 - creation of a VM on a PM
534 - migration of a VM from a PM to another PM
536 - MSG_process_join(msg_process_t process, double timeout)
537 - msg_bar_t MSG_barrier_init(unsigned int count)
538 - int MSG_barrier_wait(msg_bar_t barrier)
539 - void MSG_barrier_destroy(msg_bar_t barrier)
540 - msg_as_t MSG_environment_as_get_by_name(const char * name)
541 * New option "msg/debug_multiple_use" to help debugging when a task is used
544 - Improvements and finalization of MSG_storage, MSG_file APIs and their
546 - Increase code coverage in test suites
550 * Protect context stacks against stack overflow. The number of protected memory
551 pages allocated on the top of each stack (1 by default) can be configured
552 with the new command line option --cfg=contexts/guard_size:<value>.
553 * Simcalls are now generated by a python script that
554 - generates files included by SimGrid
555 - checks that all the functions exist, and proposes prototypes
557 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
558 host_set_data, host_get_data
560 - simcall_process_join(smx_process_t process, double timeout)
561 * Fix bug where sleeping processing could not be suspended.
564 * Translate surf models from C to C++
565 - Generic classes for all models: Model, Resource, Action
566 - A generic interface for each kind of model (CPU, Network, Storage
567 Workstation, WorkstationVM)
569 * Translate surf routings from C to C++
570 * Add callbacks using sigc++ or boost::signals2
571 - Add callback functions for resource creation/destruction
572 - Add callback functions for action state change
573 - Handle Energy as a plugin
574 * Replace swag by boost::intrusive
575 * Add new routing models for clusters. For documentation, see
576 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
577 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
578 parameters for cluster tag
579 - Fat trees, with topology="FAT_TREE" and
580 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
581 - see examples/platforms/torus_cluster.xml and
582 examples/platforms/fat_tree_cluster.xml
586 * Hostfiles support host:nb_processes construct to deploy several processes on
588 * Collective communication algorithms should not crash if used with
589 improper number of nodes and report the error.
590 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
591 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
592 MPI_Dims_create, MPI_Cart_sub are supported.
593 * New interface to use SMPI programmatically (still depends on MSG for
594 some parts, see examples/smpi/smpi_msg_masterslave) :
595 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
599 * Global variables privatization in MPI executables is now performed at runtime
600 with the option smpi/privatize_global_variables (default:no).
601 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
602 dynamic libraries loaded by the application are not privatized (static
603 linking with these libraries is advised in this case)
606 * Options defined in XML work correctly now.
609 * New cmake option, enable_lib_in_jar, to control whether native libraries are
610 copied into simgrid.jar or not (ON by default). Use this option if you want
611 to reduce the size of the installed simgrid.jar, *and* the native libraries
612 are kept installed elsewhere.
613 * Surf binding with SWIG (code generated in maintainer mode only):
614 - plugin to handle callbacks
615 - CPU model only for the moment
618 * Supernovae build mode is definitively removed. It was used to improve
619 inlining and inter-module optimizations. It is nowadays superseded by
620 link-time optimizations commonly available in compilers.
621 * Update ns-3 find lib. Bindings for ns-3 should work again now.
622 * Add boost dependency for surf++
623 * Add new macro for tests
624 - ADD_TESH(name <tesh_args>)
625 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
629 - xbt_bar_t XBT_barrier_init(unsigned int count)
630 - int XBT_barrier_wait(xbt_bar_t barrier)
631 - void XBT_barrier_destroy(xbt_bar_t barrier)
632 * Make the xbt_os_time module public
634 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
636 SimGrid (3.10) stable; urgency=low
638 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
641 * Reintegrate Java to the main archive as desynchronizing these
642 package is not acceptable anymore (Java is now considered stable)
643 * Add explicit synchronization facilities through semaphores
644 * Bug fix: Task.setDataSize() only changed the C world, not the value
645 cached in the Java world
648 * Dramatically change the way files are handled. API and internals changed, but
649 this part of MSG was not considered as production grade either.
650 * Add explicit synchronization facilities through semaphores
651 * Add a new function MSG_host_get_process_list()
652 * Preliminary DVFS support (see examples/msg/energy/ for details)
655 * SMPI is now included directly in the libsimgrid as the windows
656 linker doesn't force us on splitting it anymore.
657 * Improvements of the SMPI replay tool:
658 - Most of the collective communications are now rooted in the same process as
659 in the original application.
660 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
661 used until now). Multiple data types can now be used in a trace.
662 - The replay tool now supports traces produce either by TAU or a modified
664 - Bug Fix: the compute part of the reduce action is now taken into account.
665 - Gatherv collective is now supported
666 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
667 of smpirun outputs time independent traces for the current run. One file
668 is created per process. If too many processes are simulated, this behavior
669 can be changed to one file for all processes by using the
670 tracing/smpi/format/ti_one_file flag
671 * smpirun generates the host file if needed (with given host count and platform)
672 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
673 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
674 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
675 --cfg=smpi/coll_selector:(mpich/ompi)
676 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
678 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
679 happening inside a collective SMPI call.
680 * Fix the behavior of complex data types handling.
681 * Make MPI_Wtime another synchronization point to take computations into
683 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
684 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
685 * Add all missing Fortran bindings, SMPI should work with Fortran 90
686 (no privatization of global variables yet)
687 * Preliminary DVFS support (see examples/smpi/energy/ for details)
690 * Verification of liveness properties is now available for SMPI applications
691 (in addition to MSG applications)
692 * Bugged examples using SMPI in examples/smpi/mc/
693 * Add --cfg=model-check/visited option. Allows the verification of infinite
694 programs. Detection of loops in the execution thanks to the system state
695 comparison and reduction of the state space to explore. Can be combined with
696 DPOR for safety properties.
699 * Allow to change SimGrid configuration (see --help) within the code
700 thanks to SD_config() as it can be done in MSG.
701 * Add a new function SD_task_set_amount() upon user request.
704 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
705 * Remove rule based routing (no more PCRE dependency)
706 * Add a limiter_link option to cluster tag, to specify a maximum reachable
707 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
708 * Add a loopback_bw and loopback_lat options to cluster tag.
709 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
710 routing. Now peers are encapsulated in an AS and have their own private
711 router but this is transparent.
714 * Our own implementation of getline is renamed xbt_getline, and gets
715 used even if the OS provide a getline(). This should reduce the
716 configuration complexity by using the same code on all platforms.
717 * New type: xbt_cfg_elm_boolean.
718 * Allow to use yes/no for boolean configuration options in the command line.
719 * Allow to disable SimGrid cleanups at exit from command line option.
720 There are situations where one may want a simulation to end with an exit.
721 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
722 annoying when scripting around the simulator. Adding a
723 --cfg=clean_atexit:no allows to circumvent this issue.
726 * Lots of memory leaks were corrected in this release.
727 There is no known memory leaks anymore, in all of our 600+ tests.
728 * New command line option --version, to get SimGrid version information.
729 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
731 * Supernovae builds are deprecated, and expected to be removed in the next
734 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
736 SimGrid (3.9) stable; urgency=low
738 The Grasgory release: GRAS is really dead now.
740 * Complete overhaul of the internal host structures scheme.
743 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
744 was considered as experimental and badly maintained since then.
745 * Keeping it was thus a trap to our potential users, that could take
746 it instead of MSG or SMPI by mistake despite is pity state.
747 * GRAS seems to have very few users (if any), and no one volunteered
748 to maintain it further. It also induces a lot of XBT code (for
749 portability sake), that must be maintained too.
750 * For all these reasons, we killed GRAS. If someone wants to revive it
751 in the future, don't cry, our git history still remembers of GRAS.
754 * Major overhaul. Merge our documentation again as time proved that
755 splitting it was really not helping our users.
756 * Further improve the developer documentation to help newcomers
757 hacking on SimGrid itself. The user documentation (and in
758 particular, the beginner documentation) is still in a sorry state.
761 * Now works on Windows too!
762 * Much more extensive test suite, from MPICH
765 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
766 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
767 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
768 * Bug fix: let task be scheduled when the last dependency to be solved is
769 a control dependency.
770 * Remove SD_load_environment_script function.
771 Use the C sg_platf function if you want to declare a platform
775 * New function: MSG_process_get_number()
776 * Old function documented: MSG_config()
777 * Remove MSG_load_platform_script function
778 Use the C sg_platf function if you want to declare a platform
782 * Change the default value of the TCP_gamma constant (maximal size of TCP
783 congestion window) to a more realistic 4MiB value. If you notice changes in
784 your simulation results, you can fall back to the previous 20k tiny window
785 by adding --cfg=network/TCP_gamma:20000 on command line.
786 * (Hopefully) fix a bug wrt periodic availability and state traces
787 * Bug fix: use default values at start when first event in availability/state
788 trace is not at time 0.
791 * remove the "new_" part of function name sg_platf_new_trace_connect
792 (resulting in sg_platf_trace_connect), since it does not create
796 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
797 use case where it's really mandatory, and maintaining it was a pain in
799 * New: xbt_fifo_search(), search an item with a user-provided
800 comparison function instead of dumb pointer comparison.
803 * Fix the lua deployment:
804 Use `simgrid.init_application()` before deployment instead of
805 `simgrid.msg_register_application()` after.
808 * Transfer the tracing files into the corresponding modules.
810 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
812 SimGrid (3.8.1) stable; urgency=low
814 The "we are told that some people want to also *install* the simgrid
817 * Add missing file "tesh.1" to the archive.
819 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
821 SimGrid (3.8) stable; urgency=low
823 The Psssshiiiit release: SimGrid jumps into the Cloud.
826 * Add an experimental interface to manipulate VMs. They are mainly
827 process groups with very few intrinsic semantic, but they should
828 allow you to build the semantic you want easily.
829 * New function: MSG_host_set_property_value()
830 * New function: MSG_process_on_exit(). To clean memory in all cases.
831 * Bug fixes that made the host (and link) failures unusable.
832 * Add a way to auto-restart process when the host in which they are
833 executing comes back (ON_FAILURE="RESTART" on deployment file,
834 MSG_process_auto_restart_set).
835 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
836 please stop using the old ones, they are DEPRECATED.
838 * Deprecate functions MSG_global_init() / MSG_global_init_args()
839 Please use MSG_init() instead. (reducing the amount of entry
840 points in the library helps us).
841 * Make it impossible to link against the wrong version of the lib
842 * Deprecate MSG_clean(). No need to call it anymore.
843 * Function MSG_get_host_number() is not deprecated anymore.
846 * Split the doc into a user guide and a reference guide.
847 * Start a developper guide to help people hacking on SimGrid.
850 * Enable tracing by default. This modules rocks you should use it.
851 * Remove option custom_flags. Now use environment variables CFLAGS
853 * Use default cmake things to detect lua instead of home grown ones.
854 * New option "enable_mallocators" to disable mallocators, for debugging
855 purpose ("on" by default).
858 * Bug fixes around the resource failures: don't let the processes
859 survive the host they are running onto.
860 * Add an interface to auto-restart processes when the host in which they are
861 executing comes back.
862 * Ensures that SIMIX_clean is called automatically. It's not part of
863 the public interface anymore (bindings should be updated).
866 * Bug fix for when SD_Simulate is called with a positive value: be careful
867 when comparing doubles. Sometimes they are different for non significant
869 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
870 parallel task whose initial work is distributed among host according
871 to the Amdahl's law. Such tasks are created with a parameter alpha
872 that corresponds to the non-parallelizable part of the computation.
873 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
874 between two sets of workstations assuming a 1D block distribution (each
875 workstation owns a similar share of data) on both sides.
877 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
878 Data redistribution will be automatically scheduled once parent and child
879 are both scheduled. The filling of computation_amount and
880 communication_amount structures is now done seamlessly thanks to the chosen
882 * New function SD_workstation_dump to display various information
883 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
884 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
885 bandwidth on the route onto which the task is scheduled and the amount of
887 To divide the nominal bandwidth by 2, the rate then has to be :
888 rate = bandwidth/(2*amount)
889 * Compute tasks that have failed can now be rescheduled and executed again
890 (from their beginning)
891 * Increasing source code coverage (src/simdag is now covered at 95.8%
895 * Re-implement time-independent trace replay using SMPI (at the
896 smpi_smp_* level) instead of MSG. This should replace
897 examples/msg/actions/actions.c
898 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
899 hindexed and structs)
900 * Implement the exchange of non-contiguous data.
901 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
902 * Correct behavior of smpi/sender_gap and set its default value to 0
903 * Add option to asynchronously send small messages to allow better
904 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
905 specifies the size in bytes under which messages will be asynchronously sent.
906 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
907 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
910 * New C interface to define a platform: XML is now optional.
911 For more info, please check include/simgrid/platf.h.
912 * New interface to define random platforms from the C:
913 For more info, please check include/simgrid/platf_generator.h and
914 examples/msg/masterslave/masterslave_platfgen.c
915 * Export a sg_cmdline dynar containing all the arguments we got from
919 * Two new tracing options for adding comments to trace file so you
920 can track your experiments (see --help-tracing for details).
921 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
922 * Adding the SimGrid version that generated the trace file as a comment.
923 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
924 * Fix to avoid key clashes on Paje links
925 * Other minor fixes related to the Paje specification
928 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
929 and renamed to xbt_str_hash() and xbt_str_hash_ext().
930 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
931 * Greatly improve the robustness of mmalloc to user errors (such as
932 using an area after freeing it, or freeing it twice)
934 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
936 SimGrid-java (3.8.1) stable; urgency=low
938 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
939 Warning: all methods are visible, but only some of them are
940 implemented so far. Check the source (src/jtrace.c)
941 for further information.
942 * New module: org.simgrid.msg.File (SimGrid File management functions)
943 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
944 * Change the meaning of Process.restart: now restart the process from
945 the begining, like MSG_process_restart in C.
946 * Add Process.setAutoRestart: handling of process restart when failed
948 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
949 you to retrieve the properties of the processes/hosts
950 * Deprecate Msg.clean(): you can just forget about it now.
951 * New function Process.getCount(), that only works when compiling
952 with the not yet released version 3.9 of the C library.
954 * New context factory based on Coroutines. It mandates a modified JVM
955 but then, the simulations run about five times faster, and there is
956 no limit to the amount of processes (beside of the available memory).
958 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
960 SimGrid (3.7.1) stable; urgency=low
963 * Restore the prototype of MSG_process_create_with_environment() to
964 the pre-3.7 situation by removing the kill_time argument.
965 * Add a MSG_process_set_kill_time() function instead.
968 * Fix weird behaviors when dealing with parallel tasks.
971 * Simgrid is now built as a dll.
972 * Simgrid-java now works on Windows.
973 * Simgrid-Java is now included into Windows package.
976 * First pre-build package for MacOSX.
979 * Fix compilation when using MSG_USE_DEPRECATED.
980 * Fix some compilation issues on Macs and Windows.
981 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
982 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
984 We discovered that the Lua console is broken, but we are missing the
985 manpower to fix it right now. The problem existed in 3.7 too, so we
986 are not blocking the release for that. Sorry if you depended on this
987 feature, any help would be really welcome.
989 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
992 SimGrid-java (3.7.1) stable; urgency=low
994 The "Java aint got to be bloated and slow" release
997 * Various internal cleanups and performance improvement
998 Simulations are expected to run up to twice faster or so
999 * Make Process.kill(process) an instance method, not a static one
1000 * User processes are not java.lang.Thread subclasses.
1001 This breaks the compatibility (sorry), but previous API was
1002 brain-dead, making it impossible to have non-trivial
1003 initializations in the process constructor.
1004 * Require a full constructor per Process sub-class.
1005 Kinda breaks the compatibility (sorry), but this allows a much more
1006 efficient way to launch the processes at simulation startup.
1007 * Do not embeed our version of semaphores, java 1.5 can be considered
1008 as sufficiently prevalent for us to not dupplicate its features.
1012 * Add examples for almost every part of the API
1013 We spotted and fixed a lot of bugs in the process
1015 * New module: asynchronous communication API
1016 * New function: Process.sleep()
1017 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1018 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1019 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1021 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1023 SimGrid (3.7) stable; urgency=low
1025 The "spring cleanups (before the next Big Project kicks in)" release.
1028 * We can specify the SMPI latency/bandwidth factor with command line
1029 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1030 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1031 You can also use the "config tag" from platform file by adding this line
1032 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1033 (see "example/platforms/tag_config.xml" to use "config tag").
1034 Note that the command line supersedes the platform file configuration.
1035 * Change the correction factors used in LMM model, according to
1036 the latest experiments described in INRIA RR-7821.
1037 Accuracy should be improved this way.
1038 * Use the partial invalidation optimization by default for the
1039 network too. Should produce the exact same results, only faster.
1040 * Major cleanup in surf to merge models and split some optimization
1041 mechanisms from the core of the models. As a result you can now
1042 specify which model to use (e.g., --cfg=network/model:LV08
1043 --cfg=cpu/model:Cas01) and which optimization mode to use
1044 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1045 Incompatible combinations should err at initialization. See
1046 --help-models for the list of all models and optimization modes.
1047 * The CLM03 workstation model was dropped for simplicity because it
1048 used the deprecated CM02 network model. Use default instead.
1049 * Rename the TCP_gamma configuration option to network/TCP_gamma
1050 * Rename the coordinates configuration option to
1051 network/coordinates, and document it
1052 * Use now crosstraffic keyword instead of the terribly misleading
1053 fullduplex keyword. It is activated by default now in the current
1054 default model, use --cfg=network/crosstraffic:0 to turn it off.
1055 * Ongoing refactoring the model parsing to make XML files optional
1056 See include/simgrid/platf.h for details (still to be completed)
1059 * Major overhaul of the documentation. Almost instructive now :/
1060 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1061 functions and friends. This interface was considered as
1062 deprecated since over 2 years, it's time to inform our users that it is.
1063 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1064 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1065 use these (crappy) functions in your code.
1066 These functions will be removed soon. Stop using them now.
1067 * Deprecate MSG_get_host_{table,number}
1068 Implement MSG_hosts_as_dynar() instead.
1069 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1070 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1073 * Stabilize the parallel execution mode of user contexts
1074 * Introduce configuration variables to control parallel execution:
1075 - contexts/synchro: Synchronization mode to use when running
1076 contexts in parallel (either futex, posix or busy_wait)
1077 - contexts/parallel_threshold: Minimal number of user contexts
1078 that must be part of a scheduling round to switch to parallel
1079 execution mode (raw contexts only)
1080 * Fix bugs that prevented to use suspend/resume along with
1081 synchronization structures.
1082 * Fix bugs in process termination that lead to invalid memory access
1083 in very specific conditions.
1086 * Introduce a parallel mode for the models (controlled by surf/nthreads
1087 configuration item). In our tests, running the models in parallel
1088 never lead to any speedups because they are so fast that the gain
1089 of computing each model in parallel does not amortizes the
1090 synchronization costs, even when ultra fast futexes are used.
1091 This is released anyway because YMMV.
1094 * Performance boost by using a swag internally to compute the set of
1095 tasks that are finished and should constitute the return value of
1099 * Enable it by default now that it is considered rather stable.
1102 * Documentation of the tracing functions.
1103 * Performance gains when tracing categorized/uncategorized resource
1104 utilization by avoiding calls to get route when updating resource
1105 variables. LMM constraints are being used instead.
1106 * API changed to set task categories. Use MSG_task_set_category instead
1107 of TRACE_msg_set_task_category, and SD_task_set_category instead
1108 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1109 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1110 * Examples for link user variables added (see at examples/msg/tracing/)
1111 * Deprecated function TRACE_msg_set_process_category completely removed
1112 * Trace header updated according to the latest Paje file format
1113 * Tracing network lazy updates, no longer obligate users to use full updates
1114 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1115 * Experimental: let user code declare/set/push/pop application states for hosts
1116 * API changed to allow the manual creation of graph configuration files
1117 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1120 * Improve the API of Lua MSG bindings, using the Lua spirit.
1121 * Each simulated process now lives in its own Lua world (globals are
1122 automatically duplicated). It helps writing simulators. Will allow
1123 to run Splay programs within SimGrid in the future.
1124 * Add a Chord example in Lua, equivalent to the MSG one.
1127 * Start the implementation of a solution to express temporal
1128 properties, not only local assertions. This is still an
1129 experimental work in progress, stay clear from it to be safe.
1133 - Add new runtime parameters --help-logs and --help-log-categories
1134 to display information about supported logging parameters and
1136 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1137 don't exists anymore.
1138 * Mallocators: allow value NULL for the reset function.
1140 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1141 create homogeneous dictionaries, where all the elements share the
1142 same free function. Non homogeneous dictionaries will be
1143 deprecated in the next release.
1144 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1145 - Multi-level dictionaries are deprecated.
1147 - new function xbt_dynar_search_or_negative() that is useful when
1148 you have less than 2 million elements in your dynar and don't
1149 want of the extra complexity of catching exceptions when the
1150 element is not found.
1152 - Make xbt_os_thread module (for thread portability) public.
1153 Documentation is still to come, sorry.
1155 - Cleanups and simplifications to make it maintainable again.
1156 - Exotic features (such as memalign and valloc) were removed.
1157 - The metadata were extended and improved so that the
1158 model-checker becomes able to explore and inspect the heaps.
1159 - This may induce a performance drop when enable_model-checking is
1160 ON in cmake (even if it's not used in the simulation), but it is
1161 necessary at this point to get MC working.
1163 Turn model-checking OFF if simulation performance matters to you.
1164 Not enabling it at runtime is not enough, disable it in cmake.
1166 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1169 SimGrid (3.6.2) stable; urgency=low
1171 The "Not coding new stuff allows to polish old things" release.
1174 * New bindings to the NS3 packet level simulator (experimental)
1175 * Use the raw (efficient) execution contexts instead of the sysv
1176 (portable) ones when possible.
1177 * libpcre is now mandatory in any cases since not using it led to
1178 severe performance loss and possibly other issues
1179 * Update the XML platforms:
1180 - G5K: include the latest machine in Nancy
1181 - GridPP and LCG: new platforms
1182 * Documentation was partially updated, at least (more to come)
1184 Bug fixes, cosmetics and small improvements
1185 * Free terminated processes before the end of the simulation to avoid
1186 exhausting the memory of users having very dynamic amount of
1188 * Bug fix and cosmetics about canceling non-running tasks
1189 * Bug fix about the dot loader's issues when using libcgraph
1192 * Create an installer for windows with nsis (amd64 and win32)
1193 - Add an hello world project to illustrate simgrid project creation.
1194 - Embed libpcre into the Simgrid installer to avoid
1195 its compilation burden
1196 * The raw execution contexts should work on Apple now
1197 * Port to Windows 64 bits
1198 - Sysv contexts now have an implementation for this arch
1199 - GRAS communication features now support this arch
1200 * Drop support for borland compiler on windows
1201 - this code was not maintained, and we kinda depend on gcc nowadays
1202 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1203 * Fix portability issue on unstable ubuntu: linker became picky on
1206 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1208 SimGrid (3.6.1) stable; urgency=low
1210 The "Oops, we broke Macs too" release
1213 * Fixed contexts detection so that raw ones are used when possible
1214 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1215 leads to a strange error, with user code segfaulting sometimes when
1216 the generated code is not perfectly aligned (which is not
1217 controllable from the user side, depends on the amount of code)
1220 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1222 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1224 SimGrid-java (3.6) unstable; urgency=low
1227 * Split of every thing from simgrid v3.5 into a separate package.
1229 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1231 SimGrid (3.6) stable; urgency=medium
1233 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1236 * Bindings now constitute their own package, separated from the main one.
1237 Rationale: reduce our maintenance nightmare by reducing the module coupling
1238 They will soon be released on their own on gforge.
1240 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1241 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1243 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1244 * It's not quite deprecated for now because we have no replacement,
1245 but it may soon become the case.
1248 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1249 * Fortran: New user-level cache variable to store the rank of the running
1250 process. This improves performance by an order of magnitude.
1251 * C: New coccinelle script to automatically locate and modify global and
1252 local static variables.
1253 * Improved SMPI network model with a sender-side gap to account for multiple
1257 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1258 only say if a communication is finished, no matter whether it succeeded or
1259 failed. You can call MSG_comm_get_status() to know the status of a finished
1261 * New function MSG_task_dsend() to send a task and detach it. When a
1262 communication is detached, you are never notified of its success or failure
1263 and the memory is released automatically once it is finished. This function
1264 is useful when you don't care about the end nor the success of a
1266 * Change the prototypes of action replay. Sorry for inconvenience,
1267 but this is really more efficient this way (and to adapt your code,
1268 you just have to fix the initialization, that shouldn't be too long)
1269 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1270 ever managed to do anything useful with it.
1271 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1272 takes a dynar of communications. It returns immediately and gives the
1273 index of a finished communication (if any).
1274 * New example: a basic implementation of the Chord P2P algorithm.
1277 * New model for multi-core CPUs. You can now use the core attribute to
1278 precise the number of cores of a host. This is a basic model. Every
1279 process running on the host receives at most the power provided in
1280 the DTD (throughput<=power). Total throughput of process cannot exceed
1282 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1283 router linked by an up-link and a down-link (possibly asymmetrical).
1284 This kind of pattern allows to easily build last-mile model style platforms.
1285 Aggregating such patterns in a rule-based AS is thus the technique of
1286 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1287 * New model for Vivaldi routing. We transformed the Vivaldi network model
1288 into a Vivaldi routing model (based on the rule-based model). This allows to
1289 combine Vivaldi based latencies with last-mile platforms.
1292 * Added a check for NaN of IEEE754 infinite in the double entries of
1294 * Introduce a new context factory "raw", highly inspirated from the
1295 ucontext factory, but using manually crafted functions in assembly to
1296 do the work in an efficient manner.
1297 * Allow to change the used context factory at run time, not only at
1298 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1299 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1300 contexts at runtime (only with raw contexts or ucontexts).
1301 * Completely rewrote this module to allow parallel execution of user
1302 processes. Use --cfg=contexts/nthreads:N to execute user processes
1303 with N parallel threads (the default is 1, meaning no parallelism).
1304 * Allow to decide dynamically between sequential and parallel modes.
1305 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1306 processes in parallel only when their number is greater than or equal to P
1308 * Added a check for NaN of IEEE754 infinite in the double entries of
1312 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1313 won't output the state of processes when interrupted with Ctrl-C
1314 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1315 NULL-terminated array. This may solve backward compatibility issues
1316 due to the change to return type of SD_simulate. See also:
1317 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1318 * Add new macros with variable number of arguments.
1319 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1320 - in xbt/asserts.h: xbt_assert
1321 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1322 - in xbt/ex.h: THROWF and RETHROWF.
1323 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1325 * Change xbt_die() to accept a format string with arguments, just like printf.
1326 * New data structure: xbt_lib_t, like a dict but more general and with better
1330 * New configuration options
1331 Options triva/categorized and triva/uncategorized can be used to generate
1332 graph configuration files for Triva visualization tool.
1333 * Configuration option tracing/platform is renamed to tracing/categorized
1334 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1335 * New TRACE_user_link_variable interface:
1336 User provides the name of the link and the tracing variable to attach to it
1337 * the declaration of tracing categories must be done after the environment
1339 * simpler tracing interface, just one way to declare categories
1340 TRACE_category or TRACE_category_with_color, it is up to you
1341 * links in the trace file are again identified by their names
1342 * trace contains the full platform hierarchy exactly as declared using the ASes
1343 * Options tracing/msg/[task|process]:1 groups the process by hosts
1344 for both cases, tasks and processes must have names that are unique during the simulation
1345 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1346 * The experimental option tracing/msg/volume is deprecated
1347 its functionality may be reincorporated if needed
1349 The tracing generates a trace file with unordered timestamped events,
1350 because of the way the core simulator (surf) works. A script available
1351 at the tools directory (fix-paje-trace.sh) can be used to put the events
1352 in order. We have changed the tracing so it can generate ordered timestamped
1353 events in the final trace, but depending on the simulator (and how much time
1354 is simulated) that can lead to a huge memory utilization. It is deactivated
1355 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1357 Build Infrastructure
1358 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1359 - We are trying hard to keep the API stable, but it may happen that
1360 some things change (we're a research project after all, not a
1361 nuclear plant operating system). If such things should happen, you
1362 could rely on that macro to adapt.
1363 - current value: 30600 for 3.06.00, aka 3.6
1364 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1365 a number that can be compared with SIMGRID_VERSION.
1366 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1367 assertions and verbose/debug logging events are disabled at compile time.
1369 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1371 SimGrid (3.5) stable; urgency=medium
1374 * New feature to any SimGrid-based simulator: Model-Checking
1375 Check SIN#1 for more details.
1378 * New Model SMPI (three-interval linear regression for correction factors)
1379 See RR-7426, available at http://hal.inria.fr/inria-00527150
1380 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1381 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1382 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1383 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1384 * See also src/smpi/README
1389 - Tracing API changes: TRACE_start and TRACE_end should not be called
1390 by user-code. They are automatically called by simulators created
1391 with SimDAG, MSG and SMPI if the toolkit is compiled with
1392 tracing_enabled=ON. Categories declaration and utilization remain the
1393 same for MSG and SimDag.
1394 - A function was added to the tracing API to declare categories with
1396 - TRACE_category_with_color (char *category, char *color)
1397 where color must be in the following format
1398 "%f %f %f", red, green, blue
1399 and red, green, blue are float values in the interval [0, 1]
1400 - User can specify NULL as color parameter, or continue calling
1401 TRACE_category (cat)
1402 On that case, the tracing system will define random colors
1403 - The following command-line options are supported:
1404 --cfg=tracing/filename:msg.trace
1405 --cfg=tracing:1 (activate tracing, needed to use others)
1406 --cfg=tracing/platform:1 (categorized resource use)
1407 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1408 --cfg=tracing/msg/task:1 (task creation)
1409 --cfg=tracing/msg/process:1 (process creation, migration)
1410 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1411 --cfg=tracing/smpi:1 (SMPI interface tracing)
1412 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1413 - examples of examples/msg/tracing updated
1415 - DAXLoader and DOTLoader functions can generate tasks with categories
1416 - A new function to attribute a category to SD tasks:
1417 TRACE_sd_set_task_category (SD_task_t task, char *category)
1418 * Tracing the MPI interface implemented by SMPI
1419 - Collective operations are traced with states
1420 - Point-to-Point operations are traced with states/links
1421 - Tracing activated by a parameter "-trace filename" passed
1422 to smpirun during execution (considering that simgrid
1423 is compiled with tracing enabled)
1424 - To run the simulation with gdb, the simulator
1425 accepts --cfg=tracing/smpi:1 to trace SMPI
1426 - tesh files to check if smpi tracing is ok
1427 - See examples/smpi/NAS/DT-trace
1428 * GTNetS tracing re-worked
1429 - adaptation to the tracing system of GTNets to cope
1430 with modifications regarding the fullduplex mode
1431 - new tesh files to check if gtnets tracing is ok
1434 * Asynchronous communications through the functions:
1435 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1436 * New function: MSG_load_platform_script()
1437 to make possible using a lua script instead of XML files to set up platforms
1438 * New function: MSG_set_function
1439 to associate functions to processes, used when bypassing the parser
1440 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1442 Platforms: Add some more examples in examples/platforms
1443 * Grid'5000: see www.grid5000.fr
1444 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1447 * Change the XML format. This is a very important modification. SimGrid
1448 3.5 introduces a new hierarchical format based on the notion of
1449 Autonomous Systems. Compatibility with old format is ensured
1450 through the perl script provided in the install bin directory
1451 bin/simgrid_update_xml.
1452 It is now possible to build platforms with specific routing
1453 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1454 connect several platforms together. We will try to provide soon
1455 set of realistic platforms exploiting these properties (have a
1456 look at examples/platforms/ for the moment).
1457 * Take the opportunity of the XML format change to be a good XML citizen:
1458 rename link:ctn to link_ctn and similar changes (also dealt with by
1460 * Add a new routing scheme (rule-based) using regular expressions. It
1461 enables to have an extremely low memory footprint when the
1462 underlying routing is simple and can be compactly described. You
1463 need to have libpcre4-dev (perl regular expressions) installed if
1464 you want to use this routing scheme.
1465 * Revive the cluster TAG and allow to easily and efficiently (both in
1466 term of memory and speed) connect clusters together. Have a look
1467 at teshsuite/simdag/platforms/ to see how this can be done. With
1468 this tag, you can create clusters with thousands of tasks at no
1469 cost (have a look at examples/platforms/).
1470 Note: clusters are implemented as ASes, so there is no need for an
1471 enclosing AS tag if you have only one cluster in your platform.
1472 * Add new generic functions in the public interface that allows the user
1473 to call SURF 'create_resource' methods from your code (same
1474 functionality as the XML bypass mechanism but with a much lighter
1476 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1477 that uses a piecewise linear approximation to produce better
1478 results when exchanging small messages.
1479 * Add a new parameter to handle correctly full duplex link and account
1480 for interferences between uplink and downlink communications
1481 (activate with --cfg=fullduplex:1).
1484 * Rename the SD_READY (all dependencies are satisfied and task is
1485 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1486 dependencies are satisfied) state.
1487 This prevents a confusion between the notion of "ready to schedule"
1488 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1489 simulated" (SD_RUNNABLE) used by the simulation kernel.
1490 * Change the way a task is considered as ready. Instead of removing
1491 dependencies when a task is done, a counter is decreased. This way,
1492 it is always possible to reach ancestors thanks to the
1493 SD_taks_get_parents function (even after the end of the simulation.)
1494 * Change the return type of SD_Simulate from (SD_task_t*) into
1495 xbt_dynar_t. This function was in handling a dynar internally and
1496 converted it into a NULL terminated array for historical reasons.
1497 * New function SD_dotload(char*) to load a DAG described in dot
1498 format. This loader and the corresponding examples require the
1499 installation of the graphviz library.
1500 * Fix a bug in the management of tasks of size 0 in the surf network
1501 models. This problem was only visible with SIMDAG and you should
1502 thus disregard results produced with earlier versions if you
1503 relied on this feature (some tasks were blocked because of this).
1504 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1505 even though your applications were doing only point-to-point
1506 communications and sequential computations. Now you can really use any
1507 model you want (of course, if you create real parallel tasks, which are
1508 not implemented in most models beside ptaskL07, this will abort).
1509 * Add an example that schedules a DAX on an heterogeneous platform
1510 using a Min-Min strategy.
1511 * New function SD_workstation_get_current_task() that returns the kind
1512 of task currently running on a workstation in the sequential access
1514 * Raise some warnings when unexecuted tasks remains at the end of the
1515 simulation. This is usually caused by cycles in the DAG.
1518 * New function: SIMIX_process_set_function() called by MSG_set_function
1519 * Change the underlying waiting queue in semaphores so that a process
1520 can wait on several of them simultaneously (as in waitany).
1521 * Fix the way to handle tokens in semaphores so that all access patterns
1522 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1523 * kill the dirty pimple SIMIX_message_sizes_output()
1524 Please use (proper) visualization instead
1527 * New data container: setset (set of sets of elements)
1528 * New module: mmalloc (mapped malloc, allowing to have several
1529 independent segments of malloc)
1530 * New function: xbt_dict_cursor_set_data()
1531 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1532 * New function: xbt_dynar_is_empty()
1533 * New function: xbt_fifo_get_last_item()
1534 * Fix xbt_dynar_shrink(): use the right element size.
1535 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1536 uninitialized areas during expand.
1537 * Fix semaphores: previous implementation was severely broken.
1538 * Use library init/fini functions for our initialization.
1539 - you can use logs and other feature as soon as you want in your
1540 code (even before the xbt_init / MSG_init)
1541 - xbt_exit is now a no-op and produce a warning when used.
1544 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1545 benefit from the latest improvement to the simulation kernel.
1546 * Kill measurement sockets for now. If you rely on them, sorry. This
1547 release is not for you. This feature will be reintroduced in the
1548 future, but we cannot delay the release any further.
1549 * New function: gras_msgtype_get_name().
1550 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1551 * Fix (at last) the pmm example: it should not randomly fail anymore.
1553 Build chain: bug fixes and overall polishing
1554 * Cmake is now stable enough. Hence, we killed the autotools.
1555 * Port to windows ( TM :)
1556 * Fix the 'make install' target.
1557 No need to use 'make install-simgrid' anymore
1558 * Introduce a 'make dist' target compiling a *source* archive
1559 'make package' compiles a binary archive
1560 * Compile java files only on need
1561 * Add --cd and --setenv command line options to tesh
1562 * Out of source builds are not fully supported yet, but we are close
1563 * Enable supernovae and optimization flags by default for our users
1566 * Add layer to set up environment directly from lua, without XML.
1567 * The effect of gras_stub_generator can be achieved through
1568 lua too (check examples/gras/console/ping_generator.lua)
1570 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1572 SimGrid (3.4.1) stable; urgency=low
1574 The "Polishing easter eggs is probably a good idea" release.
1575 This is a bug fixes release only.
1579 * Fix a bug preventing the tasks from begin garbage collected.
1582 * Fix a bug occuring when a host involved in a communication fails.
1583 This was not detected properly by the other peer involved in the
1584 communication. Now, it's reported as a network error.
1587 * Warn the user about loop dependencies in data flow of DAX files
1588 * Obey the control-flow dependencies of DAX files
1591 * Add option "enable_smpi" allowing to not compile SMPI.
1592 Probably useful for the (Mac) users experiencing a build error here
1593 * Improve the detection of lua5.1 and ruby1.8
1595 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1597 SimGrid (3.4) stable; urgency=low
1599 The "Easter in Cargese" release. Also known as (major changes):
1601 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1602 ~> bindings were greatly improved
1603 ~> new tracing infrastructure for better visualization introduced
1605 * the "Welcome to configury modernity" release.
1606 ~> we switched from autotools to cmake, and improved our cdash
1609 A more detailled list of changes follow (full detail in svn log).
1611 Java Bindings: Various Cleanups
1612 * (install java-gcj-compat-dev on debian-like to use them)
1613 * Remove put/get: no need to export deprecated interface in Java
1614 Use send/receive instead.
1615 * Cleanup the examples and add a README per directory
1616 * Remove example autoDestination (that's the only way to go now)
1617 * Remove example explicitDestination (was a plain copy of basic)
1618 * Make JniException a runtime exception, so that there is no need to
1619 declare the fact that you may encounter such a beast. I guess that
1620 nobody will ever want to survive such error.
1621 * Create specific errors for each MSG case of failure:
1622 host failure, transfer failure, timeout, task cancelled
1623 * Cleanup the exceptions that may get thrown by each function
1624 * Other internal cleanups in Java bindings. Performance still bad :/
1625 Ruby and Lua Bindings: create them
1626 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1627 * That's new and great, you should try them out.
1628 Same functionalities than Java bindings, only even less polished
1630 * Kill the useless "rate" argument of SD_task_get_execution_time()
1631 Everyone used to provide -1 as a value, it was not used, and the
1632 semantic of a possible use wasn't even clear.
1633 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1634 either as comm costs or compute costs to mean that there is no
1635 such thing for that specific task.
1636 * Add a SD_task_set_name() function
1637 * Fix SD_task_unschedule() on typed tasks
1638 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1639 * In DAX loader, accept useless 'level' attributes to <job> since
1640 LIGO DAGs have them (seem to be to ease graphical representation).
1642 * Add an example masterslave_mailbox.c using send/receive and not
1643 the deprecated put/get interface.
1644 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1645 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1647 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1648 (declare MSG_USE_DEPRECATED to still have the old name)
1649 * Add a MSG_task_set_data() function
1650 * About trace replay (see examples/msg/actions):
1652 - Allow to work with splitted trace files for each process
1653 Give the specific trace file as argument of each process,
1654 and call MSG_action_trace_run(NULL)
1655 You can still have one merged file for all processes.
1656 - Fix implementation of collective operations
1657 * Allow task_execute() on 0-sized tasks (closes #10063)
1659 * This is the first release of SimGrid where SMPI is not considered
1660 beta anymore (even if some corners should still be improved)
1661 * Port over the new SIMIX_network submodule (internal refactoring)
1662 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1663 * Implement more missing elements of the standard:
1665 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1667 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1668 MPI_Scan() MPI_Get_processor_name()
1669 - Added implementation of missing case for Alltoall (warning: it's
1670 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1671 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1672 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1673 SMPI_MPI_Allgatherv()
1674 * Bug fixes include:
1675 - MPI_Waitsome() was broken
1676 - Allow relative includes in smpicc
1677 - Command line cfg argument 'reference_speed' was ignored...
1678 - Some functions did not properly lead to auto-benching of user code
1679 - smpicc passes -O2 by default (just like openmpi one)
1681 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1682 * Bug fixes about timeouts during communications
1683 * add SIMIX_message_sizes_output() as a pimple to write to file the
1684 amount of messages per size. Use gnuplot to get histogram.
1685 Pimple because that's the only user-visible function of simix,
1686 defined directly in xbt.h (irk, sorry)
1688 - Add a SIMIX_sem_get_capacity() function
1689 - Fix interactions with processe resume/suspende
1690 - release_forever() was stupidly broken
1691 - Fix SIMIX_display_process_status() for processes in a semaphore
1692 - Make SIMIX_sem_block_onto() user-visible
1693 * Refactoring context stuff:
1694 - Use pseudo-OOP for better modularity
1695 - reimplement SIMIX_process_kill() without process_schedule() so
1696 that the latter can take as invariant that it is called from
1698 - Merge context_start into context_new for sake of simplicity
1700 * Add a Vivaldi network model, coded live during SUD'10 ;)
1701 * Rename configuration variables to start a hierarchy:
1702 o cpu_model -> cpu/model
1703 o network_model -> network/model
1704 o workstation_model -> workstation/model
1705 * New configuration variables:
1706 o network/bandwidth_factor: correction to bandwith
1707 o network/latency_factor: correction to latency
1708 o netwotk/weight_S: correction to the weight of competing streams
1709 * Add a long description to the models, that users can see with such
1710 argument on the command line: --cfg=cpu/model:help
1711 * --help-models display the long description of all known models
1713 * config: add the ability to set a default value after registration
1714 Does not override any previously set value (e.g. from cmd line)
1715 * dict: allow to have integer key and data.
1716 When so, you need to use the following functions
1717 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1718 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1719 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1720 In contrary to regular dicts, the key is not malloced before copy.
1721 Mixing scalar and regular elements in the same dict is not tested
1723 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1724 Tracing for Visualization:
1725 * SimGrid is now instrumented in order to generate a trace file for
1726 visualization analysis: to use it, need to compile SimGrid with the
1727 "tracing" option enabled, and instrument the program using SimGrid with
1728 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1729 (among other functions).
1730 * The instrumentation only traces the platform utilization for now
1731 * Documentation to use the tracing functions and how to analyze the
1732 traces with the Triva tool is written.
1733 * More information about: SimGrid FAQ (in the section Tracing Simulations
1736 * We moved to cmake as default build system. Autotools support will
1737 be dropped soon. Check the FAQ for more info about how to use it.
1738 * Greatly improved our cdash/ctest interactions
1739 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1740 * Added memory checking tests with valgrind; lot of memleak fixing.
1741 This may be the first release of simgrid with so few memory issues
1742 * Added code coverage tests.
1743 Our coverage is still improvable, but at least we see it on cdash.
1745 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1747 SimGrid (3.3.4) stable; urgency=low
1749 The "Desktop Grid needs love too" release (also called Xmas release).
1751 Models improvements:
1752 * Major speedup in the maxmin system solving by using lazy evaluation
1753 Instead of solving completely the maxmin system at each iteration,
1754 only invalidate (and recompute) the modified parts.
1755 This new feature is enabled in default models but you can try to
1756 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1757 * Cas01 IMproved as default CPU model
1758 This CPU model is the same Cas01 model, but it uses the
1759 maxmin-selective-update flag and a heap structure to manage
1760 actions on SURF kernel.
1761 It reduces the complexity to find the next action to finish and,
1762 consequently, it's faster than the old Cas01.
1763 This is the new default CPU model (Cas01).
1764 * Rename the old Cas01 model to Cas01_fullupdate
1765 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1766 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1767 * CpuTI (CPU Trace Integration)
1768 A new CPU model whose objective is simulate faster when using
1769 availability trace files.
1770 Instead of using a full featured, over engineered maxmin system for
1771 CPU modeling, this model does the pre-integration of traces files
1772 to calculate the amount of CPU power available, and so, executes
1773 faster than the old CPU models.
1774 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1775 * Use LV08 as default network model since it gives better accuracy
1776 for small messages and shouldn't change things for big ones.
1777 Use --cfg=network_model:CM02 to get the previous behavior.
1780 ******************************************
1781 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1782 ******************************************
1783 * The new CPU model may changes simulations!
1784 The point is that events occurring at the exact same timestamp
1785 are not scheduled in the same order with the old and new
1786 version. This may be enough to completely change the execution
1787 of simulations in some cases.
1788 * The new network model will change simulations!
1789 This new model is more realistic than the previous one, so you
1790 should consider redoing your old experiments with this model.
1791 Sorry for the inconvenience.
1794 * Introduce the supernovae compilation mode
1795 When compiled that way, the whole SimGrid (or almost) is put in a
1796 single compilation unit and compiled in one shoot.
1797 This is to help gcc which has difficulties to inline stuff from one
1799 The speedup seem to be above 15%, althrough more tests are needed on
1800 amd64 to confirm that gain.
1803 * Port of MSG's mailbox on top of SIMIX network
1804 The put/get mechanism was greatly simplified on the way.
1807 * New SIMIX network module. Provides:
1808 - Mailbox: rendez-vous mechanism to find with who you want to speak
1809 - Synchronous send/recv: easier and hopefully faster since the
1810 logic is handled in the maestro process directly now
1811 - Asynchronous send/recv: you dreamt of it? It's here now
1812 Too bad that nobody cared enough to propagate the change to MSG.
1813 * Add semaphores as SIMIX synchronization mechanism.
1816 * new function SD_daxload(char*) to load a DAX file
1817 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1818 * Introduce typed tasks. Specify its kind and cost at creation.
1819 At scheduling, just give where it should be placed, and the cost
1820 for each involved resource is automatically computed.
1821 Existing constructors so far (more to come of course):
1822 - SD_task_create_comm_e2e() for end-to-end communication
1823 - SD_task_create_comp_seq() for sequential computation
1824 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1825 * new function SD_task_dump() for debuging display
1826 * new function SD_task_dotty(task,FILE*) writing to file the info
1827 about the task in dotty format
1828 * SD_task_dependency_exists() can now cope with having one of its
1829 arguments NULL. If so, it tests whether the other argument has any
1831 * Add getters on list of preceding/following tasks:
1832 SD_task_get_parents(task) and SD_task_get_children(task)
1833 * Add getters on amount of workstations and list:
1834 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1835 * Add getter on task kind: SD_task_get_kind(task)
1836 * Update the start_time and finish_time of tasks on completion/failure
1837 * Bugfix: Remove task from state swags when destroyed
1840 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1843 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1844 dynar with one separate thread per value of the dynar.
1845 * Change the prototype of xbt_thread_create(), sorry.
1846 Added a boolean parameter indicating whether we want to join this
1847 thread (used in SG only for now)
1848 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1851 * GTNetS wrappers should now be usable again (and betterly tested too)
1852 * Fix a major regression from 3.2 where the timeout provided to
1853 MSG_task_put_with_timeout() was used as absolute time before which
1854 the comm should be done.
1855 * Start to fix the <cluster> tag.
1856 - Internal links should be good now (beside of the loopback, which
1857 use the private link instead)
1858 - paths to the external world is still rather broken
1859 - the <route:multi> tag is just broken. Actually that's brain-dead.
1860 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1861 to make it less stupid
1862 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1863 * Fix a source-level compatibility glitch from 3.2: after defining
1864 MSG_USE_DEPRECATED, you can use the old name
1865 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1866 * Allow to compile from the SVN with automake 1.11
1867 * Fix some problems when using the "start_time" tag in deployment XMLs.
1868 * Fix #8569: XBT/synchro.h has redundant declarations
1869 * Fix #8563: MSG return values and exceptions
1870 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1871 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1872 Thanks to Robson Peixoto.
1873 * Fix a vicious bug in dictionaries inducing that some elements were
1874 not freed on xbt_dict_free()
1876 Portability report of this version:
1877 * Main portability targets:
1878 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1879 on (amd64/i386/ia64)
1880 - mac leopard on i386
1881 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1882 but nothing critical.
1883 * Other platforms: windows, AIX and others were not tested for this release
1885 Timing report of this version:
1886 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1887 scenario parameters). From 8h to a few seconds in desktop grid settings.
1888 * Supernovae brings about 25% speedup on i386.
1890 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1892 SimGrid (3.3.3) stable; urgency=low
1894 The "Need for Speed" release.
1896 The timings done to validate the 3.3.2 were faulty.
1897 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1899 The problem was a conversion from a manually handled vector to
1900 xbt_dynar_t on the critical path.
1901 xbt_dynar_foreach calls functions, inducing stack management crap.
1903 We inlined these functions and xbt_dynar_foreach is now breath taking.
1904 We also inlined xbt_swag_belong on the way.
1906 Here are some approximate speedup measurements (on master/slaves
1907 simulations lasting between 10s and 20s each):
1908 3.3.1 -> 3.3.2: about same performance
1909 3.3.2 -> 3.3.3: 40% speedup
1910 3.3.1 -> 3.3.3: 40% speedup
1911 3.3.1 with inline patch -> 3.3.3: 30% speedup
1913 Our reading is that the refactoring which occurred in 3.3.2 made us
1914 suffer much more from the xbt_dynar_foreach low performance, but
1915 once we solved this, this refactoring proved to be very performance
1916 effective. From the 40% speedup, somehow, 10% are due to the
1917 inlining and 30% to the refactoring.
1919 That's a pitty that gcc cannot inline functions placed in other files
1920 alone. We have to choose between:
1921 - break the encapsulation (by putting private data structures and
1922 accessors in headers files to help gcc)
1923 - live with low performance
1924 - switch to a decent compiler such as icc (not quite possible).
1926 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1928 SimGrid (3.3.2) stable; urgency=low
1930 The "Simplicity does not preceed complexity, but follows it" release.
1932 The main contributors of this release were (lexical order):
1933 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1936 * Extract the routing logic into its own object.
1937 (was dupplicated in network.c and workstation_LV07.c;
1938 Allows to implement other ways of storing that info)
1939 => kill now useless network_card concept
1940 - Use dynar to represent routes (instead of void** + int*)
1941 - kill link_set (use surf_network_model->resource_set instead)
1942 - Add a command-line option to choose the routing schema to use
1943 - Add three new models:
1944 * Floyd (shortest path computed at initialization)
1945 * Dijikstra (shortest path recomputed all the time)
1946 * Cached Dijikstra (shortest path computed on need)
1947 All these models where contributed by Silas De Munck, and are
1948 described in his ICCS09 paper.
1950 * Simplify model declaration
1951 (less redirections, less function to write when defining a model)
1952 - Factorize stuff between models:
1955 surf_model_resource_set(model)
1956 surf_model_resource_by_name(model, name)
1957 - Unify the types of models in s_surf_model_t (using an union)
1958 - Embeed fields of common_public directly into s_surf_model_t
1959 - Rename model methods:
1960 action_free ~> action_unref
1961 action_change_state ~> action_state_set
1962 action_get_state ~> action_state_get
1963 - Change model methods into functions :
1964 (model)->common_public->action_use ~> surf_action_ref
1966 * Implement a generic resource; use it as ancestor to specific ones
1967 (allows to kill duplicated code in models)
1968 Drawback: timer command don't need no name nor properties;
1969 workstation_CLM03 don't need no properties
1970 (but I guess we can live with those few bytes wasted)
1972 * Improve the action object model
1973 - implement a constructor avoiding dupplicated code about field
1974 initialization in generic_action part.
1976 * Kill the SDP model: it has an external dependency, is deprecated
1977 in flavor of modern lmm models, and didn't compile since a while
1980 * Relocation of the context module from XBT to SIMIX.
1981 (the context were decoupled from the simix processes, duplicating a lot of code)
1982 => a lot of code was factorized
1983 - less overhead is introduced during scheduling
1984 - simpler API for the context factory
1985 - the logic for process creation,destruction and manipulation was simplified
1986 * Simplification of the s_smx_process_t data structure.
1987 => accesing the simix level data associated to a process is faster now,
1988 and the code is a lot more readable.
1991 * Implement some more MPI primitives:
1992 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1993 -implementation: Bcast: flat or 2-ary tree (default),
1994 Barrier: 4-ary tree,
1996 Allreduce: Reduce then Bcast
1997 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1998 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1999 Alltoallv: flat tree, like ompi
2001 * Add support for optimized collectives (Bcast is now binomial by default)
2002 * Port smpirun and smpicc to OS X
2005 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2006 Shout out if you used it.
2009 * Display the list of still queued messages in SG mode when existing
2013 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2014 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2015 * Add xbt_str_from_file(FILE*)
2016 * Add xbt_dict_get_key achieving a linear reverse search
2017 * Remove the context module
2019 Portability report of this version:
2020 * Main portability targets:
2021 - Linux(debian)/x86/context
2022 - Linux(debian)/x86/pthreads
2023 - Linux(debian)/amd64/context
2024 - Linux(debian)/amd64/pthreads
2025 On these, we still have the eratic breakages of gras/pmm and
2026 amok/saturate_sg reported in previous version. We still think
2027 that the tests are the cause of the fault, not the tested code.
2029 - Mac OSX Leopard/x86/context
2030 Still false negative in tesh autotesting.
2031 Smpi still fails, but this time because readlink does not accept -f
2032 Everything seems to work properly beside of that.
2035 - AIX version 5.3 (only tested contexts this time)
2036 Smpi still fails there because mktemp is not installed.
2037 Everything seems to work properly beside of that.
2039 I managed to compile it for the first time, but several breakages.
2040 Won't delay the release for this exotic platform.
2042 * Windows: it's still lagging behind. If you want to help, please
2045 Timing report of this version:
2046 This version seem to be more than 5% faster than 3.3.1 (on linux
2047 64bits with contextes). The gain is less than expected, we are
2048 investigating this for next release.
2050 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2052 SimGrid (3.3.1) stable; urgency=low
2055 * Implement a --cfg-help to show existing configuration variables
2056 * Build chain do not require doxygen in maintainer mode
2059 * fix a bug on struct sizeof computation, which prevented the
2060 exchange of arrays of structs in some conditions
2061 - added a regression test about this in datadesc_usage
2062 * Allow the exchange of 0-long dynamic vectors.
2063 - for that, use -1 as indicator of dynamic size instead of 0
2064 - This implied to change any size from unsigned long to long,
2065 reducing a bit communication abilities, but I guess that with
2066 64bits being quite common, this is more than enough.
2067 - This also induce a protocol change, thus bumping network protocol
2068 version from 0 to 1 (if we have external users, we have to get
2069 clean on that point too ;)
2070 - added two regression tests about this in datadesc_usage
2071 * Be more verbose when propagating local exceptions
2072 This helps debugging.
2073 * Display the status of simulated processes when receiving SIGINT in
2077 * Allow to control the simulation from a trace file.
2078 New functions MSG_action_register() and MSG_action_trace_run()
2079 The first one allows to associate a function execution to each
2080 kind of action while the second one parses a trace file and
2081 triggers the corresponding actions within the system.
2082 For now, only a toy example is provided in examples/msg/actions
2083 * Add an exemple of process migration in examples/msg/migration
2084 * Fix a bug in task exchange which broke MSG_task_get_sender()
2085 Add a teshsuite regression test for that.
2086 [Bug: if MSG_task_get_sender() is called after sender exit,
2088 * Fix a bug which prevented suspend/resume to work properly
2089 * Display the status of simulated processes when receiving SIGINT
2090 This fixes a regression of v3.3. due to the introduction of SIMIX
2091 * Bug fixing in failure management:
2092 - trace could not start by a failure at time 0
2093 - failure during communications were not working
2096 * Add SIMIX_process_set_name() to change the name of the current
2097 process in the log messages.
2098 * Store smx_hosts in a dict since we only retrieve them by name
2099 * Move the configuration infrastructure to surf
2102 * Move the configuration infrastructure to surf
2105 * Massive internal cleanups:
2106 - Store internal structures on processes instead of hosts (allows
2107 to have more than one process per host, in addition of being more
2109 - Cleanup the initialization/finalization process
2110 - Kill a whole bunch of unneeded synchronization:
2111 processes run in exclusive manner within the simulator
2112 - Move queues from global tables to process data fields
2114 - now accept -platform and -hostfile arguments
2115 - Pass the right rank value to processes according to the hostfile
2116 * Compile the examples by default, and use them as regression tests
2117 * Implement MPI_Wtime()
2118 * Change the reference speed to a command line option
2121 * TCP_gamma can now be specified as command line option using
2122 --cfg=TCP_gamma:10000000.0
2123 * Change the --surf-path cmd line option into --cfg=path:
2126 * Also include strbuff from xbt.h public header
2127 * xbt_ex_display(): do not free the exception after displaying
2128 This allows to do more with the given exception afterward.
2129 Users should call xbt_ex_free() themselves.
2133 Portability report of this version:
2134 * Main portability targets:
2135 - Linux(debian)/x86/context
2136 - Linux(debian)/x86/pthreads
2137 - Linux(debian)/amd64/context
2138 - Linux(debian)/amd64/pthreads
2139 These targets fail about 1/10 of times on gras/pmm, but we believe
2140 that this is because of the test, not because of simgrid.
2141 amok/saturate_sg fails even more rarely, and the test may not be
2144 - Mac OSX Leopard/x86/context
2145 The test suite still spits tons of errors because some obscure
2146 force prevents us from removing the temporary directories
2147 arguing that they still contain some metadata I've never heard of.
2148 Smpi fails because seq is not installed.
2149 Everything seems to work properly beside of that.
2152 - AIX version 5.3 (both contexts and pthread)
2153 Smpi still fails there because mktemp is not installed.
2154 XML inclusions seems rosty on AIX.
2156 * Windows: it's still lagging behind. If you want to help, please
2159 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2161 SimGrid (3.3) stable; urgency=high
2165 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2166 [Malek Cherier & Mt]
2168 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2171 Until now, GRAS were using MSG as an interface to SURF. It was
2172 quite difficult because both interface have several differences
2173 (MSG channels vs GRAS sockets were the most notable point).
2175 This also opens the gate to SMPI (which should occur soon) and speed
2176 up simulations by to 40% (even if it were not the main goal).
2178 **************************************
2179 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2180 **************************************
2181 The point is that events occuring at the exact same timestamp are
2182 not scheduled in the same order with the old and new version. This
2183 may be enough to completely change the execution of simulations in
2184 some cases. Sorry for the inconvenience.
2186 * Cleanup and upgrade the XML format to push further scalability
2187 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2189 * Improve the testing infrastructure with tesh. Now a very large part of
2190 the code is tested not only by being run but also by checking that the
2191 output match an expected output [Mt].
2193 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2194 is really less memory-demanding, which should allow you to use
2195 larger files in SimGrid [AL].
2197 * Inform valgrind about our contextes, so that it becomes usable
2198 with the default (and more effecient) version of SimGrid
2199 [contributed by Sékou Diakite, many thanks]
2202 * Introduce a listener thread in charge of receiving incoming
2203 messages from the network. It allows to overlap communication and
2204 computation but most notably, it removes some stupid deadlocks due
2205 to the fact that so far, a process could not send and receive at
2206 the same time. This made most non trivial communication schema
2208 * Convert the PIDs from long int to int to match the MSG ones (and
2209 linux ones too) [Mt]
2210 * New function: gras_agent_spawn() to launch a new process on
2211 current host. Only working in simulation for now. [Mt]
2212 * New function: gras_os_hostport() returning a constant form (ie,
2213 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2216 * Make the backtrace of exceptions more human readable [Mt]
2217 * New module: xbt/str [Mt]
2218 a ton of string utility functions (split, join, printf to a newly
2219 allocated buffer, trim, etc)
2220 * New module: xbt/hash [Mt]
2221 SHA1 hashing algorithm (more to come if needed)
2222 * New module: xbt/synchro [Mt]
2223 synchronization tools (mutex and conditions) working the same way
2224 in simulation and in real life (mainly useful for GRAS, but not
2226 * New module: xbt/queue [Mt]
2227 classical producer/consumer synchronization scheme
2228 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2229 (using the classical functions will get serialized) [Mt]
2230 * Make dictionary internal table dynamic. No need to specify its size
2231 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2233 * Make sure the log channels are organized as a tree under windows
2234 (because of ANSI C compatibility issue, any channel were child of
2235 root directly) [Mt].
2238 * Cleaned many thing in surf and fixed a few bugs [AL].
2239 * Add a nice command line configuration mechanism to compose models [AL].
2240 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2241 the previous one (KCCFLN05). It relies on something that looks like
2242 a max-min sharing mechanism but cannot be written as such. A new solver
2243 was thus designed [AL].
2244 * Add a new solver to lmm. Based on Lagrange optimization and
2245 gradient-based descent, it enables to efficiently maximise systems s.a
2247 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2249 This solver enables to propose two new network models for TCP Reno and
2250 TCP Vegas based on Low's work. These models still need to be fully
2251 tested though [Pedro Velho].
2254 * Bug fix in SD_simulate. Now the time bound given as argument is
2256 * Use the new parallel task model (ptask_L07) as default.
2257 * Use the SURF command line configuration mechanism.
2258 * 0-size tasks (for synchronization) should now work.
2260 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2262 SimGrid (3.2) stable; urgency=high
2266 We still experience issues on this platform, but we believe that at
2267 least MSG is usable.
2269 GRAS API BREAKAGE (for simplification purpose, sorry):
2270 * the gras_msgtype_by_name is not used anymore. Instead of
2271 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2272 you can write (and must)
2273 gras_msg_send(toserver, "request", &request);
2274 - If you still want to pass a gras_msgtype_t to the function (to cache
2275 the type and avoid the lookup time), use the gras_msg_send_() variant.
2276 - Impacted functions:
2277 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2278 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2279 * The callbacks are now expected to return 0 when everything went well
2280 (just like the main() function)
2282 GRAS new features and improvements:
2283 * New module mechanism where user code can use per process globals [Mt]
2284 This is similar to gras_userdata_*() functions, but for libraries. It
2285 factorize some code developped over and over in the examples and AMOK.
2286 It has still to be documented and used (only amok/peermanagement is
2288 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2289 loss when they were small enough to fit into the buffer and sent quickly
2290 enough so that they can all get received in one shoot.
2291 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2292 if not found. Use the *_or_null() variant for the old semantic.
2293 * In gras_msg_handle, do not discard messages without callback.
2294 They are probably messages to be explicitly awaited later (ie, proofs of
2295 mis-synchronization in userland since they are sent before being awaited)
2296 No big deal usually.
2297 * gras_socket_meas_send/recv: semantic changed!
2298 The numerical arguments used to be (1) the total amount of data to send
2299 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2300 messages. This was need for the fool willing to send more than MAXINT
2301 bytes on quite fat pipes.
2304 * Do really rename the hostmanagement module to peermanagement. [Mt]
2305 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2306 but this is rather new and this was documented in the module
2307 documentation (poor excuses, I admit)
2308 * Bandwidth measurement semantic changed! This follows the changes to
2309 gras_socket_meas_send/recv explained above.
2312 * A sequential mode has been added to the workstations. When a workstation
2313 is in sequential mode, it can execute only one task, and the other tasks
2314 are waiting in a FIFO. [Christophe Thiery]
2317 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2318 model for SIMDAG. [Christophe Thiery]
2319 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2320 equal to 0, which caused some bad behaviors in
2321 saturated_constraint_set_update. I now use a threshold mechanism like in
2325 * When running manually src/testall, you select specific units [Mt]
2326 testall is the result of our cunit mechanism, and should replace all
2327 the scripty thingy around since bash don't run easily on billware.
2329 * A mallocator system has been added. [Christophe Thiery]
2330 Mallocators allow you to recycle your unused objects instead of freeing them
2331 and allocating new ones.
2333 Documentation update:
2334 * FAQ reworking + New FAQs:
2335 - "Valgrind spits tons of errors!" [Mt]
2336 - "How to repport bugs" [Mt]
2337 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2338 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2339 - Communication time measurement within MSG [AL]
2340 - I experience weird communication times when I change the latency [AL]
2341 * GRAS tutorial [Mt]
2343 - an introduction to the framework and to the used communication model
2344 - an initiatic tour introducing the most proheminent features:
2346 . Lesson 0: Installing GRAS
2347 . Lesson 1: Setting up your own project
2348 o Part 2: Message passing
2349 . Lesson 2: Exchanging simple messages
2350 . Lesson 3: Passing arguments to the processes (in SG)
2351 . Lesson 4: Attaching callbacks to messages
2352 . Lesson 5: Using globals in processes
2353 . Lesson 6: Logging information properly
2354 . Lesson 7: Using internal timers
2355 . Lesson 8: Handling errors through exceptions
2356 . Lesson 9: Exchanging simple data
2357 . Lesson 10: Remote Procedure Calling (RPC)
2358 . Lesson 11: Explicitely waiting for messages
2359 . Recapping of message passing features in GRAS
2360 - A HOWTO section containing:
2361 o HOWTO design a GRAS application
2362 More are due, of course. They will come latter. In the meanwhile, you can
2363 check the examples which are still here.
2365 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2367 SimGrid (3.1) stable; urgency=high
2371 There was a stack corruption somewhere, visible only when optimizing
2372 with these versions. [Vince]
2375 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2376 module that enabled to play with Directed Acyclic Graphs. It is built
2377 directly on top of SURF and provides an API rather close to the old
2378 SG. Some old codes using SG are currently under rewrite to check that
2379 all needful functions are provided. [Christophe Thiery]
2382 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2383 extension of the classical CLM03 model that gracefully handles
2384 failures. This is now the default model for MSG and GRAS. It doesn't
2385 handle parallel tasks yet though. [AL]
2386 * Bug fix: Weights were not correctly set in the network part.
2387 WARNING: This may have resulted in incorrect results with simulations
2388 where there are more than one flow on a given link. [AL]
2391 * After a (long ?) discussion on simgrid-devel, we have decided that the
2392 convention we had on units was stupid. That is why it has been decided
2393 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2394 WARNING : This means that all previous platform files will not work as
2395 such with this version! A warning is issued to ask users to update
2397 A conversion script can be found in the contrib module of the CVS, under
2398 the name contrib/platform_generation/surfxml_update.pl [MQ]
2401 * Bug fix: Processes were started in reverse order, wrt deployment file.
2402 WARNING: if your code relies on this bug, please fix it. [AL]
2403 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2404 being executed on two different locations. [AL]
2405 * Bug fix: Failures are now better supported thanks to Derrick's tests
2406 (there was many failure situations I hadn't thought of and that weren't
2407 correctly handled). [AL]
2408 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2412 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2413 Exception occurring on server-side are propagated back to client (!).
2415 API CHANGE: the callback changed their prototype. Change:
2416 int my_handler(gras_socket_t expeditor, void *payload_data) {
2418 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2419 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2421 * New! function: gras_msg_handleall to deal with all messages arriving
2422 within a given period.
2423 * New! function: gras_socket_server_range to get a server socket in a
2424 range of port numbers (ease to avoid port number conflicts) [MQ]
2425 * New! gras processes display their backtrace when they get a SIGUSR1
2426 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2427 Sweet to debug RL processes [MQ]
2431 - Do not force experiment sizes to be expressed in kb, or it becomes
2432 impossible to measure the latency this way (needs one byte-long tests)
2433 WARNING: this changes the amok_bw_* function semantic. [MQ]
2434 - Implements the link saturation stuff. [MQ]
2435 * Peer management module:
2436 New! module factorizing code that we wrote over and over [MQ].
2439 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2440 - Test units are placed directly into the library code, they get extracted
2441 automatically and placed into the src/testall binary.
2442 - Convert most of the XBT tests to this system.
2443 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2444 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2445 live in the same C code. This required to change a little bit the API
2446 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2447 * New module: sparse graph structure with basic algorithms (this is work
2448 in progress and the API is not considered to be frozen yet). [AL]
2449 * Display more information on backtraces: source line & function names are
2450 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2451 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2452 * Reworked a little bit some #include statements to load only required
2453 headers. Some user code that relied on SimGrid to include stdlib or
2454 stdio may need to include it by themselves. [AL]
2455 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2457 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2458 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2459 but I guess that almost nobody used those parts. [MQ]
2461 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2463 SimGrid (3.0.1) stable; urgency=low
2466 * Unfortunately, I had missed 5 misnamed functions:
2467 xbt_fifo_item_t xbt_fifo_newitem(void);
2468 void xbt_fifo_freeitem(xbt_fifo_item_t);
2469 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2470 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2471 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2472 They're now deprecated. Please use their new versions:
2473 xbt_fifo_item_t xbt_fifo_new_item(void);
2474 void xbt_fifo_free_item(xbt_fifo_item_t);
2475 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2476 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2477 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2479 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2480 * Documentation: xbt_log module unmercifully reworked [MQ]
2481 * Bugfix: there was a problem with the ending of contexts with
2482 the pthread backend. It caused some weird deadlock or behavior
2483 depending on the pthread implementation. [AL]
2484 * Bugfix: get the exceptions raised in the simulator repport where
2485 and why they come from when they are not catched in time [AL, MQ]
2488 * Bugfix: Do repport the error when two non-connected hosts try to
2489 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2492 * Add additionnal checkings on communications. Assert that two
2493 communicating hosts are connected by a set of links... [AL]
2496 * Add additionnal checkings on channel values in communication [AL]
2497 * New: MSG_task_get_source to see on which host a task was generated [HC]
2498 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2499 the number of tasks waiting to be received on channel and sent
2501 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2502 waits for the first task coming from a given host.. [AL]
2504 GRAS new functionnalities: [MQ]
2505 * Enhance the parsing macro to allow the size of multidimentional objects
2506 to be given thru annotations.
2507 * New example (and documentation): Matrix Multiplication a la RPC
2508 (as when I was young!) and fix a bunch of bugs found on the way.
2510 GRAS performance improvements: [MQ]
2512 * Reduce the amount of cbps creation/destruction by making it static to
2513 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2516 * Change libdata to a set so that we can search for stuff by ID (and thus
2517 reduce the insane amount of dict lookups)
2520 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2521 RL communication on top of SG since it's so uneffective.
2522 It may also be used for inter-thread communication in RL, one day.
2523 * Use gras_datadesc_copy() to exchange messages on top of SG
2525 - improve message exchange performance on top of SG
2526 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2527 * Don't exchange on the network the size of the used part of buffer,
2528 instead, specify the possible buffer size to read().
2530 - reduces the amount of read/write calls (one pair per exchange)
2531 - reduces the amount of exchanged data (the size)
2532 - allows to retrieve all arrived data on receiver side, if we don't need
2533 it right now (subsequent read will peek the buffer)
2534 - allows the receiver to proceed with the begining of the stream before
2535 everything is arrived
2536 - make it possible to build an iov transport (using readv/writev)
2538 - take care of the data with non-stable storage (like stacked data),
2540 * If possible, TCP send uses vector I/O (when writev() is here)
2541 - Don't use it for receive since we send data sizes and data on the
2542 same stream, so we wouldn't be able to chain large amount of chunks
2543 before having to flush the stuff to read the size.
2544 * Rework the transport plugin mechanism to simplify it and reduce the
2545 amount of pointer dereferencement when searching for the right function
2548 * I guess that now, we do almost as few system calls as possible while
2549 doing as few data copy as possible.
2551 To improve it further, we could try to send all the sizes first and then
2552 all the data (to use iov on receiving size), but it's only a partial
2553 solution: when you have 2 dimensional data, the sizes of the second
2554 dimension is data of the first dimension, so you need 3 streams.
2556 I'm not sure the potential performance gains justify the coding burden.
2558 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2560 SimGrid (3.00) stable; urgency=high
2563 * New! Give the possibility to hijack the surf parser and thus bypass
2564 MSG_create_environment and MSG_launch_application. Have a look at
2565 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2567 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2569 SimGrid (2.96) unstable; urgency=low
2574 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2575 This deprecates the xbt_error_t mechanisms.
2576 It modifies (simplifies) all XBT and GRAS API.
2577 MSG API keeps unchanged (exceptions raised by XBT are catched from
2578 within MSG and masked with existing error handling facilities)
2581 * New! Add a FATPIPE model. [AL]
2582 * New! Add a parallel task model. [AL]
2583 * New! Add automatically a loopback interface (in the default
2584 network model) if none was precised.
2587 * Bugfix: MSG_process_resume now works with the current running process.
2589 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2590 * Modification of MSG_task_get_compute_duration. Once a task has been
2591 processed, the value returned by this function is now equal to 0. [AL]
2592 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2593 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2594 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2595 of a m_task. For now, MSG_task_cancel only works with computation
2597 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2598 of the processor (in Mflop/s) regardless of the current load on the
2600 * API Change: use proper naming convention for MSG_getClock and
2601 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2603 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2604 This function changes the priority of a computation task. This priority
2605 doesn't affect the transfer rate. A priority of 2 will make a task
2606 receive two times more cpu power than the other ones. This function
2607 has been added to suit the needs of Nguyen The Loc and hasn't been that
2608 much tested yet. So if it fails, please report it and send me your code.
2610 * API Change: removed all functions and types that were marked "deprecated"
2611 since many months. Renamed MSG_global_init_args to MSG_global_init.
2613 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2615 SimGrid (2.95) unstable; urgency=low
2618 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2619 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2620 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2623 * Bugfix: Allow absolute paths for platform description files [MQ]
2624 * Bugfix: do free the variables after use. Leads to drastic performance
2626 * Implement max_duration (ie, timeouts) on resources [AL]
2629 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2631 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2632 the needed support (timeouts) [AL]
2635 * Implement measurement sockets. You can now get the bandwidth between two
2636 hosts thanks to AMOK (see below). [MQ]
2637 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2638 dynar over the network (yeah) [MQ]
2639 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2640 * simple/static token-ring example. [Alexandre Colucci and MQ]
2641 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2642 in place before [MQ]
2645 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2646 type parsing. "short" and "long" modifiers were also ignored; other
2647 modifier (such as reference level) are still ignored. [MQ]
2648 * Embeed the buffer size within the buffer itself on SG. [MQ]
2649 That way, send() are atomic and cannot get intermixed anymore (at least
2650 the ones which are less than 100k; bigger messages still have the issue)
2651 * Array size pushed by the field, not by the field type (or each
2652 and every long int will push stuff to the cbps) [MQ]
2653 * use select() to sleep since it allows to portably sleep less than one
2656 GRAS (minor cleanups)
2657 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2660 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2661 * type category 'ignored' killed as it was never used and were difficult
2663 * whether a type can cycle or not is now a flag, leaving room for more
2664 flags (as "ignored", if we feel the need one day ;)
2665 * Rename raw sockets to measurement sockets since "raw" has another
2666 meaning in networking community.
2669 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2670 and offers features not belonging to GRAS but that most applications
2671 need. One day, it may be a set of plugins loadable at runtime.
2672 * New module: bandwidth
2673 bandwidth measurement between arbitrary nodes running this module. [MQ]
2675 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2677 SimGrid (2.94) unstable; urgency=low
2679 The first beta release of SimGrid 3 !
2683 * Update the main page and the FAQ. Adding references to gforge.
2686 * Add a gras_os_getpid function.
2689 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2690 * Extend the logs so that they also print PID, hostname, date, ... if
2692 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2693 and kill the old version which were in testsuite/
2694 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2698 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2699 for users. Instead, each programming environment (GRAS, MSG,...) use it
2700 when needed to provide such a feature to users.
2701 Don't shortcut the mechanism or you will also shortcut the virtualization
2702 you need on the simulator.
2706 * Cleanups in configury with regard to compile optimization/warning flags.
2707 Also add -fno-loop-optimize to any powerpc since it's the optimization
2708 killing gcc (< 3.4.0).
2709 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2710 needed by MSG examples complications
2711 * Borrow configury beautifications from PHP
2714 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2715 warning (thanks loris for stumbling into this one).
2716 * Bugfix: stop loading private headers (gras_config.h) from the public
2720 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2721 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2722 RL would imply a DNS resolver, which is *hard* to do in a portable way
2723 (and therefore delayed).
2724 * Implement a real timer mechanism and use it in timing macros. This allows
2725 to avoid rounding errors and get a 0.000005 sec precision in timing
2726 macros. While I was at it, various cleanups:
2727 - allow to declare more than one timed section per file (fix a stupid bug)
2728 - move some private declaration to the right place
2729 - merge conditional execution and timing macros into emulation module
2730 - document the module
2731 - make sure the module cleanups its mess on gras_exit
2732 * Documentation improvements:
2733 - (new) how to compile applications using GRAS
2734 - (new) emulation support (timing macros)
2736 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2738 SimGrid (2.93) unstable; urgency=low
2740 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2743 - Use Paje properly where used. Still to be sanitized properly.
2744 - Portability fix: Add an implementation of the contexts using pthread
2748 - Add xbt_procname(): returns the name of the current process.
2749 Use it to show the current process's name in all logging.
2751 - fix detection of older flex version and the reaction, since we do
2752 depend on modern ones (we use lex_destroy)
2753 - Better separation of SG and RL in the libs: remove all simulation code
2754 from libgras. As a result, this lib is now only 200k when stripped.
2755 Some of the xbt modules may also be duplicated (two sets and such) and
2756 should be cleaned/killed before SG3.
2757 - Insist on using xlC on AIX because of weird problems involving gcc there.
2758 - Cleanup the make remote stuff. This is now done by scripts
2759 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2760 mainly for our private use, but we're working on changing them to user
2763 - Bugfix: flush the socket on close only if there is some *output*.
2764 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2766 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2769 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2771 SimGrid (2.92) unstable; urgency=low
2773 Alpha 3 on the path to SimGrid 3
2777 - New! First try of benchmarking macros.
2778 - New! First try so that gras_stub_generator generate deployment and
2779 remote compilation helpers.
2781 - Bugfix: Initialization fix in msg_test.
2785 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2787 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2789 - API Change: gras_os_sleep to take the amount of seconds as a double.
2790 Accepting an int was error prone since it was the only location where
2791 seconds were coded as such. It leaded to damn rounding errors.
2792 - Bugfix: Hard to belive that timers ever worked before this.
2794 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2796 SimGrid (2.91) unstable; urgency=low
2798 Alpha 2 on the path to SimGrid 3
2802 - Bug fix in the lmm_solver.
2804 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2805 through the function MSG_paje_output.
2806 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2807 - It is possible to bound the rate of a communication in MSG with
2808 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2809 to write it in the changelog).
2810 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2815 - Various cleanups to the autotools stuff
2816 - Begin to move Gras examples to examples/gras/
2817 - Let make distcheck work again (yeah!)
2819 - documentation overhauled using doxygen.
2820 gtk-doc-tools is dead in SimGrid now.
2821 - Automatically extract all existing logging categories, and add the list
2822 to the documentation (long standing one, to say the less)
2824 - Cleanup the known architecture table. Reorder the entries to group what
2825 should be, and use a more consistent naming scheme.
2826 (some of the test dataset are still to be regenerated)
2827 - New! Allow library to register globals on each process just as userdata
2829 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2830 lookup time (for now).
2831 Use it in msg and trp.
2832 This cleans a lot the internals and helps enforcing privacy of the
2833 headers between the gras components.
2834 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2835 - Bugfix: gras_os_time was delirious in RL.
2836 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2838 - Reenable GRAS now that it works.
2840 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2842 SimGrid (2.90) unstable; urgency=low
2844 Alpha 1 on the path to SimGrid 3
2846 * It is a long time since the last release of SimGrid. I'm sorry about
2847 that but as I had told you, I was rewriting a lot of things. I apologize
2848 to those who had been reporting bugs to me and that I had not answered.
2849 If your bug is still in the new version, please tell me. Here is a
2850 summary of the main changes.
2852 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2853 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2854 portability, speed, and a lot more but you'll figure it out later.
2855 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2856 I wanted to release the new MSG as soon as possible and I have
2857 broken GRAS, which is the reason why, for now, only MSG is fully
2858 functional. A laconic description of these projects is available
2859 in the documentation.
2861 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2862 kernel optimized for our needs (called SURF but only the developers
2863 should use it). Hence, MSG is now roughly 30 times faster and I think
2864 that by rewriting a little bit MSG, I could event speed it up a little
2865 bit more. Beside the gain in speed, it is also much easier to encode a
2866 new platform model with SURF than it was with SG. More to come...
2868 * REVOLUTION 3: I have tried to change a little as possible the API of
2869 MSG but a few things really had to disappear. The main differences
2870 with the previous version are :
2871 1) no more m_links_t and the corresponding functions. Platforms are
2872 directly read from a XML description and cannot be hard-coded
2873 anymore. The same format is used for application deployment
2874 description. The new format is described in the documentation.
2875 Have a look in tools/platform_generation. There is a tiny script
2876 that converts from the old platform format to the new one. Concerning
2877 the application deployment format, parsing the old one is tricky.
2878 I think most of you should however be able to convert your files. If
2879 it is really an issue, I can write a C code that does the conversion.
2881 2) the toolbox tbx does not exist anymore. We now have a library
2882 with much more data-structures but without the hash-tables (we have
2883 dictionaries that are much faster).
2885 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2887 *****************************************************************************
2888 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2889 * don't feel like converting the oldies. (Mt) *
2890 *****************************************************************************
2893 Version 2.90: "the long awaited one"
2894 - Finished rewriting and debugging MSG. Rewrote the documentation.
2895 - disable GRAS for now since it needs to be ported to the newest SG
2898 - Finish the port to windows (using mingw32 for cross-compile)
2901 - Main loop and datastructures of SURF. A cpu resource object is
2902 functional. Surf can thus be used to create cpu's with variable
2903 performance on which you can execute some actions.
2905 2004-11-15 Martin Quinson
2906 - Port to ARM. Simply added the alignment and size descriptions. Should
2907 work, but the ARM machines are so slow that I didn't had the opportunity
2908 to 'make check' over there yet.
2910 2004-11-15 Arnaud Legrand
2911 - Trace manager now written. It uses a heap structure and is therefore
2912 expected to be efficient. It may however be speeded up (particularly
2913 when many events occur at the same date) by using red and black
2914 trees. One day maybe...
2915 - Max-min linear system solver written. It uses a sparse matrix
2916 structure taking advantage of its expected use. Most operations are
2917 O(1) and free/calloc are called as few as possible. The computation of
2918 the minimum could however be improved by using a red and black tree
2921 2004-11-03 Arnaud Legrand
2922 - Rename every gras_* function that was in xbt/ to its xbt_
2924 - Add a heap and a doubly-linked list to xbt
2925 - Added a dichotomy to the dictionaries. make check works as well before
2926 so I assume that the patch is correct. I do not know however if things
2927 run effectively faster than before now. :)
2929 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2930 SimGrid, and the version number is bumped to 2.x
2932 2004-10-29 Martin Quinson
2933 - Introduction of the remote errors.
2934 They are the result of a RMI/RPC on the remote machine.
2935 ErrCodes being scalar values, you can't get the host on which those
2936 errors did happen. Extending the error mechanism as in Gnome is possible.
2937 No idea yet whether it is a good idea.
2939 2004-10-28 Martin Quinson
2940 - Interface revolution: the Starred Structure Eradication.
2941 I used to do typedef struct {} toto_t; and then handle *toto_t.
2942 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2943 - ???_t is a valid type (builded with typedef)
2944 - s_toto_t is a structure (access to fields with .)
2945 - s_toto is a structure needing 'struct' keyword to be used
2946 - e_toto_t is an enum
2947 - toto_t is an 'object' (struct*)
2949 typedef struct s_toto {} s_toto_t, *toto_t;
2950 typedef enum {} e_toto_t;
2951 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2952 s_toto_t) is private.
2954 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2955 so that it can set the variable to NULL. It was so for dicts and sets,
2956 it changed for dynars.
2958 - Fix a bunch of memleaks in dict_remove
2959 - Fix a bug in sg/server_socket opening: it failed all the time.
2961 2004-10-07 Martin Quinson
2962 - Speed up dynar lookup operation a bit.
2964 gras_dynar_get is dead.
2966 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2967 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2968 (which gives you the address of the stored data).
2970 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2971 copy of the data using an affectation to do the job and not a memcpy.
2973 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2974 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2976 It does not really speedup the dynar test because they are
2977 setting elements all the time (and look them seldom). But the dict does
2978 far more lookup than setting.
2980 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2982 2004-10-05 Martin Quinson
2983 - Allow to (en/dis)able the cycle detection at run time.
2985 Whether we should check for cycle or not is now a property of each
2986 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2987 datadesc_cycle_unset allow to remove this property when previously set.
2989 Note that the cycle detection is off by default since it impacts the
2990 performance. Watch the data you feed GRAS with ;)
2992 This property is hereditary. Any element embedded in a structure having it
2993 set have it set for the time of this data exchange.
2995 You should set it both on sender and receiver side. If you don't set it on
2996 sender side, it will enter an endless loop. If you forget on receiver
2997 side, the cycles won't be recreated after communication.
2999 - Header reorganization.
3000 Kill gras_private.h, each submodule must load the headers it needs.
3002 2004-10-04 Martin Quinson
3003 - Interface revolution: do not try to survive to malloc failure.
3005 Now, gras_malloc and friends call gras_abort() on failure.
3006 As a conclusion, malloc_error is not a valid error anymore, and all
3007 functions for which it was the only gras_error_t return value are
3008 changed. They now return void, or there result directly.
3009 This simplify the API a lot.
3011 2004-09-29 Martin Quinson
3012 - Re-enable raw sockets.
3013 Created by gras_socket_{client,server}_ext;
3014 Used with gras_raw_{send,recv}
3017 It should allow to kill the last bits of gras first version soon.
3019 This is not completely satisfactory yet (duplicate code with
3020 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3023 - Simplify transport plugin (internal) interface by not passing any
3024 argument to _server and _client, but embedding them in the socket
3027 2004-09-28 Martin Quinson
3028 - Finish the port to AIX.
3029 autoconf was my problem (segfault within the malloc replacement
3030 function. No idea why)
3032 2004-09-16 Martin Quinson
3033 - Fix some size_t madness on 64bit architectures.
3035 2004-09-08 Martin Quinson
3036 - Reduce the number of system headers loaded, overload some more system
3037 calls (such as malloc to cast the result of the system one, and work
3039 - Fix and reintroduce the config support
3041 2004-09-07 Martin Quinson
3042 - Source code reorganization to allow Arnaud to surf all over there.
3043 - Allow to document the logging categories.
3044 - Remove all uppercase from logging categories and useless cleanup in names.
3046 2004-08-18 Martin Quinson
3047 Version 0.6.2 (protocol not changed; API changed)
3048 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3049 gras_error_t), and NULL when not found. Functions expecting a msgtype
3050 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3051 hopefully usefull message.
3052 - Portability to prehistoric sparcs again
3054 2004-08-17 Martin Quinson
3055 Version 0.6.1 (protocol not changed; ABI not changed)
3056 - prealloc some buffers to speed things up
3058 2004-08-11 Martin Quinson
3059 Version 0.6 (protocol not changed; ABI expended)
3060 - The parsing macro can deal with the references, provided that you add
3061 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3063 2004-08-09 Martin Quinson
3064 Version 0.5 (protocol not changed; ABI changed)
3065 - Allow to off turn the cycle detection code in data exchange at
3066 compilation time. It should be at run time, but I'm short of time (and
3067 the config stuff is still broken). That way, we keep dict out of the
3068 critical path, which is good because the performance is poor:
3069 - search not dichotomial yet
3070 - dynar give no way to access their content and memcpy everytime
3071 - In composed data description (struct, ref and so on), stop foolness of
3072 keeping the subtype's ID, but store the type itself. This keeps sets out
3073 of the critical path, which is good since they rely on dynar and
3074 dictionnaries. The only loose of that is that we cannot detect the
3075 redeclaration of a structure/union with another content (but I'm not sure
3076 the code detected well this error before anyway). We still can detect
3077 the redefinition discrepancy for the other types.
3078 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3079 it breaks the code because of type-punning used all over the place).
3080 This breaks on all non-gcc architectures (for now).
3082 All those changes (plus the buffer of last time) allow me to gain 2 order
3083 of magnitude on cruel tests consisting of 800000 array of integers on two
3084 level of a hierarchical structure (200 secondes -> 4 secondes)
3087 - the selector of reference must now return the type it points to, not
3088 the ID of this type.
3090 2004-08-06 Martin Quinson
3091 Version 0.4 (protocol changed; ABI not changed)
3092 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3093 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3094 - Implement a buffer transport, and use it by default (it relies on tcp in
3095 real life and on sg in simulation).
3096 That's a bit hackish since I had a new field to the structure to store
3097 its data without interfering with the subtype ones. Inheritance
3098 is tricky in C. And that's a kind of reverse inheritance with one class
3099 derivating two classes. Or maybe a game with java interfaces. Anyway,
3100 that's damn hard in C (at least).
3101 Moreover, I got tired while trying to ensure plugin separation and
3102 genericity in SG mode. MSG wants me to do weird things, so let's go for
3103 cruel hacks (temporarily of course ;).
3104 See comment in transport_private.h:71
3105 - do not include all the _interface headers in private but in the files
3106 which really need them (to cut the compilation time when they are
3109 2004-07-26 Martin Quinson
3110 Version 0.3 (protocol not changed; ABI changed)
3111 - Major overhault of the datadesc interface to simplify it:
3112 - shorted the function names:
3113 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3114 - add a trivial way to push/pop integers into the cbps without malloc.
3115 This allows to make really generic sub_type description, which simply
3116 pop their size of the stack.
3117 - add a function gras_datadesc_ref_pop_arr() which does what users want
3118 most of the time: Declare a dynamic array (which pops its size of the
3119 stack) and declare a reference to it. Poor name, but anyway.
3120 - kill the post-send callback, add a post-receive one
3122 2004-07-23 Martin Quinson
3123 Version 0.2 (protocol changed; ABI changed)
3124 - add some testing for cpbs in the test cases, and fix some more bugs.
3125 This invalidate again the little64 data file, since I cannot regenerate
3127 - remove an awfull optimization in the logging stuff, allowing me to:
3128 - understand it again
3129 - learn gcc how to check that the argument match the provided format
3130 - fix all errors revealed by gcc after that
3131 - internal keys of dict are not \0 terminated. Deal with it properly in
3132 loggings instead of segfaulting when the user want to see the logs :-/
3134 2004-07-22 Martin Quinson
3135 - Fix some stupid bug preventing cbps (callback postit) from working
3137 2004-07-21 Martin Quinson
3138 - Some documentation cleanups
3139 - remove the useless last argument of msgtype_declare
3140 - rename the Virtu functions to fit into the 'os' namespace
3141 - move headers src/include -> src/include/gras/ and stop fooling with
3142 gras -> . symbolic link
3143 - make distcheck is now successful
3145 2004-07-19 Martin Quinson
3147 - Build shared library also
3148 - Install html doc to the right location
3149 - stop removing maintainer files in make clean
3150 - build tests only on make check
3152 2004-07-13 Martin Quinson
3154 - No major issue in previous version => change versioning schema
3155 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3156 corresponding dataset.
3158 2004-07-11 Martin Quinson
3160 - Get it working with any kind of structure (we can compute the padding
3161 bytes remotely for all the architectures I have access to)
3162 - Implement the structure parsing macro (still not quite robust/complete)
3163 - Improvement to the remote testing toysuite
3165 2004-07-10 Martin Quinson
3166 [autoconf mechanism]
3167 - get ride of a bunch of deprecated macros
3168 - actually run the test for two-compliment, not only compile it :-/
3169 - test whether the structures get packed (and bail out if yes. Damn.
3170 Alignment is a serious matter)
3171 - test whether the structures get compacted (but respecting the alignment
3172 constraints of each types)
3173 - test whether the array fields of structures can straddle alignment boundaries
3175 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3176 (took me 2 hours to find that bug, looking at the wrong place)
3177 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3178 before sending/receiving and are used to compute the offsets of fields
3179 - Given that padding size depend even on compiler options, keep track of
3180 alignment and aligned_size only for the current architecture. Not a big
3181 deal since we send structure fields one after the other (seems
3183 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3184 program, let it work on linux/gcc/little32. portability todo.
3186 2004-07-08 Martin Quinson
3187 - import and improve remote compilation support from FAST
3188 - make sure make check works on half a dozen of machines out there
3190 2004-07-07 Martin Quinson
3191 Let's say it's version 0.0.3 ;)
3192 - Implement conversions (yuhu!)
3193 - Let it work on solaris (beside conversion, of course)
3194 - Stupid me, using rand() to generate the conversion datatests in not wise.
3196 2004-07-06 Martin Quinson
3197 - Let make dist work, since I'm gonna need it to compile on remote hosts
3198 - Let Tests/datadesc_usage write the architecture on which the file was
3199 generated as first byte.
3200 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3201 architecture descriptions.
3202 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3203 on those architectures.
3204 - Optimization: send/recv array of scalar in one shoot
3206 2004-07-05 Martin Quinson
3207 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3209 - Plug a whole bunch of memleaks
3210 - each process now have to call gras_{init,exit}. One day, their log
3211 settings will be separated
3212 - Continue the code factorisation between SG, RL and common in Transport.
3214 2004-07-04 Martin Quinson
3216 - Redistribution between SG and RL.
3217 We wanna have to accept in SG, so move accepted related parts of RL in
3218 the common part. (more precisely, the dynar of all known sockets is no
3219 more a static in transport.c, but part of the process_data)
3221 [gras_stub_generator]
3222 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3224 2004-07-03 Martin Quinson
3225 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3226 them all up in one shot)
3228 - Fix the ugly case of reference to dynamic array.
3229 - New (semi-public) function gras_datadesc_size to allow the messaging
3230 layer to malloc the needed space for the buffer.
3232 - gras_socket_close now expect the socket to close (and not its address to
3233 put NULL in it after it). This is because the socket passed to handlers
3234 is one of their argument (=> not writable).
3236 - propagate the interface cleanup from last week in datadesc, ie remove a
3237 superfluous level of indirection. User pass adress of variable
3238 containing data (both when sending and receiving), and not of a variable
3239 being a pointer to the data. Let's say that I like it better ;)
3240 The price for that is constructs like "int msg=*(int*)payload" in
3241 handlers, but it's a fine price, IMHO.
3243 - Let it work in RL (yuhu)
3245 2004-06-21 Martin Quinson
3247 - porting SG plugin and SG select to new standards (works almost).
3248 - plug memleaks and fix bugs around.
3251 - cleanup the prototype of data recv and force users to specify when they
3252 want to handle references to objects. Test case working even for cycles.
3253 - plug memleaks. Valgrind is perfectly ok with this.
3255 2004-06-12 Martin Quinson
3257 - cleanup the separation between plugin and main code in plugin creation
3259 2004-06-11 Martin Quinson
3261 - Reput hook for raw sockets, needed for BW experiments
3262 - kill a few lines of dead code
3263 [Data description] Interface cleanup
3264 - gras_datadesc_by_name returns the searched type or NULL.
3265 That way, no variable is needed to use a type desc once, which makes
3267 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3268 two parameters were strings (field name, type name), leading to
3270 [Dicos] Interface cleanup
3271 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3272 This is consistant with the dynar API.
3274 2004-04-21 Martin Quinson
3276 - Porting to new standards.
3278 - interface cleanup.
3279 There is no bag anymore, no need to take extra provision to mask the
3280 pointers behind "ID".
3281 Better splitup of functions between files create/exchange/convert.
3282 This is still a bit artificial since convert and receive are so
3283 interleaved, but anyway.
3285 - add a queued message list to procdata (the ones not matching criteria
3287 - factorize some more code between SG and RL wrt procdata
3289 - use gras_exit in example to track memleaks
3290 - get rid of gs_example now that GS is properly integrated into gras
3291 - update run_test to integrate the latest tests (datadesc)
3293 - rename WARNINGn macros to WARNn since it prooved error-prone
3295 2004-04-19 Martin Quinson
3297 - register init/exit functions within gras module mechanism
3298 - send/receive function.
3299 Convertion is not implemented, but short-cutted if not needed.
3300 struct/array elements are sent one by one (instead of block-wise), but
3301 nobody really cares (yet). Get a prototype before optimizing.
3302 - tests (using a file socket) for DD send/receive on:
3303 - base types: int, float
3304 - array: fixed size, string (ie ref to dynamic string)
3305 - structure: homogeneous, heterogeneous
3306 - chained list, graph with cycle
3307 Believe it or not, valgrind is not too unhappy with the results. The
3308 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3312 - Bugfix when using a filename explicitely (instead of '-')
3314 2004-04-09 Martin Quinson
3316 - factorize more code between RL and SG in socket creation
3317 - Complete the implementation and tests of:
3319 o file (only in RL, and mainly for debugging)
3321 I lost 3 days to design a portable address resolver, and then decided
3322 that the prototype mainly have to run on my box.
3323 Addressing portability too early may be like optimizing too early :-/
3325 - use gras_init in the Tests instead of the crappy parse_log_opt
3326 (the latter function is removed)
3327 [Conditional execution]
3328 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3329 [Code reorganisation]
3330 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3331 Build examples against the RL library, since there is no way to disable
3332 its creation for now.
3334 For information, the beginning of coding on GRAS was back in june
3335 2003. I guess that every line has been rewritten at least twice since