1 SimGrid (3.17) UNRELEASED (release target: September 22 2017)
3 SimGrid (3.16) Released June 22. 2017.
5 The Blooming Spring Release: developments are budding.
8 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
9 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
10 - Java JAR file should be finally fully working on Mac OSX too.
12 XML platforms: Switch to platform v4.1 format.
13 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
14 - <zone> can be used as a synonym for the now deprecated <as>
15 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
16 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
17 - <actor> can be used as a synonym for the now deprecated <process>
18 - state_file and avail_file periodicity is now easier to express
19 (check the documentation) the old behavior should still work.
20 * Storage: not backward compatible, but it seems that nobody ever used it
21 - Remove attribute 'content_type' of <storage_type>: was never used
22 - Make attribute 'model' of <storage_type> optional: for future usage
23 - Remove Bconnection model property: was never (in)validated. Replaced by
24 the maximum of the read and write bandwidth as a resource constraint.
27 - New and Backwards Compatibility break:
28 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
29 When one wants to get the list of tasks whose states have changed during a
30 simulation round, s/he has to allocate and free a dynar and use it as
31 argument to this function. The former SD_simulate (double how_long)
35 - Allow multicore VMs, with the correct sharing (unless you overcommit)
36 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
39 - New function to update the consumption of all hosts at once.
40 - Fix the model for multi-core hosts, linear in the amount of busy
41 cores with an abnormality for pIdle (see paper).
42 BUG: Still not satisfactory for ptask on mono-cores.
45 - The netzone are now available from the MSG API.
46 The old names still work, but are now deprecated.
49 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
50 - New option: smpi/keep-temps to not cleanup temp files
51 - New option : smpi/shared-malloc-blocksize . Relevant only when global shared
52 mallocs mode is used, allows to change the size of the fake file used
53 (default 1MB), to potentially limit the number of mappings for large runs.
54 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
55 - Fortran ifort and flang compilers support
56 - New RMA calls supported (experimental) :
57 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
58 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
59 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
60 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
61 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
62 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
65 - New: this_actor::isMaestro() returns whether we are in kernel mode.
66 - Behind the scene, ActivityImpl are now automatically refcounted.
67 Removing this nasty bottleneck will greatly simplify our internals.
70 - Replay: New function xbt_replay_action_get():
71 Retrieve the function previously associated to an event type.
72 - DROPPED FUNCTION: xbt_str_varsubst()
73 - DROPPED MODULE: strbuff. We don't need it anymore.
74 - DROPPED MODULE: matrix. We don't need it anymore.
75 - DROPPED MODULE: lib. We don't need it anymore.
77 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
79 SimGrid (3.15) Released March 22. 2017
81 The Spring Release: continuous integration servers become green
82 We fixed even the transient bugs on all target architectures:
83 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
84 Windows, FreeBSD, NetBSD.
86 - Rename NetCards to NetPoints.
87 This was intended to help NS3 users, but that's not a netcard.
88 That's a point in the routing algorithm, let's avoid wrong simplifications.
91 - New: MSG_process_yield(). Stop and yield to other processes.
92 - New: MSG_process_daemon(). Daemon processes are automatically killed
93 when the last non-daemon process terminates
94 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
95 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
96 to make room for the upcoming network energy plugin.
97 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
100 - Ensure that an actor can kill itself with Process::exit()
101 - Kill the obscure NativeException. Nobody want to survive the issues
102 it denotes, so use JniException that is a RuntimeException (not to
103 be caught explicitly).
104 - Partial bug fix in initialization. SimGrid flags on command line were
105 consumed at C level but stayed in the original Java String[] args.
106 This could mess users' args[i] if SG flags were not put at the end of
108 The SimGrid flags are now removed from the Java arguments. However,
109 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
110 if args.length is greater than the number of YOUR OWN ARGUMENTS.
111 It might be if you have --log or --cfg flags in the command line.
112 - Fix numerous memleaks all around the place. In particular, around VMs.
116 - s4u::onPlatformCreated: right before the simulation starts
117 - s4u::onSimulationEnd: right after the main simulation loop
118 - s4u::onTimeAdvance: right after a clock change (time is discrete)
119 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
120 event from the availability_file changes the avail speed.
121 - Links are now usable from s4u
122 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
123 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
124 - New examples: The conversion of MSG examples to S4U has begun
125 - Actors: create, kill, migration, and suspend
126 - Applications: master-worker and token-ring
127 - Action replay: communications and storage
128 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
131 - Backwards Compatibility breaks
132 - The SD_route_get_list and SD_route_get_size functions have been removed.
133 They are replaced by the unique
134 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
135 The route from src to dst is built in the links dynar whose size can
136 be obtained with xbt_dynar_length.
137 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
138 sg_host_route_bandwidth and sg_host_route_latency.
139 Macros ensure the backwards compatibility, but you should fix your code
142 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
143 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
144 - Bump our claim of support from MPI 1.1 to MPI 2.2.
145 We don't support 100% of it, but it should be enough. Contact us if not.
146 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
147 naive and probably inaccurate)
148 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
149 mapped onto a unique small file using some system magic.
150 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
151 - Fortran: cleanups, fixes, support of user-added operations
152 - RMA: MPI_Accumulate are applied in correct order.
153 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
154 - Better support for MPI_IN_PLACE
155 - Support for MPI_Win attrs and keyvals.
156 - Support MPI_Comm_group_create, MPI_Type_size_x
157 - MPI ops should be properly applied to all allowed types
158 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
159 algorithm by a non failing one from mpich.
162 - Kill the fifo data container: we don't use it anymore.
165 - A <cluster> can now be created with different speed values to represent pstates.
166 see examples/platforms/energy_cluster.xml for an example.
168 SimGrid (3.14.159) Released December 28. 2016
170 The previous release was only a crude approximation of a Pi release;
171 this one should be a bit better.
173 - Revert a last minute change that broke on Mac OSX.
174 - Fix the changelog and NEWS files.
175 - Improve the documentation, in particular of the routing module.
176 - Integrate some patches flying around in Debian.
178 SimGrid (3.14) Released December 24. 2016
181 * The whole documentation was reworked and reorganized. There is
182 still much room for improvement here, but we're on it.
185 * We now need python3 for our testing infrastructure.
186 * Model checking needs libevent
189 * Now works on FreeBSD
192 * Privatization now works on FreeBSD
194 * Privatization is now activated by default in smpirun
195 You can turn it off with -no-privatize if something goes wrong.
197 * Call-location tracing for SMPI.
198 You can add the exact location (filename / linenumber) of an MPI call to
199 your trace files and slow down or speed up the simulation between two
200 consecutive calls by using an adjustment file (see the documentation).
202 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
203 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
204 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
205 values were ignored in some cases. The timings of these functions can now
206 be significantly different.
208 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
209 smpi/running-power is renamed to smpi/host-speed
211 * smpi/grow-injected-times option to enable or disable multiplication of the
212 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
213 make simulation less precise (but also much faster).
215 * smpirun script should be (much) faster for large deployments.
217 * SMPI tracing : fixed issue with poor matching of send/receives.
219 * Replay : Fix broken waitall
221 New functions and features
222 * MSG_parallel_task_execute_with_timeout, to timeout computations.
224 Dropped / renamed functions and features
225 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
226 - MSG_mailbox_is_empty() -> Mailbox::empty()
227 - MSG_mailbox_front() -> Mailbox::front()
228 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
229 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
230 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
231 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
232 * Don't pass the free_f parameter to property related functions:
233 - MSG_host_set_property_value()
234 - MSG_as_router_set_property_value()
235 - MSG_storage_set_property_value()
236 * VM properties. Since msg_vm_t are msg_host_t, just use
237 MSG_host_get_property_value() and friends
238 * VM I/O related things:
239 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
240 - Unimplemented save/restore methods
241 * MSG_as_router_get_property_value() was redundent with
242 MSG_environment_as_get_property_value().
243 - Removed MSG_as_router_*propert*() functions
244 - Added MSG_environment_as_set_property_value() for consistency
245 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
246 Well, they are still there for now, but deprecated with a warning.
247 Please switch to xbt_dict_new_homogeneous() before this is removed
249 * Task affinity. Its intended behavior (that was very badly tested
250 and probably not really working) was deceiving what most users
251 would have hoped here.
252 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
253 * Option network/coordinates is now useless and should be dropped.
256 * Added option storage/max_file_descriptors to allow more than 1024 files opened
259 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
260 * Added Dragonfly topology support
263 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
264 EiBps, EBps, Eibps, Ebps for bandwidth.
265 They may become useful to some lucky ones.
268 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
269 * ex/app_tokenring: new example, very similar to the MSG Token Ring
270 * ex/async_waitAll: new example, on asynchronous communications
273 * Memory usage should be decreased for simulations with a large number
274 of processes. This also helps for SMPI.
276 SimGrid (3.13) stable; urgency=low
278 The Half Release, a.k.a. the Zealous Easter Trim.
280 - We removed half of the lines, that were mostly experimental cruft.
281 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
282 (not counting blanks and comments -- according to openhub.net)
283 - The internals are now compiled in C++ (and will soon be clean C++)
284 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
285 59 klines of C, etc. We added only 29 klines of C++ in replacement.
287 * Backwards Compatibility breaks
288 - Removed Lua simulation bindings (switch to C or Java for that).
289 Lua can still be used to describe platforms
290 - Removed Java kernel plug-ins.
291 Will be reintroduced after the ongoing major internals reorg.
293 - the following functions were removed.
294 They were too specific and should be reimplemented in a generic
295 way, with filter function.
296 - MSG_task_listen_from_host
297 - MSG_mailbox_get_count_host_waiting_tasks
298 - MSG_mailbox_put_with_timeout was removed.
299 Please use MSG_task_send_with_timeout instead.
301 - the SD_application_reinit function was removed. It has been a noop for a while.
302 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
303 replaced by a more flexible mechanism.
304 - The following functions thus do not exist anymore
305 - SD_workstation_get_access_mode
306 - SD_workstation_set_access_mode
307 - SD_workstation_get_current_task
308 - Basic estimation functions have been removed but can easily be replaced
309 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
310 - SD_workstation_get_computation_time => amount / sg_host_speed()
312 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
313 Use VM.getSpeed()*load/100 for the legacy behavior.
315 - option enable_tracing was removed. It was not doing anything for a while.
316 - In the ModelChecker:
317 - the model-checker now ptraces the model-checked process which means
318 you cannot use a debugger on the latter anymore (we might make this
319 optional in the feature);
320 - removed soft-dirty page tracking;
321 - remove model-checked side snapshot management,
322 MC_snapshot() and MC_compare_snapshot();
323 - keep the MC_cut() function as a stub (it was not really working
324 in the previous release).
327 * All options are consistently in kebab-case. Old names are kept as alias.
330 * Switch to platform v4 format.
331 - Rename from 'power' to 'speed' the attributes describing the amount of
332 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
333 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
334 - In <host> and <link>, attributes availability and state are gone.
335 It was redundent with state and availability traces, and with peak values.
336 - In <cluster>, attributes availability_file and state_file are gone.
337 It was too complex and unused.
338 - Kill <gpu>. Was not doing anything.
339 - The DOCTYPE points to the right URL:
340 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
341 (the file at this address now documents the changelog since its v1)
342 - A warning is emitted for unit-less values (they are still accepted).
343 - speed. Default: 'f' or 'flops'. Also defined:
344 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
345 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
346 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
347 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
348 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
349 - latency. Default: 's' second. Also defined:
350 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
352 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
353 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
356 * s4u::Host is now the preferred public interface to the Host features.
357 sg_host_* functions are C bindings to the exact same behavior
358 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
361 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
362 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
363 They were deprecated since a few years, and probably did not even compile anymore.
366 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
367 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
368 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
369 encouraged to update their codes. A list of the modified functions can be found at the end of
370 include/simgrid/simdag.h
373 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
374 the lambda or closure passed as a parameter will run in kernel mode.
375 All the callback functions should be rewritten to that interface at some point.
378 * Reorganizing and cleaning the internals all around the place.
381 * Remove old default barrier/bcast buggy algorithms (see #18407)
382 * Various bug fixes to handle more codes
383 * Remove the need for the --foreground option of smpirun (it is still
384 accepted for backward compatibility).
387 * Kill the setset data container: MC don't use it anymore.
388 * Kill the queue data container: it made more sense with GRAS.
389 * Kill the xbt_peer_t data type: it's useless without GRAS.
390 * Kill rm_cb feature of config sets: it was never useful.
391 * Kill graphxml parsing feature. It was not used.
392 * Kill the deprecated code protected by XBT_USE_DEPRECATED
394 - xbt_dynar_sort_strings(), when the content is char*
395 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
396 They throw exceptions on invalid input;
397 * C++ support for declaring CLI flags (simgrid::config::Flag);
398 * class for abstracting different signal backends (simgrid::xbt::signal).
399 with no external dependencies (we need very simple signals).
402 * refactoring and cleanup of the code;
403 * ongoing process to cleanly separate the model-checking algorithms
404 from the code model-checking support.
406 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
408 SimGrid (3.12) stable; urgency=low
410 The Facelift Release.
413 * Require g++ v4.7 at least to not speak prehistorical C++.
414 * Require Boost 1.48 (for signal2 component).
415 * Java must be version 7 at least when activated.
416 * Builds on Windows again (including Java bindings).
417 * Tracing is now always enabled (no way to turn it out)
418 * Remove GTNetS. It was not working anyway.
419 * Various cleanups in the cmake scripts.
420 * Move headers around to sort them out on installed systems:
421 - instr/instr.h -> simgrid/instr.h
422 - instr/jedule/* -> simgrid/jedule
423 - simdag/datatypes.h was removed
424 - simdag/simdag.h -> simgrid/simdag.h
425 - msg/datatypes.h was removed
426 - msg/msg.h -> simgrid/msg.h
429 * Interface improvement:
430 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
431 - Sanitize the interface in MSG_task_ module:
432 - Merge two functions that were close enough but misleading:
433 set_compute_duration(t) -> set_flops_amount(t)
434 get_remaining_computation(t) -> get_flops_amount(t)
435 - set_data_size(t) -> set_bytes_amount(t)
436 get_data_size(t) -> get_bytes_amount(t)
437 - Massive cleanups in the functions related to the energy
438 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
439 - New: MSG_host_get_pstate()
441 - msg/energy/onoff: switching hosts on and off
443 * Interface improvement:
444 - Rename Host.isAvail() to Host.isOn()
445 - Rename Process.currentProcess() to Process.getCurrentProcess()
446 - Rename Task.setDataSize() to Task.setBytesAmount()
447 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
449 - #18874: Actually allows the GC to reclaim tasks
453 - SIMIX_process_throw: raises an exception in a remote process
454 * Refactoring: Separate sub-modules
455 - libsmx: the public interface, as libc in a real system
456 - popping: the strange dance that converts a user request into a kernel handling
457 - smx_context_*: the virtualization mechanisms that embed the user code
458 - smx_*: the handling of each simcalls
460 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
461 * Rename smx_action_t into smx_synchro_t, making explicit that these
462 things are used to synchronize processes with their environment.
463 For example, a communication is a sort of synchronization involving
464 the communicating processes (that may block until the exchange) and
465 the platform. The same can be said from computations, etc.
467 - #18888: segfault when a process ends before its kill_time
471 - 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.
472 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
473 - MPI_Type_set_name, MPI_Type_get_name
474 - MPI_*_c2f and MPI_*_f2c functions
475 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
476 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
477 - Activate a lot of new tests from the mpich 3 testsuite
479 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
480 - InfiniBand network model added : Based on the works of Jerome Vienne
481 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
482 - When smpi/display_timing is set, also display global simulation time and application times
483 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
484 * Collective communications
485 - 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
486 - 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/).
487 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
489 - Replay now uses algorithms from wanted collective selector
490 - Replay can be used with SMP-aware algorithms
491 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
492 - 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.
494 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
495 - Correctly use loopback on fat-tree clusters
496 - Asynchronous small messages shouldn't trigger deadlocks anymore
497 * Energy/DVFS cleanup and improvement
498 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
499 - new: smpi_get_host_pstate()
502 - "Full" network optimization flag was broken since Surf++
503 - Better handling of precision flags in maxmin
504 - Fix bug causing sometimes "Impossible" errors
505 - Properly pass cluster properties to included hosts
506 * Improvement of the Energy plugin.
507 - Always update the consumption before returning that value
508 - New property: watt_off to denote the dissipation when the host is off
509 - New functions getWattMinAt and getWattMaxAt to retrieve the
510 dissipation of pstates that we are not currently at.
511 * Java: class NetworkLink renamed to Link
512 * New function: simcall_process_get_kill_time()
513 * Massive rename s/workstation/host/
514 - That's intrusive, but that's good for the project consistency. Sorry.
515 - Change config option "workstation/model" into "host/model"
518 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
519 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
520 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
521 - xbt_dirname and xbt_basename for non-POSIX systems
523 * The model checker now runs as a separate process.
524 * The model checker runs must now be launched with the new simgrid-mc program.
525 * Record/Replay: the MC can display a textual representation of a path in the
526 execution graph. It can then be replayed outside of the model checker.
528 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
530 SimGrid (3.11) stable; urgency=low
535 * Normalizing pointers addresses tool for better diff between logs
538 * Add cloud examples using new VMs
539 - examples/msg/cloud/two_tasks_vm.tesh
540 - examples/msg/cloud/simple_vm.tesh
541 - examples/java/cloud/cloud.tesh
542 - examples/java/cloud/migration/migration.tesh
543 * Add java surf examples:
544 - examples/java/surfPlugin/surf_plugin.tesh
545 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
546 - examples/java/surfCpuModel/surf_cpu_model.tesh
547 * Add SMPI+MSG example:
548 - examples/smpi/smpi_msg_masterslave/
553 - msg task destroy cancel test
554 - msg_host on/off test
555 * Move all tests in testsuite to teshsuite (adding tesh files)
556 * Restructure teshsuites
557 - one folder for each kind of test
558 * Restructure AddTests.cmake
560 - structure the order of tests (with sections)
563 * Add virtual machine
564 - creation of a VM on a PM
565 - migration of a VM from a PM to another PM
567 - MSG_process_join(msg_process_t process, double timeout)
568 - msg_bar_t MSG_barrier_init(unsigned int count)
569 - int MSG_barrier_wait(msg_bar_t barrier)
570 - void MSG_barrier_destroy(msg_bar_t barrier)
571 - msg_as_t MSG_environment_as_get_by_name(const char * name)
572 * New option "msg/debug_multiple_use" to help debugging when a task is used
575 - Improvements and finalization of MSG_storage, MSG_file APIs and their
577 - Increase code coverage in test suites
581 * Protect context stacks against stack overflow. The number of protected memory
582 pages allocated on the top of each stack (1 by default) can be configured
583 with the new command line option --cfg=contexts/guard_size:<value>.
584 * Simcalls are now generated by a python script that
585 - generates files included by SimGrid
586 - checks that all the functions exist, and proposes prototypes
588 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
589 host_set_data, host_get_data
591 - simcall_process_join(smx_process_t process, double timeout)
592 * Fix bug where sleeping processing could not be suspended.
595 * Translate surf models from C to C++
596 - Generic classes for all models: Model, Resource, Action
597 - A generic interface for each kind of model (CPU, Network, Storage
598 Workstation, WorkstationVM)
600 * Translate surf routings from C to C++
601 * Add callbacks using sigc++ or boost::signals2
602 - Add callback functions for resource creation/destruction
603 - Add callback functions for action state change
604 - Handle Energy as a plugin
605 * Replace swag by boost::intrusive
606 * Add new routing models for clusters. For documentation, see
607 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
608 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
609 parameters for cluster tag
610 - Fat trees, with topology="FAT_TREE" and
611 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
612 - see examples/platforms/torus_cluster.xml and
613 examples/platforms/fat_tree_cluster.xml
617 * Hostfiles support host:nb_processes construct to deploy several processes on
619 * Collective communication algorithms should not crash if used with
620 improper number of nodes and report the error.
621 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
622 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
623 MPI_Dims_create, MPI_Cart_sub are supported.
624 * New interface to use SMPI programmatically (still depends on MSG for
625 some parts, see examples/smpi/smpi_msg_masterslave) :
626 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
630 * Global variables privatization in MPI executables is now performed at runtime
631 with the option smpi/privatize_global_variables (default:no).
632 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
633 dynamic libraries loaded by the application are not privatized (static
634 linking with these libraries is advised in this case)
637 * Options defined in XML work correctly now.
640 * New cmake option, enable_lib_in_jar, to control whether native libraries are
641 copied into simgrid.jar or not (ON by default). Use this option if you want
642 to reduce the size of the installed simgrid.jar, *and* the native libraries
643 are kept installed elsewhere.
644 * Surf binding with SWIG (code generated in maintainer mode only):
645 - plugin to handle callbacks
646 - CPU model only for the moment
649 * Supernovae build mode is definitively removed. It was used to improve
650 inlining and inter-module optimizations. It is nowadays superseded by
651 link-time optimizations commonly available in compilers.
652 * Update ns-3 find lib. Bindings for ns-3 should work again now.
653 * Add boost dependency for surf++
654 * Add new macro for tests
655 - ADD_TESH(name <tesh_args>)
656 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
660 - xbt_bar_t XBT_barrier_init(unsigned int count)
661 - int XBT_barrier_wait(xbt_bar_t barrier)
662 - void XBT_barrier_destroy(xbt_bar_t barrier)
663 * Make the xbt_os_time module public
665 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
667 SimGrid (3.10) stable; urgency=low
669 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
672 * Reintegrate Java to the main archive as desynchronizing these
673 package is not acceptable anymore (Java is now considered stable)
674 * Add explicit synchronization facilities through semaphores
675 * Bug fix: Task.setDataSize() only changed the C world, not the value
676 cached in the Java world
679 * Dramatically change the way files are handled. API and internals changed, but
680 this part of MSG was not considered as production grade either.
681 * Add explicit synchronization facilities through semaphores
682 * Add a new function MSG_host_get_process_list()
683 * Preliminary DVFS support (see examples/msg/energy/ for details)
686 * SMPI is now included directly in the libsimgrid as the windows
687 linker doesn't force us on splitting it anymore.
688 * Improvements of the SMPI replay tool:
689 - Most of the collective communications are now rooted in the same process as
690 in the original application.
691 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
692 used until now). Multiple data types can now be used in a trace.
693 - The replay tool now supports traces produce either by TAU or a modified
695 - Bug Fix: the compute part of the reduce action is now taken into account.
696 - Gatherv collective is now supported
697 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
698 of smpirun outputs time independent traces for the current run. One file
699 is created per process. If too many processes are simulated, this behavior
700 can be changed to one file for all processes by using the
701 tracing/smpi/format/ti_one_file flag
702 * smpirun generates the host file if needed (with given host count and platform)
703 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
704 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
705 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
706 --cfg=smpi/coll_selector:(mpich/ompi)
707 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
709 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
710 happening inside a collective SMPI call.
711 * Fix the behavior of complex data types handling.
712 * Make MPI_Wtime another synchronization point to take computations into
714 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
715 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
716 * Add all missing Fortran bindings, SMPI should work with Fortran 90
717 (no privatization of global variables yet)
718 * Preliminary DVFS support (see examples/smpi/energy/ for details)
721 * Verification of liveness properties is now available for SMPI applications
722 (in addition to MSG applications)
723 * Bugged examples using SMPI in examples/smpi/mc/
724 * Add --cfg=model-check/visited option. Allows the verification of infinite
725 programs. Detection of loops in the execution thanks to the system state
726 comparison and reduction of the state space to explore. Can be combined with
727 DPOR for safety properties.
730 * Allow to change SimGrid configuration (see --help) within the code
731 thanks to SD_config() as it can be done in MSG.
732 * Add a new function SD_task_set_amount() upon user request.
735 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
736 * Remove rule based routing (no more PCRE dependency)
737 * Add a limiter_link option to cluster tag, to specify a maximum reachable
738 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
739 * Add a loopback_bw and loopback_lat options to cluster tag.
740 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
741 routing. Now peers are encapsulated in an AS and have their own private
742 router but this is transparent.
745 * Our own implementation of getline is renamed xbt_getline, and gets
746 used even if the OS provide a getline(). This should reduce the
747 configuration complexity by using the same code on all platforms.
748 * New type: xbt_cfg_elm_boolean.
749 * Allow to use yes/no for boolean configuration options in the command line.
750 * Allow to disable SimGrid cleanups at exit from command line option.
751 There are situations where one may want a simulation to end with an exit.
752 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
753 annoying when scripting around the simulator. Adding a
754 --cfg=clean_atexit:no allows to circumvent this issue.
757 * Lots of memory leaks were corrected in this release.
758 There is no known memory leaks anymore, in all of our 600+ tests.
759 * New command line option --version, to get SimGrid version information.
760 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
762 * Supernovae builds are deprecated, and expected to be removed in the next
765 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
767 SimGrid (3.9) stable; urgency=low
769 The Grasgory release: GRAS is really dead now.
771 * Complete overhaul of the internal host structures scheme.
774 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
775 was considered as experimental and badly maintained since then.
776 * Keeping it was thus a trap to our potential users, that could take
777 it instead of MSG or SMPI by mistake despite is pity state.
778 * GRAS seems to have very few users (if any), and no one volunteered
779 to maintain it further. It also induces a lot of XBT code (for
780 portability sake), that must be maintained too.
781 * For all these reasons, we killed GRAS. If someone wants to revive it
782 in the future, don't cry, our git history still remembers of GRAS.
785 * Major overhaul. Merge our documentation again as time proved that
786 splitting it was really not helping our users.
787 * Further improve the developer documentation to help newcomers
788 hacking on SimGrid itself. The user documentation (and in
789 particular, the beginner documentation) is still in a sorry state.
792 * Now works on Windows too!
793 * Much more extensive test suite, from MPICH
796 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
797 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
798 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
799 * Bug fix: let task be scheduled when the last dependency to be solved is
800 a control dependency.
801 * Remove SD_load_environment_script function.
802 Use the C sg_platf function if you want to declare a platform
806 * New function: MSG_process_get_number()
807 * Old function documented: MSG_config()
808 * Remove MSG_load_platform_script function
809 Use the C sg_platf function if you want to declare a platform
813 * Change the default value of the TCP_gamma constant (maximal size of TCP
814 congestion window) to a more realistic 4MiB value. If you notice changes in
815 your simulation results, you can fall back to the previous 20k tiny window
816 by adding --cfg=network/TCP_gamma:20000 on command line.
817 * (Hopefully) fix a bug wrt periodic availability and state traces
818 * Bug fix: use default values at start when first event in availability/state
819 trace is not at time 0.
822 * remove the "new_" part of function name sg_platf_new_trace_connect
823 (resulting in sg_platf_trace_connect), since it does not create
827 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
828 use case where it's really mandatory, and maintaining it was a pain in
830 * New: xbt_fifo_search(), search an item with a user-provided
831 comparison function instead of dumb pointer comparison.
834 * Fix the lua deployment:
835 Use `simgrid.init_application()` before deployment instead of
836 `simgrid.msg_register_application()` after.
839 * Transfer the tracing files into the corresponding modules.
841 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
843 SimGrid (3.8.1) stable; urgency=low
845 The "we are told that some people want to also *install* the simgrid
848 * Add missing file "tesh.1" to the archive.
850 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
852 SimGrid (3.8) stable; urgency=low
854 The Psssshiiiit release: SimGrid jumps into the Cloud.
857 * Add an experimental interface to manipulate VMs. They are mainly
858 process groups with very few intrinsic semantic, but they should
859 allow you to build the semantic you want easily.
860 * New function: MSG_host_set_property_value()
861 * New function: MSG_process_on_exit(). To clean memory in all cases.
862 * Bug fixes that made the host (and link) failures unusable.
863 * Add a way to auto-restart process when the host in which they are
864 executing comes back (ON_FAILURE="RESTART" on deployment file,
865 MSG_process_auto_restart_set).
866 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
867 please stop using the old ones, they are DEPRECATED.
869 * Deprecate functions MSG_global_init() / MSG_global_init_args()
870 Please use MSG_init() instead. (reducing the amount of entry
871 points in the library helps us).
872 * Make it impossible to link against the wrong version of the lib
873 * Deprecate MSG_clean(). No need to call it anymore.
874 * Function MSG_get_host_number() is not deprecated anymore.
877 * Split the doc into a user guide and a reference guide.
878 * Start a developper guide to help people hacking on SimGrid.
881 * Enable tracing by default. This modules rocks you should use it.
882 * Remove option custom_flags. Now use environment variables CFLAGS
884 * Use default cmake things to detect lua instead of home grown ones.
885 * New option "enable_mallocators" to disable mallocators, for debugging
886 purpose ("on" by default).
889 * Bug fixes around the resource failures: don't let the processes
890 survive the host they are running onto.
891 * Add an interface to auto-restart processes when the host in which they are
892 executing comes back.
893 * Ensures that SIMIX_clean is called automatically. It's not part of
894 the public interface anymore (bindings should be updated).
897 * Bug fix for when SD_Simulate is called with a positive value: be careful
898 when comparing doubles. Sometimes they are different for non significant
900 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
901 parallel task whose initial work is distributed among host according
902 to the Amdahl's law. Such tasks are created with a parameter alpha
903 that corresponds to the non-parallelizable part of the computation.
904 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
905 between two sets of workstations assuming a 1D block distribution (each
906 workstation owns a similar share of data) on both sides.
908 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
909 Data redistribution will be automatically scheduled once parent and child
910 are both scheduled. The filling of computation_amount and
911 communication_amount structures is now done seamlessly thanks to the chosen
913 * New function SD_workstation_dump to display various information
914 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
915 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
916 bandwidth on the route onto which the task is scheduled and the amount of
918 To divide the nominal bandwidth by 2, the rate then has to be :
919 rate = bandwidth/(2*amount)
920 * Compute tasks that have failed can now be rescheduled and executed again
921 (from their beginning)
922 * Increasing source code coverage (src/simdag is now covered at 95.8%
926 * Re-implement time-independent trace replay using SMPI (at the
927 smpi_smp_* level) instead of MSG. This should replace
928 examples/msg/actions/actions.c
929 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
930 hindexed and structs)
931 * Implement the exchange of non-contiguous data.
932 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
933 * Correct behavior of smpi/sender_gap and set its default value to 0
934 * Add option to asynchronously send small messages to allow better
935 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
936 specifies the size in bytes under which messages will be asynchronously sent.
937 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
938 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
941 * New C interface to define a platform: XML is now optional.
942 For more info, please check include/simgrid/platf.h.
943 * New interface to define random platforms from the C:
944 For more info, please check include/simgrid/platf_generator.h and
945 examples/msg/masterslave/masterslave_platfgen.c
946 * Export a sg_cmdline dynar containing all the arguments we got from
950 * Two new tracing options for adding comments to trace file so you
951 can track your experiments (see --help-tracing for details).
952 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
953 * Adding the SimGrid version that generated the trace file as a comment.
954 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
955 * Fix to avoid key clashes on Paje links
956 * Other minor fixes related to the Paje specification
959 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
960 and renamed to xbt_str_hash() and xbt_str_hash_ext().
961 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
962 * Greatly improve the robustness of mmalloc to user errors (such as
963 using an area after freeing it, or freeing it twice)
965 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
967 SimGrid-java (3.8.1) stable; urgency=low
969 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
970 Warning: all methods are visible, but only some of them are
971 implemented so far. Check the source (src/jtrace.c)
972 for further information.
973 * New module: org.simgrid.msg.File (SimGrid File management functions)
974 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
975 * Change the meaning of Process.restart: now restart the process from
976 the begining, like MSG_process_restart in C.
977 * Add Process.setAutoRestart: handling of process restart when failed
979 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
980 you to retrieve the properties of the processes/hosts
981 * Deprecate Msg.clean(): you can just forget about it now.
982 * New function Process.getCount(), that only works when compiling
983 with the not yet released version 3.9 of the C library.
985 * New context factory based on Coroutines. It mandates a modified JVM
986 but then, the simulations run about five times faster, and there is
987 no limit to the amount of processes (beside of the available memory).
989 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
991 SimGrid (3.7.1) stable; urgency=low
994 * Restore the prototype of MSG_process_create_with_environment() to
995 the pre-3.7 situation by removing the kill_time argument.
996 * Add a MSG_process_set_kill_time() function instead.
999 * Fix weird behaviors when dealing with parallel tasks.
1002 * Simgrid is now built as a dll.
1003 * Simgrid-java now works on Windows.
1004 * Simgrid-Java is now included into Windows package.
1007 * First pre-build package for MacOSX.
1010 * Fix compilation when using MSG_USE_DEPRECATED.
1011 * Fix some compilation issues on Macs and Windows.
1012 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1013 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1015 We discovered that the Lua console is broken, but we are missing the
1016 manpower to fix it right now. The problem existed in 3.7 too, so we
1017 are not blocking the release for that. Sorry if you depended on this
1018 feature, any help would be really welcome.
1020 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1023 SimGrid-java (3.7.1) stable; urgency=low
1025 The "Java aint got to be bloated and slow" release
1028 * Various internal cleanups and performance improvement
1029 Simulations are expected to run up to twice faster or so
1030 * Make Process.kill(process) an instance method, not a static one
1031 * User processes are not java.lang.Thread subclasses.
1032 This breaks the compatibility (sorry), but previous API was
1033 brain-dead, making it impossible to have non-trivial
1034 initializations in the process constructor.
1035 * Require a full constructor per Process sub-class.
1036 Kinda breaks the compatibility (sorry), but this allows a much more
1037 efficient way to launch the processes at simulation startup.
1038 * Do not embeed our version of semaphores, java 1.5 can be considered
1039 as sufficiently prevalent for us to not dupplicate its features.
1043 * Add examples for almost every part of the API
1044 We spotted and fixed a lot of bugs in the process
1046 * New module: asynchronous communication API
1047 * New function: Process.sleep()
1048 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1049 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1050 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1052 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1054 SimGrid (3.7) stable; urgency=low
1056 The "spring cleanups (before the next Big Project kicks in)" release.
1059 * We can specify the SMPI latency/bandwidth factor with command line
1060 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1061 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1062 You can also use the "config tag" from platform file by adding this line
1063 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1064 (see "example/platforms/tag_config.xml" to use "config tag").
1065 Note that the command line supersedes the platform file configuration.
1066 * Change the correction factors used in LMM model, according to
1067 the latest experiments described in INRIA RR-7821.
1068 Accuracy should be improved this way.
1069 * Use the partial invalidation optimization by default for the
1070 network too. Should produce the exact same results, only faster.
1071 * Major cleanup in surf to merge models and split some optimization
1072 mechanisms from the core of the models. As a result you can now
1073 specify which model to use (e.g., --cfg=network/model:LV08
1074 --cfg=cpu/model:Cas01) and which optimization mode to use
1075 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1076 Incompatible combinations should err at initialization. See
1077 --help-models for the list of all models and optimization modes.
1078 * The CLM03 workstation model was dropped for simplicity because it
1079 used the deprecated CM02 network model. Use default instead.
1080 * Rename the TCP_gamma configuration option to network/TCP_gamma
1081 * Rename the coordinates configuration option to
1082 network/coordinates, and document it
1083 * Use now crosstraffic keyword instead of the terribly misleading
1084 fullduplex keyword. It is activated by default now in the current
1085 default model, use --cfg=network/crosstraffic:0 to turn it off.
1086 * Ongoing refactoring the model parsing to make XML files optional
1087 See include/simgrid/platf.h for details (still to be completed)
1090 * Major overhaul of the documentation. Almost instructive now :/
1091 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1092 functions and friends. This interface was considered as
1093 deprecated since over 2 years, it's time to inform our users that it is.
1094 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1095 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1096 use these (crappy) functions in your code.
1097 These functions will be removed soon. Stop using them now.
1098 * Deprecate MSG_get_host_{table,number}
1099 Implement MSG_hosts_as_dynar() instead.
1100 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1101 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1104 * Stabilize the parallel execution mode of user contexts
1105 * Introduce configuration variables to control parallel execution:
1106 - contexts/synchro: Synchronization mode to use when running
1107 contexts in parallel (either futex, posix or busy_wait)
1108 - contexts/parallel_threshold: Minimal number of user contexts
1109 that must be part of a scheduling round to switch to parallel
1110 execution mode (raw contexts only)
1111 * Fix bugs that prevented to use suspend/resume along with
1112 synchronization structures.
1113 * Fix bugs in process termination that lead to invalid memory access
1114 in very specific conditions.
1117 * Introduce a parallel mode for the models (controlled by surf/nthreads
1118 configuration item). In our tests, running the models in parallel
1119 never lead to any speedups because they are so fast that the gain
1120 of computing each model in parallel does not amortizes the
1121 synchronization costs, even when ultra fast futexes are used.
1122 This is released anyway because YMMV.
1125 * Performance boost by using a swag internally to compute the set of
1126 tasks that are finished and should constitute the return value of
1130 * Enable it by default now that it is considered rather stable.
1133 * Documentation of the tracing functions.
1134 * Performance gains when tracing categorized/uncategorized resource
1135 utilization by avoiding calls to get route when updating resource
1136 variables. LMM constraints are being used instead.
1137 * API changed to set task categories. Use MSG_task_set_category instead
1138 of TRACE_msg_set_task_category, and SD_task_set_category instead
1139 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1140 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1141 * Examples for link user variables added (see at examples/msg/tracing/)
1142 * Deprecated function TRACE_msg_set_process_category completely removed
1143 * Trace header updated according to the latest Paje file format
1144 * Tracing network lazy updates, no longer obligate users to use full updates
1145 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1146 * Experimental: let user code declare/set/push/pop application states for hosts
1147 * API changed to allow the manual creation of graph configuration files
1148 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1151 * Improve the API of Lua MSG bindings, using the Lua spirit.
1152 * Each simulated process now lives in its own Lua world (globals are
1153 automatically duplicated). It helps writing simulators. Will allow
1154 to run Splay programs within SimGrid in the future.
1155 * Add a Chord example in Lua, equivalent to the MSG one.
1158 * Start the implementation of a solution to express temporal
1159 properties, not only local assertions. This is still an
1160 experimental work in progress, stay clear from it to be safe.
1164 - Add new runtime parameters --help-logs and --help-log-categories
1165 to display information about supported logging parameters and
1167 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1168 don't exists anymore.
1169 * Mallocators: allow value NULL for the reset function.
1171 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1172 create homogeneous dictionaries, where all the elements share the
1173 same free function. Non homogeneous dictionaries will be
1174 deprecated in the next release.
1175 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1176 - Multi-level dictionaries are deprecated.
1178 - new function xbt_dynar_search_or_negative() that is useful when
1179 you have less than 2 million elements in your dynar and don't
1180 want of the extra complexity of catching exceptions when the
1181 element is not found.
1183 - Make xbt_os_thread module (for thread portability) public.
1184 Documentation is still to come, sorry.
1186 - Cleanups and simplifications to make it maintainable again.
1187 - Exotic features (such as memalign and valloc) were removed.
1188 - The metadata were extended and improved so that the
1189 model-checker becomes able to explore and inspect the heaps.
1190 - This may induce a performance drop when enable_model-checking is
1191 ON in cmake (even if it's not used in the simulation), but it is
1192 necessary at this point to get MC working.
1194 Turn model-checking OFF if simulation performance matters to you.
1195 Not enabling it at runtime is not enough, disable it in cmake.
1197 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1200 SimGrid (3.6.2) stable; urgency=low
1202 The "Not coding new stuff allows to polish old things" release.
1205 * New bindings to the NS3 packet level simulator (experimental)
1206 * Use the raw (efficient) execution contexts instead of the sysv
1207 (portable) ones when possible.
1208 * libpcre is now mandatory in any cases since not using it led to
1209 severe performance loss and possibly other issues
1210 * Update the XML platforms:
1211 - G5K: include the latest machine in Nancy
1212 - GridPP and LCG: new platforms
1213 * Documentation was partially updated, at least (more to come)
1215 Bug fixes, cosmetics and small improvements
1216 * Free terminated processes before the end of the simulation to avoid
1217 exhausting the memory of users having very dynamic amount of
1219 * Bug fix and cosmetics about canceling non-running tasks
1220 * Bug fix about the dot loader's issues when using libcgraph
1223 * Create an installer for windows with nsis (amd64 and win32)
1224 - Add an hello world project to illustrate simgrid project creation.
1225 - Embed libpcre into the Simgrid installer to avoid
1226 its compilation burden
1227 * The raw execution contexts should work on Apple now
1228 * Port to Windows 64 bits
1229 - Sysv contexts now have an implementation for this arch
1230 - GRAS communication features now support this arch
1231 * Drop support for borland compiler on windows
1232 - this code was not maintained, and we kinda depend on gcc nowadays
1233 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1234 * Fix portability issue on unstable ubuntu: linker became picky on
1237 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1239 SimGrid (3.6.1) stable; urgency=low
1241 The "Oops, we broke Macs too" release
1244 * Fixed contexts detection so that raw ones are used when possible
1245 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1246 leads to a strange error, with user code segfaulting sometimes when
1247 the generated code is not perfectly aligned (which is not
1248 controllable from the user side, depends on the amount of code)
1251 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1253 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1255 SimGrid-java (3.6) unstable; urgency=low
1258 * Split of every thing from simgrid v3.5 into a separate package.
1260 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1262 SimGrid (3.6) stable; urgency=medium
1264 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1267 * Bindings now constitute their own package, separated from the main one.
1268 Rationale: reduce our maintenance nightmare by reducing the module coupling
1269 They will soon be released on their own on gforge.
1271 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1272 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1274 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1275 * It's not quite deprecated for now because we have no replacement,
1276 but it may soon become the case.
1279 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1280 * Fortran: New user-level cache variable to store the rank of the running
1281 process. This improves performance by an order of magnitude.
1282 * C: New coccinelle script to automatically locate and modify global and
1283 local static variables.
1284 * Improved SMPI network model with a sender-side gap to account for multiple
1288 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1289 only say if a communication is finished, no matter whether it succeeded or
1290 failed. You can call MSG_comm_get_status() to know the status of a finished
1292 * New function MSG_task_dsend() to send a task and detach it. When a
1293 communication is detached, you are never notified of its success or failure
1294 and the memory is released automatically once it is finished. This function
1295 is useful when you don't care about the end nor the success of a
1297 * Change the prototypes of action replay. Sorry for inconvenience,
1298 but this is really more efficient this way (and to adapt your code,
1299 you just have to fix the initialization, that shouldn't be too long)
1300 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1301 ever managed to do anything useful with it.
1302 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1303 takes a dynar of communications. It returns immediately and gives the
1304 index of a finished communication (if any).
1305 * New example: a basic implementation of the Chord P2P algorithm.
1308 * New model for multi-core CPUs. You can now use the core attribute to
1309 precise the number of cores of a host. This is a basic model. Every
1310 process running on the host receives at most the power provided in
1311 the DTD (throughput<=power). Total throughput of process cannot exceed
1313 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1314 router linked by an up-link and a down-link (possibly asymmetrical).
1315 This kind of pattern allows to easily build last-mile model style platforms.
1316 Aggregating such patterns in a rule-based AS is thus the technique of
1317 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1318 * New model for Vivaldi routing. We transformed the Vivaldi network model
1319 into a Vivaldi routing model (based on the rule-based model). This allows to
1320 combine Vivaldi based latencies with last-mile platforms.
1323 * Added a check for NaN of IEEE754 infinite in the double entries of
1325 * Introduce a new context factory "raw", highly inspirated from the
1326 ucontext factory, but using manually crafted functions in assembly to
1327 do the work in an efficient manner.
1328 * Allow to change the used context factory at run time, not only at
1329 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1330 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1331 contexts at runtime (only with raw contexts or ucontexts).
1332 * Completely rewrote this module to allow parallel execution of user
1333 processes. Use --cfg=contexts/nthreads:N to execute user processes
1334 with N parallel threads (the default is 1, meaning no parallelism).
1335 * Allow to decide dynamically between sequential and parallel modes.
1336 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1337 processes in parallel only when their number is greater than or equal to P
1339 * Added a check for NaN of IEEE754 infinite in the double entries of
1343 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1344 won't output the state of processes when interrupted with Ctrl-C
1345 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1346 NULL-terminated array. This may solve backward compatibility issues
1347 due to the change to return type of SD_simulate. See also:
1348 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1349 * Add new macros with variable number of arguments.
1350 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1351 - in xbt/asserts.h: xbt_assert
1352 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1353 - in xbt/ex.h: THROWF and RETHROWF.
1354 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1356 * Change xbt_die() to accept a format string with arguments, just like printf.
1357 * New data structure: xbt_lib_t, like a dict but more general and with better
1361 * New configuration options
1362 Options triva/categorized and triva/uncategorized can be used to generate
1363 graph configuration files for Triva visualization tool.
1364 * Configuration option tracing/platform is renamed to tracing/categorized
1365 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1366 * New TRACE_user_link_variable interface:
1367 User provides the name of the link and the tracing variable to attach to it
1368 * the declaration of tracing categories must be done after the environment
1370 * simpler tracing interface, just one way to declare categories
1371 TRACE_category or TRACE_category_with_color, it is up to you
1372 * links in the trace file are again identified by their names
1373 * trace contains the full platform hierarchy exactly as declared using the ASes
1374 * Options tracing/msg/[task|process]:1 groups the process by hosts
1375 for both cases, tasks and processes must have names that are unique during the simulation
1376 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1377 * The experimental option tracing/msg/volume is deprecated
1378 its functionality may be reincorporated if needed
1380 The tracing generates a trace file with unordered timestamped events,
1381 because of the way the core simulator (surf) works. A script available
1382 at the tools directory (fix-paje-trace.sh) can be used to put the events
1383 in order. We have changed the tracing so it can generate ordered timestamped
1384 events in the final trace, but depending on the simulator (and how much time
1385 is simulated) that can lead to a huge memory utilization. It is deactivated
1386 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1388 Build Infrastructure
1389 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1390 - We are trying hard to keep the API stable, but it may happen that
1391 some things change (we're a research project after all, not a
1392 nuclear plant operating system). If such things should happen, you
1393 could rely on that macro to adapt.
1394 - current value: 30600 for 3.06.00, aka 3.6
1395 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1396 a number that can be compared with SIMGRID_VERSION.
1397 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1398 assertions and verbose/debug logging events are disabled at compile time.
1400 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1402 SimGrid (3.5) stable; urgency=medium
1405 * New feature to any SimGrid-based simulator: Model-Checking
1406 Check SIN#1 for more details.
1409 * New Model SMPI (three-interval linear regression for correction factors)
1410 See RR-7426, available at http://hal.inria.fr/inria-00527150
1411 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1412 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1413 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1414 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1415 * See also src/smpi/README
1420 - Tracing API changes: TRACE_start and TRACE_end should not be called
1421 by user-code. They are automatically called by simulators created
1422 with SimDAG, MSG and SMPI if the toolkit is compiled with
1423 tracing_enabled=ON. Categories declaration and utilization remain the
1424 same for MSG and SimDag.
1425 - A function was added to the tracing API to declare categories with
1427 - TRACE_category_with_color (char *category, char *color)
1428 where color must be in the following format
1429 "%f %f %f", red, green, blue
1430 and red, green, blue are float values in the interval [0, 1]
1431 - User can specify NULL as color parameter, or continue calling
1432 TRACE_category (cat)
1433 On that case, the tracing system will define random colors
1434 - The following command-line options are supported:
1435 --cfg=tracing/filename:msg.trace
1436 --cfg=tracing:1 (activate tracing, needed to use others)
1437 --cfg=tracing/platform:1 (categorized resource use)
1438 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1439 --cfg=tracing/msg/task:1 (task creation)
1440 --cfg=tracing/msg/process:1 (process creation, migration)
1441 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1442 --cfg=tracing/smpi:1 (SMPI interface tracing)
1443 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1444 - examples of examples/msg/tracing updated
1446 - DAXLoader and DOTLoader functions can generate tasks with categories
1447 - A new function to attribute a category to SD tasks:
1448 TRACE_sd_set_task_category (SD_task_t task, char *category)
1449 * Tracing the MPI interface implemented by SMPI
1450 - Collective operations are traced with states
1451 - Point-to-Point operations are traced with states/links
1452 - Tracing activated by a parameter "-trace filename" passed
1453 to smpirun during execution (considering that simgrid
1454 is compiled with tracing enabled)
1455 - To run the simulation with gdb, the simulator
1456 accepts --cfg=tracing/smpi:1 to trace SMPI
1457 - tesh files to check if smpi tracing is ok
1458 - See examples/smpi/NAS/DT-trace
1459 * GTNetS tracing re-worked
1460 - adaptation to the tracing system of GTNets to cope
1461 with modifications regarding the fullduplex mode
1462 - new tesh files to check if gtnets tracing is ok
1465 * Asynchronous communications through the functions:
1466 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1467 * New function: MSG_load_platform_script()
1468 to make possible using a lua script instead of XML files to set up platforms
1469 * New function: MSG_set_function
1470 to associate functions to processes, used when bypassing the parser
1471 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1473 Platforms: Add some more examples in examples/platforms
1474 * Grid'5000: see www.grid5000.fr
1475 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1478 * Change the XML format. This is a very important modification. SimGrid
1479 3.5 introduces a new hierarchical format based on the notion of
1480 Autonomous Systems. Compatibility with old format is ensured
1481 through the perl script provided in the install bin directory
1482 bin/simgrid_update_xml.
1483 It is now possible to build platforms with specific routing
1484 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1485 connect several platforms together. We will try to provide soon
1486 set of realistic platforms exploiting these properties (have a
1487 look at examples/platforms/ for the moment).
1488 * Take the opportunity of the XML format change to be a good XML citizen:
1489 rename link:ctn to link_ctn and similar changes (also dealt with by
1491 * Add a new routing scheme (rule-based) using regular expressions. It
1492 enables to have an extremely low memory footprint when the
1493 underlying routing is simple and can be compactly described. You
1494 need to have libpcre4-dev (perl regular expressions) installed if
1495 you want to use this routing scheme.
1496 * Revive the cluster TAG and allow to easily and efficiently (both in
1497 term of memory and speed) connect clusters together. Have a look
1498 at teshsuite/simdag/platforms/ to see how this can be done. With
1499 this tag, you can create clusters with thousands of tasks at no
1500 cost (have a look at examples/platforms/).
1501 Note: clusters are implemented as ASes, so there is no need for an
1502 enclosing AS tag if you have only one cluster in your platform.
1503 * Add new generic functions in the public interface that allows the user
1504 to call SURF 'create_resource' methods from your code (same
1505 functionality as the XML bypass mechanism but with a much lighter
1507 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1508 that uses a piecewise linear approximation to produce better
1509 results when exchanging small messages.
1510 * Add a new parameter to handle correctly full duplex link and account
1511 for interferences between uplink and downlink communications
1512 (activate with --cfg=fullduplex:1).
1515 * Rename the SD_READY (all dependencies are satisfied and task is
1516 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1517 dependencies are satisfied) state.
1518 This prevents a confusion between the notion of "ready to schedule"
1519 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1520 simulated" (SD_RUNNABLE) used by the simulation kernel.
1521 * Change the way a task is considered as ready. Instead of removing
1522 dependencies when a task is done, a counter is decreased. This way,
1523 it is always possible to reach ancestors thanks to the
1524 SD_taks_get_parents function (even after the end of the simulation.)
1525 * Change the return type of SD_Simulate from (SD_task_t*) into
1526 xbt_dynar_t. This function was in handling a dynar internally and
1527 converted it into a NULL terminated array for historical reasons.
1528 * New function SD_dotload(char*) to load a DAG described in dot
1529 format. This loader and the corresponding examples require the
1530 installation of the graphviz library.
1531 * Fix a bug in the management of tasks of size 0 in the surf network
1532 models. This problem was only visible with SIMDAG and you should
1533 thus disregard results produced with earlier versions if you
1534 relied on this feature (some tasks were blocked because of this).
1535 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1536 even though your applications were doing only point-to-point
1537 communications and sequential computations. Now you can really use any
1538 model you want (of course, if you create real parallel tasks, which are
1539 not implemented in most models beside ptaskL07, this will abort).
1540 * Add an example that schedules a DAX on an heterogeneous platform
1541 using a Min-Min strategy.
1542 * New function SD_workstation_get_current_task() that returns the kind
1543 of task currently running on a workstation in the sequential access
1545 * Raise some warnings when unexecuted tasks remains at the end of the
1546 simulation. This is usually caused by cycles in the DAG.
1549 * New function: SIMIX_process_set_function() called by MSG_set_function
1550 * Change the underlying waiting queue in semaphores so that a process
1551 can wait on several of them simultaneously (as in waitany).
1552 * Fix the way to handle tokens in semaphores so that all access patterns
1553 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1554 * kill the dirty pimple SIMIX_message_sizes_output()
1555 Please use (proper) visualization instead
1558 * New data container: setset (set of sets of elements)
1559 * New module: mmalloc (mapped malloc, allowing to have several
1560 independent segments of malloc)
1561 * New function: xbt_dict_cursor_set_data()
1562 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1563 * New function: xbt_dynar_is_empty()
1564 * New function: xbt_fifo_get_last_item()
1565 * Fix xbt_dynar_shrink(): use the right element size.
1566 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1567 uninitialized areas during expand.
1568 * Fix semaphores: previous implementation was severely broken.
1569 * Use library init/fini functions for our initialization.
1570 - you can use logs and other feature as soon as you want in your
1571 code (even before the xbt_init / MSG_init)
1572 - xbt_exit is now a no-op and produce a warning when used.
1575 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1576 benefit from the latest improvement to the simulation kernel.
1577 * Kill measurement sockets for now. If you rely on them, sorry. This
1578 release is not for you. This feature will be reintroduced in the
1579 future, but we cannot delay the release any further.
1580 * New function: gras_msgtype_get_name().
1581 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1582 * Fix (at last) the pmm example: it should not randomly fail anymore.
1584 Build chain: bug fixes and overall polishing
1585 * Cmake is now stable enough. Hence, we killed the autotools.
1586 * Port to windows ( TM :)
1587 * Fix the 'make install' target.
1588 No need to use 'make install-simgrid' anymore
1589 * Introduce a 'make dist' target compiling a *source* archive
1590 'make package' compiles a binary archive
1591 * Compile java files only on need
1592 * Add --cd and --setenv command line options to tesh
1593 * Out of source builds are not fully supported yet, but we are close
1594 * Enable supernovae and optimization flags by default for our users
1597 * Add layer to set up environment directly from lua, without XML.
1598 * The effect of gras_stub_generator can be achieved through
1599 lua too (check examples/gras/console/ping_generator.lua)
1601 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1603 SimGrid (3.4.1) stable; urgency=low
1605 The "Polishing easter eggs is probably a good idea" release.
1606 This is a bug fixes release only.
1610 * Fix a bug preventing the tasks from begin garbage collected.
1613 * Fix a bug occuring when a host involved in a communication fails.
1614 This was not detected properly by the other peer involved in the
1615 communication. Now, it's reported as a network error.
1618 * Warn the user about loop dependencies in data flow of DAX files
1619 * Obey the control-flow dependencies of DAX files
1622 * Add option "enable_smpi" allowing to not compile SMPI.
1623 Probably useful for the (Mac) users experiencing a build error here
1624 * Improve the detection of lua5.1 and ruby1.8
1626 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1628 SimGrid (3.4) stable; urgency=low
1630 The "Easter in Cargese" release. Also known as (major changes):
1632 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1633 ~> bindings were greatly improved
1634 ~> new tracing infrastructure for better visualization introduced
1636 * the "Welcome to configury modernity" release.
1637 ~> we switched from autotools to cmake, and improved our cdash
1640 A more detailled list of changes follow (full detail in svn log).
1642 Java Bindings: Various Cleanups
1643 * (install java-gcj-compat-dev on debian-like to use them)
1644 * Remove put/get: no need to export deprecated interface in Java
1645 Use send/receive instead.
1646 * Cleanup the examples and add a README per directory
1647 * Remove example autoDestination (that's the only way to go now)
1648 * Remove example explicitDestination (was a plain copy of basic)
1649 * Make JniException a runtime exception, so that there is no need to
1650 declare the fact that you may encounter such a beast. I guess that
1651 nobody will ever want to survive such error.
1652 * Create specific errors for each MSG case of failure:
1653 host failure, transfer failure, timeout, task cancelled
1654 * Cleanup the exceptions that may get thrown by each function
1655 * Other internal cleanups in Java bindings. Performance still bad :/
1656 Ruby and Lua Bindings: create them
1657 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1658 * That's new and great, you should try them out.
1659 Same functionalities than Java bindings, only even less polished
1661 * Kill the useless "rate" argument of SD_task_get_execution_time()
1662 Everyone used to provide -1 as a value, it was not used, and the
1663 semantic of a possible use wasn't even clear.
1664 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1665 either as comm costs or compute costs to mean that there is no
1666 such thing for that specific task.
1667 * Add a SD_task_set_name() function
1668 * Fix SD_task_unschedule() on typed tasks
1669 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1670 * In DAX loader, accept useless 'level' attributes to <job> since
1671 LIGO DAGs have them (seem to be to ease graphical representation).
1673 * Add an example masterslave_mailbox.c using send/receive and not
1674 the deprecated put/get interface.
1675 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1676 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1678 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1679 (declare MSG_USE_DEPRECATED to still have the old name)
1680 * Add a MSG_task_set_data() function
1681 * About trace replay (see examples/msg/actions):
1683 - Allow to work with splitted trace files for each process
1684 Give the specific trace file as argument of each process,
1685 and call MSG_action_trace_run(NULL)
1686 You can still have one merged file for all processes.
1687 - Fix implementation of collective operations
1688 * Allow task_execute() on 0-sized tasks (closes #10063)
1690 * This is the first release of SimGrid where SMPI is not considered
1691 beta anymore (even if some corners should still be improved)
1692 * Port over the new SIMIX_network submodule (internal refactoring)
1693 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1694 * Implement more missing elements of the standard:
1696 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1698 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1699 MPI_Scan() MPI_Get_processor_name()
1700 - Added implementation of missing case for Alltoall (warning: it's
1701 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1702 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1703 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1704 SMPI_MPI_Allgatherv()
1705 * Bug fixes include:
1706 - MPI_Waitsome() was broken
1707 - Allow relative includes in smpicc
1708 - Command line cfg argument 'reference_speed' was ignored...
1709 - Some functions did not properly lead to auto-benching of user code
1710 - smpicc passes -O2 by default (just like openmpi one)
1712 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1713 * Bug fixes about timeouts during communications
1714 * add SIMIX_message_sizes_output() as a pimple to write to file the
1715 amount of messages per size. Use gnuplot to get histogram.
1716 Pimple because that's the only user-visible function of simix,
1717 defined directly in xbt.h (irk, sorry)
1719 - Add a SIMIX_sem_get_capacity() function
1720 - Fix interactions with processe resume/suspende
1721 - release_forever() was stupidly broken
1722 - Fix SIMIX_display_process_status() for processes in a semaphore
1723 - Make SIMIX_sem_block_onto() user-visible
1724 * Refactoring context stuff:
1725 - Use pseudo-OOP for better modularity
1726 - reimplement SIMIX_process_kill() without process_schedule() so
1727 that the latter can take as invariant that it is called from
1729 - Merge context_start into context_new for sake of simplicity
1731 * Add a Vivaldi network model, coded live during SUD'10 ;)
1732 * Rename configuration variables to start a hierarchy:
1733 o cpu_model -> cpu/model
1734 o network_model -> network/model
1735 o workstation_model -> workstation/model
1736 * New configuration variables:
1737 o network/bandwidth_factor: correction to bandwith
1738 o network/latency_factor: correction to latency
1739 o netwotk/weight_S: correction to the weight of competing streams
1740 * Add a long description to the models, that users can see with such
1741 argument on the command line: --cfg=cpu/model:help
1742 * --help-models display the long description of all known models
1744 * config: add the ability to set a default value after registration
1745 Does not override any previously set value (e.g. from cmd line)
1746 * dict: allow to have integer key and data.
1747 When so, you need to use the following functions
1748 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1749 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1750 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1751 In contrary to regular dicts, the key is not malloced before copy.
1752 Mixing scalar and regular elements in the same dict is not tested
1754 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1755 Tracing for Visualization:
1756 * SimGrid is now instrumented in order to generate a trace file for
1757 visualization analysis: to use it, need to compile SimGrid with the
1758 "tracing" option enabled, and instrument the program using SimGrid with
1759 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1760 (among other functions).
1761 * The instrumentation only traces the platform utilization for now
1762 * Documentation to use the tracing functions and how to analyze the
1763 traces with the Triva tool is written.
1764 * More information about: SimGrid FAQ (in the section Tracing Simulations
1767 * We moved to cmake as default build system. Autotools support will
1768 be dropped soon. Check the FAQ for more info about how to use it.
1769 * Greatly improved our cdash/ctest interactions
1770 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1771 * Added memory checking tests with valgrind; lot of memleak fixing.
1772 This may be the first release of simgrid with so few memory issues
1773 * Added code coverage tests.
1774 Our coverage is still improvable, but at least we see it on cdash.
1776 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1778 SimGrid (3.3.4) stable; urgency=low
1780 The "Desktop Grid needs love too" release (also called Xmas release).
1782 Models improvements:
1783 * Major speedup in the maxmin system solving by using lazy evaluation
1784 Instead of solving completely the maxmin system at each iteration,
1785 only invalidate (and recompute) the modified parts.
1786 This new feature is enabled in default models but you can try to
1787 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1788 * Cas01 IMproved as default CPU model
1789 This CPU model is the same Cas01 model, but it uses the
1790 maxmin-selective-update flag and a heap structure to manage
1791 actions on SURF kernel.
1792 It reduces the complexity to find the next action to finish and,
1793 consequently, it's faster than the old Cas01.
1794 This is the new default CPU model (Cas01).
1795 * Rename the old Cas01 model to Cas01_fullupdate
1796 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1797 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1798 * CpuTI (CPU Trace Integration)
1799 A new CPU model whose objective is simulate faster when using
1800 availability trace files.
1801 Instead of using a full featured, over engineered maxmin system for
1802 CPU modeling, this model does the pre-integration of traces files
1803 to calculate the amount of CPU power available, and so, executes
1804 faster than the old CPU models.
1805 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1806 * Use LV08 as default network model since it gives better accuracy
1807 for small messages and shouldn't change things for big ones.
1808 Use --cfg=network_model:CM02 to get the previous behavior.
1811 ******************************************
1812 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1813 ******************************************
1814 * The new CPU model may changes simulations!
1815 The point is that events occurring at the exact same timestamp
1816 are not scheduled in the same order with the old and new
1817 version. This may be enough to completely change the execution
1818 of simulations in some cases.
1819 * The new network model will change simulations!
1820 This new model is more realistic than the previous one, so you
1821 should consider redoing your old experiments with this model.
1822 Sorry for the inconvenience.
1825 * Introduce the supernovae compilation mode
1826 When compiled that way, the whole SimGrid (or almost) is put in a
1827 single compilation unit and compiled in one shoot.
1828 This is to help gcc which has difficulties to inline stuff from one
1830 The speedup seem to be above 15%, althrough more tests are needed on
1831 amd64 to confirm that gain.
1834 * Port of MSG's mailbox on top of SIMIX network
1835 The put/get mechanism was greatly simplified on the way.
1838 * New SIMIX network module. Provides:
1839 - Mailbox: rendez-vous mechanism to find with who you want to speak
1840 - Synchronous send/recv: easier and hopefully faster since the
1841 logic is handled in the maestro process directly now
1842 - Asynchronous send/recv: you dreamt of it? It's here now
1843 Too bad that nobody cared enough to propagate the change to MSG.
1844 * Add semaphores as SIMIX synchronization mechanism.
1847 * new function SD_daxload(char*) to load a DAX file
1848 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1849 * Introduce typed tasks. Specify its kind and cost at creation.
1850 At scheduling, just give where it should be placed, and the cost
1851 for each involved resource is automatically computed.
1852 Existing constructors so far (more to come of course):
1853 - SD_task_create_comm_e2e() for end-to-end communication
1854 - SD_task_create_comp_seq() for sequential computation
1855 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1856 * new function SD_task_dump() for debuging display
1857 * new function SD_task_dotty(task,FILE*) writing to file the info
1858 about the task in dotty format
1859 * SD_task_dependency_exists() can now cope with having one of its
1860 arguments NULL. If so, it tests whether the other argument has any
1862 * Add getters on list of preceding/following tasks:
1863 SD_task_get_parents(task) and SD_task_get_children(task)
1864 * Add getters on amount of workstations and list:
1865 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1866 * Add getter on task kind: SD_task_get_kind(task)
1867 * Update the start_time and finish_time of tasks on completion/failure
1868 * Bugfix: Remove task from state swags when destroyed
1871 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1874 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1875 dynar with one separate thread per value of the dynar.
1876 * Change the prototype of xbt_thread_create(), sorry.
1877 Added a boolean parameter indicating whether we want to join this
1878 thread (used in SG only for now)
1879 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1882 * GTNetS wrappers should now be usable again (and betterly tested too)
1883 * Fix a major regression from 3.2 where the timeout provided to
1884 MSG_task_put_with_timeout() was used as absolute time before which
1885 the comm should be done.
1886 * Start to fix the <cluster> tag.
1887 - Internal links should be good now (beside of the loopback, which
1888 use the private link instead)
1889 - paths to the external world is still rather broken
1890 - the <route:multi> tag is just broken. Actually that's brain-dead.
1891 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1892 to make it less stupid
1893 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1894 * Fix a source-level compatibility glitch from 3.2: after defining
1895 MSG_USE_DEPRECATED, you can use the old name
1896 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1897 * Allow to compile from the SVN with automake 1.11
1898 * Fix some problems when using the "start_time" tag in deployment XMLs.
1899 * Fix #8569: XBT/synchro.h has redundant declarations
1900 * Fix #8563: MSG return values and exceptions
1901 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1902 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1903 Thanks to Robson Peixoto.
1904 * Fix a vicious bug in dictionaries inducing that some elements were
1905 not freed on xbt_dict_free()
1907 Portability report of this version:
1908 * Main portability targets:
1909 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1910 on (amd64/i386/ia64)
1911 - mac leopard on i386
1912 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1913 but nothing critical.
1914 * Other platforms: windows, AIX and others were not tested for this release
1916 Timing report of this version:
1917 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1918 scenario parameters). From 8h to a few seconds in desktop grid settings.
1919 * Supernovae brings about 25% speedup on i386.
1921 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1923 SimGrid (3.3.3) stable; urgency=low
1925 The "Need for Speed" release.
1927 The timings done to validate the 3.3.2 were faulty.
1928 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1930 The problem was a conversion from a manually handled vector to
1931 xbt_dynar_t on the critical path.
1932 xbt_dynar_foreach calls functions, inducing stack management crap.
1934 We inlined these functions and xbt_dynar_foreach is now breath taking.
1935 We also inlined xbt_swag_belong on the way.
1937 Here are some approximate speedup measurements (on master/slaves
1938 simulations lasting between 10s and 20s each):
1939 3.3.1 -> 3.3.2: about same performance
1940 3.3.2 -> 3.3.3: 40% speedup
1941 3.3.1 -> 3.3.3: 40% speedup
1942 3.3.1 with inline patch -> 3.3.3: 30% speedup
1944 Our reading is that the refactoring which occurred in 3.3.2 made us
1945 suffer much more from the xbt_dynar_foreach low performance, but
1946 once we solved this, this refactoring proved to be very performance
1947 effective. From the 40% speedup, somehow, 10% are due to the
1948 inlining and 30% to the refactoring.
1950 That's a pitty that gcc cannot inline functions placed in other files
1951 alone. We have to choose between:
1952 - break the encapsulation (by putting private data structures and
1953 accessors in headers files to help gcc)
1954 - live with low performance
1955 - switch to a decent compiler such as icc (not quite possible).
1957 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1959 SimGrid (3.3.2) stable; urgency=low
1961 The "Simplicity does not preceed complexity, but follows it" release.
1963 The main contributors of this release were (lexical order):
1964 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1967 * Extract the routing logic into its own object.
1968 (was dupplicated in network.c and workstation_LV07.c;
1969 Allows to implement other ways of storing that info)
1970 => kill now useless network_card concept
1971 - Use dynar to represent routes (instead of void** + int*)
1972 - kill link_set (use surf_network_model->resource_set instead)
1973 - Add a command-line option to choose the routing schema to use
1974 - Add three new models:
1975 * Floyd (shortest path computed at initialization)
1976 * Dijikstra (shortest path recomputed all the time)
1977 * Cached Dijikstra (shortest path computed on need)
1978 All these models where contributed by Silas De Munck, and are
1979 described in his ICCS09 paper.
1981 * Simplify model declaration
1982 (less redirections, less function to write when defining a model)
1983 - Factorize stuff between models:
1986 surf_model_resource_set(model)
1987 surf_model_resource_by_name(model, name)
1988 - Unify the types of models in s_surf_model_t (using an union)
1989 - Embeed fields of common_public directly into s_surf_model_t
1990 - Rename model methods:
1991 action_free ~> action_unref
1992 action_change_state ~> action_state_set
1993 action_get_state ~> action_state_get
1994 - Change model methods into functions :
1995 (model)->common_public->action_use ~> surf_action_ref
1997 * Implement a generic resource; use it as ancestor to specific ones
1998 (allows to kill duplicated code in models)
1999 Drawback: timer command don't need no name nor properties;
2000 workstation_CLM03 don't need no properties
2001 (but I guess we can live with those few bytes wasted)
2003 * Improve the action object model
2004 - implement a constructor avoiding dupplicated code about field
2005 initialization in generic_action part.
2007 * Kill the SDP model: it has an external dependency, is deprecated
2008 in flavor of modern lmm models, and didn't compile since a while
2011 * Relocation of the context module from XBT to SIMIX.
2012 (the context were decoupled from the simix processes, duplicating a lot of code)
2013 => a lot of code was factorized
2014 - less overhead is introduced during scheduling
2015 - simpler API for the context factory
2016 - the logic for process creation,destruction and manipulation was simplified
2017 * Simplification of the s_smx_process_t data structure.
2018 => accesing the simix level data associated to a process is faster now,
2019 and the code is a lot more readable.
2022 * Implement some more MPI primitives:
2023 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2024 -implementation: Bcast: flat or 2-ary tree (default),
2025 Barrier: 4-ary tree,
2027 Allreduce: Reduce then Bcast
2028 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2029 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2030 Alltoallv: flat tree, like ompi
2032 * Add support for optimized collectives (Bcast is now binomial by default)
2033 * Port smpirun and smpicc to OS X
2036 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2037 Shout out if you used it.
2040 * Display the list of still queued messages in SG mode when existing
2044 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2045 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2046 * Add xbt_str_from_file(FILE*)
2047 * Add xbt_dict_get_key achieving a linear reverse search
2048 * Remove the context module
2050 Portability report of this version:
2051 * Main portability targets:
2052 - Linux(debian)/x86/context
2053 - Linux(debian)/x86/pthreads
2054 - Linux(debian)/amd64/context
2055 - Linux(debian)/amd64/pthreads
2056 On these, we still have the eratic breakages of gras/pmm and
2057 amok/saturate_sg reported in previous version. We still think
2058 that the tests are the cause of the fault, not the tested code.
2060 - Mac OSX Leopard/x86/context
2061 Still false negative in tesh autotesting.
2062 Smpi still fails, but this time because readlink does not accept -f
2063 Everything seems to work properly beside of that.
2066 - AIX version 5.3 (only tested contexts this time)
2067 Smpi still fails there because mktemp is not installed.
2068 Everything seems to work properly beside of that.
2070 I managed to compile it for the first time, but several breakages.
2071 Won't delay the release for this exotic platform.
2073 * Windows: it's still lagging behind. If you want to help, please
2076 Timing report of this version:
2077 This version seem to be more than 5% faster than 3.3.1 (on linux
2078 64bits with contextes). The gain is less than expected, we are
2079 investigating this for next release.
2081 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2083 SimGrid (3.3.1) stable; urgency=low
2086 * Implement a --cfg-help to show existing configuration variables
2087 * Build chain do not require doxygen in maintainer mode
2090 * fix a bug on struct sizeof computation, which prevented the
2091 exchange of arrays of structs in some conditions
2092 - added a regression test about this in datadesc_usage
2093 * Allow the exchange of 0-long dynamic vectors.
2094 - for that, use -1 as indicator of dynamic size instead of 0
2095 - This implied to change any size from unsigned long to long,
2096 reducing a bit communication abilities, but I guess that with
2097 64bits being quite common, this is more than enough.
2098 - This also induce a protocol change, thus bumping network protocol
2099 version from 0 to 1 (if we have external users, we have to get
2100 clean on that point too ;)
2101 - added two regression tests about this in datadesc_usage
2102 * Be more verbose when propagating local exceptions
2103 This helps debugging.
2104 * Display the status of simulated processes when receiving SIGINT in
2108 * Allow to control the simulation from a trace file.
2109 New functions MSG_action_register() and MSG_action_trace_run()
2110 The first one allows to associate a function execution to each
2111 kind of action while the second one parses a trace file and
2112 triggers the corresponding actions within the system.
2113 For now, only a toy example is provided in examples/msg/actions
2114 * Add an exemple of process migration in examples/msg/migration
2115 * Fix a bug in task exchange which broke MSG_task_get_sender()
2116 Add a teshsuite regression test for that.
2117 [Bug: if MSG_task_get_sender() is called after sender exit,
2119 * Fix a bug which prevented suspend/resume to work properly
2120 * Display the status of simulated processes when receiving SIGINT
2121 This fixes a regression of v3.3. due to the introduction of SIMIX
2122 * Bug fixing in failure management:
2123 - trace could not start by a failure at time 0
2124 - failure during communications were not working
2127 * Add SIMIX_process_set_name() to change the name of the current
2128 process in the log messages.
2129 * Store smx_hosts in a dict since we only retrieve them by name
2130 * Move the configuration infrastructure to surf
2133 * Move the configuration infrastructure to surf
2136 * Massive internal cleanups:
2137 - Store internal structures on processes instead of hosts (allows
2138 to have more than one process per host, in addition of being more
2140 - Cleanup the initialization/finalization process
2141 - Kill a whole bunch of unneeded synchronization:
2142 processes run in exclusive manner within the simulator
2143 - Move queues from global tables to process data fields
2145 - now accept -platform and -hostfile arguments
2146 - Pass the right rank value to processes according to the hostfile
2147 * Compile the examples by default, and use them as regression tests
2148 * Implement MPI_Wtime()
2149 * Change the reference speed to a command line option
2152 * TCP_gamma can now be specified as command line option using
2153 --cfg=TCP_gamma:10000000.0
2154 * Change the --surf-path cmd line option into --cfg=path:
2157 * Also include strbuff from xbt.h public header
2158 * xbt_ex_display(): do not free the exception after displaying
2159 This allows to do more with the given exception afterward.
2160 Users should call xbt_ex_free() themselves.
2164 Portability report of this version:
2165 * Main portability targets:
2166 - Linux(debian)/x86/context
2167 - Linux(debian)/x86/pthreads
2168 - Linux(debian)/amd64/context
2169 - Linux(debian)/amd64/pthreads
2170 These targets fail about 1/10 of times on gras/pmm, but we believe
2171 that this is because of the test, not because of simgrid.
2172 amok/saturate_sg fails even more rarely, and the test may not be
2175 - Mac OSX Leopard/x86/context
2176 The test suite still spits tons of errors because some obscure
2177 force prevents us from removing the temporary directories
2178 arguing that they still contain some metadata I've never heard of.
2179 Smpi fails because seq is not installed.
2180 Everything seems to work properly beside of that.
2183 - AIX version 5.3 (both contexts and pthread)
2184 Smpi still fails there because mktemp is not installed.
2185 XML inclusions seems rosty on AIX.
2187 * Windows: it's still lagging behind. If you want to help, please
2190 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2192 SimGrid (3.3) stable; urgency=high
2196 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2197 [Malek Cherier & Mt]
2199 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2202 Until now, GRAS were using MSG as an interface to SURF. It was
2203 quite difficult because both interface have several differences
2204 (MSG channels vs GRAS sockets were the most notable point).
2206 This also opens the gate to SMPI (which should occur soon) and speed
2207 up simulations by to 40% (even if it were not the main goal).
2209 **************************************
2210 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2211 **************************************
2212 The point is that events occuring at the exact same timestamp are
2213 not scheduled in the same order with the old and new version. This
2214 may be enough to completely change the execution of simulations in
2215 some cases. Sorry for the inconvenience.
2217 * Cleanup and upgrade the XML format to push further scalability
2218 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2220 * Improve the testing infrastructure with tesh. Now a very large part of
2221 the code is tested not only by being run but also by checking that the
2222 output match an expected output [Mt].
2224 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2225 is really less memory-demanding, which should allow you to use
2226 larger files in SimGrid [AL].
2228 * Inform valgrind about our contextes, so that it becomes usable
2229 with the default (and more effecient) version of SimGrid
2230 [contributed by Sékou Diakite, many thanks]
2233 * Introduce a listener thread in charge of receiving incoming
2234 messages from the network. It allows to overlap communication and
2235 computation but most notably, it removes some stupid deadlocks due
2236 to the fact that so far, a process could not send and receive at
2237 the same time. This made most non trivial communication schema
2239 * Convert the PIDs from long int to int to match the MSG ones (and
2240 linux ones too) [Mt]
2241 * New function: gras_agent_spawn() to launch a new process on
2242 current host. Only working in simulation for now. [Mt]
2243 * New function: gras_os_hostport() returning a constant form (ie,
2244 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2247 * Make the backtrace of exceptions more human readable [Mt]
2248 * New module: xbt/str [Mt]
2249 a ton of string utility functions (split, join, printf to a newly
2250 allocated buffer, trim, etc)
2251 * New module: xbt/hash [Mt]
2252 SHA1 hashing algorithm (more to come if needed)
2253 * New module: xbt/synchro [Mt]
2254 synchronization tools (mutex and conditions) working the same way
2255 in simulation and in real life (mainly useful for GRAS, but not
2257 * New module: xbt/queue [Mt]
2258 classical producer/consumer synchronization scheme
2259 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2260 (using the classical functions will get serialized) [Mt]
2261 * Make dictionary internal table dynamic. No need to specify its size
2262 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2264 * Make sure the log channels are organized as a tree under windows
2265 (because of ANSI C compatibility issue, any channel were child of
2266 root directly) [Mt].
2269 * Cleaned many thing in surf and fixed a few bugs [AL].
2270 * Add a nice command line configuration mechanism to compose models [AL].
2271 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2272 the previous one (KCCFLN05). It relies on something that looks like
2273 a max-min sharing mechanism but cannot be written as such. A new solver
2274 was thus designed [AL].
2275 * Add a new solver to lmm. Based on Lagrange optimization and
2276 gradient-based descent, it enables to efficiently maximise systems s.a
2278 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2280 This solver enables to propose two new network models for TCP Reno and
2281 TCP Vegas based on Low's work. These models still need to be fully
2282 tested though [Pedro Velho].
2285 * Bug fix in SD_simulate. Now the time bound given as argument is
2287 * Use the new parallel task model (ptask_L07) as default.
2288 * Use the SURF command line configuration mechanism.
2289 * 0-size tasks (for synchronization) should now work.
2291 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2293 SimGrid (3.2) stable; urgency=high
2297 We still experience issues on this platform, but we believe that at
2298 least MSG is usable.
2300 GRAS API BREAKAGE (for simplification purpose, sorry):
2301 * the gras_msgtype_by_name is not used anymore. Instead of
2302 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2303 you can write (and must)
2304 gras_msg_send(toserver, "request", &request);
2305 - If you still want to pass a gras_msgtype_t to the function (to cache
2306 the type and avoid the lookup time), use the gras_msg_send_() variant.
2307 - Impacted functions:
2308 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2309 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2310 * The callbacks are now expected to return 0 when everything went well
2311 (just like the main() function)
2313 GRAS new features and improvements:
2314 * New module mechanism where user code can use per process globals [Mt]
2315 This is similar to gras_userdata_*() functions, but for libraries. It
2316 factorize some code developped over and over in the examples and AMOK.
2317 It has still to be documented and used (only amok/peermanagement is
2319 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2320 loss when they were small enough to fit into the buffer and sent quickly
2321 enough so that they can all get received in one shoot.
2322 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2323 if not found. Use the *_or_null() variant for the old semantic.
2324 * In gras_msg_handle, do not discard messages without callback.
2325 They are probably messages to be explicitly awaited later (ie, proofs of
2326 mis-synchronization in userland since they are sent before being awaited)
2327 No big deal usually.
2328 * gras_socket_meas_send/recv: semantic changed!
2329 The numerical arguments used to be (1) the total amount of data to send
2330 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2331 messages. This was need for the fool willing to send more than MAXINT
2332 bytes on quite fat pipes.
2335 * Do really rename the hostmanagement module to peermanagement. [Mt]
2336 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2337 but this is rather new and this was documented in the module
2338 documentation (poor excuses, I admit)
2339 * Bandwidth measurement semantic changed! This follows the changes to
2340 gras_socket_meas_send/recv explained above.
2343 * A sequential mode has been added to the workstations. When a workstation
2344 is in sequential mode, it can execute only one task, and the other tasks
2345 are waiting in a FIFO. [Christophe Thiery]
2348 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2349 model for SIMDAG. [Christophe Thiery]
2350 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2351 equal to 0, which caused some bad behaviors in
2352 saturated_constraint_set_update. I now use a threshold mechanism like in
2356 * When running manually src/testall, you select specific units [Mt]
2357 testall is the result of our cunit mechanism, and should replace all
2358 the scripty thingy around since bash don't run easily on billware.
2360 * A mallocator system has been added. [Christophe Thiery]
2361 Mallocators allow you to recycle your unused objects instead of freeing them
2362 and allocating new ones.
2364 Documentation update:
2365 * FAQ reworking + New FAQs:
2366 - "Valgrind spits tons of errors!" [Mt]
2367 - "How to repport bugs" [Mt]
2368 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2369 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2370 - Communication time measurement within MSG [AL]
2371 - I experience weird communication times when I change the latency [AL]
2372 * GRAS tutorial [Mt]
2374 - an introduction to the framework and to the used communication model
2375 - an initiatic tour introducing the most proheminent features:
2377 . Lesson 0: Installing GRAS
2378 . Lesson 1: Setting up your own project
2379 o Part 2: Message passing
2380 . Lesson 2: Exchanging simple messages
2381 . Lesson 3: Passing arguments to the processes (in SG)
2382 . Lesson 4: Attaching callbacks to messages
2383 . Lesson 5: Using globals in processes
2384 . Lesson 6: Logging information properly
2385 . Lesson 7: Using internal timers
2386 . Lesson 8: Handling errors through exceptions
2387 . Lesson 9: Exchanging simple data
2388 . Lesson 10: Remote Procedure Calling (RPC)
2389 . Lesson 11: Explicitely waiting for messages
2390 . Recapping of message passing features in GRAS
2391 - A HOWTO section containing:
2392 o HOWTO design a GRAS application
2393 More are due, of course. They will come latter. In the meanwhile, you can
2394 check the examples which are still here.
2396 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2398 SimGrid (3.1) stable; urgency=high
2402 There was a stack corruption somewhere, visible only when optimizing
2403 with these versions. [Vince]
2406 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2407 module that enabled to play with Directed Acyclic Graphs. It is built
2408 directly on top of SURF and provides an API rather close to the old
2409 SG. Some old codes using SG are currently under rewrite to check that
2410 all needful functions are provided. [Christophe Thiery]
2413 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2414 extension of the classical CLM03 model that gracefully handles
2415 failures. This is now the default model for MSG and GRAS. It doesn't
2416 handle parallel tasks yet though. [AL]
2417 * Bug fix: Weights were not correctly set in the network part.
2418 WARNING: This may have resulted in incorrect results with simulations
2419 where there are more than one flow on a given link. [AL]
2422 * After a (long ?) discussion on simgrid-devel, we have decided that the
2423 convention we had on units was stupid. That is why it has been decided
2424 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2425 WARNING : This means that all previous platform files will not work as
2426 such with this version! A warning is issued to ask users to update
2428 A conversion script can be found in the contrib module of the CVS, under
2429 the name contrib/platform_generation/surfxml_update.pl [MQ]
2432 * Bug fix: Processes were started in reverse order, wrt deployment file.
2433 WARNING: if your code relies on this bug, please fix it. [AL]
2434 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2435 being executed on two different locations. [AL]
2436 * Bug fix: Failures are now better supported thanks to Derrick's tests
2437 (there was many failure situations I hadn't thought of and that weren't
2438 correctly handled). [AL]
2439 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2443 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2444 Exception occurring on server-side are propagated back to client (!).
2446 API CHANGE: the callback changed their prototype. Change:
2447 int my_handler(gras_socket_t expeditor, void *payload_data) {
2449 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2450 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2452 * New! function: gras_msg_handleall to deal with all messages arriving
2453 within a given period.
2454 * New! function: gras_socket_server_range to get a server socket in a
2455 range of port numbers (ease to avoid port number conflicts) [MQ]
2456 * New! gras processes display their backtrace when they get a SIGUSR1
2457 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2458 Sweet to debug RL processes [MQ]
2462 - Do not force experiment sizes to be expressed in kb, or it becomes
2463 impossible to measure the latency this way (needs one byte-long tests)
2464 WARNING: this changes the amok_bw_* function semantic. [MQ]
2465 - Implements the link saturation stuff. [MQ]
2466 * Peer management module:
2467 New! module factorizing code that we wrote over and over [MQ].
2470 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2471 - Test units are placed directly into the library code, they get extracted
2472 automatically and placed into the src/testall binary.
2473 - Convert most of the XBT tests to this system.
2474 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2475 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2476 live in the same C code. This required to change a little bit the API
2477 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2478 * New module: sparse graph structure with basic algorithms (this is work
2479 in progress and the API is not considered to be frozen yet). [AL]
2480 * Display more information on backtraces: source line & function names are
2481 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2482 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2483 * Reworked a little bit some #include statements to load only required
2484 headers. Some user code that relied on SimGrid to include stdlib or
2485 stdio may need to include it by themselves. [AL]
2486 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2488 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2489 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2490 but I guess that almost nobody used those parts. [MQ]
2492 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2494 SimGrid (3.0.1) stable; urgency=low
2497 * Unfortunately, I had missed 5 misnamed functions:
2498 xbt_fifo_item_t xbt_fifo_newitem(void);
2499 void xbt_fifo_freeitem(xbt_fifo_item_t);
2500 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2501 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2502 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2503 They're now deprecated. Please use their new versions:
2504 xbt_fifo_item_t xbt_fifo_new_item(void);
2505 void xbt_fifo_free_item(xbt_fifo_item_t);
2506 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2507 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2508 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2510 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2511 * Documentation: xbt_log module unmercifully reworked [MQ]
2512 * Bugfix: there was a problem with the ending of contexts with
2513 the pthread backend. It caused some weird deadlock or behavior
2514 depending on the pthread implementation. [AL]
2515 * Bugfix: get the exceptions raised in the simulator repport where
2516 and why they come from when they are not catched in time [AL, MQ]
2519 * Bugfix: Do repport the error when two non-connected hosts try to
2520 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2523 * Add additionnal checkings on communications. Assert that two
2524 communicating hosts are connected by a set of links... [AL]
2527 * Add additionnal checkings on channel values in communication [AL]
2528 * New: MSG_task_get_source to see on which host a task was generated [HC]
2529 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2530 the number of tasks waiting to be received on channel and sent
2532 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2533 waits for the first task coming from a given host.. [AL]
2535 GRAS new functionnalities: [MQ]
2536 * Enhance the parsing macro to allow the size of multidimentional objects
2537 to be given thru annotations.
2538 * New example (and documentation): Matrix Multiplication a la RPC
2539 (as when I was young!) and fix a bunch of bugs found on the way.
2541 GRAS performance improvements: [MQ]
2543 * Reduce the amount of cbps creation/destruction by making it static to
2544 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2547 * Change libdata to a set so that we can search for stuff by ID (and thus
2548 reduce the insane amount of dict lookups)
2551 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2552 RL communication on top of SG since it's so uneffective.
2553 It may also be used for inter-thread communication in RL, one day.
2554 * Use gras_datadesc_copy() to exchange messages on top of SG
2556 - improve message exchange performance on top of SG
2557 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2558 * Don't exchange on the network the size of the used part of buffer,
2559 instead, specify the possible buffer size to read().
2561 - reduces the amount of read/write calls (one pair per exchange)
2562 - reduces the amount of exchanged data (the size)
2563 - allows to retrieve all arrived data on receiver side, if we don't need
2564 it right now (subsequent read will peek the buffer)
2565 - allows the receiver to proceed with the begining of the stream before
2566 everything is arrived
2567 - make it possible to build an iov transport (using readv/writev)
2569 - take care of the data with non-stable storage (like stacked data),
2571 * If possible, TCP send uses vector I/O (when writev() is here)
2572 - Don't use it for receive since we send data sizes and data on the
2573 same stream, so we wouldn't be able to chain large amount of chunks
2574 before having to flush the stuff to read the size.
2575 * Rework the transport plugin mechanism to simplify it and reduce the
2576 amount of pointer dereferencement when searching for the right function
2579 * I guess that now, we do almost as few system calls as possible while
2580 doing as few data copy as possible.
2582 To improve it further, we could try to send all the sizes first and then
2583 all the data (to use iov on receiving size), but it's only a partial
2584 solution: when you have 2 dimensional data, the sizes of the second
2585 dimension is data of the first dimension, so you need 3 streams.
2587 I'm not sure the potential performance gains justify the coding burden.
2589 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2591 SimGrid (3.00) stable; urgency=high
2594 * New! Give the possibility to hijack the surf parser and thus bypass
2595 MSG_create_environment and MSG_launch_application. Have a look at
2596 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2598 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2600 SimGrid (2.96) unstable; urgency=low
2605 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2606 This deprecates the xbt_error_t mechanisms.
2607 It modifies (simplifies) all XBT and GRAS API.
2608 MSG API keeps unchanged (exceptions raised by XBT are catched from
2609 within MSG and masked with existing error handling facilities)
2612 * New! Add a FATPIPE model. [AL]
2613 * New! Add a parallel task model. [AL]
2614 * New! Add automatically a loopback interface (in the default
2615 network model) if none was precised.
2618 * Bugfix: MSG_process_resume now works with the current running process.
2620 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2621 * Modification of MSG_task_get_compute_duration. Once a task has been
2622 processed, the value returned by this function is now equal to 0. [AL]
2623 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2624 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2625 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2626 of a m_task. For now, MSG_task_cancel only works with computation
2628 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2629 of the processor (in Mflop/s) regardless of the current load on the
2631 * API Change: use proper naming convention for MSG_getClock and
2632 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2634 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2635 This function changes the priority of a computation task. This priority
2636 doesn't affect the transfer rate. A priority of 2 will make a task
2637 receive two times more cpu power than the other ones. This function
2638 has been added to suit the needs of Nguyen The Loc and hasn't been that
2639 much tested yet. So if it fails, please report it and send me your code.
2641 * API Change: removed all functions and types that were marked "deprecated"
2642 since many months. Renamed MSG_global_init_args to MSG_global_init.
2644 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2646 SimGrid (2.95) unstable; urgency=low
2649 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2650 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2651 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2654 * Bugfix: Allow absolute paths for platform description files [MQ]
2655 * Bugfix: do free the variables after use. Leads to drastic performance
2657 * Implement max_duration (ie, timeouts) on resources [AL]
2660 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2662 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2663 the needed support (timeouts) [AL]
2666 * Implement measurement sockets. You can now get the bandwidth between two
2667 hosts thanks to AMOK (see below). [MQ]
2668 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2669 dynar over the network (yeah) [MQ]
2670 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2671 * simple/static token-ring example. [Alexandre Colucci and MQ]
2672 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2673 in place before [MQ]
2676 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2677 type parsing. "short" and "long" modifiers were also ignored; other
2678 modifier (such as reference level) are still ignored. [MQ]
2679 * Embeed the buffer size within the buffer itself on SG. [MQ]
2680 That way, send() are atomic and cannot get intermixed anymore (at least
2681 the ones which are less than 100k; bigger messages still have the issue)
2682 * Array size pushed by the field, not by the field type (or each
2683 and every long int will push stuff to the cbps) [MQ]
2684 * use select() to sleep since it allows to portably sleep less than one
2687 GRAS (minor cleanups)
2688 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2691 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2692 * type category 'ignored' killed as it was never used and were difficult
2694 * whether a type can cycle or not is now a flag, leaving room for more
2695 flags (as "ignored", if we feel the need one day ;)
2696 * Rename raw sockets to measurement sockets since "raw" has another
2697 meaning in networking community.
2700 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2701 and offers features not belonging to GRAS but that most applications
2702 need. One day, it may be a set of plugins loadable at runtime.
2703 * New module: bandwidth
2704 bandwidth measurement between arbitrary nodes running this module. [MQ]
2706 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2708 SimGrid (2.94) unstable; urgency=low
2710 The first beta release of SimGrid 3 !
2714 * Update the main page and the FAQ. Adding references to gforge.
2717 * Add a gras_os_getpid function.
2720 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2721 * Extend the logs so that they also print PID, hostname, date, ... if
2723 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2724 and kill the old version which were in testsuite/
2725 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2729 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2730 for users. Instead, each programming environment (GRAS, MSG,...) use it
2731 when needed to provide such a feature to users.
2732 Don't shortcut the mechanism or you will also shortcut the virtualization
2733 you need on the simulator.
2737 * Cleanups in configury with regard to compile optimization/warning flags.
2738 Also add -fno-loop-optimize to any powerpc since it's the optimization
2739 killing gcc (< 3.4.0).
2740 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2741 needed by MSG examples complications
2742 * Borrow configury beautifications from PHP
2745 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2746 warning (thanks loris for stumbling into this one).
2747 * Bugfix: stop loading private headers (gras_config.h) from the public
2751 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2752 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2753 RL would imply a DNS resolver, which is *hard* to do in a portable way
2754 (and therefore delayed).
2755 * Implement a real timer mechanism and use it in timing macros. This allows
2756 to avoid rounding errors and get a 0.000005 sec precision in timing
2757 macros. While I was at it, various cleanups:
2758 - allow to declare more than one timed section per file (fix a stupid bug)
2759 - move some private declaration to the right place
2760 - merge conditional execution and timing macros into emulation module
2761 - document the module
2762 - make sure the module cleanups its mess on gras_exit
2763 * Documentation improvements:
2764 - (new) how to compile applications using GRAS
2765 - (new) emulation support (timing macros)
2767 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2769 SimGrid (2.93) unstable; urgency=low
2771 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2774 - Use Paje properly where used. Still to be sanitized properly.
2775 - Portability fix: Add an implementation of the contexts using pthread
2779 - Add xbt_procname(): returns the name of the current process.
2780 Use it to show the current process's name in all logging.
2782 - fix detection of older flex version and the reaction, since we do
2783 depend on modern ones (we use lex_destroy)
2784 - Better separation of SG and RL in the libs: remove all simulation code
2785 from libgras. As a result, this lib is now only 200k when stripped.
2786 Some of the xbt modules may also be duplicated (two sets and such) and
2787 should be cleaned/killed before SG3.
2788 - Insist on using xlC on AIX because of weird problems involving gcc there.
2789 - Cleanup the make remote stuff. This is now done by scripts
2790 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2791 mainly for our private use, but we're working on changing them to user
2794 - Bugfix: flush the socket on close only if there is some *output*.
2795 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2797 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2800 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2802 SimGrid (2.92) unstable; urgency=low
2804 Alpha 3 on the path to SimGrid 3
2808 - New! First try of benchmarking macros.
2809 - New! First try so that gras_stub_generator generate deployment and
2810 remote compilation helpers.
2812 - Bugfix: Initialization fix in msg_test.
2816 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2818 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2820 - API Change: gras_os_sleep to take the amount of seconds as a double.
2821 Accepting an int was error prone since it was the only location where
2822 seconds were coded as such. It leaded to damn rounding errors.
2823 - Bugfix: Hard to belive that timers ever worked before this.
2825 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2827 SimGrid (2.91) unstable; urgency=low
2829 Alpha 2 on the path to SimGrid 3
2833 - Bug fix in the lmm_solver.
2835 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2836 through the function MSG_paje_output.
2837 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2838 - It is possible to bound the rate of a communication in MSG with
2839 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2840 to write it in the changelog).
2841 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2846 - Various cleanups to the autotools stuff
2847 - Begin to move Gras examples to examples/gras/
2848 - Let make distcheck work again (yeah!)
2850 - documentation overhauled using doxygen.
2851 gtk-doc-tools is dead in SimGrid now.
2852 - Automatically extract all existing logging categories, and add the list
2853 to the documentation (long standing one, to say the less)
2855 - Cleanup the known architecture table. Reorder the entries to group what
2856 should be, and use a more consistent naming scheme.
2857 (some of the test dataset are still to be regenerated)
2858 - New! Allow library to register globals on each process just as userdata
2860 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2861 lookup time (for now).
2862 Use it in msg and trp.
2863 This cleans a lot the internals and helps enforcing privacy of the
2864 headers between the gras components.
2865 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2866 - Bugfix: gras_os_time was delirious in RL.
2867 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2869 - Reenable GRAS now that it works.
2871 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2873 SimGrid (2.90) unstable; urgency=low
2875 Alpha 1 on the path to SimGrid 3
2877 * It is a long time since the last release of SimGrid. I'm sorry about
2878 that but as I had told you, I was rewriting a lot of things. I apologize
2879 to those who had been reporting bugs to me and that I had not answered.
2880 If your bug is still in the new version, please tell me. Here is a
2881 summary of the main changes.
2883 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2884 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2885 portability, speed, and a lot more but you'll figure it out later.
2886 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2887 I wanted to release the new MSG as soon as possible and I have
2888 broken GRAS, which is the reason why, for now, only MSG is fully
2889 functional. A laconic description of these projects is available
2890 in the documentation.
2892 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2893 kernel optimized for our needs (called SURF but only the developers
2894 should use it). Hence, MSG is now roughly 30 times faster and I think
2895 that by rewriting a little bit MSG, I could event speed it up a little
2896 bit more. Beside the gain in speed, it is also much easier to encode a
2897 new platform model with SURF than it was with SG. More to come...
2899 * REVOLUTION 3: I have tried to change a little as possible the API of
2900 MSG but a few things really had to disappear. The main differences
2901 with the previous version are :
2902 1) no more m_links_t and the corresponding functions. Platforms are
2903 directly read from a XML description and cannot be hard-coded
2904 anymore. The same format is used for application deployment
2905 description. The new format is described in the documentation.
2906 Have a look in tools/platform_generation. There is a tiny script
2907 that converts from the old platform format to the new one. Concerning
2908 the application deployment format, parsing the old one is tricky.
2909 I think most of you should however be able to convert your files. If
2910 it is really an issue, I can write a C code that does the conversion.
2912 2) the toolbox tbx does not exist anymore. We now have a library
2913 with much more data-structures but without the hash-tables (we have
2914 dictionaries that are much faster).
2916 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2918 *****************************************************************************
2919 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2920 * don't feel like converting the oldies. (Mt) *
2921 *****************************************************************************
2924 Version 2.90: "the long awaited one"
2925 - Finished rewriting and debugging MSG. Rewrote the documentation.
2926 - disable GRAS for now since it needs to be ported to the newest SG
2929 - Finish the port to windows (using mingw32 for cross-compile)
2932 - Main loop and datastructures of SURF. A cpu resource object is
2933 functional. Surf can thus be used to create cpu's with variable
2934 performance on which you can execute some actions.
2936 2004-11-15 Martin Quinson
2937 - Port to ARM. Simply added the alignment and size descriptions. Should
2938 work, but the ARM machines are so slow that I didn't had the opportunity
2939 to 'make check' over there yet.
2941 2004-11-15 Arnaud Legrand
2942 - Trace manager now written. It uses a heap structure and is therefore
2943 expected to be efficient. It may however be speeded up (particularly
2944 when many events occur at the same date) by using red and black
2945 trees. One day maybe...
2946 - Max-min linear system solver written. It uses a sparse matrix
2947 structure taking advantage of its expected use. Most operations are
2948 O(1) and free/calloc are called as few as possible. The computation of
2949 the minimum could however be improved by using a red and black tree
2952 2004-11-03 Arnaud Legrand
2953 - Rename every gras_* function that was in xbt/ to its xbt_
2955 - Add a heap and a doubly-linked list to xbt
2956 - Added a dichotomy to the dictionaries. make check works as well before
2957 so I assume that the patch is correct. I do not know however if things
2958 run effectively faster than before now. :)
2960 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2961 SimGrid, and the version number is bumped to 2.x
2963 2004-10-29 Martin Quinson
2964 - Introduction of the remote errors.
2965 They are the result of a RMI/RPC on the remote machine.
2966 ErrCodes being scalar values, you can't get the host on which those
2967 errors did happen. Extending the error mechanism as in Gnome is possible.
2968 No idea yet whether it is a good idea.
2970 2004-10-28 Martin Quinson
2971 - Interface revolution: the Starred Structure Eradication.
2972 I used to do typedef struct {} toto_t; and then handle *toto_t.
2973 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2974 - ???_t is a valid type (builded with typedef)
2975 - s_toto_t is a structure (access to fields with .)
2976 - s_toto is a structure needing 'struct' keyword to be used
2977 - e_toto_t is an enum
2978 - toto_t is an 'object' (struct*)
2980 typedef struct s_toto {} s_toto_t, *toto_t;
2981 typedef enum {} e_toto_t;
2982 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2983 s_toto_t) is private.
2985 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2986 so that it can set the variable to NULL. It was so for dicts and sets,
2987 it changed for dynars.
2989 - Fix a bunch of memleaks in dict_remove
2990 - Fix a bug in sg/server_socket opening: it failed all the time.
2992 2004-10-07 Martin Quinson
2993 - Speed up dynar lookup operation a bit.
2995 gras_dynar_get is dead.
2997 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2998 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2999 (which gives you the address of the stored data).
3001 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3002 copy of the data using an affectation to do the job and not a memcpy.
3004 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3005 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3007 It does not really speedup the dynar test because they are
3008 setting elements all the time (and look them seldom). But the dict does
3009 far more lookup than setting.
3011 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3013 2004-10-05 Martin Quinson
3014 - Allow to (en/dis)able the cycle detection at run time.
3016 Whether we should check for cycle or not is now a property of each
3017 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3018 datadesc_cycle_unset allow to remove this property when previously set.
3020 Note that the cycle detection is off by default since it impacts the
3021 performance. Watch the data you feed GRAS with ;)
3023 This property is hereditary. Any element embedded in a structure having it
3024 set have it set for the time of this data exchange.
3026 You should set it both on sender and receiver side. If you don't set it on
3027 sender side, it will enter an endless loop. If you forget on receiver
3028 side, the cycles won't be recreated after communication.
3030 - Header reorganization.
3031 Kill gras_private.h, each submodule must load the headers it needs.
3033 2004-10-04 Martin Quinson
3034 - Interface revolution: do not try to survive to malloc failure.
3036 Now, gras_malloc and friends call gras_abort() on failure.
3037 As a conclusion, malloc_error is not a valid error anymore, and all
3038 functions for which it was the only gras_error_t return value are
3039 changed. They now return void, or there result directly.
3040 This simplify the API a lot.
3042 2004-09-29 Martin Quinson
3043 - Re-enable raw sockets.
3044 Created by gras_socket_{client,server}_ext;
3045 Used with gras_raw_{send,recv}
3048 It should allow to kill the last bits of gras first version soon.
3050 This is not completely satisfactory yet (duplicate code with
3051 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3054 - Simplify transport plugin (internal) interface by not passing any
3055 argument to _server and _client, but embedding them in the socket
3058 2004-09-28 Martin Quinson
3059 - Finish the port to AIX.
3060 autoconf was my problem (segfault within the malloc replacement
3061 function. No idea why)
3063 2004-09-16 Martin Quinson
3064 - Fix some size_t madness on 64bit architectures.
3066 2004-09-08 Martin Quinson
3067 - Reduce the number of system headers loaded, overload some more system
3068 calls (such as malloc to cast the result of the system one, and work
3070 - Fix and reintroduce the config support
3072 2004-09-07 Martin Quinson
3073 - Source code reorganization to allow Arnaud to surf all over there.
3074 - Allow to document the logging categories.
3075 - Remove all uppercase from logging categories and useless cleanup in names.
3077 2004-08-18 Martin Quinson
3078 Version 0.6.2 (protocol not changed; API changed)
3079 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3080 gras_error_t), and NULL when not found. Functions expecting a msgtype
3081 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3082 hopefully usefull message.
3083 - Portability to prehistoric sparcs again
3085 2004-08-17 Martin Quinson
3086 Version 0.6.1 (protocol not changed; ABI not changed)
3087 - prealloc some buffers to speed things up
3089 2004-08-11 Martin Quinson
3090 Version 0.6 (protocol not changed; ABI expended)
3091 - The parsing macro can deal with the references, provided that you add
3092 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3094 2004-08-09 Martin Quinson
3095 Version 0.5 (protocol not changed; ABI changed)
3096 - Allow to off turn the cycle detection code in data exchange at
3097 compilation time. It should be at run time, but I'm short of time (and
3098 the config stuff is still broken). That way, we keep dict out of the
3099 critical path, which is good because the performance is poor:
3100 - search not dichotomial yet
3101 - dynar give no way to access their content and memcpy everytime
3102 - In composed data description (struct, ref and so on), stop foolness of
3103 keeping the subtype's ID, but store the type itself. This keeps sets out
3104 of the critical path, which is good since they rely on dynar and
3105 dictionnaries. The only loose of that is that we cannot detect the
3106 redeclaration of a structure/union with another content (but I'm not sure
3107 the code detected well this error before anyway). We still can detect
3108 the redefinition discrepancy for the other types.
3109 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3110 it breaks the code because of type-punning used all over the place).
3111 This breaks on all non-gcc architectures (for now).
3113 All those changes (plus the buffer of last time) allow me to gain 2 order
3114 of magnitude on cruel tests consisting of 800000 array of integers on two
3115 level of a hierarchical structure (200 secondes -> 4 secondes)
3118 - the selector of reference must now return the type it points to, not
3119 the ID of this type.
3121 2004-08-06 Martin Quinson
3122 Version 0.4 (protocol changed; ABI not changed)
3123 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3124 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3125 - Implement a buffer transport, and use it by default (it relies on tcp in
3126 real life and on sg in simulation).
3127 That's a bit hackish since I had a new field to the structure to store
3128 its data without interfering with the subtype ones. Inheritance
3129 is tricky in C. And that's a kind of reverse inheritance with one class
3130 derivating two classes. Or maybe a game with java interfaces. Anyway,
3131 that's damn hard in C (at least).
3132 Moreover, I got tired while trying to ensure plugin separation and
3133 genericity in SG mode. MSG wants me to do weird things, so let's go for
3134 cruel hacks (temporarily of course ;).
3135 See comment in transport_private.h:71
3136 - do not include all the _interface headers in private but in the files
3137 which really need them (to cut the compilation time when they are
3140 2004-07-26 Martin Quinson
3141 Version 0.3 (protocol not changed; ABI changed)
3142 - Major overhault of the datadesc interface to simplify it:
3143 - shorted the function names:
3144 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3145 - add a trivial way to push/pop integers into the cbps without malloc.
3146 This allows to make really generic sub_type description, which simply
3147 pop their size of the stack.
3148 - add a function gras_datadesc_ref_pop_arr() which does what users want
3149 most of the time: Declare a dynamic array (which pops its size of the
3150 stack) and declare a reference to it. Poor name, but anyway.
3151 - kill the post-send callback, add a post-receive one
3153 2004-07-23 Martin Quinson
3154 Version 0.2 (protocol changed; ABI changed)
3155 - add some testing for cpbs in the test cases, and fix some more bugs.
3156 This invalidate again the little64 data file, since I cannot regenerate
3158 - remove an awfull optimization in the logging stuff, allowing me to:
3159 - understand it again
3160 - learn gcc how to check that the argument match the provided format
3161 - fix all errors revealed by gcc after that
3162 - internal keys of dict are not \0 terminated. Deal with it properly in
3163 loggings instead of segfaulting when the user want to see the logs :-/
3165 2004-07-22 Martin Quinson
3166 - Fix some stupid bug preventing cbps (callback postit) from working
3168 2004-07-21 Martin Quinson
3169 - Some documentation cleanups
3170 - remove the useless last argument of msgtype_declare
3171 - rename the Virtu functions to fit into the 'os' namespace
3172 - move headers src/include -> src/include/gras/ and stop fooling with
3173 gras -> . symbolic link
3174 - make distcheck is now successful
3176 2004-07-19 Martin Quinson
3178 - Build shared library also
3179 - Install html doc to the right location
3180 - stop removing maintainer files in make clean
3181 - build tests only on make check
3183 2004-07-13 Martin Quinson
3185 - No major issue in previous version => change versioning schema
3186 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3187 corresponding dataset.
3189 2004-07-11 Martin Quinson
3191 - Get it working with any kind of structure (we can compute the padding
3192 bytes remotely for all the architectures I have access to)
3193 - Implement the structure parsing macro (still not quite robust/complete)
3194 - Improvement to the remote testing toysuite
3196 2004-07-10 Martin Quinson
3197 [autoconf mechanism]
3198 - get ride of a bunch of deprecated macros
3199 - actually run the test for two-compliment, not only compile it :-/
3200 - test whether the structures get packed (and bail out if yes. Damn.
3201 Alignment is a serious matter)
3202 - test whether the structures get compacted (but respecting the alignment
3203 constraints of each types)
3204 - test whether the array fields of structures can straddle alignment boundaries
3206 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3207 (took me 2 hours to find that bug, looking at the wrong place)
3208 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3209 before sending/receiving and are used to compute the offsets of fields
3210 - Given that padding size depend even on compiler options, keep track of
3211 alignment and aligned_size only for the current architecture. Not a big
3212 deal since we send structure fields one after the other (seems
3214 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3215 program, let it work on linux/gcc/little32. portability todo.
3217 2004-07-08 Martin Quinson
3218 - import and improve remote compilation support from FAST
3219 - make sure make check works on half a dozen of machines out there
3221 2004-07-07 Martin Quinson
3222 Let's say it's version 0.0.3 ;)
3223 - Implement conversions (yuhu!)
3224 - Let it work on solaris (beside conversion, of course)
3225 - Stupid me, using rand() to generate the conversion datatests in not wise.
3227 2004-07-06 Martin Quinson
3228 - Let make dist work, since I'm gonna need it to compile on remote hosts
3229 - Let Tests/datadesc_usage write the architecture on which the file was
3230 generated as first byte.
3231 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3232 architecture descriptions.
3233 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3234 on those architectures.
3235 - Optimization: send/recv array of scalar in one shoot
3237 2004-07-05 Martin Quinson
3238 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3240 - Plug a whole bunch of memleaks
3241 - each process now have to call gras_{init,exit}. One day, their log
3242 settings will be separated
3243 - Continue the code factorisation between SG, RL and common in Transport.
3245 2004-07-04 Martin Quinson
3247 - Redistribution between SG and RL.
3248 We wanna have to accept in SG, so move accepted related parts of RL in
3249 the common part. (more precisely, the dynar of all known sockets is no
3250 more a static in transport.c, but part of the process_data)
3252 [gras_stub_generator]
3253 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3255 2004-07-03 Martin Quinson
3256 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3257 them all up in one shot)
3259 - Fix the ugly case of reference to dynamic array.
3260 - New (semi-public) function gras_datadesc_size to allow the messaging
3261 layer to malloc the needed space for the buffer.
3263 - gras_socket_close now expect the socket to close (and not its address to
3264 put NULL in it after it). This is because the socket passed to handlers
3265 is one of their argument (=> not writable).
3267 - propagate the interface cleanup from last week in datadesc, ie remove a
3268 superfluous level of indirection. User pass adress of variable
3269 containing data (both when sending and receiving), and not of a variable
3270 being a pointer to the data. Let's say that I like it better ;)
3271 The price for that is constructs like "int msg=*(int*)payload" in
3272 handlers, but it's a fine price, IMHO.
3274 - Let it work in RL (yuhu)
3276 2004-06-21 Martin Quinson
3278 - porting SG plugin and SG select to new standards (works almost).
3279 - plug memleaks and fix bugs around.
3282 - cleanup the prototype of data recv and force users to specify when they
3283 want to handle references to objects. Test case working even for cycles.
3284 - plug memleaks. Valgrind is perfectly ok with this.
3286 2004-06-12 Martin Quinson
3288 - cleanup the separation between plugin and main code in plugin creation
3290 2004-06-11 Martin Quinson
3292 - Reput hook for raw sockets, needed for BW experiments
3293 - kill a few lines of dead code
3294 [Data description] Interface cleanup
3295 - gras_datadesc_by_name returns the searched type or NULL.
3296 That way, no variable is needed to use a type desc once, which makes
3298 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3299 two parameters were strings (field name, type name), leading to
3301 [Dicos] Interface cleanup
3302 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3303 This is consistant with the dynar API.
3305 2004-04-21 Martin Quinson
3307 - Porting to new standards.
3309 - interface cleanup.
3310 There is no bag anymore, no need to take extra provision to mask the
3311 pointers behind "ID".
3312 Better splitup of functions between files create/exchange/convert.
3313 This is still a bit artificial since convert and receive are so
3314 interleaved, but anyway.
3316 - add a queued message list to procdata (the ones not matching criteria
3318 - factorize some more code between SG and RL wrt procdata
3320 - use gras_exit in example to track memleaks
3321 - get rid of gs_example now that GS is properly integrated into gras
3322 - update run_test to integrate the latest tests (datadesc)
3324 - rename WARNINGn macros to WARNn since it prooved error-prone
3326 2004-04-19 Martin Quinson
3328 - register init/exit functions within gras module mechanism
3329 - send/receive function.
3330 Convertion is not implemented, but short-cutted if not needed.
3331 struct/array elements are sent one by one (instead of block-wise), but
3332 nobody really cares (yet). Get a prototype before optimizing.
3333 - tests (using a file socket) for DD send/receive on:
3334 - base types: int, float
3335 - array: fixed size, string (ie ref to dynamic string)
3336 - structure: homogeneous, heterogeneous
3337 - chained list, graph with cycle
3338 Believe it or not, valgrind is not too unhappy with the results. The
3339 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3343 - Bugfix when using a filename explicitely (instead of '-')
3345 2004-04-09 Martin Quinson
3347 - factorize more code between RL and SG in socket creation
3348 - Complete the implementation and tests of:
3350 o file (only in RL, and mainly for debugging)
3352 I lost 3 days to design a portable address resolver, and then decided
3353 that the prototype mainly have to run on my box.
3354 Addressing portability too early may be like optimizing too early :-/
3356 - use gras_init in the Tests instead of the crappy parse_log_opt
3357 (the latter function is removed)
3358 [Conditional execution]
3359 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3360 [Code reorganisation]
3361 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3362 Build examples against the RL library, since there is no way to disable
3363 its creation for now.
3365 For information, the beginning of coding on GRAS was back in june
3366 2003. I guess that every line has been rewritten at least twice since