1 SimGrid (3.17) UNRELEASED (release target: September 22 2017)
4 - start and forget about isend comms with detach()
6 SimGrid (3.16) Released June 22. 2017.
8 The Blooming Spring Release: developments are budding.
11 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
12 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
13 - Java JAR file should be finally fully working on Mac OSX too.
15 XML platforms: Switch to platform v4.1 format.
16 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
17 - <zone> can be used as a synonym for the now deprecated <as>
18 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
19 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
20 - <actor> can be used as a synonym for the now deprecated <process>
21 - state_file and avail_file periodicity is now easier to express
22 (check the documentation) the old behavior should still work.
23 * Storage: not backward compatible, but it seems that nobody ever used it
24 - Remove attribute 'content_type' of <storage_type>: was never used
25 - Make attribute 'model' of <storage_type> optional: for future usage
26 - Remove Bconnection model property: was never (in)validated. Replaced by
27 the maximum of the read and write bandwidth as a resource constraint.
30 - New and Backwards Compatibility break:
31 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
32 When one wants to get the list of tasks whose states have changed during a
33 simulation round, s/he has to allocate and free a dynar and use it as
34 argument to this function. The former SD_simulate (double how_long)
38 - Allow multicore VMs, with the correct sharing (unless you overcommit)
39 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
42 - New function to update the consumption of all hosts at once.
43 - Fix the model for multi-core hosts, linear in the amount of busy
44 cores with an abnormality for pIdle (see paper).
45 BUG: Still not satisfactory for ptask on mono-cores.
48 - The netzone are now available from the MSG API.
49 The old names still work, but are now deprecated.
52 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
53 - New option: smpi/keep-temps to not cleanup temp files
54 - New option : smpi/shared-malloc-blocksize . Relevant only when global shared
55 mallocs mode is used, allows to change the size of the fake file used
56 (default 1MB), to potentially limit the number of mappings for large runs.
57 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
58 - Fortran ifort and flang compilers support
59 - New RMA calls supported (experimental) :
60 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
61 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
62 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
63 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
64 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
65 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
68 - New: this_actor::isMaestro() returns whether we are in kernel mode.
69 - Behind the scene, ActivityImpl are now automatically refcounted.
70 Removing this nasty bottleneck will greatly simplify our internals.
73 - Replay: New function xbt_replay_action_get():
74 Retrieve the function previously associated to an event type.
75 - DROPPED FUNCTION: xbt_str_varsubst()
76 - DROPPED MODULE: strbuff. We don't need it anymore.
77 - DROPPED MODULE: matrix. We don't need it anymore.
78 - DROPPED MODULE: lib. We don't need it anymore.
80 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
82 SimGrid (3.15) Released March 22. 2017
84 The Spring Release: continuous integration servers become green
85 We fixed even the transient bugs on all target architectures:
86 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
87 Windows, FreeBSD, NetBSD.
89 - Rename NetCards to NetPoints.
90 This was intended to help NS3 users, but that's not a netcard.
91 That's a point in the routing algorithm, let's avoid wrong simplifications.
94 - New: MSG_process_yield(). Stop and yield to other processes.
95 - New: MSG_process_daemon(). Daemon processes are automatically killed
96 when the last non-daemon process terminates
97 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
98 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
99 to make room for the upcoming network energy plugin.
100 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
103 - Ensure that an actor can kill itself with Process::exit()
104 - Kill the obscure NativeException. Nobody want to survive the issues
105 it denotes, so use JniException that is a RuntimeException (not to
106 be caught explicitly).
107 - Partial bug fix in initialization. SimGrid flags on command line were
108 consumed at C level but stayed in the original Java String[] args.
109 This could mess users' args[i] if SG flags were not put at the end of
111 The SimGrid flags are now removed from the Java arguments. However,
112 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
113 if args.length is greater than the number of YOUR OWN ARGUMENTS.
114 It might be if you have --log or --cfg flags in the command line.
115 - Fix numerous memleaks all around the place. In particular, around VMs.
119 - s4u::onPlatformCreated: right before the simulation starts
120 - s4u::onSimulationEnd: right after the main simulation loop
121 - s4u::onTimeAdvance: right after a clock change (time is discrete)
122 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
123 event from the availability_file changes the avail speed.
124 - Links are now usable from s4u
125 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
126 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
127 - New examples: The conversion of MSG examples to S4U has begun
128 - Actors: create, kill, migration, and suspend
129 - Applications: master-worker and token-ring
130 - Action replay: communications and storage
131 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
134 - Backwards Compatibility breaks
135 - The SD_route_get_list and SD_route_get_size functions have been removed.
136 They are replaced by the unique
137 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
138 The route from src to dst is built in the links dynar whose size can
139 be obtained with xbt_dynar_length.
140 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
141 sg_host_route_bandwidth and sg_host_route_latency.
142 Macros ensure the backwards compatibility, but you should fix your code
145 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
146 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
147 - Bump our claim of support from MPI 1.1 to MPI 2.2.
148 We don't support 100% of it, but it should be enough. Contact us if not.
149 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
150 naive and probably inaccurate)
151 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
152 mapped onto a unique small file using some system magic.
153 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
154 - Fortran: cleanups, fixes, support of user-added operations
155 - RMA: MPI_Accumulate are applied in correct order.
156 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
157 - Better support for MPI_IN_PLACE
158 - Support for MPI_Win attrs and keyvals.
159 - Support MPI_Comm_group_create, MPI_Type_size_x
160 - MPI ops should be properly applied to all allowed types
161 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
162 algorithm by a non failing one from mpich.
165 - Kill the fifo data container: we don't use it anymore.
168 - A <cluster> can now be created with different speed values to represent pstates.
169 see examples/platforms/energy_cluster.xml for an example.
171 SimGrid (3.14.159) Released December 28. 2016
173 The previous release was only a crude approximation of a Pi release;
174 this one should be a bit better.
176 - Revert a last minute change that broke on Mac OSX.
177 - Fix the changelog and NEWS files.
178 - Improve the documentation, in particular of the routing module.
179 - Integrate some patches flying around in Debian.
181 SimGrid (3.14) Released December 24. 2016
184 * The whole documentation was reworked and reorganized. There is
185 still much room for improvement here, but we're on it.
188 * We now need python3 for our testing infrastructure.
189 * Model checking needs libevent
192 * Now works on FreeBSD
195 * Privatization now works on FreeBSD
197 * Privatization is now activated by default in smpirun
198 You can turn it off with -no-privatize if something goes wrong.
200 * Call-location tracing for SMPI.
201 You can add the exact location (filename / linenumber) of an MPI call to
202 your trace files and slow down or speed up the simulation between two
203 consecutive calls by using an adjustment file (see the documentation).
205 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
206 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
207 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
208 values were ignored in some cases. The timings of these functions can now
209 be significantly different.
211 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
212 smpi/running-power is renamed to smpi/host-speed
214 * smpi/grow-injected-times option to enable or disable multiplication of the
215 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
216 make simulation less precise (but also much faster).
218 * smpirun script should be (much) faster for large deployments.
220 * SMPI tracing : fixed issue with poor matching of send/receives.
222 * Replay : Fix broken waitall
224 New functions and features
225 * MSG_parallel_task_execute_with_timeout, to timeout computations.
227 Dropped / renamed functions and features
228 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
229 - MSG_mailbox_is_empty() -> Mailbox::empty()
230 - MSG_mailbox_front() -> Mailbox::front()
231 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
232 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
233 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
234 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
235 * Don't pass the free_f parameter to property related functions:
236 - MSG_host_set_property_value()
237 - MSG_as_router_set_property_value()
238 - MSG_storage_set_property_value()
239 * VM properties. Since msg_vm_t are msg_host_t, just use
240 MSG_host_get_property_value() and friends
241 * VM I/O related things:
242 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
243 - Unimplemented save/restore methods
244 * MSG_as_router_get_property_value() was redundent with
245 MSG_environment_as_get_property_value().
246 - Removed MSG_as_router_*propert*() functions
247 - Added MSG_environment_as_set_property_value() for consistency
248 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
249 Well, they are still there for now, but deprecated with a warning.
250 Please switch to xbt_dict_new_homogeneous() before this is removed
252 * Task affinity. Its intended behavior (that was very badly tested
253 and probably not really working) was deceiving what most users
254 would have hoped here.
255 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
256 * Option network/coordinates is now useless and should be dropped.
259 * Added option storage/max_file_descriptors to allow more than 1024 files opened
262 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
263 * Added Dragonfly topology support
266 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
267 EiBps, EBps, Eibps, Ebps for bandwidth.
268 They may become useful to some lucky ones.
271 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
272 * ex/app_tokenring: new example, very similar to the MSG Token Ring
273 * ex/async_waitAll: new example, on asynchronous communications
276 * Memory usage should be decreased for simulations with a large number
277 of processes. This also helps for SMPI.
279 SimGrid (3.13) stable; urgency=low
281 The Half Release, a.k.a. the Zealous Easter Trim.
283 - We removed half of the lines, that were mostly experimental cruft.
284 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
285 (not counting blanks and comments -- according to openhub.net)
286 - The internals are now compiled in C++ (and will soon be clean C++)
287 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
288 59 klines of C, etc. We added only 29 klines of C++ in replacement.
290 * Backwards Compatibility breaks
291 - Removed Lua simulation bindings (switch to C or Java for that).
292 Lua can still be used to describe platforms
293 - Removed Java kernel plug-ins.
294 Will be reintroduced after the ongoing major internals reorg.
296 - the following functions were removed.
297 They were too specific and should be reimplemented in a generic
298 way, with filter function.
299 - MSG_task_listen_from_host
300 - MSG_mailbox_get_count_host_waiting_tasks
301 - MSG_mailbox_put_with_timeout was removed.
302 Please use MSG_task_send_with_timeout instead.
304 - the SD_application_reinit function was removed. It has been a noop for a while.
305 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
306 replaced by a more flexible mechanism.
307 - The following functions thus do not exist anymore
308 - SD_workstation_get_access_mode
309 - SD_workstation_set_access_mode
310 - SD_workstation_get_current_task
311 - Basic estimation functions have been removed but can easily be replaced
312 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
313 - SD_workstation_get_computation_time => amount / sg_host_speed()
315 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
316 Use VM.getSpeed()*load/100 for the legacy behavior.
318 - option enable_tracing was removed. It was not doing anything for a while.
319 - In the ModelChecker:
320 - the model-checker now ptraces the model-checked process which means
321 you cannot use a debugger on the latter anymore (we might make this
322 optional in the feature);
323 - removed soft-dirty page tracking;
324 - remove model-checked side snapshot management,
325 MC_snapshot() and MC_compare_snapshot();
326 - keep the MC_cut() function as a stub (it was not really working
327 in the previous release).
330 * All options are consistently in kebab-case. Old names are kept as alias.
333 * Switch to platform v4 format.
334 - Rename from 'power' to 'speed' the attributes describing the amount of
335 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
336 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
337 - In <host> and <link>, attributes availability and state are gone.
338 It was redundent with state and availability traces, and with peak values.
339 - In <cluster>, attributes availability_file and state_file are gone.
340 It was too complex and unused.
341 - Kill <gpu>. Was not doing anything.
342 - The DOCTYPE points to the right URL:
343 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
344 (the file at this address now documents the changelog since its v1)
345 - A warning is emitted for unit-less values (they are still accepted).
346 - speed. Default: 'f' or 'flops'. Also defined:
347 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
348 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
349 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
350 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
351 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
352 - latency. Default: 's' second. Also defined:
353 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
355 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
356 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
359 * s4u::Host is now the preferred public interface to the Host features.
360 sg_host_* functions are C bindings to the exact same behavior
361 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
364 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
365 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
366 They were deprecated since a few years, and probably did not even compile anymore.
369 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
370 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
371 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
372 encouraged to update their codes. A list of the modified functions can be found at the end of
373 include/simgrid/simdag.h
376 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
377 the lambda or closure passed as a parameter will run in kernel mode.
378 All the callback functions should be rewritten to that interface at some point.
381 * Reorganizing and cleaning the internals all around the place.
384 * Remove old default barrier/bcast buggy algorithms (see #18407)
385 * Various bug fixes to handle more codes
386 * Remove the need for the --foreground option of smpirun (it is still
387 accepted for backward compatibility).
390 * Kill the setset data container: MC don't use it anymore.
391 * Kill the queue data container: it made more sense with GRAS.
392 * Kill the xbt_peer_t data type: it's useless without GRAS.
393 * Kill rm_cb feature of config sets: it was never useful.
394 * Kill graphxml parsing feature. It was not used.
395 * Kill the deprecated code protected by XBT_USE_DEPRECATED
397 - xbt_dynar_sort_strings(), when the content is char*
398 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
399 They throw exceptions on invalid input;
400 * C++ support for declaring CLI flags (simgrid::config::Flag);
401 * class for abstracting different signal backends (simgrid::xbt::signal).
402 with no external dependencies (we need very simple signals).
405 * refactoring and cleanup of the code;
406 * ongoing process to cleanly separate the model-checking algorithms
407 from the code model-checking support.
409 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
411 SimGrid (3.12) stable; urgency=low
413 The Facelift Release.
416 * Require g++ v4.7 at least to not speak prehistorical C++.
417 * Require Boost 1.48 (for signal2 component).
418 * Java must be version 7 at least when activated.
419 * Builds on Windows again (including Java bindings).
420 * Tracing is now always enabled (no way to turn it out)
421 * Remove GTNetS. It was not working anyway.
422 * Various cleanups in the cmake scripts.
423 * Move headers around to sort them out on installed systems:
424 - instr/instr.h -> simgrid/instr.h
425 - instr/jedule/* -> simgrid/jedule
426 - simdag/datatypes.h was removed
427 - simdag/simdag.h -> simgrid/simdag.h
428 - msg/datatypes.h was removed
429 - msg/msg.h -> simgrid/msg.h
432 * Interface improvement:
433 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
434 - Sanitize the interface in MSG_task_ module:
435 - Merge two functions that were close enough but misleading:
436 set_compute_duration(t) -> set_flops_amount(t)
437 get_remaining_computation(t) -> get_flops_amount(t)
438 - set_data_size(t) -> set_bytes_amount(t)
439 get_data_size(t) -> get_bytes_amount(t)
440 - Massive cleanups in the functions related to the energy
441 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
442 - New: MSG_host_get_pstate()
444 - msg/energy/onoff: switching hosts on and off
446 * Interface improvement:
447 - Rename Host.isAvail() to Host.isOn()
448 - Rename Process.currentProcess() to Process.getCurrentProcess()
449 - Rename Task.setDataSize() to Task.setBytesAmount()
450 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
452 - #18874: Actually allows the GC to reclaim tasks
456 - SIMIX_process_throw: raises an exception in a remote process
457 * Refactoring: Separate sub-modules
458 - libsmx: the public interface, as libc in a real system
459 - popping: the strange dance that converts a user request into a kernel handling
460 - smx_context_*: the virtualization mechanisms that embed the user code
461 - smx_*: the handling of each simcalls
463 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
464 * Rename smx_action_t into smx_synchro_t, making explicit that these
465 things are used to synchronize processes with their environment.
466 For example, a communication is a sort of synchronization involving
467 the communicating processes (that may block until the exchange) and
468 the platform. The same can be said from computations, etc.
470 - #18888: segfault when a process ends before its kill_time
474 - 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.
475 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
476 - MPI_Type_set_name, MPI_Type_get_name
477 - MPI_*_c2f and MPI_*_f2c functions
478 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
479 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
480 - Activate a lot of new tests from the mpich 3 testsuite
482 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
483 - InfiniBand network model added : Based on the works of Jerome Vienne
484 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
485 - When smpi/display_timing is set, also display global simulation time and application times
486 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
487 * Collective communications
488 - 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
489 - 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/).
490 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
492 - Replay now uses algorithms from wanted collective selector
493 - Replay can be used with SMP-aware algorithms
494 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
495 - 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.
497 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
498 - Correctly use loopback on fat-tree clusters
499 - Asynchronous small messages shouldn't trigger deadlocks anymore
500 * Energy/DVFS cleanup and improvement
501 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
502 - new: smpi_get_host_pstate()
505 - "Full" network optimization flag was broken since Surf++
506 - Better handling of precision flags in maxmin
507 - Fix bug causing sometimes "Impossible" errors
508 - Properly pass cluster properties to included hosts
509 * Improvement of the Energy plugin.
510 - Always update the consumption before returning that value
511 - New property: watt_off to denote the dissipation when the host is off
512 - New functions getWattMinAt and getWattMaxAt to retrieve the
513 dissipation of pstates that we are not currently at.
514 * Java: class NetworkLink renamed to Link
515 * New function: simcall_process_get_kill_time()
516 * Massive rename s/workstation/host/
517 - That's intrusive, but that's good for the project consistency. Sorry.
518 - Change config option "workstation/model" into "host/model"
521 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
522 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
523 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
524 - xbt_dirname and xbt_basename for non-POSIX systems
526 * The model checker now runs as a separate process.
527 * The model checker runs must now be launched with the new simgrid-mc program.
528 * Record/Replay: the MC can display a textual representation of a path in the
529 execution graph. It can then be replayed outside of the model checker.
531 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
533 SimGrid (3.11) stable; urgency=low
538 * Normalizing pointers addresses tool for better diff between logs
541 * Add cloud examples using new VMs
542 - examples/msg/cloud/two_tasks_vm.tesh
543 - examples/msg/cloud/simple_vm.tesh
544 - examples/java/cloud/cloud.tesh
545 - examples/java/cloud/migration/migration.tesh
546 * Add java surf examples:
547 - examples/java/surfPlugin/surf_plugin.tesh
548 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
549 - examples/java/surfCpuModel/surf_cpu_model.tesh
550 * Add SMPI+MSG example:
551 - examples/smpi/smpi_msg_masterslave/
556 - msg task destroy cancel test
557 - msg_host on/off test
558 * Move all tests in testsuite to teshsuite (adding tesh files)
559 * Restructure teshsuites
560 - one folder for each kind of test
561 * Restructure AddTests.cmake
563 - structure the order of tests (with sections)
566 * Add virtual machine
567 - creation of a VM on a PM
568 - migration of a VM from a PM to another PM
570 - MSG_process_join(msg_process_t process, double timeout)
571 - msg_bar_t MSG_barrier_init(unsigned int count)
572 - int MSG_barrier_wait(msg_bar_t barrier)
573 - void MSG_barrier_destroy(msg_bar_t barrier)
574 - msg_as_t MSG_environment_as_get_by_name(const char * name)
575 * New option "msg/debug_multiple_use" to help debugging when a task is used
578 - Improvements and finalization of MSG_storage, MSG_file APIs and their
580 - Increase code coverage in test suites
584 * Protect context stacks against stack overflow. The number of protected memory
585 pages allocated on the top of each stack (1 by default) can be configured
586 with the new command line option --cfg=contexts/guard_size:<value>.
587 * Simcalls are now generated by a python script that
588 - generates files included by SimGrid
589 - checks that all the functions exist, and proposes prototypes
591 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
592 host_set_data, host_get_data
594 - simcall_process_join(smx_process_t process, double timeout)
595 * Fix bug where sleeping processing could not be suspended.
598 * Translate surf models from C to C++
599 - Generic classes for all models: Model, Resource, Action
600 - A generic interface for each kind of model (CPU, Network, Storage
601 Workstation, WorkstationVM)
603 * Translate surf routings from C to C++
604 * Add callbacks using sigc++ or boost::signals2
605 - Add callback functions for resource creation/destruction
606 - Add callback functions for action state change
607 - Handle Energy as a plugin
608 * Replace swag by boost::intrusive
609 * Add new routing models for clusters. For documentation, see
610 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
611 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
612 parameters for cluster tag
613 - Fat trees, with topology="FAT_TREE" and
614 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
615 - see examples/platforms/torus_cluster.xml and
616 examples/platforms/fat_tree_cluster.xml
620 * Hostfiles support host:nb_processes construct to deploy several processes on
622 * Collective communication algorithms should not crash if used with
623 improper number of nodes and report the error.
624 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
625 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
626 MPI_Dims_create, MPI_Cart_sub are supported.
627 * New interface to use SMPI programmatically (still depends on MSG for
628 some parts, see examples/smpi/smpi_msg_masterslave) :
629 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
633 * Global variables privatization in MPI executables is now performed at runtime
634 with the option smpi/privatize_global_variables (default:no).
635 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
636 dynamic libraries loaded by the application are not privatized (static
637 linking with these libraries is advised in this case)
640 * Options defined in XML work correctly now.
643 * New cmake option, enable_lib_in_jar, to control whether native libraries are
644 copied into simgrid.jar or not (ON by default). Use this option if you want
645 to reduce the size of the installed simgrid.jar, *and* the native libraries
646 are kept installed elsewhere.
647 * Surf binding with SWIG (code generated in maintainer mode only):
648 - plugin to handle callbacks
649 - CPU model only for the moment
652 * Supernovae build mode is definitively removed. It was used to improve
653 inlining and inter-module optimizations. It is nowadays superseded by
654 link-time optimizations commonly available in compilers.
655 * Update ns-3 find lib. Bindings for ns-3 should work again now.
656 * Add boost dependency for surf++
657 * Add new macro for tests
658 - ADD_TESH(name <tesh_args>)
659 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
663 - xbt_bar_t XBT_barrier_init(unsigned int count)
664 - int XBT_barrier_wait(xbt_bar_t barrier)
665 - void XBT_barrier_destroy(xbt_bar_t barrier)
666 * Make the xbt_os_time module public
668 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
670 SimGrid (3.10) stable; urgency=low
672 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
675 * Reintegrate Java to the main archive as desynchronizing these
676 package is not acceptable anymore (Java is now considered stable)
677 * Add explicit synchronization facilities through semaphores
678 * Bug fix: Task.setDataSize() only changed the C world, not the value
679 cached in the Java world
682 * Dramatically change the way files are handled. API and internals changed, but
683 this part of MSG was not considered as production grade either.
684 * Add explicit synchronization facilities through semaphores
685 * Add a new function MSG_host_get_process_list()
686 * Preliminary DVFS support (see examples/msg/energy/ for details)
689 * SMPI is now included directly in the libsimgrid as the windows
690 linker doesn't force us on splitting it anymore.
691 * Improvements of the SMPI replay tool:
692 - Most of the collective communications are now rooted in the same process as
693 in the original application.
694 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
695 used until now). Multiple data types can now be used in a trace.
696 - The replay tool now supports traces produce either by TAU or a modified
698 - Bug Fix: the compute part of the reduce action is now taken into account.
699 - Gatherv collective is now supported
700 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
701 of smpirun outputs time independent traces for the current run. One file
702 is created per process. If too many processes are simulated, this behavior
703 can be changed to one file for all processes by using the
704 tracing/smpi/format/ti_one_file flag
705 * smpirun generates the host file if needed (with given host count and platform)
706 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
707 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
708 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
709 --cfg=smpi/coll_selector:(mpich/ompi)
710 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
712 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
713 happening inside a collective SMPI call.
714 * Fix the behavior of complex data types handling.
715 * Make MPI_Wtime another synchronization point to take computations into
717 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
718 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
719 * Add all missing Fortran bindings, SMPI should work with Fortran 90
720 (no privatization of global variables yet)
721 * Preliminary DVFS support (see examples/smpi/energy/ for details)
724 * Verification of liveness properties is now available for SMPI applications
725 (in addition to MSG applications)
726 * Bugged examples using SMPI in examples/smpi/mc/
727 * Add --cfg=model-check/visited option. Allows the verification of infinite
728 programs. Detection of loops in the execution thanks to the system state
729 comparison and reduction of the state space to explore. Can be combined with
730 DPOR for safety properties.
733 * Allow to change SimGrid configuration (see --help) within the code
734 thanks to SD_config() as it can be done in MSG.
735 * Add a new function SD_task_set_amount() upon user request.
738 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
739 * Remove rule based routing (no more PCRE dependency)
740 * Add a limiter_link option to cluster tag, to specify a maximum reachable
741 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
742 * Add a loopback_bw and loopback_lat options to cluster tag.
743 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
744 routing. Now peers are encapsulated in an AS and have their own private
745 router but this is transparent.
748 * Our own implementation of getline is renamed xbt_getline, and gets
749 used even if the OS provide a getline(). This should reduce the
750 configuration complexity by using the same code on all platforms.
751 * New type: xbt_cfg_elm_boolean.
752 * Allow to use yes/no for boolean configuration options in the command line.
753 * Allow to disable SimGrid cleanups at exit from command line option.
754 There are situations where one may want a simulation to end with an exit.
755 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
756 annoying when scripting around the simulator. Adding a
757 --cfg=clean_atexit:no allows to circumvent this issue.
760 * Lots of memory leaks were corrected in this release.
761 There is no known memory leaks anymore, in all of our 600+ tests.
762 * New command line option --version, to get SimGrid version information.
763 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
765 * Supernovae builds are deprecated, and expected to be removed in the next
768 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
770 SimGrid (3.9) stable; urgency=low
772 The Grasgory release: GRAS is really dead now.
774 * Complete overhaul of the internal host structures scheme.
777 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
778 was considered as experimental and badly maintained since then.
779 * Keeping it was thus a trap to our potential users, that could take
780 it instead of MSG or SMPI by mistake despite is pity state.
781 * GRAS seems to have very few users (if any), and no one volunteered
782 to maintain it further. It also induces a lot of XBT code (for
783 portability sake), that must be maintained too.
784 * For all these reasons, we killed GRAS. If someone wants to revive it
785 in the future, don't cry, our git history still remembers of GRAS.
788 * Major overhaul. Merge our documentation again as time proved that
789 splitting it was really not helping our users.
790 * Further improve the developer documentation to help newcomers
791 hacking on SimGrid itself. The user documentation (and in
792 particular, the beginner documentation) is still in a sorry state.
795 * Now works on Windows too!
796 * Much more extensive test suite, from MPICH
799 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
800 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
801 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
802 * Bug fix: let task be scheduled when the last dependency to be solved is
803 a control dependency.
804 * Remove SD_load_environment_script function.
805 Use the C sg_platf function if you want to declare a platform
809 * New function: MSG_process_get_number()
810 * Old function documented: MSG_config()
811 * Remove MSG_load_platform_script function
812 Use the C sg_platf function if you want to declare a platform
816 * Change the default value of the TCP_gamma constant (maximal size of TCP
817 congestion window) to a more realistic 4MiB value. If you notice changes in
818 your simulation results, you can fall back to the previous 20k tiny window
819 by adding --cfg=network/TCP_gamma:20000 on command line.
820 * (Hopefully) fix a bug wrt periodic availability and state traces
821 * Bug fix: use default values at start when first event in availability/state
822 trace is not at time 0.
825 * remove the "new_" part of function name sg_platf_new_trace_connect
826 (resulting in sg_platf_trace_connect), since it does not create
830 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
831 use case where it's really mandatory, and maintaining it was a pain in
833 * New: xbt_fifo_search(), search an item with a user-provided
834 comparison function instead of dumb pointer comparison.
837 * Fix the lua deployment:
838 Use `simgrid.init_application()` before deployment instead of
839 `simgrid.msg_register_application()` after.
842 * Transfer the tracing files into the corresponding modules.
844 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
846 SimGrid (3.8.1) stable; urgency=low
848 The "we are told that some people want to also *install* the simgrid
851 * Add missing file "tesh.1" to the archive.
853 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
855 SimGrid (3.8) stable; urgency=low
857 The Psssshiiiit release: SimGrid jumps into the Cloud.
860 * Add an experimental interface to manipulate VMs. They are mainly
861 process groups with very few intrinsic semantic, but they should
862 allow you to build the semantic you want easily.
863 * New function: MSG_host_set_property_value()
864 * New function: MSG_process_on_exit(). To clean memory in all cases.
865 * Bug fixes that made the host (and link) failures unusable.
866 * Add a way to auto-restart process when the host in which they are
867 executing comes back (ON_FAILURE="RESTART" on deployment file,
868 MSG_process_auto_restart_set).
869 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
870 please stop using the old ones, they are DEPRECATED.
872 * Deprecate functions MSG_global_init() / MSG_global_init_args()
873 Please use MSG_init() instead. (reducing the amount of entry
874 points in the library helps us).
875 * Make it impossible to link against the wrong version of the lib
876 * Deprecate MSG_clean(). No need to call it anymore.
877 * Function MSG_get_host_number() is not deprecated anymore.
880 * Split the doc into a user guide and a reference guide.
881 * Start a developper guide to help people hacking on SimGrid.
884 * Enable tracing by default. This modules rocks you should use it.
885 * Remove option custom_flags. Now use environment variables CFLAGS
887 * Use default cmake things to detect lua instead of home grown ones.
888 * New option "enable_mallocators" to disable mallocators, for debugging
889 purpose ("on" by default).
892 * Bug fixes around the resource failures: don't let the processes
893 survive the host they are running onto.
894 * Add an interface to auto-restart processes when the host in which they are
895 executing comes back.
896 * Ensures that SIMIX_clean is called automatically. It's not part of
897 the public interface anymore (bindings should be updated).
900 * Bug fix for when SD_Simulate is called with a positive value: be careful
901 when comparing doubles. Sometimes they are different for non significant
903 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
904 parallel task whose initial work is distributed among host according
905 to the Amdahl's law. Such tasks are created with a parameter alpha
906 that corresponds to the non-parallelizable part of the computation.
907 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
908 between two sets of workstations assuming a 1D block distribution (each
909 workstation owns a similar share of data) on both sides.
911 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
912 Data redistribution will be automatically scheduled once parent and child
913 are both scheduled. The filling of computation_amount and
914 communication_amount structures is now done seamlessly thanks to the chosen
916 * New function SD_workstation_dump to display various information
917 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
918 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
919 bandwidth on the route onto which the task is scheduled and the amount of
921 To divide the nominal bandwidth by 2, the rate then has to be :
922 rate = bandwidth/(2*amount)
923 * Compute tasks that have failed can now be rescheduled and executed again
924 (from their beginning)
925 * Increasing source code coverage (src/simdag is now covered at 95.8%
929 * Re-implement time-independent trace replay using SMPI (at the
930 smpi_smp_* level) instead of MSG. This should replace
931 examples/msg/actions/actions.c
932 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
933 hindexed and structs)
934 * Implement the exchange of non-contiguous data.
935 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
936 * Correct behavior of smpi/sender_gap and set its default value to 0
937 * Add option to asynchronously send small messages to allow better
938 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
939 specifies the size in bytes under which messages will be asynchronously sent.
940 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
941 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
944 * New C interface to define a platform: XML is now optional.
945 For more info, please check include/simgrid/platf.h.
946 * New interface to define random platforms from the C:
947 For more info, please check include/simgrid/platf_generator.h and
948 examples/msg/masterslave/masterslave_platfgen.c
949 * Export a sg_cmdline dynar containing all the arguments we got from
953 * Two new tracing options for adding comments to trace file so you
954 can track your experiments (see --help-tracing for details).
955 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
956 * Adding the SimGrid version that generated the trace file as a comment.
957 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
958 * Fix to avoid key clashes on Paje links
959 * Other minor fixes related to the Paje specification
962 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
963 and renamed to xbt_str_hash() and xbt_str_hash_ext().
964 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
965 * Greatly improve the robustness of mmalloc to user errors (such as
966 using an area after freeing it, or freeing it twice)
968 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
970 SimGrid-java (3.8.1) stable; urgency=low
972 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
973 Warning: all methods are visible, but only some of them are
974 implemented so far. Check the source (src/jtrace.c)
975 for further information.
976 * New module: org.simgrid.msg.File (SimGrid File management functions)
977 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
978 * Change the meaning of Process.restart: now restart the process from
979 the begining, like MSG_process_restart in C.
980 * Add Process.setAutoRestart: handling of process restart when failed
982 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
983 you to retrieve the properties of the processes/hosts
984 * Deprecate Msg.clean(): you can just forget about it now.
985 * New function Process.getCount(), that only works when compiling
986 with the not yet released version 3.9 of the C library.
988 * New context factory based on Coroutines. It mandates a modified JVM
989 but then, the simulations run about five times faster, and there is
990 no limit to the amount of processes (beside of the available memory).
992 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
994 SimGrid (3.7.1) stable; urgency=low
997 * Restore the prototype of MSG_process_create_with_environment() to
998 the pre-3.7 situation by removing the kill_time argument.
999 * Add a MSG_process_set_kill_time() function instead.
1002 * Fix weird behaviors when dealing with parallel tasks.
1005 * Simgrid is now built as a dll.
1006 * Simgrid-java now works on Windows.
1007 * Simgrid-Java is now included into Windows package.
1010 * First pre-build package for MacOSX.
1013 * Fix compilation when using MSG_USE_DEPRECATED.
1014 * Fix some compilation issues on Macs and Windows.
1015 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1016 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1018 We discovered that the Lua console is broken, but we are missing the
1019 manpower to fix it right now. The problem existed in 3.7 too, so we
1020 are not blocking the release for that. Sorry if you depended on this
1021 feature, any help would be really welcome.
1023 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1026 SimGrid-java (3.7.1) stable; urgency=low
1028 The "Java aint got to be bloated and slow" release
1031 * Various internal cleanups and performance improvement
1032 Simulations are expected to run up to twice faster or so
1033 * Make Process.kill(process) an instance method, not a static one
1034 * User processes are not java.lang.Thread subclasses.
1035 This breaks the compatibility (sorry), but previous API was
1036 brain-dead, making it impossible to have non-trivial
1037 initializations in the process constructor.
1038 * Require a full constructor per Process sub-class.
1039 Kinda breaks the compatibility (sorry), but this allows a much more
1040 efficient way to launch the processes at simulation startup.
1041 * Do not embeed our version of semaphores, java 1.5 can be considered
1042 as sufficiently prevalent for us to not dupplicate its features.
1046 * Add examples for almost every part of the API
1047 We spotted and fixed a lot of bugs in the process
1049 * New module: asynchronous communication API
1050 * New function: Process.sleep()
1051 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1052 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1053 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1055 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1057 SimGrid (3.7) stable; urgency=low
1059 The "spring cleanups (before the next Big Project kicks in)" release.
1062 * We can specify the SMPI latency/bandwidth factor with command line
1063 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1064 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1065 You can also use the "config tag" from platform file by adding this line
1066 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1067 (see "example/platforms/tag_config.xml" to use "config tag").
1068 Note that the command line supersedes the platform file configuration.
1069 * Change the correction factors used in LMM model, according to
1070 the latest experiments described in INRIA RR-7821.
1071 Accuracy should be improved this way.
1072 * Use the partial invalidation optimization by default for the
1073 network too. Should produce the exact same results, only faster.
1074 * Major cleanup in surf to merge models and split some optimization
1075 mechanisms from the core of the models. As a result you can now
1076 specify which model to use (e.g., --cfg=network/model:LV08
1077 --cfg=cpu/model:Cas01) and which optimization mode to use
1078 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1079 Incompatible combinations should err at initialization. See
1080 --help-models for the list of all models and optimization modes.
1081 * The CLM03 workstation model was dropped for simplicity because it
1082 used the deprecated CM02 network model. Use default instead.
1083 * Rename the TCP_gamma configuration option to network/TCP_gamma
1084 * Rename the coordinates configuration option to
1085 network/coordinates, and document it
1086 * Use now crosstraffic keyword instead of the terribly misleading
1087 fullduplex keyword. It is activated by default now in the current
1088 default model, use --cfg=network/crosstraffic:0 to turn it off.
1089 * Ongoing refactoring the model parsing to make XML files optional
1090 See include/simgrid/platf.h for details (still to be completed)
1093 * Major overhaul of the documentation. Almost instructive now :/
1094 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1095 functions and friends. This interface was considered as
1096 deprecated since over 2 years, it's time to inform our users that it is.
1097 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1098 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1099 use these (crappy) functions in your code.
1100 These functions will be removed soon. Stop using them now.
1101 * Deprecate MSG_get_host_{table,number}
1102 Implement MSG_hosts_as_dynar() instead.
1103 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1104 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1107 * Stabilize the parallel execution mode of user contexts
1108 * Introduce configuration variables to control parallel execution:
1109 - contexts/synchro: Synchronization mode to use when running
1110 contexts in parallel (either futex, posix or busy_wait)
1111 - contexts/parallel_threshold: Minimal number of user contexts
1112 that must be part of a scheduling round to switch to parallel
1113 execution mode (raw contexts only)
1114 * Fix bugs that prevented to use suspend/resume along with
1115 synchronization structures.
1116 * Fix bugs in process termination that lead to invalid memory access
1117 in very specific conditions.
1120 * Introduce a parallel mode for the models (controlled by surf/nthreads
1121 configuration item). In our tests, running the models in parallel
1122 never lead to any speedups because they are so fast that the gain
1123 of computing each model in parallel does not amortizes the
1124 synchronization costs, even when ultra fast futexes are used.
1125 This is released anyway because YMMV.
1128 * Performance boost by using a swag internally to compute the set of
1129 tasks that are finished and should constitute the return value of
1133 * Enable it by default now that it is considered rather stable.
1136 * Documentation of the tracing functions.
1137 * Performance gains when tracing categorized/uncategorized resource
1138 utilization by avoiding calls to get route when updating resource
1139 variables. LMM constraints are being used instead.
1140 * API changed to set task categories. Use MSG_task_set_category instead
1141 of TRACE_msg_set_task_category, and SD_task_set_category instead
1142 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1143 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1144 * Examples for link user variables added (see at examples/msg/tracing/)
1145 * Deprecated function TRACE_msg_set_process_category completely removed
1146 * Trace header updated according to the latest Paje file format
1147 * Tracing network lazy updates, no longer obligate users to use full updates
1148 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1149 * Experimental: let user code declare/set/push/pop application states for hosts
1150 * API changed to allow the manual creation of graph configuration files
1151 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1154 * Improve the API of Lua MSG bindings, using the Lua spirit.
1155 * Each simulated process now lives in its own Lua world (globals are
1156 automatically duplicated). It helps writing simulators. Will allow
1157 to run Splay programs within SimGrid in the future.
1158 * Add a Chord example in Lua, equivalent to the MSG one.
1161 * Start the implementation of a solution to express temporal
1162 properties, not only local assertions. This is still an
1163 experimental work in progress, stay clear from it to be safe.
1167 - Add new runtime parameters --help-logs and --help-log-categories
1168 to display information about supported logging parameters and
1170 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1171 don't exists anymore.
1172 * Mallocators: allow value NULL for the reset function.
1174 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1175 create homogeneous dictionaries, where all the elements share the
1176 same free function. Non homogeneous dictionaries will be
1177 deprecated in the next release.
1178 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1179 - Multi-level dictionaries are deprecated.
1181 - new function xbt_dynar_search_or_negative() that is useful when
1182 you have less than 2 million elements in your dynar and don't
1183 want of the extra complexity of catching exceptions when the
1184 element is not found.
1186 - Make xbt_os_thread module (for thread portability) public.
1187 Documentation is still to come, sorry.
1189 - Cleanups and simplifications to make it maintainable again.
1190 - Exotic features (such as memalign and valloc) were removed.
1191 - The metadata were extended and improved so that the
1192 model-checker becomes able to explore and inspect the heaps.
1193 - This may induce a performance drop when enable_model-checking is
1194 ON in cmake (even if it's not used in the simulation), but it is
1195 necessary at this point to get MC working.
1197 Turn model-checking OFF if simulation performance matters to you.
1198 Not enabling it at runtime is not enough, disable it in cmake.
1200 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1203 SimGrid (3.6.2) stable; urgency=low
1205 The "Not coding new stuff allows to polish old things" release.
1208 * New bindings to the NS3 packet level simulator (experimental)
1209 * Use the raw (efficient) execution contexts instead of the sysv
1210 (portable) ones when possible.
1211 * libpcre is now mandatory in any cases since not using it led to
1212 severe performance loss and possibly other issues
1213 * Update the XML platforms:
1214 - G5K: include the latest machine in Nancy
1215 - GridPP and LCG: new platforms
1216 * Documentation was partially updated, at least (more to come)
1218 Bug fixes, cosmetics and small improvements
1219 * Free terminated processes before the end of the simulation to avoid
1220 exhausting the memory of users having very dynamic amount of
1222 * Bug fix and cosmetics about canceling non-running tasks
1223 * Bug fix about the dot loader's issues when using libcgraph
1226 * Create an installer for windows with nsis (amd64 and win32)
1227 - Add an hello world project to illustrate simgrid project creation.
1228 - Embed libpcre into the Simgrid installer to avoid
1229 its compilation burden
1230 * The raw execution contexts should work on Apple now
1231 * Port to Windows 64 bits
1232 - Sysv contexts now have an implementation for this arch
1233 - GRAS communication features now support this arch
1234 * Drop support for borland compiler on windows
1235 - this code was not maintained, and we kinda depend on gcc nowadays
1236 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1237 * Fix portability issue on unstable ubuntu: linker became picky on
1240 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1242 SimGrid (3.6.1) stable; urgency=low
1244 The "Oops, we broke Macs too" release
1247 * Fixed contexts detection so that raw ones are used when possible
1248 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1249 leads to a strange error, with user code segfaulting sometimes when
1250 the generated code is not perfectly aligned (which is not
1251 controllable from the user side, depends on the amount of code)
1254 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1256 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1258 SimGrid-java (3.6) unstable; urgency=low
1261 * Split of every thing from simgrid v3.5 into a separate package.
1263 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1265 SimGrid (3.6) stable; urgency=medium
1267 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1270 * Bindings now constitute their own package, separated from the main one.
1271 Rationale: reduce our maintenance nightmare by reducing the module coupling
1272 They will soon be released on their own on gforge.
1274 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1275 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1277 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1278 * It's not quite deprecated for now because we have no replacement,
1279 but it may soon become the case.
1282 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1283 * Fortran: New user-level cache variable to store the rank of the running
1284 process. This improves performance by an order of magnitude.
1285 * C: New coccinelle script to automatically locate and modify global and
1286 local static variables.
1287 * Improved SMPI network model with a sender-side gap to account for multiple
1291 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1292 only say if a communication is finished, no matter whether it succeeded or
1293 failed. You can call MSG_comm_get_status() to know the status of a finished
1295 * New function MSG_task_dsend() to send a task and detach it. When a
1296 communication is detached, you are never notified of its success or failure
1297 and the memory is released automatically once it is finished. This function
1298 is useful when you don't care about the end nor the success of a
1300 * Change the prototypes of action replay. Sorry for inconvenience,
1301 but this is really more efficient this way (and to adapt your code,
1302 you just have to fix the initialization, that shouldn't be too long)
1303 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1304 ever managed to do anything useful with it.
1305 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1306 takes a dynar of communications. It returns immediately and gives the
1307 index of a finished communication (if any).
1308 * New example: a basic implementation of the Chord P2P algorithm.
1311 * New model for multi-core CPUs. You can now use the core attribute to
1312 precise the number of cores of a host. This is a basic model. Every
1313 process running on the host receives at most the power provided in
1314 the DTD (throughput<=power). Total throughput of process cannot exceed
1316 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1317 router linked by an up-link and a down-link (possibly asymmetrical).
1318 This kind of pattern allows to easily build last-mile model style platforms.
1319 Aggregating such patterns in a rule-based AS is thus the technique of
1320 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1321 * New model for Vivaldi routing. We transformed the Vivaldi network model
1322 into a Vivaldi routing model (based on the rule-based model). This allows to
1323 combine Vivaldi based latencies with last-mile platforms.
1326 * Added a check for NaN of IEEE754 infinite in the double entries of
1328 * Introduce a new context factory "raw", highly inspirated from the
1329 ucontext factory, but using manually crafted functions in assembly to
1330 do the work in an efficient manner.
1331 * Allow to change the used context factory at run time, not only at
1332 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1333 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1334 contexts at runtime (only with raw contexts or ucontexts).
1335 * Completely rewrote this module to allow parallel execution of user
1336 processes. Use --cfg=contexts/nthreads:N to execute user processes
1337 with N parallel threads (the default is 1, meaning no parallelism).
1338 * Allow to decide dynamically between sequential and parallel modes.
1339 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1340 processes in parallel only when their number is greater than or equal to P
1342 * Added a check for NaN of IEEE754 infinite in the double entries of
1346 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1347 won't output the state of processes when interrupted with Ctrl-C
1348 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1349 NULL-terminated array. This may solve backward compatibility issues
1350 due to the change to return type of SD_simulate. See also:
1351 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1352 * Add new macros with variable number of arguments.
1353 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1354 - in xbt/asserts.h: xbt_assert
1355 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1356 - in xbt/ex.h: THROWF and RETHROWF.
1357 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1359 * Change xbt_die() to accept a format string with arguments, just like printf.
1360 * New data structure: xbt_lib_t, like a dict but more general and with better
1364 * New configuration options
1365 Options triva/categorized and triva/uncategorized can be used to generate
1366 graph configuration files for Triva visualization tool.
1367 * Configuration option tracing/platform is renamed to tracing/categorized
1368 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1369 * New TRACE_user_link_variable interface:
1370 User provides the name of the link and the tracing variable to attach to it
1371 * the declaration of tracing categories must be done after the environment
1373 * simpler tracing interface, just one way to declare categories
1374 TRACE_category or TRACE_category_with_color, it is up to you
1375 * links in the trace file are again identified by their names
1376 * trace contains the full platform hierarchy exactly as declared using the ASes
1377 * Options tracing/msg/[task|process]:1 groups the process by hosts
1378 for both cases, tasks and processes must have names that are unique during the simulation
1379 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1380 * The experimental option tracing/msg/volume is deprecated
1381 its functionality may be reincorporated if needed
1383 The tracing generates a trace file with unordered timestamped events,
1384 because of the way the core simulator (surf) works. A script available
1385 at the tools directory (fix-paje-trace.sh) can be used to put the events
1386 in order. We have changed the tracing so it can generate ordered timestamped
1387 events in the final trace, but depending on the simulator (and how much time
1388 is simulated) that can lead to a huge memory utilization. It is deactivated
1389 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1391 Build Infrastructure
1392 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1393 - We are trying hard to keep the API stable, but it may happen that
1394 some things change (we're a research project after all, not a
1395 nuclear plant operating system). If such things should happen, you
1396 could rely on that macro to adapt.
1397 - current value: 30600 for 3.06.00, aka 3.6
1398 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1399 a number that can be compared with SIMGRID_VERSION.
1400 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1401 assertions and verbose/debug logging events are disabled at compile time.
1403 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1405 SimGrid (3.5) stable; urgency=medium
1408 * New feature to any SimGrid-based simulator: Model-Checking
1409 Check SIN#1 for more details.
1412 * New Model SMPI (three-interval linear regression for correction factors)
1413 See RR-7426, available at http://hal.inria.fr/inria-00527150
1414 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1415 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1416 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1417 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1418 * See also src/smpi/README
1423 - Tracing API changes: TRACE_start and TRACE_end should not be called
1424 by user-code. They are automatically called by simulators created
1425 with SimDAG, MSG and SMPI if the toolkit is compiled with
1426 tracing_enabled=ON. Categories declaration and utilization remain the
1427 same for MSG and SimDag.
1428 - A function was added to the tracing API to declare categories with
1430 - TRACE_category_with_color (char *category, char *color)
1431 where color must be in the following format
1432 "%f %f %f", red, green, blue
1433 and red, green, blue are float values in the interval [0, 1]
1434 - User can specify NULL as color parameter, or continue calling
1435 TRACE_category (cat)
1436 On that case, the tracing system will define random colors
1437 - The following command-line options are supported:
1438 --cfg=tracing/filename:msg.trace
1439 --cfg=tracing:1 (activate tracing, needed to use others)
1440 --cfg=tracing/platform:1 (categorized resource use)
1441 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1442 --cfg=tracing/msg/task:1 (task creation)
1443 --cfg=tracing/msg/process:1 (process creation, migration)
1444 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1445 --cfg=tracing/smpi:1 (SMPI interface tracing)
1446 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1447 - examples of examples/msg/tracing updated
1449 - DAXLoader and DOTLoader functions can generate tasks with categories
1450 - A new function to attribute a category to SD tasks:
1451 TRACE_sd_set_task_category (SD_task_t task, char *category)
1452 * Tracing the MPI interface implemented by SMPI
1453 - Collective operations are traced with states
1454 - Point-to-Point operations are traced with states/links
1455 - Tracing activated by a parameter "-trace filename" passed
1456 to smpirun during execution (considering that simgrid
1457 is compiled with tracing enabled)
1458 - To run the simulation with gdb, the simulator
1459 accepts --cfg=tracing/smpi:1 to trace SMPI
1460 - tesh files to check if smpi tracing is ok
1461 - See examples/smpi/NAS/DT-trace
1462 * GTNetS tracing re-worked
1463 - adaptation to the tracing system of GTNets to cope
1464 with modifications regarding the fullduplex mode
1465 - new tesh files to check if gtnets tracing is ok
1468 * Asynchronous communications through the functions:
1469 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1470 * New function: MSG_load_platform_script()
1471 to make possible using a lua script instead of XML files to set up platforms
1472 * New function: MSG_set_function
1473 to associate functions to processes, used when bypassing the parser
1474 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1476 Platforms: Add some more examples in examples/platforms
1477 * Grid'5000: see www.grid5000.fr
1478 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1481 * Change the XML format. This is a very important modification. SimGrid
1482 3.5 introduces a new hierarchical format based on the notion of
1483 Autonomous Systems. Compatibility with old format is ensured
1484 through the perl script provided in the install bin directory
1485 bin/simgrid_update_xml.
1486 It is now possible to build platforms with specific routing
1487 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1488 connect several platforms together. We will try to provide soon
1489 set of realistic platforms exploiting these properties (have a
1490 look at examples/platforms/ for the moment).
1491 * Take the opportunity of the XML format change to be a good XML citizen:
1492 rename link:ctn to link_ctn and similar changes (also dealt with by
1494 * Add a new routing scheme (rule-based) using regular expressions. It
1495 enables to have an extremely low memory footprint when the
1496 underlying routing is simple and can be compactly described. You
1497 need to have libpcre4-dev (perl regular expressions) installed if
1498 you want to use this routing scheme.
1499 * Revive the cluster TAG and allow to easily and efficiently (both in
1500 term of memory and speed) connect clusters together. Have a look
1501 at teshsuite/simdag/platforms/ to see how this can be done. With
1502 this tag, you can create clusters with thousands of tasks at no
1503 cost (have a look at examples/platforms/).
1504 Note: clusters are implemented as ASes, so there is no need for an
1505 enclosing AS tag if you have only one cluster in your platform.
1506 * Add new generic functions in the public interface that allows the user
1507 to call SURF 'create_resource' methods from your code (same
1508 functionality as the XML bypass mechanism but with a much lighter
1510 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1511 that uses a piecewise linear approximation to produce better
1512 results when exchanging small messages.
1513 * Add a new parameter to handle correctly full duplex link and account
1514 for interferences between uplink and downlink communications
1515 (activate with --cfg=fullduplex:1).
1518 * Rename the SD_READY (all dependencies are satisfied and task is
1519 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1520 dependencies are satisfied) state.
1521 This prevents a confusion between the notion of "ready to schedule"
1522 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1523 simulated" (SD_RUNNABLE) used by the simulation kernel.
1524 * Change the way a task is considered as ready. Instead of removing
1525 dependencies when a task is done, a counter is decreased. This way,
1526 it is always possible to reach ancestors thanks to the
1527 SD_taks_get_parents function (even after the end of the simulation.)
1528 * Change the return type of SD_Simulate from (SD_task_t*) into
1529 xbt_dynar_t. This function was in handling a dynar internally and
1530 converted it into a NULL terminated array for historical reasons.
1531 * New function SD_dotload(char*) to load a DAG described in dot
1532 format. This loader and the corresponding examples require the
1533 installation of the graphviz library.
1534 * Fix a bug in the management of tasks of size 0 in the surf network
1535 models. This problem was only visible with SIMDAG and you should
1536 thus disregard results produced with earlier versions if you
1537 relied on this feature (some tasks were blocked because of this).
1538 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1539 even though your applications were doing only point-to-point
1540 communications and sequential computations. Now you can really use any
1541 model you want (of course, if you create real parallel tasks, which are
1542 not implemented in most models beside ptaskL07, this will abort).
1543 * Add an example that schedules a DAX on an heterogeneous platform
1544 using a Min-Min strategy.
1545 * New function SD_workstation_get_current_task() that returns the kind
1546 of task currently running on a workstation in the sequential access
1548 * Raise some warnings when unexecuted tasks remains at the end of the
1549 simulation. This is usually caused by cycles in the DAG.
1552 * New function: SIMIX_process_set_function() called by MSG_set_function
1553 * Change the underlying waiting queue in semaphores so that a process
1554 can wait on several of them simultaneously (as in waitany).
1555 * Fix the way to handle tokens in semaphores so that all access patterns
1556 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1557 * kill the dirty pimple SIMIX_message_sizes_output()
1558 Please use (proper) visualization instead
1561 * New data container: setset (set of sets of elements)
1562 * New module: mmalloc (mapped malloc, allowing to have several
1563 independent segments of malloc)
1564 * New function: xbt_dict_cursor_set_data()
1565 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1566 * New function: xbt_dynar_is_empty()
1567 * New function: xbt_fifo_get_last_item()
1568 * Fix xbt_dynar_shrink(): use the right element size.
1569 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1570 uninitialized areas during expand.
1571 * Fix semaphores: previous implementation was severely broken.
1572 * Use library init/fini functions for our initialization.
1573 - you can use logs and other feature as soon as you want in your
1574 code (even before the xbt_init / MSG_init)
1575 - xbt_exit is now a no-op and produce a warning when used.
1578 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1579 benefit from the latest improvement to the simulation kernel.
1580 * Kill measurement sockets for now. If you rely on them, sorry. This
1581 release is not for you. This feature will be reintroduced in the
1582 future, but we cannot delay the release any further.
1583 * New function: gras_msgtype_get_name().
1584 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1585 * Fix (at last) the pmm example: it should not randomly fail anymore.
1587 Build chain: bug fixes and overall polishing
1588 * Cmake is now stable enough. Hence, we killed the autotools.
1589 * Port to windows ( TM :)
1590 * Fix the 'make install' target.
1591 No need to use 'make install-simgrid' anymore
1592 * Introduce a 'make dist' target compiling a *source* archive
1593 'make package' compiles a binary archive
1594 * Compile java files only on need
1595 * Add --cd and --setenv command line options to tesh
1596 * Out of source builds are not fully supported yet, but we are close
1597 * Enable supernovae and optimization flags by default for our users
1600 * Add layer to set up environment directly from lua, without XML.
1601 * The effect of gras_stub_generator can be achieved through
1602 lua too (check examples/gras/console/ping_generator.lua)
1604 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1606 SimGrid (3.4.1) stable; urgency=low
1608 The "Polishing easter eggs is probably a good idea" release.
1609 This is a bug fixes release only.
1613 * Fix a bug preventing the tasks from begin garbage collected.
1616 * Fix a bug occuring when a host involved in a communication fails.
1617 This was not detected properly by the other peer involved in the
1618 communication. Now, it's reported as a network error.
1621 * Warn the user about loop dependencies in data flow of DAX files
1622 * Obey the control-flow dependencies of DAX files
1625 * Add option "enable_smpi" allowing to not compile SMPI.
1626 Probably useful for the (Mac) users experiencing a build error here
1627 * Improve the detection of lua5.1 and ruby1.8
1629 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1631 SimGrid (3.4) stable; urgency=low
1633 The "Easter in Cargese" release. Also known as (major changes):
1635 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1636 ~> bindings were greatly improved
1637 ~> new tracing infrastructure for better visualization introduced
1639 * the "Welcome to configury modernity" release.
1640 ~> we switched from autotools to cmake, and improved our cdash
1643 A more detailled list of changes follow (full detail in svn log).
1645 Java Bindings: Various Cleanups
1646 * (install java-gcj-compat-dev on debian-like to use them)
1647 * Remove put/get: no need to export deprecated interface in Java
1648 Use send/receive instead.
1649 * Cleanup the examples and add a README per directory
1650 * Remove example autoDestination (that's the only way to go now)
1651 * Remove example explicitDestination (was a plain copy of basic)
1652 * Make JniException a runtime exception, so that there is no need to
1653 declare the fact that you may encounter such a beast. I guess that
1654 nobody will ever want to survive such error.
1655 * Create specific errors for each MSG case of failure:
1656 host failure, transfer failure, timeout, task cancelled
1657 * Cleanup the exceptions that may get thrown by each function
1658 * Other internal cleanups in Java bindings. Performance still bad :/
1659 Ruby and Lua Bindings: create them
1660 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1661 * That's new and great, you should try them out.
1662 Same functionalities than Java bindings, only even less polished
1664 * Kill the useless "rate" argument of SD_task_get_execution_time()
1665 Everyone used to provide -1 as a value, it was not used, and the
1666 semantic of a possible use wasn't even clear.
1667 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1668 either as comm costs or compute costs to mean that there is no
1669 such thing for that specific task.
1670 * Add a SD_task_set_name() function
1671 * Fix SD_task_unschedule() on typed tasks
1672 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1673 * In DAX loader, accept useless 'level' attributes to <job> since
1674 LIGO DAGs have them (seem to be to ease graphical representation).
1676 * Add an example masterslave_mailbox.c using send/receive and not
1677 the deprecated put/get interface.
1678 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1679 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1681 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1682 (declare MSG_USE_DEPRECATED to still have the old name)
1683 * Add a MSG_task_set_data() function
1684 * About trace replay (see examples/msg/actions):
1686 - Allow to work with splitted trace files for each process
1687 Give the specific trace file as argument of each process,
1688 and call MSG_action_trace_run(NULL)
1689 You can still have one merged file for all processes.
1690 - Fix implementation of collective operations
1691 * Allow task_execute() on 0-sized tasks (closes #10063)
1693 * This is the first release of SimGrid where SMPI is not considered
1694 beta anymore (even if some corners should still be improved)
1695 * Port over the new SIMIX_network submodule (internal refactoring)
1696 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1697 * Implement more missing elements of the standard:
1699 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1701 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1702 MPI_Scan() MPI_Get_processor_name()
1703 - Added implementation of missing case for Alltoall (warning: it's
1704 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1705 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1706 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1707 SMPI_MPI_Allgatherv()
1708 * Bug fixes include:
1709 - MPI_Waitsome() was broken
1710 - Allow relative includes in smpicc
1711 - Command line cfg argument 'reference_speed' was ignored...
1712 - Some functions did not properly lead to auto-benching of user code
1713 - smpicc passes -O2 by default (just like openmpi one)
1715 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1716 * Bug fixes about timeouts during communications
1717 * add SIMIX_message_sizes_output() as a pimple to write to file the
1718 amount of messages per size. Use gnuplot to get histogram.
1719 Pimple because that's the only user-visible function of simix,
1720 defined directly in xbt.h (irk, sorry)
1722 - Add a SIMIX_sem_get_capacity() function
1723 - Fix interactions with processe resume/suspende
1724 - release_forever() was stupidly broken
1725 - Fix SIMIX_display_process_status() for processes in a semaphore
1726 - Make SIMIX_sem_block_onto() user-visible
1727 * Refactoring context stuff:
1728 - Use pseudo-OOP for better modularity
1729 - reimplement SIMIX_process_kill() without process_schedule() so
1730 that the latter can take as invariant that it is called from
1732 - Merge context_start into context_new for sake of simplicity
1734 * Add a Vivaldi network model, coded live during SUD'10 ;)
1735 * Rename configuration variables to start a hierarchy:
1736 o cpu_model -> cpu/model
1737 o network_model -> network/model
1738 o workstation_model -> workstation/model
1739 * New configuration variables:
1740 o network/bandwidth_factor: correction to bandwith
1741 o network/latency_factor: correction to latency
1742 o netwotk/weight_S: correction to the weight of competing streams
1743 * Add a long description to the models, that users can see with such
1744 argument on the command line: --cfg=cpu/model:help
1745 * --help-models display the long description of all known models
1747 * config: add the ability to set a default value after registration
1748 Does not override any previously set value (e.g. from cmd line)
1749 * dict: allow to have integer key and data.
1750 When so, you need to use the following functions
1751 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1752 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1753 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1754 In contrary to regular dicts, the key is not malloced before copy.
1755 Mixing scalar and regular elements in the same dict is not tested
1757 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1758 Tracing for Visualization:
1759 * SimGrid is now instrumented in order to generate a trace file for
1760 visualization analysis: to use it, need to compile SimGrid with the
1761 "tracing" option enabled, and instrument the program using SimGrid with
1762 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1763 (among other functions).
1764 * The instrumentation only traces the platform utilization for now
1765 * Documentation to use the tracing functions and how to analyze the
1766 traces with the Triva tool is written.
1767 * More information about: SimGrid FAQ (in the section Tracing Simulations
1770 * We moved to cmake as default build system. Autotools support will
1771 be dropped soon. Check the FAQ for more info about how to use it.
1772 * Greatly improved our cdash/ctest interactions
1773 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1774 * Added memory checking tests with valgrind; lot of memleak fixing.
1775 This may be the first release of simgrid with so few memory issues
1776 * Added code coverage tests.
1777 Our coverage is still improvable, but at least we see it on cdash.
1779 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1781 SimGrid (3.3.4) stable; urgency=low
1783 The "Desktop Grid needs love too" release (also called Xmas release).
1785 Models improvements:
1786 * Major speedup in the maxmin system solving by using lazy evaluation
1787 Instead of solving completely the maxmin system at each iteration,
1788 only invalidate (and recompute) the modified parts.
1789 This new feature is enabled in default models but you can try to
1790 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1791 * Cas01 IMproved as default CPU model
1792 This CPU model is the same Cas01 model, but it uses the
1793 maxmin-selective-update flag and a heap structure to manage
1794 actions on SURF kernel.
1795 It reduces the complexity to find the next action to finish and,
1796 consequently, it's faster than the old Cas01.
1797 This is the new default CPU model (Cas01).
1798 * Rename the old Cas01 model to Cas01_fullupdate
1799 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1800 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1801 * CpuTI (CPU Trace Integration)
1802 A new CPU model whose objective is simulate faster when using
1803 availability trace files.
1804 Instead of using a full featured, over engineered maxmin system for
1805 CPU modeling, this model does the pre-integration of traces files
1806 to calculate the amount of CPU power available, and so, executes
1807 faster than the old CPU models.
1808 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1809 * Use LV08 as default network model since it gives better accuracy
1810 for small messages and shouldn't change things for big ones.
1811 Use --cfg=network_model:CM02 to get the previous behavior.
1814 ******************************************
1815 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1816 ******************************************
1817 * The new CPU model may changes simulations!
1818 The point is that events occurring at the exact same timestamp
1819 are not scheduled in the same order with the old and new
1820 version. This may be enough to completely change the execution
1821 of simulations in some cases.
1822 * The new network model will change simulations!
1823 This new model is more realistic than the previous one, so you
1824 should consider redoing your old experiments with this model.
1825 Sorry for the inconvenience.
1828 * Introduce the supernovae compilation mode
1829 When compiled that way, the whole SimGrid (or almost) is put in a
1830 single compilation unit and compiled in one shoot.
1831 This is to help gcc which has difficulties to inline stuff from one
1833 The speedup seem to be above 15%, althrough more tests are needed on
1834 amd64 to confirm that gain.
1837 * Port of MSG's mailbox on top of SIMIX network
1838 The put/get mechanism was greatly simplified on the way.
1841 * New SIMIX network module. Provides:
1842 - Mailbox: rendez-vous mechanism to find with who you want to speak
1843 - Synchronous send/recv: easier and hopefully faster since the
1844 logic is handled in the maestro process directly now
1845 - Asynchronous send/recv: you dreamt of it? It's here now
1846 Too bad that nobody cared enough to propagate the change to MSG.
1847 * Add semaphores as SIMIX synchronization mechanism.
1850 * new function SD_daxload(char*) to load a DAX file
1851 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1852 * Introduce typed tasks. Specify its kind and cost at creation.
1853 At scheduling, just give where it should be placed, and the cost
1854 for each involved resource is automatically computed.
1855 Existing constructors so far (more to come of course):
1856 - SD_task_create_comm_e2e() for end-to-end communication
1857 - SD_task_create_comp_seq() for sequential computation
1858 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1859 * new function SD_task_dump() for debuging display
1860 * new function SD_task_dotty(task,FILE*) writing to file the info
1861 about the task in dotty format
1862 * SD_task_dependency_exists() can now cope with having one of its
1863 arguments NULL. If so, it tests whether the other argument has any
1865 * Add getters on list of preceding/following tasks:
1866 SD_task_get_parents(task) and SD_task_get_children(task)
1867 * Add getters on amount of workstations and list:
1868 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1869 * Add getter on task kind: SD_task_get_kind(task)
1870 * Update the start_time and finish_time of tasks on completion/failure
1871 * Bugfix: Remove task from state swags when destroyed
1874 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1877 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1878 dynar with one separate thread per value of the dynar.
1879 * Change the prototype of xbt_thread_create(), sorry.
1880 Added a boolean parameter indicating whether we want to join this
1881 thread (used in SG only for now)
1882 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1885 * GTNetS wrappers should now be usable again (and betterly tested too)
1886 * Fix a major regression from 3.2 where the timeout provided to
1887 MSG_task_put_with_timeout() was used as absolute time before which
1888 the comm should be done.
1889 * Start to fix the <cluster> tag.
1890 - Internal links should be good now (beside of the loopback, which
1891 use the private link instead)
1892 - paths to the external world is still rather broken
1893 - the <route:multi> tag is just broken. Actually that's brain-dead.
1894 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1895 to make it less stupid
1896 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1897 * Fix a source-level compatibility glitch from 3.2: after defining
1898 MSG_USE_DEPRECATED, you can use the old name
1899 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1900 * Allow to compile from the SVN with automake 1.11
1901 * Fix some problems when using the "start_time" tag in deployment XMLs.
1902 * Fix #8569: XBT/synchro.h has redundant declarations
1903 * Fix #8563: MSG return values and exceptions
1904 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1905 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1906 Thanks to Robson Peixoto.
1907 * Fix a vicious bug in dictionaries inducing that some elements were
1908 not freed on xbt_dict_free()
1910 Portability report of this version:
1911 * Main portability targets:
1912 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1913 on (amd64/i386/ia64)
1914 - mac leopard on i386
1915 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1916 but nothing critical.
1917 * Other platforms: windows, AIX and others were not tested for this release
1919 Timing report of this version:
1920 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1921 scenario parameters). From 8h to a few seconds in desktop grid settings.
1922 * Supernovae brings about 25% speedup on i386.
1924 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1926 SimGrid (3.3.3) stable; urgency=low
1928 The "Need for Speed" release.
1930 The timings done to validate the 3.3.2 were faulty.
1931 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1933 The problem was a conversion from a manually handled vector to
1934 xbt_dynar_t on the critical path.
1935 xbt_dynar_foreach calls functions, inducing stack management crap.
1937 We inlined these functions and xbt_dynar_foreach is now breath taking.
1938 We also inlined xbt_swag_belong on the way.
1940 Here are some approximate speedup measurements (on master/slaves
1941 simulations lasting between 10s and 20s each):
1942 3.3.1 -> 3.3.2: about same performance
1943 3.3.2 -> 3.3.3: 40% speedup
1944 3.3.1 -> 3.3.3: 40% speedup
1945 3.3.1 with inline patch -> 3.3.3: 30% speedup
1947 Our reading is that the refactoring which occurred in 3.3.2 made us
1948 suffer much more from the xbt_dynar_foreach low performance, but
1949 once we solved this, this refactoring proved to be very performance
1950 effective. From the 40% speedup, somehow, 10% are due to the
1951 inlining and 30% to the refactoring.
1953 That's a pitty that gcc cannot inline functions placed in other files
1954 alone. We have to choose between:
1955 - break the encapsulation (by putting private data structures and
1956 accessors in headers files to help gcc)
1957 - live with low performance
1958 - switch to a decent compiler such as icc (not quite possible).
1960 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1962 SimGrid (3.3.2) stable; urgency=low
1964 The "Simplicity does not preceed complexity, but follows it" release.
1966 The main contributors of this release were (lexical order):
1967 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1970 * Extract the routing logic into its own object.
1971 (was dupplicated in network.c and workstation_LV07.c;
1972 Allows to implement other ways of storing that info)
1973 => kill now useless network_card concept
1974 - Use dynar to represent routes (instead of void** + int*)
1975 - kill link_set (use surf_network_model->resource_set instead)
1976 - Add a command-line option to choose the routing schema to use
1977 - Add three new models:
1978 * Floyd (shortest path computed at initialization)
1979 * Dijikstra (shortest path recomputed all the time)
1980 * Cached Dijikstra (shortest path computed on need)
1981 All these models where contributed by Silas De Munck, and are
1982 described in his ICCS09 paper.
1984 * Simplify model declaration
1985 (less redirections, less function to write when defining a model)
1986 - Factorize stuff between models:
1989 surf_model_resource_set(model)
1990 surf_model_resource_by_name(model, name)
1991 - Unify the types of models in s_surf_model_t (using an union)
1992 - Embeed fields of common_public directly into s_surf_model_t
1993 - Rename model methods:
1994 action_free ~> action_unref
1995 action_change_state ~> action_state_set
1996 action_get_state ~> action_state_get
1997 - Change model methods into functions :
1998 (model)->common_public->action_use ~> surf_action_ref
2000 * Implement a generic resource; use it as ancestor to specific ones
2001 (allows to kill duplicated code in models)
2002 Drawback: timer command don't need no name nor properties;
2003 workstation_CLM03 don't need no properties
2004 (but I guess we can live with those few bytes wasted)
2006 * Improve the action object model
2007 - implement a constructor avoiding dupplicated code about field
2008 initialization in generic_action part.
2010 * Kill the SDP model: it has an external dependency, is deprecated
2011 in flavor of modern lmm models, and didn't compile since a while
2014 * Relocation of the context module from XBT to SIMIX.
2015 (the context were decoupled from the simix processes, duplicating a lot of code)
2016 => a lot of code was factorized
2017 - less overhead is introduced during scheduling
2018 - simpler API for the context factory
2019 - the logic for process creation,destruction and manipulation was simplified
2020 * Simplification of the s_smx_process_t data structure.
2021 => accesing the simix level data associated to a process is faster now,
2022 and the code is a lot more readable.
2025 * Implement some more MPI primitives:
2026 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2027 -implementation: Bcast: flat or 2-ary tree (default),
2028 Barrier: 4-ary tree,
2030 Allreduce: Reduce then Bcast
2031 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2032 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2033 Alltoallv: flat tree, like ompi
2035 * Add support for optimized collectives (Bcast is now binomial by default)
2036 * Port smpirun and smpicc to OS X
2039 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2040 Shout out if you used it.
2043 * Display the list of still queued messages in SG mode when existing
2047 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2048 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2049 * Add xbt_str_from_file(FILE*)
2050 * Add xbt_dict_get_key achieving a linear reverse search
2051 * Remove the context module
2053 Portability report of this version:
2054 * Main portability targets:
2055 - Linux(debian)/x86/context
2056 - Linux(debian)/x86/pthreads
2057 - Linux(debian)/amd64/context
2058 - Linux(debian)/amd64/pthreads
2059 On these, we still have the eratic breakages of gras/pmm and
2060 amok/saturate_sg reported in previous version. We still think
2061 that the tests are the cause of the fault, not the tested code.
2063 - Mac OSX Leopard/x86/context
2064 Still false negative in tesh autotesting.
2065 Smpi still fails, but this time because readlink does not accept -f
2066 Everything seems to work properly beside of that.
2069 - AIX version 5.3 (only tested contexts this time)
2070 Smpi still fails there because mktemp is not installed.
2071 Everything seems to work properly beside of that.
2073 I managed to compile it for the first time, but several breakages.
2074 Won't delay the release for this exotic platform.
2076 * Windows: it's still lagging behind. If you want to help, please
2079 Timing report of this version:
2080 This version seem to be more than 5% faster than 3.3.1 (on linux
2081 64bits with contextes). The gain is less than expected, we are
2082 investigating this for next release.
2084 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2086 SimGrid (3.3.1) stable; urgency=low
2089 * Implement a --cfg-help to show existing configuration variables
2090 * Build chain do not require doxygen in maintainer mode
2093 * fix a bug on struct sizeof computation, which prevented the
2094 exchange of arrays of structs in some conditions
2095 - added a regression test about this in datadesc_usage
2096 * Allow the exchange of 0-long dynamic vectors.
2097 - for that, use -1 as indicator of dynamic size instead of 0
2098 - This implied to change any size from unsigned long to long,
2099 reducing a bit communication abilities, but I guess that with
2100 64bits being quite common, this is more than enough.
2101 - This also induce a protocol change, thus bumping network protocol
2102 version from 0 to 1 (if we have external users, we have to get
2103 clean on that point too ;)
2104 - added two regression tests about this in datadesc_usage
2105 * Be more verbose when propagating local exceptions
2106 This helps debugging.
2107 * Display the status of simulated processes when receiving SIGINT in
2111 * Allow to control the simulation from a trace file.
2112 New functions MSG_action_register() and MSG_action_trace_run()
2113 The first one allows to associate a function execution to each
2114 kind of action while the second one parses a trace file and
2115 triggers the corresponding actions within the system.
2116 For now, only a toy example is provided in examples/msg/actions
2117 * Add an exemple of process migration in examples/msg/migration
2118 * Fix a bug in task exchange which broke MSG_task_get_sender()
2119 Add a teshsuite regression test for that.
2120 [Bug: if MSG_task_get_sender() is called after sender exit,
2122 * Fix a bug which prevented suspend/resume to work properly
2123 * Display the status of simulated processes when receiving SIGINT
2124 This fixes a regression of v3.3. due to the introduction of SIMIX
2125 * Bug fixing in failure management:
2126 - trace could not start by a failure at time 0
2127 - failure during communications were not working
2130 * Add SIMIX_process_set_name() to change the name of the current
2131 process in the log messages.
2132 * Store smx_hosts in a dict since we only retrieve them by name
2133 * Move the configuration infrastructure to surf
2136 * Move the configuration infrastructure to surf
2139 * Massive internal cleanups:
2140 - Store internal structures on processes instead of hosts (allows
2141 to have more than one process per host, in addition of being more
2143 - Cleanup the initialization/finalization process
2144 - Kill a whole bunch of unneeded synchronization:
2145 processes run in exclusive manner within the simulator
2146 - Move queues from global tables to process data fields
2148 - now accept -platform and -hostfile arguments
2149 - Pass the right rank value to processes according to the hostfile
2150 * Compile the examples by default, and use them as regression tests
2151 * Implement MPI_Wtime()
2152 * Change the reference speed to a command line option
2155 * TCP_gamma can now be specified as command line option using
2156 --cfg=TCP_gamma:10000000.0
2157 * Change the --surf-path cmd line option into --cfg=path:
2160 * Also include strbuff from xbt.h public header
2161 * xbt_ex_display(): do not free the exception after displaying
2162 This allows to do more with the given exception afterward.
2163 Users should call xbt_ex_free() themselves.
2167 Portability report of this version:
2168 * Main portability targets:
2169 - Linux(debian)/x86/context
2170 - Linux(debian)/x86/pthreads
2171 - Linux(debian)/amd64/context
2172 - Linux(debian)/amd64/pthreads
2173 These targets fail about 1/10 of times on gras/pmm, but we believe
2174 that this is because of the test, not because of simgrid.
2175 amok/saturate_sg fails even more rarely, and the test may not be
2178 - Mac OSX Leopard/x86/context
2179 The test suite still spits tons of errors because some obscure
2180 force prevents us from removing the temporary directories
2181 arguing that they still contain some metadata I've never heard of.
2182 Smpi fails because seq is not installed.
2183 Everything seems to work properly beside of that.
2186 - AIX version 5.3 (both contexts and pthread)
2187 Smpi still fails there because mktemp is not installed.
2188 XML inclusions seems rosty on AIX.
2190 * Windows: it's still lagging behind. If you want to help, please
2193 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2195 SimGrid (3.3) stable; urgency=high
2199 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2200 [Malek Cherier & Mt]
2202 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2205 Until now, GRAS were using MSG as an interface to SURF. It was
2206 quite difficult because both interface have several differences
2207 (MSG channels vs GRAS sockets were the most notable point).
2209 This also opens the gate to SMPI (which should occur soon) and speed
2210 up simulations by to 40% (even if it were not the main goal).
2212 **************************************
2213 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2214 **************************************
2215 The point is that events occuring at the exact same timestamp are
2216 not scheduled in the same order with the old and new version. This
2217 may be enough to completely change the execution of simulations in
2218 some cases. Sorry for the inconvenience.
2220 * Cleanup and upgrade the XML format to push further scalability
2221 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2223 * Improve the testing infrastructure with tesh. Now a very large part of
2224 the code is tested not only by being run but also by checking that the
2225 output match an expected output [Mt].
2227 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2228 is really less memory-demanding, which should allow you to use
2229 larger files in SimGrid [AL].
2231 * Inform valgrind about our contextes, so that it becomes usable
2232 with the default (and more effecient) version of SimGrid
2233 [contributed by Sékou Diakite, many thanks]
2236 * Introduce a listener thread in charge of receiving incoming
2237 messages from the network. It allows to overlap communication and
2238 computation but most notably, it removes some stupid deadlocks due
2239 to the fact that so far, a process could not send and receive at
2240 the same time. This made most non trivial communication schema
2242 * Convert the PIDs from long int to int to match the MSG ones (and
2243 linux ones too) [Mt]
2244 * New function: gras_agent_spawn() to launch a new process on
2245 current host. Only working in simulation for now. [Mt]
2246 * New function: gras_os_hostport() returning a constant form (ie,
2247 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2250 * Make the backtrace of exceptions more human readable [Mt]
2251 * New module: xbt/str [Mt]
2252 a ton of string utility functions (split, join, printf to a newly
2253 allocated buffer, trim, etc)
2254 * New module: xbt/hash [Mt]
2255 SHA1 hashing algorithm (more to come if needed)
2256 * New module: xbt/synchro [Mt]
2257 synchronization tools (mutex and conditions) working the same way
2258 in simulation and in real life (mainly useful for GRAS, but not
2260 * New module: xbt/queue [Mt]
2261 classical producer/consumer synchronization scheme
2262 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2263 (using the classical functions will get serialized) [Mt]
2264 * Make dictionary internal table dynamic. No need to specify its size
2265 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2267 * Make sure the log channels are organized as a tree under windows
2268 (because of ANSI C compatibility issue, any channel were child of
2269 root directly) [Mt].
2272 * Cleaned many thing in surf and fixed a few bugs [AL].
2273 * Add a nice command line configuration mechanism to compose models [AL].
2274 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2275 the previous one (KCCFLN05). It relies on something that looks like
2276 a max-min sharing mechanism but cannot be written as such. A new solver
2277 was thus designed [AL].
2278 * Add a new solver to lmm. Based on Lagrange optimization and
2279 gradient-based descent, it enables to efficiently maximise systems s.a
2281 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2283 This solver enables to propose two new network models for TCP Reno and
2284 TCP Vegas based on Low's work. These models still need to be fully
2285 tested though [Pedro Velho].
2288 * Bug fix in SD_simulate. Now the time bound given as argument is
2290 * Use the new parallel task model (ptask_L07) as default.
2291 * Use the SURF command line configuration mechanism.
2292 * 0-size tasks (for synchronization) should now work.
2294 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2296 SimGrid (3.2) stable; urgency=high
2300 We still experience issues on this platform, but we believe that at
2301 least MSG is usable.
2303 GRAS API BREAKAGE (for simplification purpose, sorry):
2304 * the gras_msgtype_by_name is not used anymore. Instead of
2305 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2306 you can write (and must)
2307 gras_msg_send(toserver, "request", &request);
2308 - If you still want to pass a gras_msgtype_t to the function (to cache
2309 the type and avoid the lookup time), use the gras_msg_send_() variant.
2310 - Impacted functions:
2311 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2312 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2313 * The callbacks are now expected to return 0 when everything went well
2314 (just like the main() function)
2316 GRAS new features and improvements:
2317 * New module mechanism where user code can use per process globals [Mt]
2318 This is similar to gras_userdata_*() functions, but for libraries. It
2319 factorize some code developped over and over in the examples and AMOK.
2320 It has still to be documented and used (only amok/peermanagement is
2322 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2323 loss when they were small enough to fit into the buffer and sent quickly
2324 enough so that they can all get received in one shoot.
2325 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2326 if not found. Use the *_or_null() variant for the old semantic.
2327 * In gras_msg_handle, do not discard messages without callback.
2328 They are probably messages to be explicitly awaited later (ie, proofs of
2329 mis-synchronization in userland since they are sent before being awaited)
2330 No big deal usually.
2331 * gras_socket_meas_send/recv: semantic changed!
2332 The numerical arguments used to be (1) the total amount of data to send
2333 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2334 messages. This was need for the fool willing to send more than MAXINT
2335 bytes on quite fat pipes.
2338 * Do really rename the hostmanagement module to peermanagement. [Mt]
2339 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2340 but this is rather new and this was documented in the module
2341 documentation (poor excuses, I admit)
2342 * Bandwidth measurement semantic changed! This follows the changes to
2343 gras_socket_meas_send/recv explained above.
2346 * A sequential mode has been added to the workstations. When a workstation
2347 is in sequential mode, it can execute only one task, and the other tasks
2348 are waiting in a FIFO. [Christophe Thiery]
2351 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2352 model for SIMDAG. [Christophe Thiery]
2353 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2354 equal to 0, which caused some bad behaviors in
2355 saturated_constraint_set_update. I now use a threshold mechanism like in
2359 * When running manually src/testall, you select specific units [Mt]
2360 testall is the result of our cunit mechanism, and should replace all
2361 the scripty thingy around since bash don't run easily on billware.
2363 * A mallocator system has been added. [Christophe Thiery]
2364 Mallocators allow you to recycle your unused objects instead of freeing them
2365 and allocating new ones.
2367 Documentation update:
2368 * FAQ reworking + New FAQs:
2369 - "Valgrind spits tons of errors!" [Mt]
2370 - "How to repport bugs" [Mt]
2371 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2372 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2373 - Communication time measurement within MSG [AL]
2374 - I experience weird communication times when I change the latency [AL]
2375 * GRAS tutorial [Mt]
2377 - an introduction to the framework and to the used communication model
2378 - an initiatic tour introducing the most proheminent features:
2380 . Lesson 0: Installing GRAS
2381 . Lesson 1: Setting up your own project
2382 o Part 2: Message passing
2383 . Lesson 2: Exchanging simple messages
2384 . Lesson 3: Passing arguments to the processes (in SG)
2385 . Lesson 4: Attaching callbacks to messages
2386 . Lesson 5: Using globals in processes
2387 . Lesson 6: Logging information properly
2388 . Lesson 7: Using internal timers
2389 . Lesson 8: Handling errors through exceptions
2390 . Lesson 9: Exchanging simple data
2391 . Lesson 10: Remote Procedure Calling (RPC)
2392 . Lesson 11: Explicitely waiting for messages
2393 . Recapping of message passing features in GRAS
2394 - A HOWTO section containing:
2395 o HOWTO design a GRAS application
2396 More are due, of course. They will come latter. In the meanwhile, you can
2397 check the examples which are still here.
2399 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2401 SimGrid (3.1) stable; urgency=high
2405 There was a stack corruption somewhere, visible only when optimizing
2406 with these versions. [Vince]
2409 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2410 module that enabled to play with Directed Acyclic Graphs. It is built
2411 directly on top of SURF and provides an API rather close to the old
2412 SG. Some old codes using SG are currently under rewrite to check that
2413 all needful functions are provided. [Christophe Thiery]
2416 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2417 extension of the classical CLM03 model that gracefully handles
2418 failures. This is now the default model for MSG and GRAS. It doesn't
2419 handle parallel tasks yet though. [AL]
2420 * Bug fix: Weights were not correctly set in the network part.
2421 WARNING: This may have resulted in incorrect results with simulations
2422 where there are more than one flow on a given link. [AL]
2425 * After a (long ?) discussion on simgrid-devel, we have decided that the
2426 convention we had on units was stupid. That is why it has been decided
2427 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2428 WARNING : This means that all previous platform files will not work as
2429 such with this version! A warning is issued to ask users to update
2431 A conversion script can be found in the contrib module of the CVS, under
2432 the name contrib/platform_generation/surfxml_update.pl [MQ]
2435 * Bug fix: Processes were started in reverse order, wrt deployment file.
2436 WARNING: if your code relies on this bug, please fix it. [AL]
2437 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2438 being executed on two different locations. [AL]
2439 * Bug fix: Failures are now better supported thanks to Derrick's tests
2440 (there was many failure situations I hadn't thought of and that weren't
2441 correctly handled). [AL]
2442 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2446 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2447 Exception occurring on server-side are propagated back to client (!).
2449 API CHANGE: the callback changed their prototype. Change:
2450 int my_handler(gras_socket_t expeditor, void *payload_data) {
2452 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2453 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2455 * New! function: gras_msg_handleall to deal with all messages arriving
2456 within a given period.
2457 * New! function: gras_socket_server_range to get a server socket in a
2458 range of port numbers (ease to avoid port number conflicts) [MQ]
2459 * New! gras processes display their backtrace when they get a SIGUSR1
2460 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2461 Sweet to debug RL processes [MQ]
2465 - Do not force experiment sizes to be expressed in kb, or it becomes
2466 impossible to measure the latency this way (needs one byte-long tests)
2467 WARNING: this changes the amok_bw_* function semantic. [MQ]
2468 - Implements the link saturation stuff. [MQ]
2469 * Peer management module:
2470 New! module factorizing code that we wrote over and over [MQ].
2473 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2474 - Test units are placed directly into the library code, they get extracted
2475 automatically and placed into the src/testall binary.
2476 - Convert most of the XBT tests to this system.
2477 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2478 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2479 live in the same C code. This required to change a little bit the API
2480 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2481 * New module: sparse graph structure with basic algorithms (this is work
2482 in progress and the API is not considered to be frozen yet). [AL]
2483 * Display more information on backtraces: source line & function names are
2484 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2485 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2486 * Reworked a little bit some #include statements to load only required
2487 headers. Some user code that relied on SimGrid to include stdlib or
2488 stdio may need to include it by themselves. [AL]
2489 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2491 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2492 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2493 but I guess that almost nobody used those parts. [MQ]
2495 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2497 SimGrid (3.0.1) stable; urgency=low
2500 * Unfortunately, I had missed 5 misnamed functions:
2501 xbt_fifo_item_t xbt_fifo_newitem(void);
2502 void xbt_fifo_freeitem(xbt_fifo_item_t);
2503 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2504 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2505 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2506 They're now deprecated. Please use their new versions:
2507 xbt_fifo_item_t xbt_fifo_new_item(void);
2508 void xbt_fifo_free_item(xbt_fifo_item_t);
2509 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2510 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2511 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2513 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2514 * Documentation: xbt_log module unmercifully reworked [MQ]
2515 * Bugfix: there was a problem with the ending of contexts with
2516 the pthread backend. It caused some weird deadlock or behavior
2517 depending on the pthread implementation. [AL]
2518 * Bugfix: get the exceptions raised in the simulator repport where
2519 and why they come from when they are not catched in time [AL, MQ]
2522 * Bugfix: Do repport the error when two non-connected hosts try to
2523 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2526 * Add additionnal checkings on communications. Assert that two
2527 communicating hosts are connected by a set of links... [AL]
2530 * Add additionnal checkings on channel values in communication [AL]
2531 * New: MSG_task_get_source to see on which host a task was generated [HC]
2532 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2533 the number of tasks waiting to be received on channel and sent
2535 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2536 waits for the first task coming from a given host.. [AL]
2538 GRAS new functionnalities: [MQ]
2539 * Enhance the parsing macro to allow the size of multidimentional objects
2540 to be given thru annotations.
2541 * New example (and documentation): Matrix Multiplication a la RPC
2542 (as when I was young!) and fix a bunch of bugs found on the way.
2544 GRAS performance improvements: [MQ]
2546 * Reduce the amount of cbps creation/destruction by making it static to
2547 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2550 * Change libdata to a set so that we can search for stuff by ID (and thus
2551 reduce the insane amount of dict lookups)
2554 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2555 RL communication on top of SG since it's so uneffective.
2556 It may also be used for inter-thread communication in RL, one day.
2557 * Use gras_datadesc_copy() to exchange messages on top of SG
2559 - improve message exchange performance on top of SG
2560 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2561 * Don't exchange on the network the size of the used part of buffer,
2562 instead, specify the possible buffer size to read().
2564 - reduces the amount of read/write calls (one pair per exchange)
2565 - reduces the amount of exchanged data (the size)
2566 - allows to retrieve all arrived data on receiver side, if we don't need
2567 it right now (subsequent read will peek the buffer)
2568 - allows the receiver to proceed with the begining of the stream before
2569 everything is arrived
2570 - make it possible to build an iov transport (using readv/writev)
2572 - take care of the data with non-stable storage (like stacked data),
2574 * If possible, TCP send uses vector I/O (when writev() is here)
2575 - Don't use it for receive since we send data sizes and data on the
2576 same stream, so we wouldn't be able to chain large amount of chunks
2577 before having to flush the stuff to read the size.
2578 * Rework the transport plugin mechanism to simplify it and reduce the
2579 amount of pointer dereferencement when searching for the right function
2582 * I guess that now, we do almost as few system calls as possible while
2583 doing as few data copy as possible.
2585 To improve it further, we could try to send all the sizes first and then
2586 all the data (to use iov on receiving size), but it's only a partial
2587 solution: when you have 2 dimensional data, the sizes of the second
2588 dimension is data of the first dimension, so you need 3 streams.
2590 I'm not sure the potential performance gains justify the coding burden.
2592 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2594 SimGrid (3.00) stable; urgency=high
2597 * New! Give the possibility to hijack the surf parser and thus bypass
2598 MSG_create_environment and MSG_launch_application. Have a look at
2599 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2601 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2603 SimGrid (2.96) unstable; urgency=low
2608 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2609 This deprecates the xbt_error_t mechanisms.
2610 It modifies (simplifies) all XBT and GRAS API.
2611 MSG API keeps unchanged (exceptions raised by XBT are catched from
2612 within MSG and masked with existing error handling facilities)
2615 * New! Add a FATPIPE model. [AL]
2616 * New! Add a parallel task model. [AL]
2617 * New! Add automatically a loopback interface (in the default
2618 network model) if none was precised.
2621 * Bugfix: MSG_process_resume now works with the current running process.
2623 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2624 * Modification of MSG_task_get_compute_duration. Once a task has been
2625 processed, the value returned by this function is now equal to 0. [AL]
2626 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2627 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2628 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2629 of a m_task. For now, MSG_task_cancel only works with computation
2631 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2632 of the processor (in Mflop/s) regardless of the current load on the
2634 * API Change: use proper naming convention for MSG_getClock and
2635 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2637 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2638 This function changes the priority of a computation task. This priority
2639 doesn't affect the transfer rate. A priority of 2 will make a task
2640 receive two times more cpu power than the other ones. This function
2641 has been added to suit the needs of Nguyen The Loc and hasn't been that
2642 much tested yet. So if it fails, please report it and send me your code.
2644 * API Change: removed all functions and types that were marked "deprecated"
2645 since many months. Renamed MSG_global_init_args to MSG_global_init.
2647 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2649 SimGrid (2.95) unstable; urgency=low
2652 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2653 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2654 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2657 * Bugfix: Allow absolute paths for platform description files [MQ]
2658 * Bugfix: do free the variables after use. Leads to drastic performance
2660 * Implement max_duration (ie, timeouts) on resources [AL]
2663 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2665 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2666 the needed support (timeouts) [AL]
2669 * Implement measurement sockets. You can now get the bandwidth between two
2670 hosts thanks to AMOK (see below). [MQ]
2671 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2672 dynar over the network (yeah) [MQ]
2673 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2674 * simple/static token-ring example. [Alexandre Colucci and MQ]
2675 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2676 in place before [MQ]
2679 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2680 type parsing. "short" and "long" modifiers were also ignored; other
2681 modifier (such as reference level) are still ignored. [MQ]
2682 * Embeed the buffer size within the buffer itself on SG. [MQ]
2683 That way, send() are atomic and cannot get intermixed anymore (at least
2684 the ones which are less than 100k; bigger messages still have the issue)
2685 * Array size pushed by the field, not by the field type (or each
2686 and every long int will push stuff to the cbps) [MQ]
2687 * use select() to sleep since it allows to portably sleep less than one
2690 GRAS (minor cleanups)
2691 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2694 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2695 * type category 'ignored' killed as it was never used and were difficult
2697 * whether a type can cycle or not is now a flag, leaving room for more
2698 flags (as "ignored", if we feel the need one day ;)
2699 * Rename raw sockets to measurement sockets since "raw" has another
2700 meaning in networking community.
2703 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2704 and offers features not belonging to GRAS but that most applications
2705 need. One day, it may be a set of plugins loadable at runtime.
2706 * New module: bandwidth
2707 bandwidth measurement between arbitrary nodes running this module. [MQ]
2709 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2711 SimGrid (2.94) unstable; urgency=low
2713 The first beta release of SimGrid 3 !
2717 * Update the main page and the FAQ. Adding references to gforge.
2720 * Add a gras_os_getpid function.
2723 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2724 * Extend the logs so that they also print PID, hostname, date, ... if
2726 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2727 and kill the old version which were in testsuite/
2728 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2732 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2733 for users. Instead, each programming environment (GRAS, MSG,...) use it
2734 when needed to provide such a feature to users.
2735 Don't shortcut the mechanism or you will also shortcut the virtualization
2736 you need on the simulator.
2740 * Cleanups in configury with regard to compile optimization/warning flags.
2741 Also add -fno-loop-optimize to any powerpc since it's the optimization
2742 killing gcc (< 3.4.0).
2743 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2744 needed by MSG examples complications
2745 * Borrow configury beautifications from PHP
2748 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2749 warning (thanks loris for stumbling into this one).
2750 * Bugfix: stop loading private headers (gras_config.h) from the public
2754 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2755 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2756 RL would imply a DNS resolver, which is *hard* to do in a portable way
2757 (and therefore delayed).
2758 * Implement a real timer mechanism and use it in timing macros. This allows
2759 to avoid rounding errors and get a 0.000005 sec precision in timing
2760 macros. While I was at it, various cleanups:
2761 - allow to declare more than one timed section per file (fix a stupid bug)
2762 - move some private declaration to the right place
2763 - merge conditional execution and timing macros into emulation module
2764 - document the module
2765 - make sure the module cleanups its mess on gras_exit
2766 * Documentation improvements:
2767 - (new) how to compile applications using GRAS
2768 - (new) emulation support (timing macros)
2770 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2772 SimGrid (2.93) unstable; urgency=low
2774 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2777 - Use Paje properly where used. Still to be sanitized properly.
2778 - Portability fix: Add an implementation of the contexts using pthread
2782 - Add xbt_procname(): returns the name of the current process.
2783 Use it to show the current process's name in all logging.
2785 - fix detection of older flex version and the reaction, since we do
2786 depend on modern ones (we use lex_destroy)
2787 - Better separation of SG and RL in the libs: remove all simulation code
2788 from libgras. As a result, this lib is now only 200k when stripped.
2789 Some of the xbt modules may also be duplicated (two sets and such) and
2790 should be cleaned/killed before SG3.
2791 - Insist on using xlC on AIX because of weird problems involving gcc there.
2792 - Cleanup the make remote stuff. This is now done by scripts
2793 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2794 mainly for our private use, but we're working on changing them to user
2797 - Bugfix: flush the socket on close only if there is some *output*.
2798 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2800 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2803 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2805 SimGrid (2.92) unstable; urgency=low
2807 Alpha 3 on the path to SimGrid 3
2811 - New! First try of benchmarking macros.
2812 - New! First try so that gras_stub_generator generate deployment and
2813 remote compilation helpers.
2815 - Bugfix: Initialization fix in msg_test.
2819 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2821 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2823 - API Change: gras_os_sleep to take the amount of seconds as a double.
2824 Accepting an int was error prone since it was the only location where
2825 seconds were coded as such. It leaded to damn rounding errors.
2826 - Bugfix: Hard to belive that timers ever worked before this.
2828 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2830 SimGrid (2.91) unstable; urgency=low
2832 Alpha 2 on the path to SimGrid 3
2836 - Bug fix in the lmm_solver.
2838 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2839 through the function MSG_paje_output.
2840 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2841 - It is possible to bound the rate of a communication in MSG with
2842 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2843 to write it in the changelog).
2844 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2849 - Various cleanups to the autotools stuff
2850 - Begin to move Gras examples to examples/gras/
2851 - Let make distcheck work again (yeah!)
2853 - documentation overhauled using doxygen.
2854 gtk-doc-tools is dead in SimGrid now.
2855 - Automatically extract all existing logging categories, and add the list
2856 to the documentation (long standing one, to say the less)
2858 - Cleanup the known architecture table. Reorder the entries to group what
2859 should be, and use a more consistent naming scheme.
2860 (some of the test dataset are still to be regenerated)
2861 - New! Allow library to register globals on each process just as userdata
2863 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2864 lookup time (for now).
2865 Use it in msg and trp.
2866 This cleans a lot the internals and helps enforcing privacy of the
2867 headers between the gras components.
2868 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2869 - Bugfix: gras_os_time was delirious in RL.
2870 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2872 - Reenable GRAS now that it works.
2874 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2876 SimGrid (2.90) unstable; urgency=low
2878 Alpha 1 on the path to SimGrid 3
2880 * It is a long time since the last release of SimGrid. I'm sorry about
2881 that but as I had told you, I was rewriting a lot of things. I apologize
2882 to those who had been reporting bugs to me and that I had not answered.
2883 If your bug is still in the new version, please tell me. Here is a
2884 summary of the main changes.
2886 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2887 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2888 portability, speed, and a lot more but you'll figure it out later.
2889 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2890 I wanted to release the new MSG as soon as possible and I have
2891 broken GRAS, which is the reason why, for now, only MSG is fully
2892 functional. A laconic description of these projects is available
2893 in the documentation.
2895 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2896 kernel optimized for our needs (called SURF but only the developers
2897 should use it). Hence, MSG is now roughly 30 times faster and I think
2898 that by rewriting a little bit MSG, I could event speed it up a little
2899 bit more. Beside the gain in speed, it is also much easier to encode a
2900 new platform model with SURF than it was with SG. More to come...
2902 * REVOLUTION 3: I have tried to change a little as possible the API of
2903 MSG but a few things really had to disappear. The main differences
2904 with the previous version are :
2905 1) no more m_links_t and the corresponding functions. Platforms are
2906 directly read from a XML description and cannot be hard-coded
2907 anymore. The same format is used for application deployment
2908 description. The new format is described in the documentation.
2909 Have a look in tools/platform_generation. There is a tiny script
2910 that converts from the old platform format to the new one. Concerning
2911 the application deployment format, parsing the old one is tricky.
2912 I think most of you should however be able to convert your files. If
2913 it is really an issue, I can write a C code that does the conversion.
2915 2) the toolbox tbx does not exist anymore. We now have a library
2916 with much more data-structures but without the hash-tables (we have
2917 dictionaries that are much faster).
2919 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2921 *****************************************************************************
2922 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2923 * don't feel like converting the oldies. (Mt) *
2924 *****************************************************************************
2927 Version 2.90: "the long awaited one"
2928 - Finished rewriting and debugging MSG. Rewrote the documentation.
2929 - disable GRAS for now since it needs to be ported to the newest SG
2932 - Finish the port to windows (using mingw32 for cross-compile)
2935 - Main loop and datastructures of SURF. A cpu resource object is
2936 functional. Surf can thus be used to create cpu's with variable
2937 performance on which you can execute some actions.
2939 2004-11-15 Martin Quinson
2940 - Port to ARM. Simply added the alignment and size descriptions. Should
2941 work, but the ARM machines are so slow that I didn't had the opportunity
2942 to 'make check' over there yet.
2944 2004-11-15 Arnaud Legrand
2945 - Trace manager now written. It uses a heap structure and is therefore
2946 expected to be efficient. It may however be speeded up (particularly
2947 when many events occur at the same date) by using red and black
2948 trees. One day maybe...
2949 - Max-min linear system solver written. It uses a sparse matrix
2950 structure taking advantage of its expected use. Most operations are
2951 O(1) and free/calloc are called as few as possible. The computation of
2952 the minimum could however be improved by using a red and black tree
2955 2004-11-03 Arnaud Legrand
2956 - Rename every gras_* function that was in xbt/ to its xbt_
2958 - Add a heap and a doubly-linked list to xbt
2959 - Added a dichotomy to the dictionaries. make check works as well before
2960 so I assume that the patch is correct. I do not know however if things
2961 run effectively faster than before now. :)
2963 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2964 SimGrid, and the version number is bumped to 2.x
2966 2004-10-29 Martin Quinson
2967 - Introduction of the remote errors.
2968 They are the result of a RMI/RPC on the remote machine.
2969 ErrCodes being scalar values, you can't get the host on which those
2970 errors did happen. Extending the error mechanism as in Gnome is possible.
2971 No idea yet whether it is a good idea.
2973 2004-10-28 Martin Quinson
2974 - Interface revolution: the Starred Structure Eradication.
2975 I used to do typedef struct {} toto_t; and then handle *toto_t.
2976 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2977 - ???_t is a valid type (builded with typedef)
2978 - s_toto_t is a structure (access to fields with .)
2979 - s_toto is a structure needing 'struct' keyword to be used
2980 - e_toto_t is an enum
2981 - toto_t is an 'object' (struct*)
2983 typedef struct s_toto {} s_toto_t, *toto_t;
2984 typedef enum {} e_toto_t;
2985 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2986 s_toto_t) is private.
2988 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2989 so that it can set the variable to NULL. It was so for dicts and sets,
2990 it changed for dynars.
2992 - Fix a bunch of memleaks in dict_remove
2993 - Fix a bug in sg/server_socket opening: it failed all the time.
2995 2004-10-07 Martin Quinson
2996 - Speed up dynar lookup operation a bit.
2998 gras_dynar_get is dead.
3000 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3001 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3002 (which gives you the address of the stored data).
3004 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3005 copy of the data using an affectation to do the job and not a memcpy.
3007 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3008 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3010 It does not really speedup the dynar test because they are
3011 setting elements all the time (and look them seldom). But the dict does
3012 far more lookup than setting.
3014 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3016 2004-10-05 Martin Quinson
3017 - Allow to (en/dis)able the cycle detection at run time.
3019 Whether we should check for cycle or not is now a property of each
3020 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3021 datadesc_cycle_unset allow to remove this property when previously set.
3023 Note that the cycle detection is off by default since it impacts the
3024 performance. Watch the data you feed GRAS with ;)
3026 This property is hereditary. Any element embedded in a structure having it
3027 set have it set for the time of this data exchange.
3029 You should set it both on sender and receiver side. If you don't set it on
3030 sender side, it will enter an endless loop. If you forget on receiver
3031 side, the cycles won't be recreated after communication.
3033 - Header reorganization.
3034 Kill gras_private.h, each submodule must load the headers it needs.
3036 2004-10-04 Martin Quinson
3037 - Interface revolution: do not try to survive to malloc failure.
3039 Now, gras_malloc and friends call gras_abort() on failure.
3040 As a conclusion, malloc_error is not a valid error anymore, and all
3041 functions for which it was the only gras_error_t return value are
3042 changed. They now return void, or there result directly.
3043 This simplify the API a lot.
3045 2004-09-29 Martin Quinson
3046 - Re-enable raw sockets.
3047 Created by gras_socket_{client,server}_ext;
3048 Used with gras_raw_{send,recv}
3051 It should allow to kill the last bits of gras first version soon.
3053 This is not completely satisfactory yet (duplicate code with
3054 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3057 - Simplify transport plugin (internal) interface by not passing any
3058 argument to _server and _client, but embedding them in the socket
3061 2004-09-28 Martin Quinson
3062 - Finish the port to AIX.
3063 autoconf was my problem (segfault within the malloc replacement
3064 function. No idea why)
3066 2004-09-16 Martin Quinson
3067 - Fix some size_t madness on 64bit architectures.
3069 2004-09-08 Martin Quinson
3070 - Reduce the number of system headers loaded, overload some more system
3071 calls (such as malloc to cast the result of the system one, and work
3073 - Fix and reintroduce the config support
3075 2004-09-07 Martin Quinson
3076 - Source code reorganization to allow Arnaud to surf all over there.
3077 - Allow to document the logging categories.
3078 - Remove all uppercase from logging categories and useless cleanup in names.
3080 2004-08-18 Martin Quinson
3081 Version 0.6.2 (protocol not changed; API changed)
3082 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3083 gras_error_t), and NULL when not found. Functions expecting a msgtype
3084 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3085 hopefully usefull message.
3086 - Portability to prehistoric sparcs again
3088 2004-08-17 Martin Quinson
3089 Version 0.6.1 (protocol not changed; ABI not changed)
3090 - prealloc some buffers to speed things up
3092 2004-08-11 Martin Quinson
3093 Version 0.6 (protocol not changed; ABI expended)
3094 - The parsing macro can deal with the references, provided that you add
3095 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3097 2004-08-09 Martin Quinson
3098 Version 0.5 (protocol not changed; ABI changed)
3099 - Allow to off turn the cycle detection code in data exchange at
3100 compilation time. It should be at run time, but I'm short of time (and
3101 the config stuff is still broken). That way, we keep dict out of the
3102 critical path, which is good because the performance is poor:
3103 - search not dichotomial yet
3104 - dynar give no way to access their content and memcpy everytime
3105 - In composed data description (struct, ref and so on), stop foolness of
3106 keeping the subtype's ID, but store the type itself. This keeps sets out
3107 of the critical path, which is good since they rely on dynar and
3108 dictionnaries. The only loose of that is that we cannot detect the
3109 redeclaration of a structure/union with another content (but I'm not sure
3110 the code detected well this error before anyway). We still can detect
3111 the redefinition discrepancy for the other types.
3112 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3113 it breaks the code because of type-punning used all over the place).
3114 This breaks on all non-gcc architectures (for now).
3116 All those changes (plus the buffer of last time) allow me to gain 2 order
3117 of magnitude on cruel tests consisting of 800000 array of integers on two
3118 level of a hierarchical structure (200 secondes -> 4 secondes)
3121 - the selector of reference must now return the type it points to, not
3122 the ID of this type.
3124 2004-08-06 Martin Quinson
3125 Version 0.4 (protocol changed; ABI not changed)
3126 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3127 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3128 - Implement a buffer transport, and use it by default (it relies on tcp in
3129 real life and on sg in simulation).
3130 That's a bit hackish since I had a new field to the structure to store
3131 its data without interfering with the subtype ones. Inheritance
3132 is tricky in C. And that's a kind of reverse inheritance with one class
3133 derivating two classes. Or maybe a game with java interfaces. Anyway,
3134 that's damn hard in C (at least).
3135 Moreover, I got tired while trying to ensure plugin separation and
3136 genericity in SG mode. MSG wants me to do weird things, so let's go for
3137 cruel hacks (temporarily of course ;).
3138 See comment in transport_private.h:71
3139 - do not include all the _interface headers in private but in the files
3140 which really need them (to cut the compilation time when they are
3143 2004-07-26 Martin Quinson
3144 Version 0.3 (protocol not changed; ABI changed)
3145 - Major overhault of the datadesc interface to simplify it:
3146 - shorted the function names:
3147 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3148 - add a trivial way to push/pop integers into the cbps without malloc.
3149 This allows to make really generic sub_type description, which simply
3150 pop their size of the stack.
3151 - add a function gras_datadesc_ref_pop_arr() which does what users want
3152 most of the time: Declare a dynamic array (which pops its size of the
3153 stack) and declare a reference to it. Poor name, but anyway.
3154 - kill the post-send callback, add a post-receive one
3156 2004-07-23 Martin Quinson
3157 Version 0.2 (protocol changed; ABI changed)
3158 - add some testing for cpbs in the test cases, and fix some more bugs.
3159 This invalidate again the little64 data file, since I cannot regenerate
3161 - remove an awfull optimization in the logging stuff, allowing me to:
3162 - understand it again
3163 - learn gcc how to check that the argument match the provided format
3164 - fix all errors revealed by gcc after that
3165 - internal keys of dict are not \0 terminated. Deal with it properly in
3166 loggings instead of segfaulting when the user want to see the logs :-/
3168 2004-07-22 Martin Quinson
3169 - Fix some stupid bug preventing cbps (callback postit) from working
3171 2004-07-21 Martin Quinson
3172 - Some documentation cleanups
3173 - remove the useless last argument of msgtype_declare
3174 - rename the Virtu functions to fit into the 'os' namespace
3175 - move headers src/include -> src/include/gras/ and stop fooling with
3176 gras -> . symbolic link
3177 - make distcheck is now successful
3179 2004-07-19 Martin Quinson
3181 - Build shared library also
3182 - Install html doc to the right location
3183 - stop removing maintainer files in make clean
3184 - build tests only on make check
3186 2004-07-13 Martin Quinson
3188 - No major issue in previous version => change versioning schema
3189 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3190 corresponding dataset.
3192 2004-07-11 Martin Quinson
3194 - Get it working with any kind of structure (we can compute the padding
3195 bytes remotely for all the architectures I have access to)
3196 - Implement the structure parsing macro (still not quite robust/complete)
3197 - Improvement to the remote testing toysuite
3199 2004-07-10 Martin Quinson
3200 [autoconf mechanism]
3201 - get ride of a bunch of deprecated macros
3202 - actually run the test for two-compliment, not only compile it :-/
3203 - test whether the structures get packed (and bail out if yes. Damn.
3204 Alignment is a serious matter)
3205 - test whether the structures get compacted (but respecting the alignment
3206 constraints of each types)
3207 - test whether the array fields of structures can straddle alignment boundaries
3209 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3210 (took me 2 hours to find that bug, looking at the wrong place)
3211 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3212 before sending/receiving and are used to compute the offsets of fields
3213 - Given that padding size depend even on compiler options, keep track of
3214 alignment and aligned_size only for the current architecture. Not a big
3215 deal since we send structure fields one after the other (seems
3217 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3218 program, let it work on linux/gcc/little32. portability todo.
3220 2004-07-08 Martin Quinson
3221 - import and improve remote compilation support from FAST
3222 - make sure make check works on half a dozen of machines out there
3224 2004-07-07 Martin Quinson
3225 Let's say it's version 0.0.3 ;)
3226 - Implement conversions (yuhu!)
3227 - Let it work on solaris (beside conversion, of course)
3228 - Stupid me, using rand() to generate the conversion datatests in not wise.
3230 2004-07-06 Martin Quinson
3231 - Let make dist work, since I'm gonna need it to compile on remote hosts
3232 - Let Tests/datadesc_usage write the architecture on which the file was
3233 generated as first byte.
3234 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3235 architecture descriptions.
3236 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3237 on those architectures.
3238 - Optimization: send/recv array of scalar in one shoot
3240 2004-07-05 Martin Quinson
3241 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3243 - Plug a whole bunch of memleaks
3244 - each process now have to call gras_{init,exit}. One day, their log
3245 settings will be separated
3246 - Continue the code factorisation between SG, RL and common in Transport.
3248 2004-07-04 Martin Quinson
3250 - Redistribution between SG and RL.
3251 We wanna have to accept in SG, so move accepted related parts of RL in
3252 the common part. (more precisely, the dynar of all known sockets is no
3253 more a static in transport.c, but part of the process_data)
3255 [gras_stub_generator]
3256 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3258 2004-07-03 Martin Quinson
3259 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3260 them all up in one shot)
3262 - Fix the ugly case of reference to dynamic array.
3263 - New (semi-public) function gras_datadesc_size to allow the messaging
3264 layer to malloc the needed space for the buffer.
3266 - gras_socket_close now expect the socket to close (and not its address to
3267 put NULL in it after it). This is because the socket passed to handlers
3268 is one of their argument (=> not writable).
3270 - propagate the interface cleanup from last week in datadesc, ie remove a
3271 superfluous level of indirection. User pass adress of variable
3272 containing data (both when sending and receiving), and not of a variable
3273 being a pointer to the data. Let's say that I like it better ;)
3274 The price for that is constructs like "int msg=*(int*)payload" in
3275 handlers, but it's a fine price, IMHO.
3277 - Let it work in RL (yuhu)
3279 2004-06-21 Martin Quinson
3281 - porting SG plugin and SG select to new standards (works almost).
3282 - plug memleaks and fix bugs around.
3285 - cleanup the prototype of data recv and force users to specify when they
3286 want to handle references to objects. Test case working even for cycles.
3287 - plug memleaks. Valgrind is perfectly ok with this.
3289 2004-06-12 Martin Quinson
3291 - cleanup the separation between plugin and main code in plugin creation
3293 2004-06-11 Martin Quinson
3295 - Reput hook for raw sockets, needed for BW experiments
3296 - kill a few lines of dead code
3297 [Data description] Interface cleanup
3298 - gras_datadesc_by_name returns the searched type or NULL.
3299 That way, no variable is needed to use a type desc once, which makes
3301 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3302 two parameters were strings (field name, type name), leading to
3304 [Dicos] Interface cleanup
3305 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3306 This is consistant with the dynar API.
3308 2004-04-21 Martin Quinson
3310 - Porting to new standards.
3312 - interface cleanup.
3313 There is no bag anymore, no need to take extra provision to mask the
3314 pointers behind "ID".
3315 Better splitup of functions between files create/exchange/convert.
3316 This is still a bit artificial since convert and receive are so
3317 interleaved, but anyway.
3319 - add a queued message list to procdata (the ones not matching criteria
3321 - factorize some more code between SG and RL wrt procdata
3323 - use gras_exit in example to track memleaks
3324 - get rid of gs_example now that GS is properly integrated into gras
3325 - update run_test to integrate the latest tests (datadesc)
3327 - rename WARNINGn macros to WARNn since it prooved error-prone
3329 2004-04-19 Martin Quinson
3331 - register init/exit functions within gras module mechanism
3332 - send/receive function.
3333 Convertion is not implemented, but short-cutted if not needed.
3334 struct/array elements are sent one by one (instead of block-wise), but
3335 nobody really cares (yet). Get a prototype before optimizing.
3336 - tests (using a file socket) for DD send/receive on:
3337 - base types: int, float
3338 - array: fixed size, string (ie ref to dynamic string)
3339 - structure: homogeneous, heterogeneous
3340 - chained list, graph with cycle
3341 Believe it or not, valgrind is not too unhappy with the results. The
3342 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3346 - Bugfix when using a filename explicitely (instead of '-')
3348 2004-04-09 Martin Quinson
3350 - factorize more code between RL and SG in socket creation
3351 - Complete the implementation and tests of:
3353 o file (only in RL, and mainly for debugging)
3355 I lost 3 days to design a portable address resolver, and then decided
3356 that the prototype mainly have to run on my box.
3357 Addressing portability too early may be like optimizing too early :-/
3359 - use gras_init in the Tests instead of the crappy parse_log_opt
3360 (the latter function is removed)
3361 [Conditional execution]
3362 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3363 [Code reorganisation]
3364 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3365 Build examples against the RL library, since there is no way to disable
3366 its creation for now.
3368 For information, the beginning of coding on GRAS was back in june
3369 2003. I guess that every line has been rewritten at least twice since