1 SimGrid (3.16) UNRELEASED
4 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
5 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
8 - New and Backwards Compatibility break:
9 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
10 When one wants to get the list of tasks whose states have changed during a
11 simulation round, s/he has to allocate and free a dynar and use it as argument
12 to this function. The former SD_simulate (double how_long) now returns void.
15 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
16 - New option: smpi/keep-temps to not cleanup temp files
19 - New function xbt_replay_action_get():
20 Retrieve the function previously associated to an event type.
22 -- Release target: June 21 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
24 SimGrid (3.15) stable; urgency=low
26 The Spring Release: continuous integration servers become green
27 We fixed even the transient bugs on all target architectures:
28 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
29 Windows, FreeBSD, NetBSD.
31 - Rename NetCards to NetPoints.
32 This was intended to help NS3 users, but that's not a netcard.
33 That's a point in the routing algorithm, let's avoid wrong simplifications.
36 - New: MSG_process_yield(). Stop and yield to other processes.
37 - New: MSG_process_daemon(). Daemon processes are automatically killed
38 when the last non-daemon process terminates
39 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
40 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
41 to make room for the upcoming network energy plugin.
42 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
45 - Ensure that an actor can kill itself with Process::exit()
46 - Kill the obscure NativeException. Nobody want to survive the issues
47 it denotes, so use JniException that is a RuntimeException (not to
48 be caught explicitly).
49 - Partial bug fix in initialization. SimGrid flags on command line were
50 consumed at C level but stayed in the original Java String[] args.
51 This could mess users' args[i] if SG flags were not put at the end of
53 The SimGrid flags are now removed from the Java arguments. However,
54 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
55 if args.length is greater than the number of YOUR OWN ARGUMENTS.
56 It might be if you have --log or --cfg flags in the command line.
57 - Fix numerous memleaks all around the place. In particular, around VMs.
61 - s4u::onPlatformCreated: right before the simulation starts
62 - s4u::onSimulationEnd: right after the main simulation loop
63 - s4u::onTimeAdvance: right after a clock change (time is discrete)
64 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
65 event from the availability_file changes the avail speed.
66 - Links are now usable from s4u
67 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
68 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
69 - New examples: The conversion of MSG examples to S4U has begun
70 - Actors: create, kill, migration, and suspend
71 - Applications: master-worker and token-ring
72 - Action replay: communications and storage
73 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
76 - Backwards Compatibility breaks
77 - The SD_route_get_list and SD_route_get_size functions have been removed.
78 They are replaced by the unique
79 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
80 The route from src to dst is built in the links dynar whose size can
81 be obtained with xbt_dynar_length.
82 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
83 sg_host_route_bandwidth and sg_host_route_latency.
84 Macros ensure the backwards compatibility, but you should fix your code
87 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
88 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
89 - Bump our claim of support from MPI 1.1 to MPI 2.2.
90 We don't support 100% of it, but it should be enough. Contact us if not.
91 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
92 naive and probably inaccurate)
93 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
94 mapped onto a unique small file using some system magic.
95 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
96 - Fortran: cleanups, fixes, support of user-added operations
97 - RMA: MPI_Accumulate are applied in correct order.
98 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
99 - Better support for MPI_IN_PLACE
100 - Support for MPI_Win attrs and keyvals.
101 - Support MPI_Comm_group_create, MPI_Type_size_x
102 - MPI ops should be properly applied to all allowed types
103 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
104 algorithm by a non failing one from mpich.
107 - Kill the fifo data container: we don't use it anymore.
110 - A <cluster> can now be created with different speed values to represent pstates.
111 see examples/platforms/energy_cluster.xml for an example.
113 -- Wed Mar 22 17:50:21 CET 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
115 SimGrid (3.14.159) stable; urgency=low
117 The previous release was only a crude approximation of a Pi release;
118 this one should be a bit better.
120 - Revert a last minute change that broke on Mac OSX.
121 - Fix the changelog and NEWS files.
122 - Improve the documentation, in particular of the routing module.
123 - Integrate some patches flying around in Debian.
125 -- Wed Dec 28 17:14:49 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
127 SimGrid (3.14) stable; urgency=low
130 * The whole documentation was reworked and reorganized. There is
131 still much room for improvement here, but we're on it.
134 * We now need python3 for our testing infrastructure.
135 * Model checking needs libevent
138 * Now works on FreeBSD
141 * Privatization now works on FreeBSD
143 * Privatization is now activated by default in smpirun
144 You can turn it off with -no-privatize if something goes wrong.
146 * Call-location tracing for SMPI.
147 You can add the exact location (filename / linenumber) of an MPI call to
148 your trace files and slow down or speed up the simulation between two
149 consecutive calls by using an adjustment file (see the documentation).
151 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
152 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
153 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
154 values were ignored in some cases. The timings of these functions can now
155 be significantly different.
157 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
158 smpi/running-power is renamed to smpi/host-speed
160 * smpi/grow-injected-times option to enable or disable multiplication of the
161 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
162 make simulation less precise (but also much faster).
164 * smpirun script should be (much) faster for large deployments.
166 * SMPI tracing : fixed issue with poor matching of send/receives.
168 * Replay : Fix broken waitall
170 New functions and features
171 * MSG_parallel_task_execute_with_timeout, to timeout computations.
173 Dropped / renamed functions and features
174 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
175 - MSG_mailbox_is_empty() -> Mailbox::empty()
176 - MSG_mailbox_front() -> Mailbox::front()
177 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
178 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
179 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
180 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
181 * Don't pass the free_f parameter to property related functions:
182 - MSG_host_set_property_value()
183 - MSG_as_router_set_property_value()
184 - MSG_storage_set_property_value()
185 * VM properties. Since msg_vm_t are msg_host_t, just use
186 MSG_host_get_property_value() and friends
187 * VM I/O related things:
188 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
189 - Unimplemented save/restore methods
190 * MSG_as_router_get_property_value() was redundent with
191 MSG_environment_as_get_property_value().
192 - Removed MSG_as_router_*propert*() functions
193 - Added MSG_environment_as_set_property_value() for consistency
194 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
195 Well, they are still there for now, but deprecated with a warning.
196 Please switch to xbt_dict_new_homogeneous() before this is removed
198 * Task affinity. Its intended behavior (that was very badly tested
199 and probably not really working) was deceiving what most users
200 would have hoped here.
201 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
202 * Option network/coordinates is now useless and should be dropped.
205 * Added option storage/max_file_descriptors to allow more than 1024 files opened
208 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
209 * Added Dragonfly topology support
212 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
213 EiBps, EBps, Eibps, Ebps for bandwidth.
214 They may become useful to some lucky ones.
217 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
218 * ex/app_tokenring: new example, very similar to the MSG Token Ring
219 * ex/async_waitAll: new example, on asynchronous communications
222 * Memory usage should be decreased for simulations with a large number
223 of processes. This also helps for SMPI.
225 -- Sat Dec 24 22:34:26 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
227 SimGrid (3.13) stable; urgency=low
229 The Half Release, a.k.a. the Zealous Easter Trim.
231 - We removed half of the lines, that were mostly experimental cruft.
232 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
233 (not counting blanks and comments -- according to openhub.net)
234 - The internals are now compiled in C++ (and will soon be clean C++)
235 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
236 59 klines of C, etc. We added only 29 klines of C++ in replacement.
238 * Backwards Compatibility breaks
239 - Removed Lua simulation bindings (switch to C or Java for that).
240 Lua can still be used to describe platforms
241 - Removed Java kernel plug-ins.
242 Will be reintroduced after the ongoing major internals reorg.
244 - the following functions were removed.
245 They were too specific and should be reimplemented in a generic
246 way, with filter function.
247 - MSG_task_listen_from_host
248 - MSG_mailbox_get_count_host_waiting_tasks
249 - MSG_mailbox_put_with_timeout was removed.
250 Please use MSG_task_send_with_timeout instead.
252 - the SD_application_reinit function was removed. It has been a noop for a while.
253 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
254 replaced by a more flexible mechanism.
255 - The following functions thus do not exist anymore
256 - SD_workstation_get_access_mode
257 - SD_workstation_set_access_mode
258 - SD_workstation_get_current_task
259 - Basic estimation functions have been removed but can easily be replaced
260 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
261 - SD_workstation_get_computation_time => amount / sg_host_speed()
263 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
264 Use VM.getSpeed()*load/100 for the legacy behavior.
266 - option enable_tracing was removed. It was not doing anything for a while.
267 - In the ModelChecker:
268 - the model-checker now ptraces the model-checked process which means
269 you cannot use a debugger on the latter anymore (we might make this
270 optional in the feature);
271 - removed soft-dirty page tracking;
272 - remove model-checked side snapshot management,
273 MC_snapshot() and MC_compare_snapshot();
274 - keep the MC_cut() function as a stub (it was not really working
275 in the previous release).
278 * All options are consistently in kebab-case. Old names are kept as alias.
281 * Switch to platform v4 format.
282 - Rename from 'power' to 'speed' the attributes describing the amount of
283 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
284 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
285 - In <host> and <link>, attributes availability and state are gone.
286 It was redundent with state and availability traces, and with peak values.
287 - In <cluster>, attributes availability_file and state_file are gone.
288 It was too complex and unused.
289 - Kill <gpu>. Was not doing anything.
290 - The DOCTYPE points to the right URL:
291 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
292 (the file at this address now documents the changelog since its v1)
293 - A warning is emitted for unit-less values (they are still accepted).
294 - speed. Default: 'f' or 'flops'. Also defined:
295 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
296 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
297 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
298 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
299 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
300 - latency. Default: 's' second. Also defined:
301 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
303 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
304 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
307 * s4u::Host is now the preferred public interface to the Host features.
308 sg_host_* functions are C bindings to the exact same behavior
309 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
312 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
313 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
314 They were deprecated since a few years, and probably did not even compile anymore.
317 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
318 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
319 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
320 encouraged to update their codes. A list of the modified functions can be found at the end of
321 include/simgrid/simdag.h
324 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
325 the lambda or closure passed as a parameter will run in kernel mode.
326 All the callback functions should be rewritten to that interface at some point.
329 * Reorganizing and cleaning the internals all around the place.
332 * Remove old default barrier/bcast buggy algorithms (see #18407)
333 * Various bug fixes to handle more codes
334 * Remove the need for the --foreground option of smpirun (it is still
335 accepted for backward compatibility).
338 * Kill the setset data container: MC don't use it anymore.
339 * Kill the queue data container: it made more sense with GRAS.
340 * Kill the xbt_peer_t data type: it's useless without GRAS.
341 * Kill rm_cb feature of config sets: it was never useful.
342 * Kill graphxml parsing feature. It was not used.
343 * Kill the deprecated code protected by XBT_USE_DEPRECATED
345 - xbt_dynar_sort_strings(), when the content is char*
346 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
347 They throw exceptions on invalid input;
348 * C++ support for declaring CLI flags (simgrid::config::Flag);
349 * class for abstracting different signal backends (simgrid::xbt::signal).
350 with no external dependencies (we need very simple signals).
353 * refactoring and cleanup of the code;
354 * ongoing process to cleanly separate the model-checking algorithms
355 from the code model-checking support.
357 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
359 SimGrid (3.12) stable; urgency=low
361 The Facelift Release.
364 * Require g++ v4.7 at least to not speak prehistorical C++.
365 * Require Boost 1.48 (for signal2 component).
366 * Java must be version 7 at least when activated.
367 * Builds on Windows again (including Java bindings).
368 * Tracing is now always enabled (no way to turn it out)
369 * Remove GTNetS. It was not working anyway.
370 * Various cleanups in the cmake scripts.
371 * Move headers around to sort them out on installed systems:
372 - instr/instr.h -> simgrid/instr.h
373 - instr/jedule/* -> simgrid/jedule
374 - simdag/datatypes.h was removed
375 - simdag/simdag.h -> simgrid/simdag.h
376 - msg/datatypes.h was removed
377 - msg/msg.h -> simgrid/msg.h
380 * Interface improvement:
381 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
382 - Sanitize the interface in MSG_task_ module:
383 - Merge two functions that were close enough but misleading:
384 set_compute_duration(t) -> set_flops_amount(t)
385 get_remaining_computation(t) -> get_flops_amount(t)
386 - set_data_size(t) -> set_bytes_amount(t)
387 get_data_size(t) -> get_bytes_amount(t)
388 - Massive cleanups in the functions related to the energy
389 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
390 - New: MSG_host_get_pstate()
392 - msg/energy/onoff: switching hosts on and off
394 * Interface improvement:
395 - Rename Host.isAvail() to Host.isOn()
396 - Rename Process.currentProcess() to Process.getCurrentProcess()
397 - Rename Task.setDataSize() to Task.setBytesAmount()
398 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
400 - #18874: Actually allows the GC to reclaim tasks
404 - SIMIX_process_throw: raises an exception in a remote process
405 * Refactoring: Separate sub-modules
406 - libsmx: the public interface, as libc in a real system
407 - popping: the strange dance that converts a user request into a kernel handling
408 - smx_context_*: the virtualization mechanisms that embed the user code
409 - smx_*: the handling of each simcalls
411 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
412 * Rename smx_action_t into smx_synchro_t, making explicit that these
413 things are used to synchronize processes with their environment.
414 For example, a communication is a sort of synchronization involving
415 the communicating processes (that may block until the exchange) and
416 the platform. The same can be said from computations, etc.
418 - #18888: segfault when a process ends before its kill_time
422 - 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.
423 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
424 - MPI_Type_set_name, MPI_Type_get_name
425 - MPI_*_c2f and MPI_*_f2c functions
426 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
427 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
428 - Activate a lot of new tests from the mpich 3 testsuite
430 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
431 - InfiniBand network model added : Based on the works of Jerome Vienne
432 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
433 - When smpi/display_timing is set, also display global simulation time and application times
434 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
435 * Collective communications
436 - 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
437 - 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/).
438 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
440 - Replay now uses algorithms from wanted collective selector
441 - Replay can be used with SMP-aware algorithms
442 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
443 - 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.
445 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
446 - Correctly use loopback on fat-tree clusters
447 - Asynchronous small messages shouldn't trigger deadlocks anymore
448 * Energy/DVFS cleanup and improvement
449 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
450 - new: smpi_get_host_pstate()
453 - "Full" network optimization flag was broken since Surf++
454 - Better handling of precision flags in maxmin
455 - Fix bug causing sometimes "Impossible" errors
456 - Properly pass cluster properties to included hosts
457 * Improvement of the Energy plugin.
458 - Always update the consumption before returning that value
459 - New property: watt_off to denote the dissipation when the host is off
460 - New functions getWattMinAt and getWattMaxAt to retrieve the
461 dissipation of pstates that we are not currently at.
462 * Java: class NetworkLink renamed to Link
463 * New function: simcall_process_get_kill_time()
464 * Massive rename s/workstation/host/
465 - That's intrusive, but that's good for the project consistency. Sorry.
466 - Change config option "workstation/model" into "host/model"
469 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
470 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
471 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
472 - xbt_dirname and xbt_basename for non-POSIX systems
474 * The model checker now runs as a separate process.
475 * The model checker runs must now be launched with the new simgrid-mc program.
476 * Record/Replay: the MC can display a textual representation of a path in the
477 execution graph. It can then be replayed outside of the model checker.
479 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
481 SimGrid (3.11) stable; urgency=low
486 * Normalizing pointers addresses tool for better diff between logs
489 * Add cloud examples using new VMs
490 - examples/msg/cloud/two_tasks_vm.tesh
491 - examples/msg/cloud/simple_vm.tesh
492 - examples/java/cloud/cloud.tesh
493 - examples/java/cloud/migration/migration.tesh
494 * Add java surf examples:
495 - examples/java/surfPlugin/surf_plugin.tesh
496 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
497 - examples/java/surfCpuModel/surf_cpu_model.tesh
498 * Add SMPI+MSG example:
499 - examples/smpi/smpi_msg_masterslave/
504 - msg task destroy cancel test
505 - msg_host on/off test
506 * Move all tests in testsuite to teshsuite (adding tesh files)
507 * Restructure teshsuites
508 - one folder for each kind of test
509 * Restructure AddTests.cmake
511 - structure the order of tests (with sections)
514 * Add virtual machine
515 - creation of a VM on a PM
516 - migration of a VM from a PM to another PM
518 - MSG_process_join(msg_process_t process, double timeout)
519 - msg_bar_t MSG_barrier_init(unsigned int count)
520 - int MSG_barrier_wait(msg_bar_t barrier)
521 - void MSG_barrier_destroy(msg_bar_t barrier)
522 - msg_as_t MSG_environment_as_get_by_name(const char * name)
523 * New option "msg/debug_multiple_use" to help debugging when a task is used
526 - Improvements and finalization of MSG_storage, MSG_file APIs and their
528 - Increase code coverage in test suites
532 * Protect context stacks against stack overflow. The number of protected memory
533 pages allocated on the top of each stack (1 by default) can be configured
534 with the new command line option --cfg=contexts/guard_size:<value>.
535 * Simcalls are now generated by a python script that
536 - generates files included by SimGrid
537 - checks that all the functions exist, and proposes prototypes
539 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
540 host_set_data, host_get_data
542 - simcall_process_join(smx_process_t process, double timeout)
543 * Fix bug where sleeping processing could not be suspended.
546 * Translate surf models from C to C++
547 - Generic classes for all models: Model, Resource, Action
548 - A generic interface for each kind of model (CPU, Network, Storage
549 Workstation, WorkstationVM)
551 * Translate surf routings from C to C++
552 * Add callbacks using sigc++ or boost::signals2
553 - Add callback functions for resource creation/destruction
554 - Add callback functions for action state change
555 - Handle Energy as a plugin
556 * Replace swag by boost::intrusive
557 * Add new routing models for clusters. For documentation, see
558 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
559 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
560 parameters for cluster tag
561 - Fat trees, with topology="FAT_TREE" and
562 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
563 - see examples/platforms/torus_cluster.xml and
564 examples/platforms/fat_tree_cluster.xml
568 * Hostfiles support host:nb_processes construct to deploy several processes on
570 * Collective communication algorithms should not crash if used with
571 improper number of nodes and report the error.
572 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
573 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
574 MPI_Dims_create, MPI_Cart_sub are supported.
575 * New interface to use SMPI programmatically (still depends on MSG for
576 some parts, see examples/smpi/smpi_msg_masterslave) :
577 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
581 * Global variables privatization in MPI executables is now performed at runtime
582 with the option smpi/privatize_global_variables (default:no).
583 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
584 dynamic libraries loaded by the application are not privatized (static
585 linking with these libraries is advised in this case)
588 * Options defined in XML work correctly now.
591 * New cmake option, enable_lib_in_jar, to control whether native libraries are
592 copied into simgrid.jar or not (ON by default). Use this option if you want
593 to reduce the size of the installed simgrid.jar, *and* the native libraries
594 are kept installed elsewhere.
595 * Surf binding with SWIG (code generated in maintainer mode only):
596 - plugin to handle callbacks
597 - CPU model only for the moment
600 * Supernovae build mode is definitively removed. It was used to improve
601 inlining and inter-module optimizations. It is nowadays superseded by
602 link-time optimizations commonly available in compilers.
603 * Update ns-3 find lib. Bindings for ns-3 should work again now.
604 * Add boost dependency for surf++
605 * Add new macro for tests
606 - ADD_TESH(name <tesh_args>)
607 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
611 - xbt_bar_t XBT_barrier_init(unsigned int count)
612 - int XBT_barrier_wait(xbt_bar_t barrier)
613 - void XBT_barrier_destroy(xbt_bar_t barrier)
614 * Make the xbt_os_time module public
616 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
618 SimGrid (3.10) stable; urgency=low
620 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
623 * Reintegrate Java to the main archive as desynchronizing these
624 package is not acceptable anymore (Java is now considered stable)
625 * Add explicit synchronization facilities through semaphores
626 * Bug fix: Task.setDataSize() only changed the C world, not the value
627 cached in the Java world
630 * Dramatically change the way files are handled. API and internals changed, but
631 this part of MSG was not considered as production grade either.
632 * Add explicit synchronization facilities through semaphores
633 * Add a new function MSG_host_get_process_list()
634 * Preliminary DVFS support (see examples/msg/energy/ for details)
637 * SMPI is now included directly in the libsimgrid as the windows
638 linker doesn't force us on splitting it anymore.
639 * Improvements of the SMPI replay tool:
640 - Most of the collective communications are now rooted in the same process as
641 in the original application.
642 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
643 used until now). Multiple data types can now be used in a trace.
644 - The replay tool now supports traces produce either by TAU or a modified
646 - Bug Fix: the compute part of the reduce action is now taken into account.
647 - Gatherv collective is now supported
648 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
649 of smpirun outputs time independent traces for the current run. One file
650 is created per process. If too many processes are simulated, this behavior
651 can be changed to one file for all processes by using the
652 tracing/smpi/format/ti_one_file flag
653 * smpirun generates the host file if needed (with given host count and platform)
654 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
655 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
656 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
657 --cfg=smpi/coll_selector:(mpich/ompi)
658 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
660 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
661 happening inside a collective SMPI call.
662 * Fix the behavior of complex data types handling.
663 * Make MPI_Wtime another synchronization point to take computations into
665 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
666 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
667 * Add all missing Fortran bindings, SMPI should work with Fortran 90
668 (no privatization of global variables yet)
669 * Preliminary DVFS support (see examples/smpi/energy/ for details)
672 * Verification of liveness properties is now available for SMPI applications
673 (in addition to MSG applications)
674 * Bugged examples using SMPI in examples/smpi/mc/
675 * Add --cfg=model-check/visited option. Allows the verification of infinite
676 programs. Detection of loops in the execution thanks to the system state
677 comparison and reduction of the state space to explore. Can be combined with
678 DPOR for safety properties.
681 * Allow to change SimGrid configuration (see --help) within the code
682 thanks to SD_config() as it can be done in MSG.
683 * Add a new function SD_task_set_amount() upon user request.
686 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
687 * Remove rule based routing (no more PCRE dependency)
688 * Add a limiter_link option to cluster tag, to specify a maximum reachable
689 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
690 * Add a loopback_bw and loopback_lat options to cluster tag.
691 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
692 routing. Now peers are encapsulated in an AS and have their own private
693 router but this is transparent.
696 * Our own implementation of getline is renamed xbt_getline, and gets
697 used even if the OS provide a getline(). This should reduce the
698 configuration complexity by using the same code on all platforms.
699 * New type: xbt_cfg_elm_boolean.
700 * Allow to use yes/no for boolean configuration options in the command line.
701 * Allow to disable SimGrid cleanups at exit from command line option.
702 There are situations where one may want a simulation to end with an exit.
703 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
704 annoying when scripting around the simulator. Adding a
705 --cfg=clean_atexit:no allows to circumvent this issue.
708 * Lots of memory leaks were corrected in this release.
709 There is no known memory leaks anymore, in all of our 600+ tests.
710 * New command line option --version, to get SimGrid version information.
711 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
713 * Supernovae builds are deprecated, and expected to be removed in the next
716 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
718 SimGrid (3.9) stable; urgency=low
720 The Grasgory release: GRAS is really dead now.
722 * Complete overhaul of the internal host structures scheme.
725 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
726 was considered as experimental and badly maintained since then.
727 * Keeping it was thus a trap to our potential users, that could take
728 it instead of MSG or SMPI by mistake despite is pity state.
729 * GRAS seems to have very few users (if any), and no one volunteered
730 to maintain it further. It also induces a lot of XBT code (for
731 portability sake), that must be maintained too.
732 * For all these reasons, we killed GRAS. If someone wants to revive it
733 in the future, don't cry, our git history still remembers of GRAS.
736 * Major overhaul. Merge our documentation again as time proved that
737 splitting it was really not helping our users.
738 * Further improve the developer documentation to help newcomers
739 hacking on SimGrid itself. The user documentation (and in
740 particular, the beginner documentation) is still in a sorry state.
743 * Now works on Windows too!
744 * Much more extensive test suite, from MPICH
747 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
748 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
749 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
750 * Bug fix: let task be scheduled when the last dependency to be solved is
751 a control dependency.
752 * Remove SD_load_environment_script function.
753 Use the C sg_platf function if you want to declare a platform
757 * New function: MSG_process_get_number()
758 * Old function documented: MSG_config()
759 * Remove MSG_load_platform_script function
760 Use the C sg_platf function if you want to declare a platform
764 * Change the default value of the TCP_gamma constant (maximal size of TCP
765 congestion window) to a more realistic 4MiB value. If you notice changes in
766 your simulation results, you can fall back to the previous 20k tiny window
767 by adding --cfg=network/TCP_gamma:20000 on command line.
768 * (Hopefully) fix a bug wrt periodic availability and state traces
769 * Bug fix: use default values at start when first event in availability/state
770 trace is not at time 0.
773 * remove the "new_" part of function name sg_platf_new_trace_connect
774 (resulting in sg_platf_trace_connect), since it does not create
778 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
779 use case where it's really mandatory, and maintaining it was a pain in
781 * New: xbt_fifo_search(), search an item with a user-provided
782 comparison function instead of dumb pointer comparison.
785 * Fix the lua deployment:
786 Use `simgrid.init_application()` before deployment instead of
787 `simgrid.msg_register_application()` after.
790 * Transfer the tracing files into the corresponding modules.
792 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
794 SimGrid (3.8.1) stable; urgency=low
796 The "we are told that some people want to also *install* the simgrid
799 * Add missing file "tesh.1" to the archive.
801 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
803 SimGrid (3.8) stable; urgency=low
805 The Psssshiiiit release: SimGrid jumps into the Cloud.
808 * Add an experimental interface to manipulate VMs. They are mainly
809 process groups with very few intrinsic semantic, but they should
810 allow you to build the semantic you want easily.
811 * New function: MSG_host_set_property_value()
812 * New function: MSG_process_on_exit(). To clean memory in all cases.
813 * Bug fixes that made the host (and link) failures unusable.
814 * Add a way to auto-restart process when the host in which they are
815 executing comes back (ON_FAILURE="RESTART" on deployment file,
816 MSG_process_auto_restart_set).
817 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
818 please stop using the old ones, they are DEPRECATED.
820 * Deprecate functions MSG_global_init() / MSG_global_init_args()
821 Please use MSG_init() instead. (reducing the amount of entry
822 points in the library helps us).
823 * Make it impossible to link against the wrong version of the lib
824 * Deprecate MSG_clean(). No need to call it anymore.
825 * Function MSG_get_host_number() is not deprecated anymore.
828 * Split the doc into a user guide and a reference guide.
829 * Start a developper guide to help people hacking on SimGrid.
832 * Enable tracing by default. This modules rocks you should use it.
833 * Remove option custom_flags. Now use environment variables CFLAGS
835 * Use default cmake things to detect lua instead of home grown ones.
836 * New option "enable_mallocators" to disable mallocators, for debugging
837 purpose ("on" by default).
840 * Bug fixes around the resource failures: don't let the processes
841 survive the host they are running onto.
842 * Add an interface to auto-restart processes when the host in which they are
843 executing comes back.
844 * Ensures that SIMIX_clean is called automatically. It's not part of
845 the public interface anymore (bindings should be updated).
848 * Bug fix for when SD_Simulate is called with a positive value: be careful
849 when comparing doubles. Sometimes they are different for non significant
851 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
852 parallel task whose initial work is distributed among host according
853 to the Amdahl's law. Such tasks are created with a parameter alpha
854 that corresponds to the non-parallelizable part of the computation.
855 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
856 between two sets of workstations assuming a 1D block distribution (each
857 workstation owns a similar share of data) on both sides.
859 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
860 Data redistribution will be automatically scheduled once parent and child
861 are both scheduled. The filling of computation_amount and
862 communication_amount structures is now done seamlessly thanks to the chosen
864 * New function SD_workstation_dump to display various information
865 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
866 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
867 bandwidth on the route onto which the task is scheduled and the amount of
869 To divide the nominal bandwidth by 2, the rate then has to be :
870 rate = bandwidth/(2*amount)
871 * Compute tasks that have failed can now be rescheduled and executed again
872 (from their beginning)
873 * Increasing source code coverage (src/simdag is now covered at 95.8%
877 * Re-implement time-independent trace replay using SMPI (at the
878 smpi_smp_* level) instead of MSG. This should replace
879 examples/msg/actions/actions.c
880 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
881 hindexed and structs)
882 * Implement the exchange of non-contiguous data.
883 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
884 * Correct behavior of smpi/sender_gap and set its default value to 0
885 * Add option to asynchronously send small messages to allow better
886 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
887 specifies the size in bytes under which messages will be asynchronously sent.
888 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
889 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
892 * New C interface to define a platform: XML is now optional.
893 For more info, please check include/simgrid/platf.h.
894 * New interface to define random platforms from the C:
895 For more info, please check include/simgrid/platf_generator.h and
896 examples/msg/masterslave/masterslave_platfgen.c
897 * Export a sg_cmdline dynar containing all the arguments we got from
901 * Two new tracing options for adding comments to trace file so you
902 can track your experiments (see --help-tracing for details).
903 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
904 * Adding the SimGrid version that generated the trace file as a comment.
905 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
906 * Fix to avoid key clashes on Paje links
907 * Other minor fixes related to the Paje specification
910 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
911 and renamed to xbt_str_hash() and xbt_str_hash_ext().
912 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
913 * Greatly improve the robustness of mmalloc to user errors (such as
914 using an area after freeing it, or freeing it twice)
916 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
918 SimGrid-java (3.8.1) stable; urgency=low
920 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
921 Warning: all methods are visible, but only some of them are
922 implemented so far. Check the source (src/jtrace.c)
923 for further information.
924 * New module: org.simgrid.msg.File (SimGrid File management functions)
925 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
926 * Change the meaning of Process.restart: now restart the process from
927 the begining, like MSG_process_restart in C.
928 * Add Process.setAutoRestart: handling of process restart when failed
930 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
931 you to retrieve the properties of the processes/hosts
932 * Deprecate Msg.clean(): you can just forget about it now.
933 * New function Process.getCount(), that only works when compiling
934 with the not yet released version 3.9 of the C library.
936 * New context factory based on Coroutines. It mandates a modified JVM
937 but then, the simulations run about five times faster, and there is
938 no limit to the amount of processes (beside of the available memory).
940 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
942 SimGrid (3.7.1) stable; urgency=low
945 * Restore the prototype of MSG_process_create_with_environment() to
946 the pre-3.7 situation by removing the kill_time argument.
947 * Add a MSG_process_set_kill_time() function instead.
950 * Fix weird behaviors when dealing with parallel tasks.
953 * Simgrid is now built as a dll.
954 * Simgrid-java now works on Windows.
955 * Simgrid-Java is now included into Windows package.
958 * First pre-build package for MacOSX.
961 * Fix compilation when using MSG_USE_DEPRECATED.
962 * Fix some compilation issues on Macs and Windows.
963 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
964 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
966 We discovered that the Lua console is broken, but we are missing the
967 manpower to fix it right now. The problem existed in 3.7 too, so we
968 are not blocking the release for that. Sorry if you depended on this
969 feature, any help would be really welcome.
971 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
974 SimGrid-java (3.7.1) stable; urgency=low
976 The "Java aint got to be bloated and slow" release
979 * Various internal cleanups and performance improvement
980 Simulations are expected to run up to twice faster or so
981 * Make Process.kill(process) an instance method, not a static one
982 * User processes are not java.lang.Thread subclasses.
983 This breaks the compatibility (sorry), but previous API was
984 brain-dead, making it impossible to have non-trivial
985 initializations in the process constructor.
986 * Require a full constructor per Process sub-class.
987 Kinda breaks the compatibility (sorry), but this allows a much more
988 efficient way to launch the processes at simulation startup.
989 * Do not embeed our version of semaphores, java 1.5 can be considered
990 as sufficiently prevalent for us to not dupplicate its features.
994 * Add examples for almost every part of the API
995 We spotted and fixed a lot of bugs in the process
997 * New module: asynchronous communication API
998 * New function: Process.sleep()
999 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1000 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1001 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1003 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1005 SimGrid (3.7) stable; urgency=low
1007 The "spring cleanups (before the next Big Project kicks in)" release.
1010 * We can specify the SMPI latency/bandwidth factor with command line
1011 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1012 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1013 You can also use the "config tag" from platform file by adding this line
1014 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1015 (see "example/platforms/tag_config.xml" to use "config tag").
1016 Note that the command line supersedes the platform file configuration.
1017 * Change the correction factors used in LMM model, according to
1018 the latest experiments described in INRIA RR-7821.
1019 Accuracy should be improved this way.
1020 * Use the partial invalidation optimization by default for the
1021 network too. Should produce the exact same results, only faster.
1022 * Major cleanup in surf to merge models and split some optimization
1023 mechanisms from the core of the models. As a result you can now
1024 specify which model to use (e.g., --cfg=network/model:LV08
1025 --cfg=cpu/model:Cas01) and which optimization mode to use
1026 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1027 Incompatible combinations should err at initialization. See
1028 --help-models for the list of all models and optimization modes.
1029 * The CLM03 workstation model was dropped for simplicity because it
1030 used the deprecated CM02 network model. Use default instead.
1031 * Rename the TCP_gamma configuration option to network/TCP_gamma
1032 * Rename the coordinates configuration option to
1033 network/coordinates, and document it
1034 * Use now crosstraffic keyword instead of the terribly misleading
1035 fullduplex keyword. It is activated by default now in the current
1036 default model, use --cfg=network/crosstraffic:0 to turn it off.
1037 * Ongoing refactoring the model parsing to make XML files optional
1038 See include/simgrid/platf.h for details (still to be completed)
1041 * Major overhaul of the documentation. Almost instructive now :/
1042 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1043 functions and friends. This interface was considered as
1044 deprecated since over 2 years, it's time to inform our users that it is.
1045 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1046 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1047 use these (crappy) functions in your code.
1048 These functions will be removed soon. Stop using them now.
1049 * Deprecate MSG_get_host_{table,number}
1050 Implement MSG_hosts_as_dynar() instead.
1051 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1052 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1055 * Stabilize the parallel execution mode of user contexts
1056 * Introduce configuration variables to control parallel execution:
1057 - contexts/synchro: Synchronization mode to use when running
1058 contexts in parallel (either futex, posix or busy_wait)
1059 - contexts/parallel_threshold: Minimal number of user contexts
1060 that must be part of a scheduling round to switch to parallel
1061 execution mode (raw contexts only)
1062 * Fix bugs that prevented to use suspend/resume along with
1063 synchronization structures.
1064 * Fix bugs in process termination that lead to invalid memory access
1065 in very specific conditions.
1068 * Introduce a parallel mode for the models (controlled by surf/nthreads
1069 configuration item). In our tests, running the models in parallel
1070 never lead to any speedups because they are so fast that the gain
1071 of computing each model in parallel does not amortizes the
1072 synchronization costs, even when ultra fast futexes are used.
1073 This is released anyway because YMMV.
1076 * Performance boost by using a swag internally to compute the set of
1077 tasks that are finished and should constitute the return value of
1081 * Enable it by default now that it is considered rather stable.
1084 * Documentation of the tracing functions.
1085 * Performance gains when tracing categorized/uncategorized resource
1086 utilization by avoiding calls to get route when updating resource
1087 variables. LMM constraints are being used instead.
1088 * API changed to set task categories. Use MSG_task_set_category instead
1089 of TRACE_msg_set_task_category, and SD_task_set_category instead
1090 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1091 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1092 * Examples for link user variables added (see at examples/msg/tracing/)
1093 * Deprecated function TRACE_msg_set_process_category completely removed
1094 * Trace header updated according to the latest Paje file format
1095 * Tracing network lazy updates, no longer obligate users to use full updates
1096 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1097 * Experimental: let user code declare/set/push/pop application states for hosts
1098 * API changed to allow the manual creation of graph configuration files
1099 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1102 * Improve the API of Lua MSG bindings, using the Lua spirit.
1103 * Each simulated process now lives in its own Lua world (globals are
1104 automatically duplicated). It helps writing simulators. Will allow
1105 to run Splay programs within SimGrid in the future.
1106 * Add a Chord example in Lua, equivalent to the MSG one.
1109 * Start the implementation of a solution to express temporal
1110 properties, not only local assertions. This is still an
1111 experimental work in progress, stay clear from it to be safe.
1115 - Add new runtime parameters --help-logs and --help-log-categories
1116 to display information about supported logging parameters and
1118 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1119 don't exists anymore.
1120 * Mallocators: allow value NULL for the reset function.
1122 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1123 create homogeneous dictionaries, where all the elements share the
1124 same free function. Non homogeneous dictionaries will be
1125 deprecated in the next release.
1126 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1127 - Multi-level dictionaries are deprecated.
1129 - new function xbt_dynar_search_or_negative() that is useful when
1130 you have less than 2 million elements in your dynar and don't
1131 want of the extra complexity of catching exceptions when the
1132 element is not found.
1134 - Make xbt_os_thread module (for thread portability) public.
1135 Documentation is still to come, sorry.
1137 - Cleanups and simplifications to make it maintainable again.
1138 - Exotic features (such as memalign and valloc) were removed.
1139 - The metadata were extended and improved so that the
1140 model-checker becomes able to explore and inspect the heaps.
1141 - This may induce a performance drop when enable_model-checking is
1142 ON in cmake (even if it's not used in the simulation), but it is
1143 necessary at this point to get MC working.
1145 Turn model-checking OFF if simulation performance matters to you.
1146 Not enabling it at runtime is not enough, disable it in cmake.
1148 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1151 SimGrid (3.6.2) stable; urgency=low
1153 The "Not coding new stuff allows to polish old things" release.
1156 * New bindings to the NS3 packet level simulator (experimental)
1157 * Use the raw (efficient) execution contexts instead of the sysv
1158 (portable) ones when possible.
1159 * libpcre is now mandatory in any cases since not using it led to
1160 severe performance loss and possibly other issues
1161 * Update the XML platforms:
1162 - G5K: include the latest machine in Nancy
1163 - GridPP and LCG: new platforms
1164 * Documentation was partially updated, at least (more to come)
1166 Bug fixes, cosmetics and small improvements
1167 * Free terminated processes before the end of the simulation to avoid
1168 exhausting the memory of users having very dynamic amount of
1170 * Bug fix and cosmetics about canceling non-running tasks
1171 * Bug fix about the dot loader's issues when using libcgraph
1174 * Create an installer for windows with nsis (amd64 and win32)
1175 - Add an hello world project to illustrate simgrid project creation.
1176 - Embed libpcre into the Simgrid installer to avoid
1177 its compilation burden
1178 * The raw execution contexts should work on Apple now
1179 * Port to Windows 64 bits
1180 - Sysv contexts now have an implementation for this arch
1181 - GRAS communication features now support this arch
1182 * Drop support for borland compiler on windows
1183 - this code was not maintained, and we kinda depend on gcc nowadays
1184 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1185 * Fix portability issue on unstable ubuntu: linker became picky on
1188 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1190 SimGrid (3.6.1) stable; urgency=low
1192 The "Oops, we broke Macs too" release
1195 * Fixed contexts detection so that raw ones are used when possible
1196 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1197 leads to a strange error, with user code segfaulting sometimes when
1198 the generated code is not perfectly aligned (which is not
1199 controllable from the user side, depends on the amount of code)
1202 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1204 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1206 SimGrid-java (3.6) unstable; urgency=low
1209 * Split of every thing from simgrid v3.5 into a separate package.
1211 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1213 SimGrid (3.6) stable; urgency=medium
1215 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1218 * Bindings now constitute their own package, separated from the main one.
1219 Rationale: reduce our maintenance nightmare by reducing the module coupling
1220 They will soon be released on their own on gforge.
1222 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1223 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1225 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1226 * It's not quite deprecated for now because we have no replacement,
1227 but it may soon become the case.
1230 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1231 * Fortran: New user-level cache variable to store the rank of the running
1232 process. This improves performance by an order of magnitude.
1233 * C: New coccinelle script to automatically locate and modify global and
1234 local static variables.
1235 * Improved SMPI network model with a sender-side gap to account for multiple
1239 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1240 only say if a communication is finished, no matter whether it succeeded or
1241 failed. You can call MSG_comm_get_status() to know the status of a finished
1243 * New function MSG_task_dsend() to send a task and detach it. When a
1244 communication is detached, you are never notified of its success or failure
1245 and the memory is released automatically once it is finished. This function
1246 is useful when you don't care about the end nor the success of a
1248 * Change the prototypes of action replay. Sorry for inconvenience,
1249 but this is really more efficient this way (and to adapt your code,
1250 you just have to fix the initialization, that shouldn't be too long)
1251 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1252 ever managed to do anything useful with it.
1253 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1254 takes a dynar of communications. It returns immediately and gives the
1255 index of a finished communication (if any).
1256 * New example: a basic implementation of the Chord P2P algorithm.
1259 * New model for multi-core CPUs. You can now use the core attribute to
1260 precise the number of cores of a host. This is a basic model. Every
1261 process running on the host receives at most the power provided in
1262 the DTD (throughput<=power). Total throughput of process cannot exceed
1264 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1265 router linked by an up-link and a down-link (possibly asymmetrical).
1266 This kind of pattern allows to easily build last-mile model style platforms.
1267 Aggregating such patterns in a rule-based AS is thus the technique of
1268 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1269 * New model for Vivaldi routing. We transformed the Vivaldi network model
1270 into a Vivaldi routing model (based on the rule-based model). This allows to
1271 combine Vivaldi based latencies with last-mile platforms.
1274 * Added a check for NaN of IEEE754 infinite in the double entries of
1276 * Introduce a new context factory "raw", highly inspirated from the
1277 ucontext factory, but using manually crafted functions in assembly to
1278 do the work in an efficient manner.
1279 * Allow to change the used context factory at run time, not only at
1280 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1281 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1282 contexts at runtime (only with raw contexts or ucontexts).
1283 * Completely rewrote this module to allow parallel execution of user
1284 processes. Use --cfg=contexts/nthreads:N to execute user processes
1285 with N parallel threads (the default is 1, meaning no parallelism).
1286 * Allow to decide dynamically between sequential and parallel modes.
1287 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1288 processes in parallel only when their number is greater than or equal to P
1290 * Added a check for NaN of IEEE754 infinite in the double entries of
1294 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1295 won't output the state of processes when interrupted with Ctrl-C
1296 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1297 NULL-terminated array. This may solve backward compatibility issues
1298 due to the change to return type of SD_simulate. See also:
1299 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1300 * Add new macros with variable number of arguments.
1301 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1302 - in xbt/asserts.h: xbt_assert
1303 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1304 - in xbt/ex.h: THROWF and RETHROWF.
1305 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1307 * Change xbt_die() to accept a format string with arguments, just like printf.
1308 * New data structure: xbt_lib_t, like a dict but more general and with better
1312 * New configuration options
1313 Options triva/categorized and triva/uncategorized can be used to generate
1314 graph configuration files for Triva visualization tool.
1315 * Configuration option tracing/platform is renamed to tracing/categorized
1316 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1317 * New TRACE_user_link_variable interface:
1318 User provides the name of the link and the tracing variable to attach to it
1319 * the declaration of tracing categories must be done after the environment
1321 * simpler tracing interface, just one way to declare categories
1322 TRACE_category or TRACE_category_with_color, it is up to you
1323 * links in the trace file are again identified by their names
1324 * trace contains the full platform hierarchy exactly as declared using the ASes
1325 * Options tracing/msg/[task|process]:1 groups the process by hosts
1326 for both cases, tasks and processes must have names that are unique during the simulation
1327 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1328 * The experimental option tracing/msg/volume is deprecated
1329 its functionality may be reincorporated if needed
1331 The tracing generates a trace file with unordered timestamped events,
1332 because of the way the core simulator (surf) works. A script available
1333 at the tools directory (fix-paje-trace.sh) can be used to put the events
1334 in order. We have changed the tracing so it can generate ordered timestamped
1335 events in the final trace, but depending on the simulator (and how much time
1336 is simulated) that can lead to a huge memory utilization. It is deactivated
1337 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1339 Build Infrastructure
1340 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1341 - We are trying hard to keep the API stable, but it may happen that
1342 some things change (we're a research project after all, not a
1343 nuclear plant operating system). If such things should happen, you
1344 could rely on that macro to adapt.
1345 - current value: 30600 for 3.06.00, aka 3.6
1346 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1347 a number that can be compared with SIMGRID_VERSION.
1348 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1349 assertions and verbose/debug logging events are disabled at compile time.
1351 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1353 SimGrid (3.5) stable; urgency=medium
1356 * New feature to any SimGrid-based simulator: Model-Checking
1357 Check SIN#1 for more details.
1360 * New Model SMPI (three-interval linear regression for correction factors)
1361 See RR-7426, available at http://hal.inria.fr/inria-00527150
1362 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1363 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1364 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1365 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1366 * See also src/smpi/README
1371 - Tracing API changes: TRACE_start and TRACE_end should not be called
1372 by user-code. They are automatically called by simulators created
1373 with SimDAG, MSG and SMPI if the toolkit is compiled with
1374 tracing_enabled=ON. Categories declaration and utilization remain the
1375 same for MSG and SimDag.
1376 - A function was added to the tracing API to declare categories with
1378 - TRACE_category_with_color (char *category, char *color)
1379 where color must be in the following format
1380 "%f %f %f", red, green, blue
1381 and red, green, blue are float values in the interval [0, 1]
1382 - User can specify NULL as color parameter, or continue calling
1383 TRACE_category (cat)
1384 On that case, the tracing system will define random colors
1385 - The following command-line options are supported:
1386 --cfg=tracing/filename:msg.trace
1387 --cfg=tracing:1 (activate tracing, needed to use others)
1388 --cfg=tracing/platform:1 (categorized resource use)
1389 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1390 --cfg=tracing/msg/task:1 (task creation)
1391 --cfg=tracing/msg/process:1 (process creation, migration)
1392 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1393 --cfg=tracing/smpi:1 (SMPI interface tracing)
1394 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1395 - examples of examples/msg/tracing updated
1397 - DAXLoader and DOTLoader functions can generate tasks with categories
1398 - A new function to attribute a category to SD tasks:
1399 TRACE_sd_set_task_category (SD_task_t task, char *category)
1400 * Tracing the MPI interface implemented by SMPI
1401 - Collective operations are traced with states
1402 - Point-to-Point operations are traced with states/links
1403 - Tracing activated by a parameter "-trace filename" passed
1404 to smpirun during execution (considering that simgrid
1405 is compiled with tracing enabled)
1406 - To run the simulation with gdb, the simulator
1407 accepts --cfg=tracing/smpi:1 to trace SMPI
1408 - tesh files to check if smpi tracing is ok
1409 - See examples/smpi/NAS/DT-trace
1410 * GTNetS tracing re-worked
1411 - adaptation to the tracing system of GTNets to cope
1412 with modifications regarding the fullduplex mode
1413 - new tesh files to check if gtnets tracing is ok
1416 * Asynchronous communications through the functions:
1417 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1418 * New function: MSG_load_platform_script()
1419 to make possible using a lua script instead of XML files to set up platforms
1420 * New function: MSG_set_function
1421 to associate functions to processes, used when bypassing the parser
1422 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1424 Platforms: Add some more examples in examples/platforms
1425 * Grid'5000: see www.grid5000.fr
1426 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1429 * Change the XML format. This is a very important modification. SimGrid
1430 3.5 introduces a new hierarchical format based on the notion of
1431 Autonomous Systems. Compatibility with old format is ensured
1432 through the perl script provided in the install bin directory
1433 bin/simgrid_update_xml.
1434 It is now possible to build platforms with specific routing
1435 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1436 connect several platforms together. We will try to provide soon
1437 set of realistic platforms exploiting these properties (have a
1438 look at examples/platforms/ for the moment).
1439 * Take the opportunity of the XML format change to be a good XML citizen:
1440 rename link:ctn to link_ctn and similar changes (also dealt with by
1442 * Add a new routing scheme (rule-based) using regular expressions. It
1443 enables to have an extremely low memory footprint when the
1444 underlying routing is simple and can be compactly described. You
1445 need to have libpcre4-dev (perl regular expressions) installed if
1446 you want to use this routing scheme.
1447 * Revive the cluster TAG and allow to easily and efficiently (both in
1448 term of memory and speed) connect clusters together. Have a look
1449 at teshsuite/simdag/platforms/ to see how this can be done. With
1450 this tag, you can create clusters with thousands of tasks at no
1451 cost (have a look at examples/platforms/).
1452 Note: clusters are implemented as ASes, so there is no need for an
1453 enclosing AS tag if you have only one cluster in your platform.
1454 * Add new generic functions in the public interface that allows the user
1455 to call SURF 'create_resource' methods from your code (same
1456 functionality as the XML bypass mechanism but with a much lighter
1458 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1459 that uses a piecewise linear approximation to produce better
1460 results when exchanging small messages.
1461 * Add a new parameter to handle correctly full duplex link and account
1462 for interferences between uplink and downlink communications
1463 (activate with --cfg=fullduplex:1).
1466 * Rename the SD_READY (all dependencies are satisfied and task is
1467 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1468 dependencies are satisfied) state.
1469 This prevents a confusion between the notion of "ready to schedule"
1470 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1471 simulated" (SD_RUNNABLE) used by the simulation kernel.
1472 * Change the way a task is considered as ready. Instead of removing
1473 dependencies when a task is done, a counter is decreased. This way,
1474 it is always possible to reach ancestors thanks to the
1475 SD_taks_get_parents function (even after the end of the simulation.)
1476 * Change the return type of SD_Simulate from (SD_task_t*) into
1477 xbt_dynar_t. This function was in handling a dynar internally and
1478 converted it into a NULL terminated array for historical reasons.
1479 * New function SD_dotload(char*) to load a DAG described in dot
1480 format. This loader and the corresponding examples require the
1481 installation of the graphviz library.
1482 * Fix a bug in the management of tasks of size 0 in the surf network
1483 models. This problem was only visible with SIMDAG and you should
1484 thus disregard results produced with earlier versions if you
1485 relied on this feature (some tasks were blocked because of this).
1486 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1487 even though your applications were doing only point-to-point
1488 communications and sequential computations. Now you can really use any
1489 model you want (of course, if you create real parallel tasks, which are
1490 not implemented in most models beside ptaskL07, this will abort).
1491 * Add an example that schedules a DAX on an heterogeneous platform
1492 using a Min-Min strategy.
1493 * New function SD_workstation_get_current_task() that returns the kind
1494 of task currently running on a workstation in the sequential access
1496 * Raise some warnings when unexecuted tasks remains at the end of the
1497 simulation. This is usually caused by cycles in the DAG.
1500 * New function: SIMIX_process_set_function() called by MSG_set_function
1501 * Change the underlying waiting queue in semaphores so that a process
1502 can wait on several of them simultaneously (as in waitany).
1503 * Fix the way to handle tokens in semaphores so that all access patterns
1504 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1505 * kill the dirty pimple SIMIX_message_sizes_output()
1506 Please use (proper) visualization instead
1509 * New data container: setset (set of sets of elements)
1510 * New module: mmalloc (mapped malloc, allowing to have several
1511 independent segments of malloc)
1512 * New function: xbt_dict_cursor_set_data()
1513 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1514 * New function: xbt_dynar_is_empty()
1515 * New function: xbt_fifo_get_last_item()
1516 * Fix xbt_dynar_shrink(): use the right element size.
1517 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1518 uninitialized areas during expand.
1519 * Fix semaphores: previous implementation was severely broken.
1520 * Use library init/fini functions for our initialization.
1521 - you can use logs and other feature as soon as you want in your
1522 code (even before the xbt_init / MSG_init)
1523 - xbt_exit is now a no-op and produce a warning when used.
1526 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1527 benefit from the latest improvement to the simulation kernel.
1528 * Kill measurement sockets for now. If you rely on them, sorry. This
1529 release is not for you. This feature will be reintroduced in the
1530 future, but we cannot delay the release any further.
1531 * New function: gras_msgtype_get_name().
1532 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1533 * Fix (at last) the pmm example: it should not randomly fail anymore.
1535 Build chain: bug fixes and overall polishing
1536 * Cmake is now stable enough. Hence, we killed the autotools.
1537 * Port to windows ( TM :)
1538 * Fix the 'make install' target.
1539 No need to use 'make install-simgrid' anymore
1540 * Introduce a 'make dist' target compiling a *source* archive
1541 'make package' compiles a binary archive
1542 * Compile java files only on need
1543 * Add --cd and --setenv command line options to tesh
1544 * Out of source builds are not fully supported yet, but we are close
1545 * Enable supernovae and optimization flags by default for our users
1548 * Add layer to set up environment directly from lua, without XML.
1549 * The effect of gras_stub_generator can be achieved through
1550 lua too (check examples/gras/console/ping_generator.lua)
1552 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1554 SimGrid (3.4.1) stable; urgency=low
1556 The "Polishing easter eggs is probably a good idea" release.
1557 This is a bug fixes release only.
1561 * Fix a bug preventing the tasks from begin garbage collected.
1564 * Fix a bug occuring when a host involved in a communication fails.
1565 This was not detected properly by the other peer involved in the
1566 communication. Now, it's reported as a network error.
1569 * Warn the user about loop dependencies in data flow of DAX files
1570 * Obey the control-flow dependencies of DAX files
1573 * Add option "enable_smpi" allowing to not compile SMPI.
1574 Probably useful for the (Mac) users experiencing a build error here
1575 * Improve the detection of lua5.1 and ruby1.8
1577 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1579 SimGrid (3.4) stable; urgency=low
1581 The "Easter in Cargese" release. Also known as (major changes):
1583 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1584 ~> bindings were greatly improved
1585 ~> new tracing infrastructure for better visualization introduced
1587 * the "Welcome to configury modernity" release.
1588 ~> we switched from autotools to cmake, and improved our cdash
1591 A more detailled list of changes follow (full detail in svn log).
1593 Java Bindings: Various Cleanups
1594 * (install java-gcj-compat-dev on debian-like to use them)
1595 * Remove put/get: no need to export deprecated interface in Java
1596 Use send/receive instead.
1597 * Cleanup the examples and add a README per directory
1598 * Remove example autoDestination (that's the only way to go now)
1599 * Remove example explicitDestination (was a plain copy of basic)
1600 * Make JniException a runtime exception, so that there is no need to
1601 declare the fact that you may encounter such a beast. I guess that
1602 nobody will ever want to survive such error.
1603 * Create specific errors for each MSG case of failure:
1604 host failure, transfer failure, timeout, task cancelled
1605 * Cleanup the exceptions that may get thrown by each function
1606 * Other internal cleanups in Java bindings. Performance still bad :/
1607 Ruby and Lua Bindings: create them
1608 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1609 * That's new and great, you should try them out.
1610 Same functionalities than Java bindings, only even less polished
1612 * Kill the useless "rate" argument of SD_task_get_execution_time()
1613 Everyone used to provide -1 as a value, it was not used, and the
1614 semantic of a possible use wasn't even clear.
1615 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1616 either as comm costs or compute costs to mean that there is no
1617 such thing for that specific task.
1618 * Add a SD_task_set_name() function
1619 * Fix SD_task_unschedule() on typed tasks
1620 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1621 * In DAX loader, accept useless 'level' attributes to <job> since
1622 LIGO DAGs have them (seem to be to ease graphical representation).
1624 * Add an example masterslave_mailbox.c using send/receive and not
1625 the deprecated put/get interface.
1626 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1627 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1629 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1630 (declare MSG_USE_DEPRECATED to still have the old name)
1631 * Add a MSG_task_set_data() function
1632 * About trace replay (see examples/msg/actions):
1634 - Allow to work with splitted trace files for each process
1635 Give the specific trace file as argument of each process,
1636 and call MSG_action_trace_run(NULL)
1637 You can still have one merged file for all processes.
1638 - Fix implementation of collective operations
1639 * Allow task_execute() on 0-sized tasks (closes #10063)
1641 * This is the first release of SimGrid where SMPI is not considered
1642 beta anymore (even if some corners should still be improved)
1643 * Port over the new SIMIX_network submodule (internal refactoring)
1644 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1645 * Implement more missing elements of the standard:
1647 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1649 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1650 MPI_Scan() MPI_Get_processor_name()
1651 - Added implementation of missing case for Alltoall (warning: it's
1652 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1653 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1654 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1655 SMPI_MPI_Allgatherv()
1656 * Bug fixes include:
1657 - MPI_Waitsome() was broken
1658 - Allow relative includes in smpicc
1659 - Command line cfg argument 'reference_speed' was ignored...
1660 - Some functions did not properly lead to auto-benching of user code
1661 - smpicc passes -O2 by default (just like openmpi one)
1663 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1664 * Bug fixes about timeouts during communications
1665 * add SIMIX_message_sizes_output() as a pimple to write to file the
1666 amount of messages per size. Use gnuplot to get histogram.
1667 Pimple because that's the only user-visible function of simix,
1668 defined directly in xbt.h (irk, sorry)
1670 - Add a SIMIX_sem_get_capacity() function
1671 - Fix interactions with processe resume/suspende
1672 - release_forever() was stupidly broken
1673 - Fix SIMIX_display_process_status() for processes in a semaphore
1674 - Make SIMIX_sem_block_onto() user-visible
1675 * Refactoring context stuff:
1676 - Use pseudo-OOP for better modularity
1677 - reimplement SIMIX_process_kill() without process_schedule() so
1678 that the latter can take as invariant that it is called from
1680 - Merge context_start into context_new for sake of simplicity
1682 * Add a Vivaldi network model, coded live during SUD'10 ;)
1683 * Rename configuration variables to start a hierarchy:
1684 o cpu_model -> cpu/model
1685 o network_model -> network/model
1686 o workstation_model -> workstation/model
1687 * New configuration variables:
1688 o network/bandwidth_factor: correction to bandwith
1689 o network/latency_factor: correction to latency
1690 o netwotk/weight_S: correction to the weight of competing streams
1691 * Add a long description to the models, that users can see with such
1692 argument on the command line: --cfg=cpu/model:help
1693 * --help-models display the long description of all known models
1695 * config: add the ability to set a default value after registration
1696 Does not override any previously set value (e.g. from cmd line)
1697 * dict: allow to have integer key and data.
1698 When so, you need to use the following functions
1699 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1700 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1701 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1702 In contrary to regular dicts, the key is not malloced before copy.
1703 Mixing scalar and regular elements in the same dict is not tested
1705 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1706 Tracing for Visualization:
1707 * SimGrid is now instrumented in order to generate a trace file for
1708 visualization analysis: to use it, need to compile SimGrid with the
1709 "tracing" option enabled, and instrument the program using SimGrid with
1710 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1711 (among other functions).
1712 * The instrumentation only traces the platform utilization for now
1713 * Documentation to use the tracing functions and how to analyze the
1714 traces with the Triva tool is written.
1715 * More information about: SimGrid FAQ (in the section Tracing Simulations
1718 * We moved to cmake as default build system. Autotools support will
1719 be dropped soon. Check the FAQ for more info about how to use it.
1720 * Greatly improved our cdash/ctest interactions
1721 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1722 * Added memory checking tests with valgrind; lot of memleak fixing.
1723 This may be the first release of simgrid with so few memory issues
1724 * Added code coverage tests.
1725 Our coverage is still improvable, but at least we see it on cdash.
1727 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1729 SimGrid (3.3.4) stable; urgency=low
1731 The "Desktop Grid needs love too" release (also called Xmas release).
1733 Models improvements:
1734 * Major speedup in the maxmin system solving by using lazy evaluation
1735 Instead of solving completely the maxmin system at each iteration,
1736 only invalidate (and recompute) the modified parts.
1737 This new feature is enabled in default models but you can try to
1738 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1739 * Cas01 IMproved as default CPU model
1740 This CPU model is the same Cas01 model, but it uses the
1741 maxmin-selective-update flag and a heap structure to manage
1742 actions on SURF kernel.
1743 It reduces the complexity to find the next action to finish and,
1744 consequently, it's faster than the old Cas01.
1745 This is the new default CPU model (Cas01).
1746 * Rename the old Cas01 model to Cas01_fullupdate
1747 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1748 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1749 * CpuTI (CPU Trace Integration)
1750 A new CPU model whose objective is simulate faster when using
1751 availability trace files.
1752 Instead of using a full featured, over engineered maxmin system for
1753 CPU modeling, this model does the pre-integration of traces files
1754 to calculate the amount of CPU power available, and so, executes
1755 faster than the old CPU models.
1756 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1757 * Use LV08 as default network model since it gives better accuracy
1758 for small messages and shouldn't change things for big ones.
1759 Use --cfg=network_model:CM02 to get the previous behavior.
1762 ******************************************
1763 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1764 ******************************************
1765 * The new CPU model may changes simulations!
1766 The point is that events occurring at the exact same timestamp
1767 are not scheduled in the same order with the old and new
1768 version. This may be enough to completely change the execution
1769 of simulations in some cases.
1770 * The new network model will change simulations!
1771 This new model is more realistic than the previous one, so you
1772 should consider redoing your old experiments with this model.
1773 Sorry for the inconvenience.
1776 * Introduce the supernovae compilation mode
1777 When compiled that way, the whole SimGrid (or almost) is put in a
1778 single compilation unit and compiled in one shoot.
1779 This is to help gcc which has difficulties to inline stuff from one
1781 The speedup seem to be above 15%, althrough more tests are needed on
1782 amd64 to confirm that gain.
1785 * Port of MSG's mailbox on top of SIMIX network
1786 The put/get mechanism was greatly simplified on the way.
1789 * New SIMIX network module. Provides:
1790 - Mailbox: rendez-vous mechanism to find with who you want to speak
1791 - Synchronous send/recv: easier and hopefully faster since the
1792 logic is handled in the maestro process directly now
1793 - Asynchronous send/recv: you dreamt of it? It's here now
1794 Too bad that nobody cared enough to propagate the change to MSG.
1795 * Add semaphores as SIMIX synchronization mechanism.
1798 * new function SD_daxload(char*) to load a DAX file
1799 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1800 * Introduce typed tasks. Specify its kind and cost at creation.
1801 At scheduling, just give where it should be placed, and the cost
1802 for each involved resource is automatically computed.
1803 Existing constructors so far (more to come of course):
1804 - SD_task_create_comm_e2e() for end-to-end communication
1805 - SD_task_create_comp_seq() for sequential computation
1806 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1807 * new function SD_task_dump() for debuging display
1808 * new function SD_task_dotty(task,FILE*) writing to file the info
1809 about the task in dotty format
1810 * SD_task_dependency_exists() can now cope with having one of its
1811 arguments NULL. If so, it tests whether the other argument has any
1813 * Add getters on list of preceding/following tasks:
1814 SD_task_get_parents(task) and SD_task_get_children(task)
1815 * Add getters on amount of workstations and list:
1816 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1817 * Add getter on task kind: SD_task_get_kind(task)
1818 * Update the start_time and finish_time of tasks on completion/failure
1819 * Bugfix: Remove task from state swags when destroyed
1822 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1825 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1826 dynar with one separate thread per value of the dynar.
1827 * Change the prototype of xbt_thread_create(), sorry.
1828 Added a boolean parameter indicating whether we want to join this
1829 thread (used in SG only for now)
1830 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1833 * GTNetS wrappers should now be usable again (and betterly tested too)
1834 * Fix a major regression from 3.2 where the timeout provided to
1835 MSG_task_put_with_timeout() was used as absolute time before which
1836 the comm should be done.
1837 * Start to fix the <cluster> tag.
1838 - Internal links should be good now (beside of the loopback, which
1839 use the private link instead)
1840 - paths to the external world is still rather broken
1841 - the <route:multi> tag is just broken. Actually that's brain-dead.
1842 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1843 to make it less stupid
1844 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1845 * Fix a source-level compatibility glitch from 3.2: after defining
1846 MSG_USE_DEPRECATED, you can use the old name
1847 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1848 * Allow to compile from the SVN with automake 1.11
1849 * Fix some problems when using the "start_time" tag in deployment XMLs.
1850 * Fix #8569: XBT/synchro.h has redundant declarations
1851 * Fix #8563: MSG return values and exceptions
1852 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1853 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1854 Thanks to Robson Peixoto.
1855 * Fix a vicious bug in dictionaries inducing that some elements were
1856 not freed on xbt_dict_free()
1858 Portability report of this version:
1859 * Main portability targets:
1860 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1861 on (amd64/i386/ia64)
1862 - mac leopard on i386
1863 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1864 but nothing critical.
1865 * Other platforms: windows, AIX and others were not tested for this release
1867 Timing report of this version:
1868 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1869 scenario parameters). From 8h to a few seconds in desktop grid settings.
1870 * Supernovae brings about 25% speedup on i386.
1872 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1874 SimGrid (3.3.3) stable; urgency=low
1876 The "Need for Speed" release.
1878 The timings done to validate the 3.3.2 were faulty.
1879 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1881 The problem was a conversion from a manually handled vector to
1882 xbt_dynar_t on the critical path.
1883 xbt_dynar_foreach calls functions, inducing stack management crap.
1885 We inlined these functions and xbt_dynar_foreach is now breath taking.
1886 We also inlined xbt_swag_belong on the way.
1888 Here are some approximate speedup measurements (on master/slaves
1889 simulations lasting between 10s and 20s each):
1890 3.3.1 -> 3.3.2: about same performance
1891 3.3.2 -> 3.3.3: 40% speedup
1892 3.3.1 -> 3.3.3: 40% speedup
1893 3.3.1 with inline patch -> 3.3.3: 30% speedup
1895 Our reading is that the refactoring which occurred in 3.3.2 made us
1896 suffer much more from the xbt_dynar_foreach low performance, but
1897 once we solved this, this refactoring proved to be very performance
1898 effective. From the 40% speedup, somehow, 10% are due to the
1899 inlining and 30% to the refactoring.
1901 That's a pitty that gcc cannot inline functions placed in other files
1902 alone. We have to choose between:
1903 - break the encapsulation (by putting private data structures and
1904 accessors in headers files to help gcc)
1905 - live with low performance
1906 - switch to a decent compiler such as icc (not quite possible).
1908 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1910 SimGrid (3.3.2) stable; urgency=low
1912 The "Simplicity does not preceed complexity, but follows it" release.
1914 The main contributors of this release were (lexical order):
1915 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1918 * Extract the routing logic into its own object.
1919 (was dupplicated in network.c and workstation_LV07.c;
1920 Allows to implement other ways of storing that info)
1921 => kill now useless network_card concept
1922 - Use dynar to represent routes (instead of void** + int*)
1923 - kill link_set (use surf_network_model->resource_set instead)
1924 - Add a command-line option to choose the routing schema to use
1925 - Add three new models:
1926 * Floyd (shortest path computed at initialization)
1927 * Dijikstra (shortest path recomputed all the time)
1928 * Cached Dijikstra (shortest path computed on need)
1929 All these models where contributed by Silas De Munck, and are
1930 described in his ICCS09 paper.
1932 * Simplify model declaration
1933 (less redirections, less function to write when defining a model)
1934 - Factorize stuff between models:
1937 surf_model_resource_set(model)
1938 surf_model_resource_by_name(model, name)
1939 - Unify the types of models in s_surf_model_t (using an union)
1940 - Embeed fields of common_public directly into s_surf_model_t
1941 - Rename model methods:
1942 action_free ~> action_unref
1943 action_change_state ~> action_state_set
1944 action_get_state ~> action_state_get
1945 - Change model methods into functions :
1946 (model)->common_public->action_use ~> surf_action_ref
1948 * Implement a generic resource; use it as ancestor to specific ones
1949 (allows to kill duplicated code in models)
1950 Drawback: timer command don't need no name nor properties;
1951 workstation_CLM03 don't need no properties
1952 (but I guess we can live with those few bytes wasted)
1954 * Improve the action object model
1955 - implement a constructor avoiding dupplicated code about field
1956 initialization in generic_action part.
1958 * Kill the SDP model: it has an external dependency, is deprecated
1959 in flavor of modern lmm models, and didn't compile since a while
1962 * Relocation of the context module from XBT to SIMIX.
1963 (the context were decoupled from the simix processes, duplicating a lot of code)
1964 => a lot of code was factorized
1965 - less overhead is introduced during scheduling
1966 - simpler API for the context factory
1967 - the logic for process creation,destruction and manipulation was simplified
1968 * Simplification of the s_smx_process_t data structure.
1969 => accesing the simix level data associated to a process is faster now,
1970 and the code is a lot more readable.
1973 * Implement some more MPI primitives:
1974 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1975 -implementation: Bcast: flat or 2-ary tree (default),
1976 Barrier: 4-ary tree,
1978 Allreduce: Reduce then Bcast
1979 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1980 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1981 Alltoallv: flat tree, like ompi
1983 * Add support for optimized collectives (Bcast is now binomial by default)
1984 * Port smpirun and smpicc to OS X
1987 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1988 Shout out if you used it.
1991 * Display the list of still queued messages in SG mode when existing
1995 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1996 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1997 * Add xbt_str_from_file(FILE*)
1998 * Add xbt_dict_get_key achieving a linear reverse search
1999 * Remove the context module
2001 Portability report of this version:
2002 * Main portability targets:
2003 - Linux(debian)/x86/context
2004 - Linux(debian)/x86/pthreads
2005 - Linux(debian)/amd64/context
2006 - Linux(debian)/amd64/pthreads
2007 On these, we still have the eratic breakages of gras/pmm and
2008 amok/saturate_sg reported in previous version. We still think
2009 that the tests are the cause of the fault, not the tested code.
2011 - Mac OSX Leopard/x86/context
2012 Still false negative in tesh autotesting.
2013 Smpi still fails, but this time because readlink does not accept -f
2014 Everything seems to work properly beside of that.
2017 - AIX version 5.3 (only tested contexts this time)
2018 Smpi still fails there because mktemp is not installed.
2019 Everything seems to work properly beside of that.
2021 I managed to compile it for the first time, but several breakages.
2022 Won't delay the release for this exotic platform.
2024 * Windows: it's still lagging behind. If you want to help, please
2027 Timing report of this version:
2028 This version seem to be more than 5% faster than 3.3.1 (on linux
2029 64bits with contextes). The gain is less than expected, we are
2030 investigating this for next release.
2032 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2034 SimGrid (3.3.1) stable; urgency=low
2037 * Implement a --cfg-help to show existing configuration variables
2038 * Build chain do not require doxygen in maintainer mode
2041 * fix a bug on struct sizeof computation, which prevented the
2042 exchange of arrays of structs in some conditions
2043 - added a regression test about this in datadesc_usage
2044 * Allow the exchange of 0-long dynamic vectors.
2045 - for that, use -1 as indicator of dynamic size instead of 0
2046 - This implied to change any size from unsigned long to long,
2047 reducing a bit communication abilities, but I guess that with
2048 64bits being quite common, this is more than enough.
2049 - This also induce a protocol change, thus bumping network protocol
2050 version from 0 to 1 (if we have external users, we have to get
2051 clean on that point too ;)
2052 - added two regression tests about this in datadesc_usage
2053 * Be more verbose when propagating local exceptions
2054 This helps debugging.
2055 * Display the status of simulated processes when receiving SIGINT in
2059 * Allow to control the simulation from a trace file.
2060 New functions MSG_action_register() and MSG_action_trace_run()
2061 The first one allows to associate a function execution to each
2062 kind of action while the second one parses a trace file and
2063 triggers the corresponding actions within the system.
2064 For now, only a toy example is provided in examples/msg/actions
2065 * Add an exemple of process migration in examples/msg/migration
2066 * Fix a bug in task exchange which broke MSG_task_get_sender()
2067 Add a teshsuite regression test for that.
2068 [Bug: if MSG_task_get_sender() is called after sender exit,
2070 * Fix a bug which prevented suspend/resume to work properly
2071 * Display the status of simulated processes when receiving SIGINT
2072 This fixes a regression of v3.3. due to the introduction of SIMIX
2073 * Bug fixing in failure management:
2074 - trace could not start by a failure at time 0
2075 - failure during communications were not working
2078 * Add SIMIX_process_set_name() to change the name of the current
2079 process in the log messages.
2080 * Store smx_hosts in a dict since we only retrieve them by name
2081 * Move the configuration infrastructure to surf
2084 * Move the configuration infrastructure to surf
2087 * Massive internal cleanups:
2088 - Store internal structures on processes instead of hosts (allows
2089 to have more than one process per host, in addition of being more
2091 - Cleanup the initialization/finalization process
2092 - Kill a whole bunch of unneeded synchronization:
2093 processes run in exclusive manner within the simulator
2094 - Move queues from global tables to process data fields
2096 - now accept -platform and -hostfile arguments
2097 - Pass the right rank value to processes according to the hostfile
2098 * Compile the examples by default, and use them as regression tests
2099 * Implement MPI_Wtime()
2100 * Change the reference speed to a command line option
2103 * TCP_gamma can now be specified as command line option using
2104 --cfg=TCP_gamma:10000000.0
2105 * Change the --surf-path cmd line option into --cfg=path:
2108 * Also include strbuff from xbt.h public header
2109 * xbt_ex_display(): do not free the exception after displaying
2110 This allows to do more with the given exception afterward.
2111 Users should call xbt_ex_free() themselves.
2115 Portability report of this version:
2116 * Main portability targets:
2117 - Linux(debian)/x86/context
2118 - Linux(debian)/x86/pthreads
2119 - Linux(debian)/amd64/context
2120 - Linux(debian)/amd64/pthreads
2121 These targets fail about 1/10 of times on gras/pmm, but we believe
2122 that this is because of the test, not because of simgrid.
2123 amok/saturate_sg fails even more rarely, and the test may not be
2126 - Mac OSX Leopard/x86/context
2127 The test suite still spits tons of errors because some obscure
2128 force prevents us from removing the temporary directories
2129 arguing that they still contain some metadata I've never heard of.
2130 Smpi fails because seq is not installed.
2131 Everything seems to work properly beside of that.
2134 - AIX version 5.3 (both contexts and pthread)
2135 Smpi still fails there because mktemp is not installed.
2136 XML inclusions seems rosty on AIX.
2138 * Windows: it's still lagging behind. If you want to help, please
2141 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2143 SimGrid (3.3) stable; urgency=high
2147 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2148 [Malek Cherier & Mt]
2150 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2153 Until now, GRAS were using MSG as an interface to SURF. It was
2154 quite difficult because both interface have several differences
2155 (MSG channels vs GRAS sockets were the most notable point).
2157 This also opens the gate to SMPI (which should occur soon) and speed
2158 up simulations by to 40% (even if it were not the main goal).
2160 **************************************
2161 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2162 **************************************
2163 The point is that events occuring at the exact same timestamp are
2164 not scheduled in the same order with the old and new version. This
2165 may be enough to completely change the execution of simulations in
2166 some cases. Sorry for the inconvenience.
2168 * Cleanup and upgrade the XML format to push further scalability
2169 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2171 * Improve the testing infrastructure with tesh. Now a very large part of
2172 the code is tested not only by being run but also by checking that the
2173 output match an expected output [Mt].
2175 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2176 is really less memory-demanding, which should allow you to use
2177 larger files in SimGrid [AL].
2179 * Inform valgrind about our contextes, so that it becomes usable
2180 with the default (and more effecient) version of SimGrid
2181 [contributed by Sékou Diakite, many thanks]
2184 * Introduce a listener thread in charge of receiving incoming
2185 messages from the network. It allows to overlap communication and
2186 computation but most notably, it removes some stupid deadlocks due
2187 to the fact that so far, a process could not send and receive at
2188 the same time. This made most non trivial communication schema
2190 * Convert the PIDs from long int to int to match the MSG ones (and
2191 linux ones too) [Mt]
2192 * New function: gras_agent_spawn() to launch a new process on
2193 current host. Only working in simulation for now. [Mt]
2194 * New function: gras_os_hostport() returning a constant form (ie,
2195 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2198 * Make the backtrace of exceptions more human readable [Mt]
2199 * New module: xbt/str [Mt]
2200 a ton of string utility functions (split, join, printf to a newly
2201 allocated buffer, trim, etc)
2202 * New module: xbt/hash [Mt]
2203 SHA1 hashing algorithm (more to come if needed)
2204 * New module: xbt/synchro [Mt]
2205 synchronization tools (mutex and conditions) working the same way
2206 in simulation and in real life (mainly useful for GRAS, but not
2208 * New module: xbt/queue [Mt]
2209 classical producer/consumer synchronization scheme
2210 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2211 (using the classical functions will get serialized) [Mt]
2212 * Make dictionary internal table dynamic. No need to specify its size
2213 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2215 * Make sure the log channels are organized as a tree under windows
2216 (because of ANSI C compatibility issue, any channel were child of
2217 root directly) [Mt].
2220 * Cleaned many thing in surf and fixed a few bugs [AL].
2221 * Add a nice command line configuration mechanism to compose models [AL].
2222 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2223 the previous one (KCCFLN05). It relies on something that looks like
2224 a max-min sharing mechanism but cannot be written as such. A new solver
2225 was thus designed [AL].
2226 * Add a new solver to lmm. Based on Lagrange optimization and
2227 gradient-based descent, it enables to efficiently maximise systems s.a
2229 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2231 This solver enables to propose two new network models for TCP Reno and
2232 TCP Vegas based on Low's work. These models still need to be fully
2233 tested though [Pedro Velho].
2236 * Bug fix in SD_simulate. Now the time bound given as argument is
2238 * Use the new parallel task model (ptask_L07) as default.
2239 * Use the SURF command line configuration mechanism.
2240 * 0-size tasks (for synchronization) should now work.
2242 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2244 SimGrid (3.2) stable; urgency=high
2248 We still experience issues on this platform, but we believe that at
2249 least MSG is usable.
2251 GRAS API BREAKAGE (for simplification purpose, sorry):
2252 * the gras_msgtype_by_name is not used anymore. Instead of
2253 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2254 you can write (and must)
2255 gras_msg_send(toserver, "request", &request);
2256 - If you still want to pass a gras_msgtype_t to the function (to cache
2257 the type and avoid the lookup time), use the gras_msg_send_() variant.
2258 - Impacted functions:
2259 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2260 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2261 * The callbacks are now expected to return 0 when everything went well
2262 (just like the main() function)
2264 GRAS new features and improvements:
2265 * New module mechanism where user code can use per process globals [Mt]
2266 This is similar to gras_userdata_*() functions, but for libraries. It
2267 factorize some code developped over and over in the examples and AMOK.
2268 It has still to be documented and used (only amok/peermanagement is
2270 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2271 loss when they were small enough to fit into the buffer and sent quickly
2272 enough so that they can all get received in one shoot.
2273 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2274 if not found. Use the *_or_null() variant for the old semantic.
2275 * In gras_msg_handle, do not discard messages without callback.
2276 They are probably messages to be explicitly awaited later (ie, proofs of
2277 mis-synchronization in userland since they are sent before being awaited)
2278 No big deal usually.
2279 * gras_socket_meas_send/recv: semantic changed!
2280 The numerical arguments used to be (1) the total amount of data to send
2281 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2282 messages. This was need for the fool willing to send more than MAXINT
2283 bytes on quite fat pipes.
2286 * Do really rename the hostmanagement module to peermanagement. [Mt]
2287 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2288 but this is rather new and this was documented in the module
2289 documentation (poor excuses, I admit)
2290 * Bandwidth measurement semantic changed! This follows the changes to
2291 gras_socket_meas_send/recv explained above.
2294 * A sequential mode has been added to the workstations. When a workstation
2295 is in sequential mode, it can execute only one task, and the other tasks
2296 are waiting in a FIFO. [Christophe Thiery]
2299 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2300 model for SIMDAG. [Christophe Thiery]
2301 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2302 equal to 0, which caused some bad behaviors in
2303 saturated_constraint_set_update. I now use a threshold mechanism like in
2307 * When running manually src/testall, you select specific units [Mt]
2308 testall is the result of our cunit mechanism, and should replace all
2309 the scripty thingy around since bash don't run easily on billware.
2311 * A mallocator system has been added. [Christophe Thiery]
2312 Mallocators allow you to recycle your unused objects instead of freeing them
2313 and allocating new ones.
2315 Documentation update:
2316 * FAQ reworking + New FAQs:
2317 - "Valgrind spits tons of errors!" [Mt]
2318 - "How to repport bugs" [Mt]
2319 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2320 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2321 - Communication time measurement within MSG [AL]
2322 - I experience weird communication times when I change the latency [AL]
2323 * GRAS tutorial [Mt]
2325 - an introduction to the framework and to the used communication model
2326 - an initiatic tour introducing the most proheminent features:
2328 . Lesson 0: Installing GRAS
2329 . Lesson 1: Setting up your own project
2330 o Part 2: Message passing
2331 . Lesson 2: Exchanging simple messages
2332 . Lesson 3: Passing arguments to the processes (in SG)
2333 . Lesson 4: Attaching callbacks to messages
2334 . Lesson 5: Using globals in processes
2335 . Lesson 6: Logging information properly
2336 . Lesson 7: Using internal timers
2337 . Lesson 8: Handling errors through exceptions
2338 . Lesson 9: Exchanging simple data
2339 . Lesson 10: Remote Procedure Calling (RPC)
2340 . Lesson 11: Explicitely waiting for messages
2341 . Recapping of message passing features in GRAS
2342 - A HOWTO section containing:
2343 o HOWTO design a GRAS application
2344 More are due, of course. They will come latter. In the meanwhile, you can
2345 check the examples which are still here.
2347 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2349 SimGrid (3.1) stable; urgency=high
2353 There was a stack corruption somewhere, visible only when optimizing
2354 with these versions. [Vince]
2357 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2358 module that enabled to play with Directed Acyclic Graphs. It is built
2359 directly on top of SURF and provides an API rather close to the old
2360 SG. Some old codes using SG are currently under rewrite to check that
2361 all needful functions are provided. [Christophe Thiery]
2364 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2365 extension of the classical CLM03 model that gracefully handles
2366 failures. This is now the default model for MSG and GRAS. It doesn't
2367 handle parallel tasks yet though. [AL]
2368 * Bug fix: Weights were not correctly set in the network part.
2369 WARNING: This may have resulted in incorrect results with simulations
2370 where there are more than one flow on a given link. [AL]
2373 * After a (long ?) discussion on simgrid-devel, we have decided that the
2374 convention we had on units was stupid. That is why it has been decided
2375 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2376 WARNING : This means that all previous platform files will not work as
2377 such with this version! A warning is issued to ask users to update
2379 A conversion script can be found in the contrib module of the CVS, under
2380 the name contrib/platform_generation/surfxml_update.pl [MQ]
2383 * Bug fix: Processes were started in reverse order, wrt deployment file.
2384 WARNING: if your code relies on this bug, please fix it. [AL]
2385 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2386 being executed on two different locations. [AL]
2387 * Bug fix: Failures are now better supported thanks to Derrick's tests
2388 (there was many failure situations I hadn't thought of and that weren't
2389 correctly handled). [AL]
2390 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2394 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2395 Exception occurring on server-side are propagated back to client (!).
2397 API CHANGE: the callback changed their prototype. Change:
2398 int my_handler(gras_socket_t expeditor, void *payload_data) {
2400 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2401 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2403 * New! function: gras_msg_handleall to deal with all messages arriving
2404 within a given period.
2405 * New! function: gras_socket_server_range to get a server socket in a
2406 range of port numbers (ease to avoid port number conflicts) [MQ]
2407 * New! gras processes display their backtrace when they get a SIGUSR1
2408 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2409 Sweet to debug RL processes [MQ]
2413 - Do not force experiment sizes to be expressed in kb, or it becomes
2414 impossible to measure the latency this way (needs one byte-long tests)
2415 WARNING: this changes the amok_bw_* function semantic. [MQ]
2416 - Implements the link saturation stuff. [MQ]
2417 * Peer management module:
2418 New! module factorizing code that we wrote over and over [MQ].
2421 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2422 - Test units are placed directly into the library code, they get extracted
2423 automatically and placed into the src/testall binary.
2424 - Convert most of the XBT tests to this system.
2425 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2426 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2427 live in the same C code. This required to change a little bit the API
2428 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2429 * New module: sparse graph structure with basic algorithms (this is work
2430 in progress and the API is not considered to be frozen yet). [AL]
2431 * Display more information on backtraces: source line & function names are
2432 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2433 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2434 * Reworked a little bit some #include statements to load only required
2435 headers. Some user code that relied on SimGrid to include stdlib or
2436 stdio may need to include it by themselves. [AL]
2437 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2439 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2440 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2441 but I guess that almost nobody used those parts. [MQ]
2443 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2445 SimGrid (3.0.1) stable; urgency=low
2448 * Unfortunately, I had missed 5 misnamed functions:
2449 xbt_fifo_item_t xbt_fifo_newitem(void);
2450 void xbt_fifo_freeitem(xbt_fifo_item_t);
2451 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2452 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2453 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2454 They're now deprecated. Please use their new versions:
2455 xbt_fifo_item_t xbt_fifo_new_item(void);
2456 void xbt_fifo_free_item(xbt_fifo_item_t);
2457 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2458 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2459 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2461 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2462 * Documentation: xbt_log module unmercifully reworked [MQ]
2463 * Bugfix: there was a problem with the ending of contexts with
2464 the pthread backend. It caused some weird deadlock or behavior
2465 depending on the pthread implementation. [AL]
2466 * Bugfix: get the exceptions raised in the simulator repport where
2467 and why they come from when they are not catched in time [AL, MQ]
2470 * Bugfix: Do repport the error when two non-connected hosts try to
2471 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2474 * Add additionnal checkings on communications. Assert that two
2475 communicating hosts are connected by a set of links... [AL]
2478 * Add additionnal checkings on channel values in communication [AL]
2479 * New: MSG_task_get_source to see on which host a task was generated [HC]
2480 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2481 the number of tasks waiting to be received on channel and sent
2483 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2484 waits for the first task coming from a given host.. [AL]
2486 GRAS new functionnalities: [MQ]
2487 * Enhance the parsing macro to allow the size of multidimentional objects
2488 to be given thru annotations.
2489 * New example (and documentation): Matrix Multiplication a la RPC
2490 (as when I was young!) and fix a bunch of bugs found on the way.
2492 GRAS performance improvements: [MQ]
2494 * Reduce the amount of cbps creation/destruction by making it static to
2495 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2498 * Change libdata to a set so that we can search for stuff by ID (and thus
2499 reduce the insane amount of dict lookups)
2502 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2503 RL communication on top of SG since it's so uneffective.
2504 It may also be used for inter-thread communication in RL, one day.
2505 * Use gras_datadesc_copy() to exchange messages on top of SG
2507 - improve message exchange performance on top of SG
2508 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2509 * Don't exchange on the network the size of the used part of buffer,
2510 instead, specify the possible buffer size to read().
2512 - reduces the amount of read/write calls (one pair per exchange)
2513 - reduces the amount of exchanged data (the size)
2514 - allows to retrieve all arrived data on receiver side, if we don't need
2515 it right now (subsequent read will peek the buffer)
2516 - allows the receiver to proceed with the begining of the stream before
2517 everything is arrived
2518 - make it possible to build an iov transport (using readv/writev)
2520 - take care of the data with non-stable storage (like stacked data),
2522 * If possible, TCP send uses vector I/O (when writev() is here)
2523 - Don't use it for receive since we send data sizes and data on the
2524 same stream, so we wouldn't be able to chain large amount of chunks
2525 before having to flush the stuff to read the size.
2526 * Rework the transport plugin mechanism to simplify it and reduce the
2527 amount of pointer dereferencement when searching for the right function
2530 * I guess that now, we do almost as few system calls as possible while
2531 doing as few data copy as possible.
2533 To improve it further, we could try to send all the sizes first and then
2534 all the data (to use iov on receiving size), but it's only a partial
2535 solution: when you have 2 dimensional data, the sizes of the second
2536 dimension is data of the first dimension, so you need 3 streams.
2538 I'm not sure the potential performance gains justify the coding burden.
2540 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2542 SimGrid (3.00) stable; urgency=high
2545 * New! Give the possibility to hijack the surf parser and thus bypass
2546 MSG_create_environment and MSG_launch_application. Have a look at
2547 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2549 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2551 SimGrid (2.96) unstable; urgency=low
2556 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2557 This deprecates the xbt_error_t mechanisms.
2558 It modifies (simplifies) all XBT and GRAS API.
2559 MSG API keeps unchanged (exceptions raised by XBT are catched from
2560 within MSG and masked with existing error handling facilities)
2563 * New! Add a FATPIPE model. [AL]
2564 * New! Add a parallel task model. [AL]
2565 * New! Add automatically a loopback interface (in the default
2566 network model) if none was precised.
2569 * Bugfix: MSG_process_resume now works with the current running process.
2571 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2572 * Modification of MSG_task_get_compute_duration. Once a task has been
2573 processed, the value returned by this function is now equal to 0. [AL]
2574 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2575 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2576 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2577 of a m_task. For now, MSG_task_cancel only works with computation
2579 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2580 of the processor (in Mflop/s) regardless of the current load on the
2582 * API Change: use proper naming convention for MSG_getClock and
2583 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2585 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2586 This function changes the priority of a computation task. This priority
2587 doesn't affect the transfer rate. A priority of 2 will make a task
2588 receive two times more cpu power than the other ones. This function
2589 has been added to suit the needs of Nguyen The Loc and hasn't been that
2590 much tested yet. So if it fails, please report it and send me your code.
2592 * API Change: removed all functions and types that were marked "deprecated"
2593 since many months. Renamed MSG_global_init_args to MSG_global_init.
2595 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2597 SimGrid (2.95) unstable; urgency=low
2600 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2601 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2602 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2605 * Bugfix: Allow absolute paths for platform description files [MQ]
2606 * Bugfix: do free the variables after use. Leads to drastic performance
2608 * Implement max_duration (ie, timeouts) on resources [AL]
2611 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2613 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2614 the needed support (timeouts) [AL]
2617 * Implement measurement sockets. You can now get the bandwidth between two
2618 hosts thanks to AMOK (see below). [MQ]
2619 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2620 dynar over the network (yeah) [MQ]
2621 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2622 * simple/static token-ring example. [Alexandre Colucci and MQ]
2623 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2624 in place before [MQ]
2627 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2628 type parsing. "short" and "long" modifiers were also ignored; other
2629 modifier (such as reference level) are still ignored. [MQ]
2630 * Embeed the buffer size within the buffer itself on SG. [MQ]
2631 That way, send() are atomic and cannot get intermixed anymore (at least
2632 the ones which are less than 100k; bigger messages still have the issue)
2633 * Array size pushed by the field, not by the field type (or each
2634 and every long int will push stuff to the cbps) [MQ]
2635 * use select() to sleep since it allows to portably sleep less than one
2638 GRAS (minor cleanups)
2639 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2642 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2643 * type category 'ignored' killed as it was never used and were difficult
2645 * whether a type can cycle or not is now a flag, leaving room for more
2646 flags (as "ignored", if we feel the need one day ;)
2647 * Rename raw sockets to measurement sockets since "raw" has another
2648 meaning in networking community.
2651 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2652 and offers features not belonging to GRAS but that most applications
2653 need. One day, it may be a set of plugins loadable at runtime.
2654 * New module: bandwidth
2655 bandwidth measurement between arbitrary nodes running this module. [MQ]
2657 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2659 SimGrid (2.94) unstable; urgency=low
2661 The first beta release of SimGrid 3 !
2665 * Update the main page and the FAQ. Adding references to gforge.
2668 * Add a gras_os_getpid function.
2671 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2672 * Extend the logs so that they also print PID, hostname, date, ... if
2674 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2675 and kill the old version which were in testsuite/
2676 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2680 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2681 for users. Instead, each programming environment (GRAS, MSG,...) use it
2682 when needed to provide such a feature to users.
2683 Don't shortcut the mechanism or you will also shortcut the virtualization
2684 you need on the simulator.
2688 * Cleanups in configury with regard to compile optimization/warning flags.
2689 Also add -fno-loop-optimize to any powerpc since it's the optimization
2690 killing gcc (< 3.4.0).
2691 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2692 needed by MSG examples complications
2693 * Borrow configury beautifications from PHP
2696 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2697 warning (thanks loris for stumbling into this one).
2698 * Bugfix: stop loading private headers (gras_config.h) from the public
2702 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2703 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2704 RL would imply a DNS resolver, which is *hard* to do in a portable way
2705 (and therefore delayed).
2706 * Implement a real timer mechanism and use it in timing macros. This allows
2707 to avoid rounding errors and get a 0.000005 sec precision in timing
2708 macros. While I was at it, various cleanups:
2709 - allow to declare more than one timed section per file (fix a stupid bug)
2710 - move some private declaration to the right place
2711 - merge conditional execution and timing macros into emulation module
2712 - document the module
2713 - make sure the module cleanups its mess on gras_exit
2714 * Documentation improvements:
2715 - (new) how to compile applications using GRAS
2716 - (new) emulation support (timing macros)
2718 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2720 SimGrid (2.93) unstable; urgency=low
2722 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2725 - Use Paje properly where used. Still to be sanitized properly.
2726 - Portability fix: Add an implementation of the contexts using pthread
2730 - Add xbt_procname(): returns the name of the current process.
2731 Use it to show the current process's name in all logging.
2733 - fix detection of older flex version and the reaction, since we do
2734 depend on modern ones (we use lex_destroy)
2735 - Better separation of SG and RL in the libs: remove all simulation code
2736 from libgras. As a result, this lib is now only 200k when stripped.
2737 Some of the xbt modules may also be duplicated (two sets and such) and
2738 should be cleaned/killed before SG3.
2739 - Insist on using xlC on AIX because of weird problems involving gcc there.
2740 - Cleanup the make remote stuff. This is now done by scripts
2741 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2742 mainly for our private use, but we're working on changing them to user
2745 - Bugfix: flush the socket on close only if there is some *output*.
2746 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2748 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2751 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2753 SimGrid (2.92) unstable; urgency=low
2755 Alpha 3 on the path to SimGrid 3
2759 - New! First try of benchmarking macros.
2760 - New! First try so that gras_stub_generator generate deployment and
2761 remote compilation helpers.
2763 - Bugfix: Initialization fix in msg_test.
2767 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2769 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2771 - API Change: gras_os_sleep to take the amount of seconds as a double.
2772 Accepting an int was error prone since it was the only location where
2773 seconds were coded as such. It leaded to damn rounding errors.
2774 - Bugfix: Hard to belive that timers ever worked before this.
2776 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2778 SimGrid (2.91) unstable; urgency=low
2780 Alpha 2 on the path to SimGrid 3
2784 - Bug fix in the lmm_solver.
2786 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2787 through the function MSG_paje_output.
2788 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2789 - It is possible to bound the rate of a communication in MSG with
2790 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2791 to write it in the changelog).
2792 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2797 - Various cleanups to the autotools stuff
2798 - Begin to move Gras examples to examples/gras/
2799 - Let make distcheck work again (yeah!)
2801 - documentation overhauled using doxygen.
2802 gtk-doc-tools is dead in SimGrid now.
2803 - Automatically extract all existing logging categories, and add the list
2804 to the documentation (long standing one, to say the less)
2806 - Cleanup the known architecture table. Reorder the entries to group what
2807 should be, and use a more consistent naming scheme.
2808 (some of the test dataset are still to be regenerated)
2809 - New! Allow library to register globals on each process just as userdata
2811 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2812 lookup time (for now).
2813 Use it in msg and trp.
2814 This cleans a lot the internals and helps enforcing privacy of the
2815 headers between the gras components.
2816 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2817 - Bugfix: gras_os_time was delirious in RL.
2818 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2820 - Reenable GRAS now that it works.
2822 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2824 SimGrid (2.90) unstable; urgency=low
2826 Alpha 1 on the path to SimGrid 3
2828 * It is a long time since the last release of SimGrid. I'm sorry about
2829 that but as I had told you, I was rewriting a lot of things. I apologize
2830 to those who had been reporting bugs to me and that I had not answered.
2831 If your bug is still in the new version, please tell me. Here is a
2832 summary of the main changes.
2834 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2835 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2836 portability, speed, and a lot more but you'll figure it out later.
2837 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2838 I wanted to release the new MSG as soon as possible and I have
2839 broken GRAS, which is the reason why, for now, only MSG is fully
2840 functional. A laconic description of these projects is available
2841 in the documentation.
2843 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2844 kernel optimized for our needs (called SURF but only the developers
2845 should use it). Hence, MSG is now roughly 30 times faster and I think
2846 that by rewriting a little bit MSG, I could event speed it up a little
2847 bit more. Beside the gain in speed, it is also much easier to encode a
2848 new platform model with SURF than it was with SG. More to come...
2850 * REVOLUTION 3: I have tried to change a little as possible the API of
2851 MSG but a few things really had to disappear. The main differences
2852 with the previous version are :
2853 1) no more m_links_t and the corresponding functions. Platforms are
2854 directly read from a XML description and cannot be hard-coded
2855 anymore. The same format is used for application deployment
2856 description. The new format is described in the documentation.
2857 Have a look in tools/platform_generation. There is a tiny script
2858 that converts from the old platform format to the new one. Concerning
2859 the application deployment format, parsing the old one is tricky.
2860 I think most of you should however be able to convert your files. If
2861 it is really an issue, I can write a C code that does the conversion.
2863 2) the toolbox tbx does not exist anymore. We now have a library
2864 with much more data-structures but without the hash-tables (we have
2865 dictionaries that are much faster).
2867 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2869 *****************************************************************************
2870 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2871 * don't feel like converting the oldies. (Mt) *
2872 *****************************************************************************
2875 Version 2.90: "the long awaited one"
2876 - Finished rewriting and debugging MSG. Rewrote the documentation.
2877 - disable GRAS for now since it needs to be ported to the newest SG
2880 - Finish the port to windows (using mingw32 for cross-compile)
2883 - Main loop and datastructures of SURF. A cpu resource object is
2884 functional. Surf can thus be used to create cpu's with variable
2885 performance on which you can execute some actions.
2887 2004-11-15 Martin Quinson
2888 - Port to ARM. Simply added the alignment and size descriptions. Should
2889 work, but the ARM machines are so slow that I didn't had the opportunity
2890 to 'make check' over there yet.
2892 2004-11-15 Arnaud Legrand
2893 - Trace manager now written. It uses a heap structure and is therefore
2894 expected to be efficient. It may however be speeded up (particularly
2895 when many events occur at the same date) by using red and black
2896 trees. One day maybe...
2897 - Max-min linear system solver written. It uses a sparse matrix
2898 structure taking advantage of its expected use. Most operations are
2899 O(1) and free/calloc are called as few as possible. The computation of
2900 the minimum could however be improved by using a red and black tree
2903 2004-11-03 Arnaud Legrand
2904 - Rename every gras_* function that was in xbt/ to its xbt_
2906 - Add a heap and a doubly-linked list to xbt
2907 - Added a dichotomy to the dictionaries. make check works as well before
2908 so I assume that the patch is correct. I do not know however if things
2909 run effectively faster than before now. :)
2911 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2912 SimGrid, and the version number is bumped to 2.x
2914 2004-10-29 Martin Quinson
2915 - Introduction of the remote errors.
2916 They are the result of a RMI/RPC on the remote machine.
2917 ErrCodes being scalar values, you can't get the host on which those
2918 errors did happen. Extending the error mechanism as in Gnome is possible.
2919 No idea yet whether it is a good idea.
2921 2004-10-28 Martin Quinson
2922 - Interface revolution: the Starred Structure Eradication.
2923 I used to do typedef struct {} toto_t; and then handle *toto_t.
2924 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2925 - ???_t is a valid type (builded with typedef)
2926 - s_toto_t is a structure (access to fields with .)
2927 - s_toto is a structure needing 'struct' keyword to be used
2928 - e_toto_t is an enum
2929 - toto_t is an 'object' (struct*)
2931 typedef struct s_toto {} s_toto_t, *toto_t;
2932 typedef enum {} e_toto_t;
2933 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2934 s_toto_t) is private.
2936 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2937 so that it can set the variable to NULL. It was so for dicts and sets,
2938 it changed for dynars.
2940 - Fix a bunch of memleaks in dict_remove
2941 - Fix a bug in sg/server_socket opening: it failed all the time.
2943 2004-10-07 Martin Quinson
2944 - Speed up dynar lookup operation a bit.
2946 gras_dynar_get is dead.
2948 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2949 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2950 (which gives you the address of the stored data).
2952 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2953 copy of the data using an affectation to do the job and not a memcpy.
2955 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2956 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2958 It does not really speedup the dynar test because they are
2959 setting elements all the time (and look them seldom). But the dict does
2960 far more lookup than setting.
2962 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2964 2004-10-05 Martin Quinson
2965 - Allow to (en/dis)able the cycle detection at run time.
2967 Whether we should check for cycle or not is now a property of each
2968 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2969 datadesc_cycle_unset allow to remove this property when previously set.
2971 Note that the cycle detection is off by default since it impacts the
2972 performance. Watch the data you feed GRAS with ;)
2974 This property is hereditary. Any element embedded in a structure having it
2975 set have it set for the time of this data exchange.
2977 You should set it both on sender and receiver side. If you don't set it on
2978 sender side, it will enter an endless loop. If you forget on receiver
2979 side, the cycles won't be recreated after communication.
2981 - Header reorganization.
2982 Kill gras_private.h, each submodule must load the headers it needs.
2984 2004-10-04 Martin Quinson
2985 - Interface revolution: do not try to survive to malloc failure.
2987 Now, gras_malloc and friends call gras_abort() on failure.
2988 As a conclusion, malloc_error is not a valid error anymore, and all
2989 functions for which it was the only gras_error_t return value are
2990 changed. They now return void, or there result directly.
2991 This simplify the API a lot.
2993 2004-09-29 Martin Quinson
2994 - Re-enable raw sockets.
2995 Created by gras_socket_{client,server}_ext;
2996 Used with gras_raw_{send,recv}
2999 It should allow to kill the last bits of gras first version soon.
3001 This is not completely satisfactory yet (duplicate code with
3002 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3005 - Simplify transport plugin (internal) interface by not passing any
3006 argument to _server and _client, but embedding them in the socket
3009 2004-09-28 Martin Quinson
3010 - Finish the port to AIX.
3011 autoconf was my problem (segfault within the malloc replacement
3012 function. No idea why)
3014 2004-09-16 Martin Quinson
3015 - Fix some size_t madness on 64bit architectures.
3017 2004-09-08 Martin Quinson
3018 - Reduce the number of system headers loaded, overload some more system
3019 calls (such as malloc to cast the result of the system one, and work
3021 - Fix and reintroduce the config support
3023 2004-09-07 Martin Quinson
3024 - Source code reorganization to allow Arnaud to surf all over there.
3025 - Allow to document the logging categories.
3026 - Remove all uppercase from logging categories and useless cleanup in names.
3028 2004-08-18 Martin Quinson
3029 Version 0.6.2 (protocol not changed; API changed)
3030 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3031 gras_error_t), and NULL when not found. Functions expecting a msgtype
3032 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3033 hopefully usefull message.
3034 - Portability to prehistoric sparcs again
3036 2004-08-17 Martin Quinson
3037 Version 0.6.1 (protocol not changed; ABI not changed)
3038 - prealloc some buffers to speed things up
3040 2004-08-11 Martin Quinson
3041 Version 0.6 (protocol not changed; ABI expended)
3042 - The parsing macro can deal with the references, provided that you add
3043 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3045 2004-08-09 Martin Quinson
3046 Version 0.5 (protocol not changed; ABI changed)
3047 - Allow to off turn the cycle detection code in data exchange at
3048 compilation time. It should be at run time, but I'm short of time (and
3049 the config stuff is still broken). That way, we keep dict out of the
3050 critical path, which is good because the performance is poor:
3051 - search not dichotomial yet
3052 - dynar give no way to access their content and memcpy everytime
3053 - In composed data description (struct, ref and so on), stop foolness of
3054 keeping the subtype's ID, but store the type itself. This keeps sets out
3055 of the critical path, which is good since they rely on dynar and
3056 dictionnaries. The only loose of that is that we cannot detect the
3057 redeclaration of a structure/union with another content (but I'm not sure
3058 the code detected well this error before anyway). We still can detect
3059 the redefinition discrepancy for the other types.
3060 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3061 it breaks the code because of type-punning used all over the place).
3062 This breaks on all non-gcc architectures (for now).
3064 All those changes (plus the buffer of last time) allow me to gain 2 order
3065 of magnitude on cruel tests consisting of 800000 array of integers on two
3066 level of a hierarchical structure (200 secondes -> 4 secondes)
3069 - the selector of reference must now return the type it points to, not
3070 the ID of this type.
3072 2004-08-06 Martin Quinson
3073 Version 0.4 (protocol changed; ABI not changed)
3074 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3075 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3076 - Implement a buffer transport, and use it by default (it relies on tcp in
3077 real life and on sg in simulation).
3078 That's a bit hackish since I had a new field to the structure to store
3079 its data without interfering with the subtype ones. Inheritance
3080 is tricky in C. And that's a kind of reverse inheritance with one class
3081 derivating two classes. Or maybe a game with java interfaces. Anyway,
3082 that's damn hard in C (at least).
3083 Moreover, I got tired while trying to ensure plugin separation and
3084 genericity in SG mode. MSG wants me to do weird things, so let's go for
3085 cruel hacks (temporarily of course ;).
3086 See comment in transport_private.h:71
3087 - do not include all the _interface headers in private but in the files
3088 which really need them (to cut the compilation time when they are
3091 2004-07-26 Martin Quinson
3092 Version 0.3 (protocol not changed; ABI changed)
3093 - Major overhault of the datadesc interface to simplify it:
3094 - shorted the function names:
3095 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3096 - add a trivial way to push/pop integers into the cbps without malloc.
3097 This allows to make really generic sub_type description, which simply
3098 pop their size of the stack.
3099 - add a function gras_datadesc_ref_pop_arr() which does what users want
3100 most of the time: Declare a dynamic array (which pops its size of the
3101 stack) and declare a reference to it. Poor name, but anyway.
3102 - kill the post-send callback, add a post-receive one
3104 2004-07-23 Martin Quinson
3105 Version 0.2 (protocol changed; ABI changed)
3106 - add some testing for cpbs in the test cases, and fix some more bugs.
3107 This invalidate again the little64 data file, since I cannot regenerate
3109 - remove an awfull optimization in the logging stuff, allowing me to:
3110 - understand it again
3111 - learn gcc how to check that the argument match the provided format
3112 - fix all errors revealed by gcc after that
3113 - internal keys of dict are not \0 terminated. Deal with it properly in
3114 loggings instead of segfaulting when the user want to see the logs :-/
3116 2004-07-22 Martin Quinson
3117 - Fix some stupid bug preventing cbps (callback postit) from working
3119 2004-07-21 Martin Quinson
3120 - Some documentation cleanups
3121 - remove the useless last argument of msgtype_declare
3122 - rename the Virtu functions to fit into the 'os' namespace
3123 - move headers src/include -> src/include/gras/ and stop fooling with
3124 gras -> . symbolic link
3125 - make distcheck is now successful
3127 2004-07-19 Martin Quinson
3129 - Build shared library also
3130 - Install html doc to the right location
3131 - stop removing maintainer files in make clean
3132 - build tests only on make check
3134 2004-07-13 Martin Quinson
3136 - No major issue in previous version => change versioning schema
3137 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3138 corresponding dataset.
3140 2004-07-11 Martin Quinson
3142 - Get it working with any kind of structure (we can compute the padding
3143 bytes remotely for all the architectures I have access to)
3144 - Implement the structure parsing macro (still not quite robust/complete)
3145 - Improvement to the remote testing toysuite
3147 2004-07-10 Martin Quinson
3148 [autoconf mechanism]
3149 - get ride of a bunch of deprecated macros
3150 - actually run the test for two-compliment, not only compile it :-/
3151 - test whether the structures get packed (and bail out if yes. Damn.
3152 Alignment is a serious matter)
3153 - test whether the structures get compacted (but respecting the alignment
3154 constraints of each types)
3155 - test whether the array fields of structures can straddle alignment boundaries
3157 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3158 (took me 2 hours to find that bug, looking at the wrong place)
3159 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3160 before sending/receiving and are used to compute the offsets of fields
3161 - Given that padding size depend even on compiler options, keep track of
3162 alignment and aligned_size only for the current architecture. Not a big
3163 deal since we send structure fields one after the other (seems
3165 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3166 program, let it work on linux/gcc/little32. portability todo.
3168 2004-07-08 Martin Quinson
3169 - import and improve remote compilation support from FAST
3170 - make sure make check works on half a dozen of machines out there
3172 2004-07-07 Martin Quinson
3173 Let's say it's version 0.0.3 ;)
3174 - Implement conversions (yuhu!)
3175 - Let it work on solaris (beside conversion, of course)
3176 - Stupid me, using rand() to generate the conversion datatests in not wise.
3178 2004-07-06 Martin Quinson
3179 - Let make dist work, since I'm gonna need it to compile on remote hosts
3180 - Let Tests/datadesc_usage write the architecture on which the file was
3181 generated as first byte.
3182 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3183 architecture descriptions.
3184 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3185 on those architectures.
3186 - Optimization: send/recv array of scalar in one shoot
3188 2004-07-05 Martin Quinson
3189 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3191 - Plug a whole bunch of memleaks
3192 - each process now have to call gras_{init,exit}. One day, their log
3193 settings will be separated
3194 - Continue the code factorisation between SG, RL and common in Transport.
3196 2004-07-04 Martin Quinson
3198 - Redistribution between SG and RL.
3199 We wanna have to accept in SG, so move accepted related parts of RL in
3200 the common part. (more precisely, the dynar of all known sockets is no
3201 more a static in transport.c, but part of the process_data)
3203 [gras_stub_generator]
3204 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3206 2004-07-03 Martin Quinson
3207 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3208 them all up in one shot)
3210 - Fix the ugly case of reference to dynamic array.
3211 - New (semi-public) function gras_datadesc_size to allow the messaging
3212 layer to malloc the needed space for the buffer.
3214 - gras_socket_close now expect the socket to close (and not its address to
3215 put NULL in it after it). This is because the socket passed to handlers
3216 is one of their argument (=> not writable).
3218 - propagate the interface cleanup from last week in datadesc, ie remove a
3219 superfluous level of indirection. User pass adress of variable
3220 containing data (both when sending and receiving), and not of a variable
3221 being a pointer to the data. Let's say that I like it better ;)
3222 The price for that is constructs like "int msg=*(int*)payload" in
3223 handlers, but it's a fine price, IMHO.
3225 - Let it work in RL (yuhu)
3227 2004-06-21 Martin Quinson
3229 - porting SG plugin and SG select to new standards (works almost).
3230 - plug memleaks and fix bugs around.
3233 - cleanup the prototype of data recv and force users to specify when they
3234 want to handle references to objects. Test case working even for cycles.
3235 - plug memleaks. Valgrind is perfectly ok with this.
3237 2004-06-12 Martin Quinson
3239 - cleanup the separation between plugin and main code in plugin creation
3241 2004-06-11 Martin Quinson
3243 - Reput hook for raw sockets, needed for BW experiments
3244 - kill a few lines of dead code
3245 [Data description] Interface cleanup
3246 - gras_datadesc_by_name returns the searched type or NULL.
3247 That way, no variable is needed to use a type desc once, which makes
3249 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3250 two parameters were strings (field name, type name), leading to
3252 [Dicos] Interface cleanup
3253 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3254 This is consistant with the dynar API.
3256 2004-04-21 Martin Quinson
3258 - Porting to new standards.
3260 - interface cleanup.
3261 There is no bag anymore, no need to take extra provision to mask the
3262 pointers behind "ID".
3263 Better splitup of functions between files create/exchange/convert.
3264 This is still a bit artificial since convert and receive are so
3265 interleaved, but anyway.
3267 - add a queued message list to procdata (the ones not matching criteria
3269 - factorize some more code between SG and RL wrt procdata
3271 - use gras_exit in example to track memleaks
3272 - get rid of gs_example now that GS is properly integrated into gras
3273 - update run_test to integrate the latest tests (datadesc)
3275 - rename WARNINGn macros to WARNn since it prooved error-prone
3277 2004-04-19 Martin Quinson
3279 - register init/exit functions within gras module mechanism
3280 - send/receive function.
3281 Convertion is not implemented, but short-cutted if not needed.
3282 struct/array elements are sent one by one (instead of block-wise), but
3283 nobody really cares (yet). Get a prototype before optimizing.
3284 - tests (using a file socket) for DD send/receive on:
3285 - base types: int, float
3286 - array: fixed size, string (ie ref to dynamic string)
3287 - structure: homogeneous, heterogeneous
3288 - chained list, graph with cycle
3289 Believe it or not, valgrind is not too unhappy with the results. The
3290 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3294 - Bugfix when using a filename explicitely (instead of '-')
3296 2004-04-09 Martin Quinson
3298 - factorize more code between RL and SG in socket creation
3299 - Complete the implementation and tests of:
3301 o file (only in RL, and mainly for debugging)
3303 I lost 3 days to design a portable address resolver, and then decided
3304 that the prototype mainly have to run on my box.
3305 Addressing portability too early may be like optimizing too early :-/
3307 - use gras_init in the Tests instead of the crappy parse_log_opt
3308 (the latter function is removed)
3309 [Conditional execution]
3310 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3311 [Code reorganisation]
3312 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3313 Build examples against the RL library, since there is no way to disable
3314 its creation for now.
3316 For information, the beginning of coding on GRAS was back in june
3317 2003. I guess that every line has been rewritten at least twice since