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.
6 - Java JAR file should be finally fully working on Mac OSX too.
8 XML platforms: Switch to platform v4.1 format.
9 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
10 - <zone> can be used as a synonym for the now deprecated <as>
11 - <zoneZoute> an be used as a synonym for the now deprecated <asroute>
12 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
13 - <actor> can be used as a synonym for the now deprecated <process>
14 - state_file and avail_file periodicity is now easier to express
15 (check the documentation) the old behavior should still work.
16 * Storage: not backward compatible, but it seems that nobody ever used it
17 - Remove attribute 'content_type' of <storage_type>: was never used
18 - Make attribute 'model' of <storage_type> optional: for future usage
19 - Remove Bconnection model property: was never (in)validated. Replaced by
20 the maximum of the read and write bandwidth as a resource constraint.
23 - New and Backwards Compatibility break:
24 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
25 When one wants to get the list of tasks whose states have changed during a
26 simulation round, s/he has to allocate and free a dynar and use it as
27 argument to this function. The former SD_simulate (double how_long)
31 - Allow multicore VMs, along with the correct sharing computations
34 - New function to update the consumption of all hosts at once.
37 - The netzone are now available from the MSG API.
38 The old names still work, but are now deprecated.
41 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
42 - New option: smpi/keep-temps to not cleanup temp files
43 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
46 - Replay: New function xbt_replay_action_get():
47 Retrieve the function previously associated to an event type.
48 - DROPPED FUNCTION: xbt_str_varsubst()
49 - DROPPED MODULE: strbuff. We don't need it anymore.
50 - DROPPED MODULE: matrix. We don't need it anymore.
51 - DROPPED MODULE: lib. We don't need it anymore.
53 -- Release target: June 21 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
55 SimGrid (3.15) stable; urgency=low
57 The Spring Release: continuous integration servers become green
58 We fixed even the transient bugs on all target architectures:
59 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
60 Windows, FreeBSD, NetBSD.
62 - Rename NetCards to NetPoints.
63 This was intended to help NS3 users, but that's not a netcard.
64 That's a point in the routing algorithm, let's avoid wrong simplifications.
67 - New: MSG_process_yield(). Stop and yield to other processes.
68 - New: MSG_process_daemon(). Daemon processes are automatically killed
69 when the last non-daemon process terminates
70 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
71 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
72 to make room for the upcoming network energy plugin.
73 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
76 - Ensure that an actor can kill itself with Process::exit()
77 - Kill the obscure NativeException. Nobody want to survive the issues
78 it denotes, so use JniException that is a RuntimeException (not to
79 be caught explicitly).
80 - Partial bug fix in initialization. SimGrid flags on command line were
81 consumed at C level but stayed in the original Java String[] args.
82 This could mess users' args[i] if SG flags were not put at the end of
84 The SimGrid flags are now removed from the Java arguments. However,
85 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
86 if args.length is greater than the number of YOUR OWN ARGUMENTS.
87 It might be if you have --log or --cfg flags in the command line.
88 - Fix numerous memleaks all around the place. In particular, around VMs.
92 - s4u::onPlatformCreated: right before the simulation starts
93 - s4u::onSimulationEnd: right after the main simulation loop
94 - s4u::onTimeAdvance: right after a clock change (time is discrete)
95 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
96 event from the availability_file changes the avail speed.
97 - Links are now usable from s4u
98 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
99 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
100 - New examples: The conversion of MSG examples to S4U has begun
101 - Actors: create, kill, migration, and suspend
102 - Applications: master-worker and token-ring
103 - Action replay: communications and storage
104 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
107 - Backwards Compatibility breaks
108 - The SD_route_get_list and SD_route_get_size functions have been removed.
109 They are replaced by the unique
110 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
111 The route from src to dst is built in the links dynar whose size can
112 be obtained with xbt_dynar_length.
113 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
114 sg_host_route_bandwidth and sg_host_route_latency.
115 Macros ensure the backwards compatibility, but you should fix your code
118 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
119 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
120 - Bump our claim of support from MPI 1.1 to MPI 2.2.
121 We don't support 100% of it, but it should be enough. Contact us if not.
122 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
123 naive and probably inaccurate)
124 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
125 mapped onto a unique small file using some system magic.
126 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
127 - Fortran: cleanups, fixes, support of user-added operations
128 - RMA: MPI_Accumulate are applied in correct order.
129 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
130 - Better support for MPI_IN_PLACE
131 - Support for MPI_Win attrs and keyvals.
132 - Support MPI_Comm_group_create, MPI_Type_size_x
133 - MPI ops should be properly applied to all allowed types
134 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
135 algorithm by a non failing one from mpich.
138 - Kill the fifo data container: we don't use it anymore.
141 - A <cluster> can now be created with different speed values to represent pstates.
142 see examples/platforms/energy_cluster.xml for an example.
144 -- Wed Mar 22 17:50:21 CET 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
146 SimGrid (3.14.159) stable; urgency=low
148 The previous release was only a crude approximation of a Pi release;
149 this one should be a bit better.
151 - Revert a last minute change that broke on Mac OSX.
152 - Fix the changelog and NEWS files.
153 - Improve the documentation, in particular of the routing module.
154 - Integrate some patches flying around in Debian.
156 -- Wed Dec 28 17:14:49 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
158 SimGrid (3.14) stable; urgency=low
161 * The whole documentation was reworked and reorganized. There is
162 still much room for improvement here, but we're on it.
165 * We now need python3 for our testing infrastructure.
166 * Model checking needs libevent
169 * Now works on FreeBSD
172 * Privatization now works on FreeBSD
174 * Privatization is now activated by default in smpirun
175 You can turn it off with -no-privatize if something goes wrong.
177 * Call-location tracing for SMPI.
178 You can add the exact location (filename / linenumber) of an MPI call to
179 your trace files and slow down or speed up the simulation between two
180 consecutive calls by using an adjustment file (see the documentation).
182 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
183 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
184 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
185 values were ignored in some cases. The timings of these functions can now
186 be significantly different.
188 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
189 smpi/running-power is renamed to smpi/host-speed
191 * smpi/grow-injected-times option to enable or disable multiplication of the
192 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
193 make simulation less precise (but also much faster).
195 * smpirun script should be (much) faster for large deployments.
197 * SMPI tracing : fixed issue with poor matching of send/receives.
199 * Replay : Fix broken waitall
201 New functions and features
202 * MSG_parallel_task_execute_with_timeout, to timeout computations.
204 Dropped / renamed functions and features
205 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
206 - MSG_mailbox_is_empty() -> Mailbox::empty()
207 - MSG_mailbox_front() -> Mailbox::front()
208 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
209 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
210 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
211 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
212 * Don't pass the free_f parameter to property related functions:
213 - MSG_host_set_property_value()
214 - MSG_as_router_set_property_value()
215 - MSG_storage_set_property_value()
216 * VM properties. Since msg_vm_t are msg_host_t, just use
217 MSG_host_get_property_value() and friends
218 * VM I/O related things:
219 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
220 - Unimplemented save/restore methods
221 * MSG_as_router_get_property_value() was redundent with
222 MSG_environment_as_get_property_value().
223 - Removed MSG_as_router_*propert*() functions
224 - Added MSG_environment_as_set_property_value() for consistency
225 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
226 Well, they are still there for now, but deprecated with a warning.
227 Please switch to xbt_dict_new_homogeneous() before this is removed
229 * Task affinity. Its intended behavior (that was very badly tested
230 and probably not really working) was deceiving what most users
231 would have hoped here.
232 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
233 * Option network/coordinates is now useless and should be dropped.
236 * Added option storage/max_file_descriptors to allow more than 1024 files opened
239 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
240 * Added Dragonfly topology support
243 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
244 EiBps, EBps, Eibps, Ebps for bandwidth.
245 They may become useful to some lucky ones.
248 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
249 * ex/app_tokenring: new example, very similar to the MSG Token Ring
250 * ex/async_waitAll: new example, on asynchronous communications
253 * Memory usage should be decreased for simulations with a large number
254 of processes. This also helps for SMPI.
256 -- Sat Dec 24 22:34:26 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
258 SimGrid (3.13) stable; urgency=low
260 The Half Release, a.k.a. the Zealous Easter Trim.
262 - We removed half of the lines, that were mostly experimental cruft.
263 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
264 (not counting blanks and comments -- according to openhub.net)
265 - The internals are now compiled in C++ (and will soon be clean C++)
266 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
267 59 klines of C, etc. We added only 29 klines of C++ in replacement.
269 * Backwards Compatibility breaks
270 - Removed Lua simulation bindings (switch to C or Java for that).
271 Lua can still be used to describe platforms
272 - Removed Java kernel plug-ins.
273 Will be reintroduced after the ongoing major internals reorg.
275 - the following functions were removed.
276 They were too specific and should be reimplemented in a generic
277 way, with filter function.
278 - MSG_task_listen_from_host
279 - MSG_mailbox_get_count_host_waiting_tasks
280 - MSG_mailbox_put_with_timeout was removed.
281 Please use MSG_task_send_with_timeout instead.
283 - the SD_application_reinit function was removed. It has been a noop for a while.
284 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
285 replaced by a more flexible mechanism.
286 - The following functions thus do not exist anymore
287 - SD_workstation_get_access_mode
288 - SD_workstation_set_access_mode
289 - SD_workstation_get_current_task
290 - Basic estimation functions have been removed but can easily be replaced
291 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
292 - SD_workstation_get_computation_time => amount / sg_host_speed()
294 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
295 Use VM.getSpeed()*load/100 for the legacy behavior.
297 - option enable_tracing was removed. It was not doing anything for a while.
298 - In the ModelChecker:
299 - the model-checker now ptraces the model-checked process which means
300 you cannot use a debugger on the latter anymore (we might make this
301 optional in the feature);
302 - removed soft-dirty page tracking;
303 - remove model-checked side snapshot management,
304 MC_snapshot() and MC_compare_snapshot();
305 - keep the MC_cut() function as a stub (it was not really working
306 in the previous release).
309 * All options are consistently in kebab-case. Old names are kept as alias.
312 * Switch to platform v4 format.
313 - Rename from 'power' to 'speed' the attributes describing the amount of
314 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
315 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
316 - In <host> and <link>, attributes availability and state are gone.
317 It was redundent with state and availability traces, and with peak values.
318 - In <cluster>, attributes availability_file and state_file are gone.
319 It was too complex and unused.
320 - Kill <gpu>. Was not doing anything.
321 - The DOCTYPE points to the right URL:
322 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
323 (the file at this address now documents the changelog since its v1)
324 - A warning is emitted for unit-less values (they are still accepted).
325 - speed. Default: 'f' or 'flops'. Also defined:
326 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
327 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
328 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
329 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
330 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
331 - latency. Default: 's' second. Also defined:
332 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
334 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
335 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
338 * s4u::Host is now the preferred public interface to the Host features.
339 sg_host_* functions are C bindings to the exact same behavior
340 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
343 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
344 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
345 They were deprecated since a few years, and probably did not even compile anymore.
348 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
349 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
350 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
351 encouraged to update their codes. A list of the modified functions can be found at the end of
352 include/simgrid/simdag.h
355 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
356 the lambda or closure passed as a parameter will run in kernel mode.
357 All the callback functions should be rewritten to that interface at some point.
360 * Reorganizing and cleaning the internals all around the place.
363 * Remove old default barrier/bcast buggy algorithms (see #18407)
364 * Various bug fixes to handle more codes
365 * Remove the need for the --foreground option of smpirun (it is still
366 accepted for backward compatibility).
369 * Kill the setset data container: MC don't use it anymore.
370 * Kill the queue data container: it made more sense with GRAS.
371 * Kill the xbt_peer_t data type: it's useless without GRAS.
372 * Kill rm_cb feature of config sets: it was never useful.
373 * Kill graphxml parsing feature. It was not used.
374 * Kill the deprecated code protected by XBT_USE_DEPRECATED
376 - xbt_dynar_sort_strings(), when the content is char*
377 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
378 They throw exceptions on invalid input;
379 * C++ support for declaring CLI flags (simgrid::config::Flag);
380 * class for abstracting different signal backends (simgrid::xbt::signal).
381 with no external dependencies (we need very simple signals).
384 * refactoring and cleanup of the code;
385 * ongoing process to cleanly separate the model-checking algorithms
386 from the code model-checking support.
388 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
390 SimGrid (3.12) stable; urgency=low
392 The Facelift Release.
395 * Require g++ v4.7 at least to not speak prehistorical C++.
396 * Require Boost 1.48 (for signal2 component).
397 * Java must be version 7 at least when activated.
398 * Builds on Windows again (including Java bindings).
399 * Tracing is now always enabled (no way to turn it out)
400 * Remove GTNetS. It was not working anyway.
401 * Various cleanups in the cmake scripts.
402 * Move headers around to sort them out on installed systems:
403 - instr/instr.h -> simgrid/instr.h
404 - instr/jedule/* -> simgrid/jedule
405 - simdag/datatypes.h was removed
406 - simdag/simdag.h -> simgrid/simdag.h
407 - msg/datatypes.h was removed
408 - msg/msg.h -> simgrid/msg.h
411 * Interface improvement:
412 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
413 - Sanitize the interface in MSG_task_ module:
414 - Merge two functions that were close enough but misleading:
415 set_compute_duration(t) -> set_flops_amount(t)
416 get_remaining_computation(t) -> get_flops_amount(t)
417 - set_data_size(t) -> set_bytes_amount(t)
418 get_data_size(t) -> get_bytes_amount(t)
419 - Massive cleanups in the functions related to the energy
420 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
421 - New: MSG_host_get_pstate()
423 - msg/energy/onoff: switching hosts on and off
425 * Interface improvement:
426 - Rename Host.isAvail() to Host.isOn()
427 - Rename Process.currentProcess() to Process.getCurrentProcess()
428 - Rename Task.setDataSize() to Task.setBytesAmount()
429 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
431 - #18874: Actually allows the GC to reclaim tasks
435 - SIMIX_process_throw: raises an exception in a remote process
436 * Refactoring: Separate sub-modules
437 - libsmx: the public interface, as libc in a real system
438 - popping: the strange dance that converts a user request into a kernel handling
439 - smx_context_*: the virtualization mechanisms that embed the user code
440 - smx_*: the handling of each simcalls
442 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
443 * Rename smx_action_t into smx_synchro_t, making explicit that these
444 things are used to synchronize processes with their environment.
445 For example, a communication is a sort of synchronization involving
446 the communicating processes (that may block until the exchange) and
447 the platform. The same can be said from computations, etc.
449 - #18888: segfault when a process ends before its kill_time
453 - 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.
454 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
455 - MPI_Type_set_name, MPI_Type_get_name
456 - MPI_*_c2f and MPI_*_f2c functions
457 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
458 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
459 - Activate a lot of new tests from the mpich 3 testsuite
461 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
462 - InfiniBand network model added : Based on the works of Jerome Vienne
463 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
464 - When smpi/display_timing is set, also display global simulation time and application times
465 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
466 * Collective communications
467 - 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
468 - 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/).
469 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
471 - Replay now uses algorithms from wanted collective selector
472 - Replay can be used with SMP-aware algorithms
473 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
474 - 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.
476 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
477 - Correctly use loopback on fat-tree clusters
478 - Asynchronous small messages shouldn't trigger deadlocks anymore
479 * Energy/DVFS cleanup and improvement
480 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
481 - new: smpi_get_host_pstate()
484 - "Full" network optimization flag was broken since Surf++
485 - Better handling of precision flags in maxmin
486 - Fix bug causing sometimes "Impossible" errors
487 - Properly pass cluster properties to included hosts
488 * Improvement of the Energy plugin.
489 - Always update the consumption before returning that value
490 - New property: watt_off to denote the dissipation when the host is off
491 - New functions getWattMinAt and getWattMaxAt to retrieve the
492 dissipation of pstates that we are not currently at.
493 * Java: class NetworkLink renamed to Link
494 * New function: simcall_process_get_kill_time()
495 * Massive rename s/workstation/host/
496 - That's intrusive, but that's good for the project consistency. Sorry.
497 - Change config option "workstation/model" into "host/model"
500 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
501 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
502 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
503 - xbt_dirname and xbt_basename for non-POSIX systems
505 * The model checker now runs as a separate process.
506 * The model checker runs must now be launched with the new simgrid-mc program.
507 * Record/Replay: the MC can display a textual representation of a path in the
508 execution graph. It can then be replayed outside of the model checker.
510 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
512 SimGrid (3.11) stable; urgency=low
517 * Normalizing pointers addresses tool for better diff between logs
520 * Add cloud examples using new VMs
521 - examples/msg/cloud/two_tasks_vm.tesh
522 - examples/msg/cloud/simple_vm.tesh
523 - examples/java/cloud/cloud.tesh
524 - examples/java/cloud/migration/migration.tesh
525 * Add java surf examples:
526 - examples/java/surfPlugin/surf_plugin.tesh
527 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
528 - examples/java/surfCpuModel/surf_cpu_model.tesh
529 * Add SMPI+MSG example:
530 - examples/smpi/smpi_msg_masterslave/
535 - msg task destroy cancel test
536 - msg_host on/off test
537 * Move all tests in testsuite to teshsuite (adding tesh files)
538 * Restructure teshsuites
539 - one folder for each kind of test
540 * Restructure AddTests.cmake
542 - structure the order of tests (with sections)
545 * Add virtual machine
546 - creation of a VM on a PM
547 - migration of a VM from a PM to another PM
549 - MSG_process_join(msg_process_t process, double timeout)
550 - msg_bar_t MSG_barrier_init(unsigned int count)
551 - int MSG_barrier_wait(msg_bar_t barrier)
552 - void MSG_barrier_destroy(msg_bar_t barrier)
553 - msg_as_t MSG_environment_as_get_by_name(const char * name)
554 * New option "msg/debug_multiple_use" to help debugging when a task is used
557 - Improvements and finalization of MSG_storage, MSG_file APIs and their
559 - Increase code coverage in test suites
563 * Protect context stacks against stack overflow. The number of protected memory
564 pages allocated on the top of each stack (1 by default) can be configured
565 with the new command line option --cfg=contexts/guard_size:<value>.
566 * Simcalls are now generated by a python script that
567 - generates files included by SimGrid
568 - checks that all the functions exist, and proposes prototypes
570 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
571 host_set_data, host_get_data
573 - simcall_process_join(smx_process_t process, double timeout)
574 * Fix bug where sleeping processing could not be suspended.
577 * Translate surf models from C to C++
578 - Generic classes for all models: Model, Resource, Action
579 - A generic interface for each kind of model (CPU, Network, Storage
580 Workstation, WorkstationVM)
582 * Translate surf routings from C to C++
583 * Add callbacks using sigc++ or boost::signals2
584 - Add callback functions for resource creation/destruction
585 - Add callback functions for action state change
586 - Handle Energy as a plugin
587 * Replace swag by boost::intrusive
588 * Add new routing models for clusters. For documentation, see
589 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
590 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
591 parameters for cluster tag
592 - Fat trees, with topology="FAT_TREE" and
593 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
594 - see examples/platforms/torus_cluster.xml and
595 examples/platforms/fat_tree_cluster.xml
599 * Hostfiles support host:nb_processes construct to deploy several processes on
601 * Collective communication algorithms should not crash if used with
602 improper number of nodes and report the error.
603 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
604 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
605 MPI_Dims_create, MPI_Cart_sub are supported.
606 * New interface to use SMPI programmatically (still depends on MSG for
607 some parts, see examples/smpi/smpi_msg_masterslave) :
608 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
612 * Global variables privatization in MPI executables is now performed at runtime
613 with the option smpi/privatize_global_variables (default:no).
614 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
615 dynamic libraries loaded by the application are not privatized (static
616 linking with these libraries is advised in this case)
619 * Options defined in XML work correctly now.
622 * New cmake option, enable_lib_in_jar, to control whether native libraries are
623 copied into simgrid.jar or not (ON by default). Use this option if you want
624 to reduce the size of the installed simgrid.jar, *and* the native libraries
625 are kept installed elsewhere.
626 * Surf binding with SWIG (code generated in maintainer mode only):
627 - plugin to handle callbacks
628 - CPU model only for the moment
631 * Supernovae build mode is definitively removed. It was used to improve
632 inlining and inter-module optimizations. It is nowadays superseded by
633 link-time optimizations commonly available in compilers.
634 * Update ns-3 find lib. Bindings for ns-3 should work again now.
635 * Add boost dependency for surf++
636 * Add new macro for tests
637 - ADD_TESH(name <tesh_args>)
638 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
642 - xbt_bar_t XBT_barrier_init(unsigned int count)
643 - int XBT_barrier_wait(xbt_bar_t barrier)
644 - void XBT_barrier_destroy(xbt_bar_t barrier)
645 * Make the xbt_os_time module public
647 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
649 SimGrid (3.10) stable; urgency=low
651 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
654 * Reintegrate Java to the main archive as desynchronizing these
655 package is not acceptable anymore (Java is now considered stable)
656 * Add explicit synchronization facilities through semaphores
657 * Bug fix: Task.setDataSize() only changed the C world, not the value
658 cached in the Java world
661 * Dramatically change the way files are handled. API and internals changed, but
662 this part of MSG was not considered as production grade either.
663 * Add explicit synchronization facilities through semaphores
664 * Add a new function MSG_host_get_process_list()
665 * Preliminary DVFS support (see examples/msg/energy/ for details)
668 * SMPI is now included directly in the libsimgrid as the windows
669 linker doesn't force us on splitting it anymore.
670 * Improvements of the SMPI replay tool:
671 - Most of the collective communications are now rooted in the same process as
672 in the original application.
673 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
674 used until now). Multiple data types can now be used in a trace.
675 - The replay tool now supports traces produce either by TAU or a modified
677 - Bug Fix: the compute part of the reduce action is now taken into account.
678 - Gatherv collective is now supported
679 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
680 of smpirun outputs time independent traces for the current run. One file
681 is created per process. If too many processes are simulated, this behavior
682 can be changed to one file for all processes by using the
683 tracing/smpi/format/ti_one_file flag
684 * smpirun generates the host file if needed (with given host count and platform)
685 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
686 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
687 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
688 --cfg=smpi/coll_selector:(mpich/ompi)
689 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
691 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
692 happening inside a collective SMPI call.
693 * Fix the behavior of complex data types handling.
694 * Make MPI_Wtime another synchronization point to take computations into
696 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
697 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
698 * Add all missing Fortran bindings, SMPI should work with Fortran 90
699 (no privatization of global variables yet)
700 * Preliminary DVFS support (see examples/smpi/energy/ for details)
703 * Verification of liveness properties is now available for SMPI applications
704 (in addition to MSG applications)
705 * Bugged examples using SMPI in examples/smpi/mc/
706 * Add --cfg=model-check/visited option. Allows the verification of infinite
707 programs. Detection of loops in the execution thanks to the system state
708 comparison and reduction of the state space to explore. Can be combined with
709 DPOR for safety properties.
712 * Allow to change SimGrid configuration (see --help) within the code
713 thanks to SD_config() as it can be done in MSG.
714 * Add a new function SD_task_set_amount() upon user request.
717 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
718 * Remove rule based routing (no more PCRE dependency)
719 * Add a limiter_link option to cluster tag, to specify a maximum reachable
720 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
721 * Add a loopback_bw and loopback_lat options to cluster tag.
722 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
723 routing. Now peers are encapsulated in an AS and have their own private
724 router but this is transparent.
727 * Our own implementation of getline is renamed xbt_getline, and gets
728 used even if the OS provide a getline(). This should reduce the
729 configuration complexity by using the same code on all platforms.
730 * New type: xbt_cfg_elm_boolean.
731 * Allow to use yes/no for boolean configuration options in the command line.
732 * Allow to disable SimGrid cleanups at exit from command line option.
733 There are situations where one may want a simulation to end with an exit.
734 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
735 annoying when scripting around the simulator. Adding a
736 --cfg=clean_atexit:no allows to circumvent this issue.
739 * Lots of memory leaks were corrected in this release.
740 There is no known memory leaks anymore, in all of our 600+ tests.
741 * New command line option --version, to get SimGrid version information.
742 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
744 * Supernovae builds are deprecated, and expected to be removed in the next
747 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
749 SimGrid (3.9) stable; urgency=low
751 The Grasgory release: GRAS is really dead now.
753 * Complete overhaul of the internal host structures scheme.
756 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
757 was considered as experimental and badly maintained since then.
758 * Keeping it was thus a trap to our potential users, that could take
759 it instead of MSG or SMPI by mistake despite is pity state.
760 * GRAS seems to have very few users (if any), and no one volunteered
761 to maintain it further. It also induces a lot of XBT code (for
762 portability sake), that must be maintained too.
763 * For all these reasons, we killed GRAS. If someone wants to revive it
764 in the future, don't cry, our git history still remembers of GRAS.
767 * Major overhaul. Merge our documentation again as time proved that
768 splitting it was really not helping our users.
769 * Further improve the developer documentation to help newcomers
770 hacking on SimGrid itself. The user documentation (and in
771 particular, the beginner documentation) is still in a sorry state.
774 * Now works on Windows too!
775 * Much more extensive test suite, from MPICH
778 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
779 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
780 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
781 * Bug fix: let task be scheduled when the last dependency to be solved is
782 a control dependency.
783 * Remove SD_load_environment_script function.
784 Use the C sg_platf function if you want to declare a platform
788 * New function: MSG_process_get_number()
789 * Old function documented: MSG_config()
790 * Remove MSG_load_platform_script function
791 Use the C sg_platf function if you want to declare a platform
795 * Change the default value of the TCP_gamma constant (maximal size of TCP
796 congestion window) to a more realistic 4MiB value. If you notice changes in
797 your simulation results, you can fall back to the previous 20k tiny window
798 by adding --cfg=network/TCP_gamma:20000 on command line.
799 * (Hopefully) fix a bug wrt periodic availability and state traces
800 * Bug fix: use default values at start when first event in availability/state
801 trace is not at time 0.
804 * remove the "new_" part of function name sg_platf_new_trace_connect
805 (resulting in sg_platf_trace_connect), since it does not create
809 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
810 use case where it's really mandatory, and maintaining it was a pain in
812 * New: xbt_fifo_search(), search an item with a user-provided
813 comparison function instead of dumb pointer comparison.
816 * Fix the lua deployment:
817 Use `simgrid.init_application()` before deployment instead of
818 `simgrid.msg_register_application()` after.
821 * Transfer the tracing files into the corresponding modules.
823 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
825 SimGrid (3.8.1) stable; urgency=low
827 The "we are told that some people want to also *install* the simgrid
830 * Add missing file "tesh.1" to the archive.
832 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
834 SimGrid (3.8) stable; urgency=low
836 The Psssshiiiit release: SimGrid jumps into the Cloud.
839 * Add an experimental interface to manipulate VMs. They are mainly
840 process groups with very few intrinsic semantic, but they should
841 allow you to build the semantic you want easily.
842 * New function: MSG_host_set_property_value()
843 * New function: MSG_process_on_exit(). To clean memory in all cases.
844 * Bug fixes that made the host (and link) failures unusable.
845 * Add a way to auto-restart process when the host in which they are
846 executing comes back (ON_FAILURE="RESTART" on deployment file,
847 MSG_process_auto_restart_set).
848 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
849 please stop using the old ones, they are DEPRECATED.
851 * Deprecate functions MSG_global_init() / MSG_global_init_args()
852 Please use MSG_init() instead. (reducing the amount of entry
853 points in the library helps us).
854 * Make it impossible to link against the wrong version of the lib
855 * Deprecate MSG_clean(). No need to call it anymore.
856 * Function MSG_get_host_number() is not deprecated anymore.
859 * Split the doc into a user guide and a reference guide.
860 * Start a developper guide to help people hacking on SimGrid.
863 * Enable tracing by default. This modules rocks you should use it.
864 * Remove option custom_flags. Now use environment variables CFLAGS
866 * Use default cmake things to detect lua instead of home grown ones.
867 * New option "enable_mallocators" to disable mallocators, for debugging
868 purpose ("on" by default).
871 * Bug fixes around the resource failures: don't let the processes
872 survive the host they are running onto.
873 * Add an interface to auto-restart processes when the host in which they are
874 executing comes back.
875 * Ensures that SIMIX_clean is called automatically. It's not part of
876 the public interface anymore (bindings should be updated).
879 * Bug fix for when SD_Simulate is called with a positive value: be careful
880 when comparing doubles. Sometimes they are different for non significant
882 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
883 parallel task whose initial work is distributed among host according
884 to the Amdahl's law. Such tasks are created with a parameter alpha
885 that corresponds to the non-parallelizable part of the computation.
886 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
887 between two sets of workstations assuming a 1D block distribution (each
888 workstation owns a similar share of data) on both sides.
890 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
891 Data redistribution will be automatically scheduled once parent and child
892 are both scheduled. The filling of computation_amount and
893 communication_amount structures is now done seamlessly thanks to the chosen
895 * New function SD_workstation_dump to display various information
896 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
897 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
898 bandwidth on the route onto which the task is scheduled and the amount of
900 To divide the nominal bandwidth by 2, the rate then has to be :
901 rate = bandwidth/(2*amount)
902 * Compute tasks that have failed can now be rescheduled and executed again
903 (from their beginning)
904 * Increasing source code coverage (src/simdag is now covered at 95.8%
908 * Re-implement time-independent trace replay using SMPI (at the
909 smpi_smp_* level) instead of MSG. This should replace
910 examples/msg/actions/actions.c
911 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
912 hindexed and structs)
913 * Implement the exchange of non-contiguous data.
914 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
915 * Correct behavior of smpi/sender_gap and set its default value to 0
916 * Add option to asynchronously send small messages to allow better
917 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
918 specifies the size in bytes under which messages will be asynchronously sent.
919 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
920 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
923 * New C interface to define a platform: XML is now optional.
924 For more info, please check include/simgrid/platf.h.
925 * New interface to define random platforms from the C:
926 For more info, please check include/simgrid/platf_generator.h and
927 examples/msg/masterslave/masterslave_platfgen.c
928 * Export a sg_cmdline dynar containing all the arguments we got from
932 * Two new tracing options for adding comments to trace file so you
933 can track your experiments (see --help-tracing for details).
934 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
935 * Adding the SimGrid version that generated the trace file as a comment.
936 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
937 * Fix to avoid key clashes on Paje links
938 * Other minor fixes related to the Paje specification
941 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
942 and renamed to xbt_str_hash() and xbt_str_hash_ext().
943 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
944 * Greatly improve the robustness of mmalloc to user errors (such as
945 using an area after freeing it, or freeing it twice)
947 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
949 SimGrid-java (3.8.1) stable; urgency=low
951 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
952 Warning: all methods are visible, but only some of them are
953 implemented so far. Check the source (src/jtrace.c)
954 for further information.
955 * New module: org.simgrid.msg.File (SimGrid File management functions)
956 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
957 * Change the meaning of Process.restart: now restart the process from
958 the begining, like MSG_process_restart in C.
959 * Add Process.setAutoRestart: handling of process restart when failed
961 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
962 you to retrieve the properties of the processes/hosts
963 * Deprecate Msg.clean(): you can just forget about it now.
964 * New function Process.getCount(), that only works when compiling
965 with the not yet released version 3.9 of the C library.
967 * New context factory based on Coroutines. It mandates a modified JVM
968 but then, the simulations run about five times faster, and there is
969 no limit to the amount of processes (beside of the available memory).
971 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
973 SimGrid (3.7.1) stable; urgency=low
976 * Restore the prototype of MSG_process_create_with_environment() to
977 the pre-3.7 situation by removing the kill_time argument.
978 * Add a MSG_process_set_kill_time() function instead.
981 * Fix weird behaviors when dealing with parallel tasks.
984 * Simgrid is now built as a dll.
985 * Simgrid-java now works on Windows.
986 * Simgrid-Java is now included into Windows package.
989 * First pre-build package for MacOSX.
992 * Fix compilation when using MSG_USE_DEPRECATED.
993 * Fix some compilation issues on Macs and Windows.
994 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
995 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
997 We discovered that the Lua console is broken, but we are missing the
998 manpower to fix it right now. The problem existed in 3.7 too, so we
999 are not blocking the release for that. Sorry if you depended on this
1000 feature, any help would be really welcome.
1002 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1005 SimGrid-java (3.7.1) stable; urgency=low
1007 The "Java aint got to be bloated and slow" release
1010 * Various internal cleanups and performance improvement
1011 Simulations are expected to run up to twice faster or so
1012 * Make Process.kill(process) an instance method, not a static one
1013 * User processes are not java.lang.Thread subclasses.
1014 This breaks the compatibility (sorry), but previous API was
1015 brain-dead, making it impossible to have non-trivial
1016 initializations in the process constructor.
1017 * Require a full constructor per Process sub-class.
1018 Kinda breaks the compatibility (sorry), but this allows a much more
1019 efficient way to launch the processes at simulation startup.
1020 * Do not embeed our version of semaphores, java 1.5 can be considered
1021 as sufficiently prevalent for us to not dupplicate its features.
1025 * Add examples for almost every part of the API
1026 We spotted and fixed a lot of bugs in the process
1028 * New module: asynchronous communication API
1029 * New function: Process.sleep()
1030 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1031 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1032 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1034 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1036 SimGrid (3.7) stable; urgency=low
1038 The "spring cleanups (before the next Big Project kicks in)" release.
1041 * We can specify the SMPI latency/bandwidth factor with command line
1042 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1043 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1044 You can also use the "config tag" from platform file by adding this line
1045 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1046 (see "example/platforms/tag_config.xml" to use "config tag").
1047 Note that the command line supersedes the platform file configuration.
1048 * Change the correction factors used in LMM model, according to
1049 the latest experiments described in INRIA RR-7821.
1050 Accuracy should be improved this way.
1051 * Use the partial invalidation optimization by default for the
1052 network too. Should produce the exact same results, only faster.
1053 * Major cleanup in surf to merge models and split some optimization
1054 mechanisms from the core of the models. As a result you can now
1055 specify which model to use (e.g., --cfg=network/model:LV08
1056 --cfg=cpu/model:Cas01) and which optimization mode to use
1057 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1058 Incompatible combinations should err at initialization. See
1059 --help-models for the list of all models and optimization modes.
1060 * The CLM03 workstation model was dropped for simplicity because it
1061 used the deprecated CM02 network model. Use default instead.
1062 * Rename the TCP_gamma configuration option to network/TCP_gamma
1063 * Rename the coordinates configuration option to
1064 network/coordinates, and document it
1065 * Use now crosstraffic keyword instead of the terribly misleading
1066 fullduplex keyword. It is activated by default now in the current
1067 default model, use --cfg=network/crosstraffic:0 to turn it off.
1068 * Ongoing refactoring the model parsing to make XML files optional
1069 See include/simgrid/platf.h for details (still to be completed)
1072 * Major overhaul of the documentation. Almost instructive now :/
1073 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1074 functions and friends. This interface was considered as
1075 deprecated since over 2 years, it's time to inform our users that it is.
1076 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1077 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1078 use these (crappy) functions in your code.
1079 These functions will be removed soon. Stop using them now.
1080 * Deprecate MSG_get_host_{table,number}
1081 Implement MSG_hosts_as_dynar() instead.
1082 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1083 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1086 * Stabilize the parallel execution mode of user contexts
1087 * Introduce configuration variables to control parallel execution:
1088 - contexts/synchro: Synchronization mode to use when running
1089 contexts in parallel (either futex, posix or busy_wait)
1090 - contexts/parallel_threshold: Minimal number of user contexts
1091 that must be part of a scheduling round to switch to parallel
1092 execution mode (raw contexts only)
1093 * Fix bugs that prevented to use suspend/resume along with
1094 synchronization structures.
1095 * Fix bugs in process termination that lead to invalid memory access
1096 in very specific conditions.
1099 * Introduce a parallel mode for the models (controlled by surf/nthreads
1100 configuration item). In our tests, running the models in parallel
1101 never lead to any speedups because they are so fast that the gain
1102 of computing each model in parallel does not amortizes the
1103 synchronization costs, even when ultra fast futexes are used.
1104 This is released anyway because YMMV.
1107 * Performance boost by using a swag internally to compute the set of
1108 tasks that are finished and should constitute the return value of
1112 * Enable it by default now that it is considered rather stable.
1115 * Documentation of the tracing functions.
1116 * Performance gains when tracing categorized/uncategorized resource
1117 utilization by avoiding calls to get route when updating resource
1118 variables. LMM constraints are being used instead.
1119 * API changed to set task categories. Use MSG_task_set_category instead
1120 of TRACE_msg_set_task_category, and SD_task_set_category instead
1121 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1122 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1123 * Examples for link user variables added (see at examples/msg/tracing/)
1124 * Deprecated function TRACE_msg_set_process_category completely removed
1125 * Trace header updated according to the latest Paje file format
1126 * Tracing network lazy updates, no longer obligate users to use full updates
1127 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1128 * Experimental: let user code declare/set/push/pop application states for hosts
1129 * API changed to allow the manual creation of graph configuration files
1130 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1133 * Improve the API of Lua MSG bindings, using the Lua spirit.
1134 * Each simulated process now lives in its own Lua world (globals are
1135 automatically duplicated). It helps writing simulators. Will allow
1136 to run Splay programs within SimGrid in the future.
1137 * Add a Chord example in Lua, equivalent to the MSG one.
1140 * Start the implementation of a solution to express temporal
1141 properties, not only local assertions. This is still an
1142 experimental work in progress, stay clear from it to be safe.
1146 - Add new runtime parameters --help-logs and --help-log-categories
1147 to display information about supported logging parameters and
1149 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1150 don't exists anymore.
1151 * Mallocators: allow value NULL for the reset function.
1153 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1154 create homogeneous dictionaries, where all the elements share the
1155 same free function. Non homogeneous dictionaries will be
1156 deprecated in the next release.
1157 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1158 - Multi-level dictionaries are deprecated.
1160 - new function xbt_dynar_search_or_negative() that is useful when
1161 you have less than 2 million elements in your dynar and don't
1162 want of the extra complexity of catching exceptions when the
1163 element is not found.
1165 - Make xbt_os_thread module (for thread portability) public.
1166 Documentation is still to come, sorry.
1168 - Cleanups and simplifications to make it maintainable again.
1169 - Exotic features (such as memalign and valloc) were removed.
1170 - The metadata were extended and improved so that the
1171 model-checker becomes able to explore and inspect the heaps.
1172 - This may induce a performance drop when enable_model-checking is
1173 ON in cmake (even if it's not used in the simulation), but it is
1174 necessary at this point to get MC working.
1176 Turn model-checking OFF if simulation performance matters to you.
1177 Not enabling it at runtime is not enough, disable it in cmake.
1179 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1182 SimGrid (3.6.2) stable; urgency=low
1184 The "Not coding new stuff allows to polish old things" release.
1187 * New bindings to the NS3 packet level simulator (experimental)
1188 * Use the raw (efficient) execution contexts instead of the sysv
1189 (portable) ones when possible.
1190 * libpcre is now mandatory in any cases since not using it led to
1191 severe performance loss and possibly other issues
1192 * Update the XML platforms:
1193 - G5K: include the latest machine in Nancy
1194 - GridPP and LCG: new platforms
1195 * Documentation was partially updated, at least (more to come)
1197 Bug fixes, cosmetics and small improvements
1198 * Free terminated processes before the end of the simulation to avoid
1199 exhausting the memory of users having very dynamic amount of
1201 * Bug fix and cosmetics about canceling non-running tasks
1202 * Bug fix about the dot loader's issues when using libcgraph
1205 * Create an installer for windows with nsis (amd64 and win32)
1206 - Add an hello world project to illustrate simgrid project creation.
1207 - Embed libpcre into the Simgrid installer to avoid
1208 its compilation burden
1209 * The raw execution contexts should work on Apple now
1210 * Port to Windows 64 bits
1211 - Sysv contexts now have an implementation for this arch
1212 - GRAS communication features now support this arch
1213 * Drop support for borland compiler on windows
1214 - this code was not maintained, and we kinda depend on gcc nowadays
1215 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1216 * Fix portability issue on unstable ubuntu: linker became picky on
1219 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1221 SimGrid (3.6.1) stable; urgency=low
1223 The "Oops, we broke Macs too" release
1226 * Fixed contexts detection so that raw ones are used when possible
1227 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1228 leads to a strange error, with user code segfaulting sometimes when
1229 the generated code is not perfectly aligned (which is not
1230 controllable from the user side, depends on the amount of code)
1233 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1235 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1237 SimGrid-java (3.6) unstable; urgency=low
1240 * Split of every thing from simgrid v3.5 into a separate package.
1242 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1244 SimGrid (3.6) stable; urgency=medium
1246 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1249 * Bindings now constitute their own package, separated from the main one.
1250 Rationale: reduce our maintenance nightmare by reducing the module coupling
1251 They will soon be released on their own on gforge.
1253 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1254 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1256 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1257 * It's not quite deprecated for now because we have no replacement,
1258 but it may soon become the case.
1261 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1262 * Fortran: New user-level cache variable to store the rank of the running
1263 process. This improves performance by an order of magnitude.
1264 * C: New coccinelle script to automatically locate and modify global and
1265 local static variables.
1266 * Improved SMPI network model with a sender-side gap to account for multiple
1270 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1271 only say if a communication is finished, no matter whether it succeeded or
1272 failed. You can call MSG_comm_get_status() to know the status of a finished
1274 * New function MSG_task_dsend() to send a task and detach it. When a
1275 communication is detached, you are never notified of its success or failure
1276 and the memory is released automatically once it is finished. This function
1277 is useful when you don't care about the end nor the success of a
1279 * Change the prototypes of action replay. Sorry for inconvenience,
1280 but this is really more efficient this way (and to adapt your code,
1281 you just have to fix the initialization, that shouldn't be too long)
1282 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1283 ever managed to do anything useful with it.
1284 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1285 takes a dynar of communications. It returns immediately and gives the
1286 index of a finished communication (if any).
1287 * New example: a basic implementation of the Chord P2P algorithm.
1290 * New model for multi-core CPUs. You can now use the core attribute to
1291 precise the number of cores of a host. This is a basic model. Every
1292 process running on the host receives at most the power provided in
1293 the DTD (throughput<=power). Total throughput of process cannot exceed
1295 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1296 router linked by an up-link and a down-link (possibly asymmetrical).
1297 This kind of pattern allows to easily build last-mile model style platforms.
1298 Aggregating such patterns in a rule-based AS is thus the technique of
1299 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1300 * New model for Vivaldi routing. We transformed the Vivaldi network model
1301 into a Vivaldi routing model (based on the rule-based model). This allows to
1302 combine Vivaldi based latencies with last-mile platforms.
1305 * Added a check for NaN of IEEE754 infinite in the double entries of
1307 * Introduce a new context factory "raw", highly inspirated from the
1308 ucontext factory, but using manually crafted functions in assembly to
1309 do the work in an efficient manner.
1310 * Allow to change the used context factory at run time, not only at
1311 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1312 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1313 contexts at runtime (only with raw contexts or ucontexts).
1314 * Completely rewrote this module to allow parallel execution of user
1315 processes. Use --cfg=contexts/nthreads:N to execute user processes
1316 with N parallel threads (the default is 1, meaning no parallelism).
1317 * Allow to decide dynamically between sequential and parallel modes.
1318 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1319 processes in parallel only when their number is greater than or equal to P
1321 * Added a check for NaN of IEEE754 infinite in the double entries of
1325 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1326 won't output the state of processes when interrupted with Ctrl-C
1327 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1328 NULL-terminated array. This may solve backward compatibility issues
1329 due to the change to return type of SD_simulate. See also:
1330 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1331 * Add new macros with variable number of arguments.
1332 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1333 - in xbt/asserts.h: xbt_assert
1334 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1335 - in xbt/ex.h: THROWF and RETHROWF.
1336 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1338 * Change xbt_die() to accept a format string with arguments, just like printf.
1339 * New data structure: xbt_lib_t, like a dict but more general and with better
1343 * New configuration options
1344 Options triva/categorized and triva/uncategorized can be used to generate
1345 graph configuration files for Triva visualization tool.
1346 * Configuration option tracing/platform is renamed to tracing/categorized
1347 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1348 * New TRACE_user_link_variable interface:
1349 User provides the name of the link and the tracing variable to attach to it
1350 * the declaration of tracing categories must be done after the environment
1352 * simpler tracing interface, just one way to declare categories
1353 TRACE_category or TRACE_category_with_color, it is up to you
1354 * links in the trace file are again identified by their names
1355 * trace contains the full platform hierarchy exactly as declared using the ASes
1356 * Options tracing/msg/[task|process]:1 groups the process by hosts
1357 for both cases, tasks and processes must have names that are unique during the simulation
1358 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1359 * The experimental option tracing/msg/volume is deprecated
1360 its functionality may be reincorporated if needed
1362 The tracing generates a trace file with unordered timestamped events,
1363 because of the way the core simulator (surf) works. A script available
1364 at the tools directory (fix-paje-trace.sh) can be used to put the events
1365 in order. We have changed the tracing so it can generate ordered timestamped
1366 events in the final trace, but depending on the simulator (and how much time
1367 is simulated) that can lead to a huge memory utilization. It is deactivated
1368 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1370 Build Infrastructure
1371 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1372 - We are trying hard to keep the API stable, but it may happen that
1373 some things change (we're a research project after all, not a
1374 nuclear plant operating system). If such things should happen, you
1375 could rely on that macro to adapt.
1376 - current value: 30600 for 3.06.00, aka 3.6
1377 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1378 a number that can be compared with SIMGRID_VERSION.
1379 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1380 assertions and verbose/debug logging events are disabled at compile time.
1382 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1384 SimGrid (3.5) stable; urgency=medium
1387 * New feature to any SimGrid-based simulator: Model-Checking
1388 Check SIN#1 for more details.
1391 * New Model SMPI (three-interval linear regression for correction factors)
1392 See RR-7426, available at http://hal.inria.fr/inria-00527150
1393 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1394 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1395 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1396 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1397 * See also src/smpi/README
1402 - Tracing API changes: TRACE_start and TRACE_end should not be called
1403 by user-code. They are automatically called by simulators created
1404 with SimDAG, MSG and SMPI if the toolkit is compiled with
1405 tracing_enabled=ON. Categories declaration and utilization remain the
1406 same for MSG and SimDag.
1407 - A function was added to the tracing API to declare categories with
1409 - TRACE_category_with_color (char *category, char *color)
1410 where color must be in the following format
1411 "%f %f %f", red, green, blue
1412 and red, green, blue are float values in the interval [0, 1]
1413 - User can specify NULL as color parameter, or continue calling
1414 TRACE_category (cat)
1415 On that case, the tracing system will define random colors
1416 - The following command-line options are supported:
1417 --cfg=tracing/filename:msg.trace
1418 --cfg=tracing:1 (activate tracing, needed to use others)
1419 --cfg=tracing/platform:1 (categorized resource use)
1420 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1421 --cfg=tracing/msg/task:1 (task creation)
1422 --cfg=tracing/msg/process:1 (process creation, migration)
1423 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1424 --cfg=tracing/smpi:1 (SMPI interface tracing)
1425 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1426 - examples of examples/msg/tracing updated
1428 - DAXLoader and DOTLoader functions can generate tasks with categories
1429 - A new function to attribute a category to SD tasks:
1430 TRACE_sd_set_task_category (SD_task_t task, char *category)
1431 * Tracing the MPI interface implemented by SMPI
1432 - Collective operations are traced with states
1433 - Point-to-Point operations are traced with states/links
1434 - Tracing activated by a parameter "-trace filename" passed
1435 to smpirun during execution (considering that simgrid
1436 is compiled with tracing enabled)
1437 - To run the simulation with gdb, the simulator
1438 accepts --cfg=tracing/smpi:1 to trace SMPI
1439 - tesh files to check if smpi tracing is ok
1440 - See examples/smpi/NAS/DT-trace
1441 * GTNetS tracing re-worked
1442 - adaptation to the tracing system of GTNets to cope
1443 with modifications regarding the fullduplex mode
1444 - new tesh files to check if gtnets tracing is ok
1447 * Asynchronous communications through the functions:
1448 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1449 * New function: MSG_load_platform_script()
1450 to make possible using a lua script instead of XML files to set up platforms
1451 * New function: MSG_set_function
1452 to associate functions to processes, used when bypassing the parser
1453 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1455 Platforms: Add some more examples in examples/platforms
1456 * Grid'5000: see www.grid5000.fr
1457 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1460 * Change the XML format. This is a very important modification. SimGrid
1461 3.5 introduces a new hierarchical format based on the notion of
1462 Autonomous Systems. Compatibility with old format is ensured
1463 through the perl script provided in the install bin directory
1464 bin/simgrid_update_xml.
1465 It is now possible to build platforms with specific routing
1466 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1467 connect several platforms together. We will try to provide soon
1468 set of realistic platforms exploiting these properties (have a
1469 look at examples/platforms/ for the moment).
1470 * Take the opportunity of the XML format change to be a good XML citizen:
1471 rename link:ctn to link_ctn and similar changes (also dealt with by
1473 * Add a new routing scheme (rule-based) using regular expressions. It
1474 enables to have an extremely low memory footprint when the
1475 underlying routing is simple and can be compactly described. You
1476 need to have libpcre4-dev (perl regular expressions) installed if
1477 you want to use this routing scheme.
1478 * Revive the cluster TAG and allow to easily and efficiently (both in
1479 term of memory and speed) connect clusters together. Have a look
1480 at teshsuite/simdag/platforms/ to see how this can be done. With
1481 this tag, you can create clusters with thousands of tasks at no
1482 cost (have a look at examples/platforms/).
1483 Note: clusters are implemented as ASes, so there is no need for an
1484 enclosing AS tag if you have only one cluster in your platform.
1485 * Add new generic functions in the public interface that allows the user
1486 to call SURF 'create_resource' methods from your code (same
1487 functionality as the XML bypass mechanism but with a much lighter
1489 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1490 that uses a piecewise linear approximation to produce better
1491 results when exchanging small messages.
1492 * Add a new parameter to handle correctly full duplex link and account
1493 for interferences between uplink and downlink communications
1494 (activate with --cfg=fullduplex:1).
1497 * Rename the SD_READY (all dependencies are satisfied and task is
1498 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1499 dependencies are satisfied) state.
1500 This prevents a confusion between the notion of "ready to schedule"
1501 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1502 simulated" (SD_RUNNABLE) used by the simulation kernel.
1503 * Change the way a task is considered as ready. Instead of removing
1504 dependencies when a task is done, a counter is decreased. This way,
1505 it is always possible to reach ancestors thanks to the
1506 SD_taks_get_parents function (even after the end of the simulation.)
1507 * Change the return type of SD_Simulate from (SD_task_t*) into
1508 xbt_dynar_t. This function was in handling a dynar internally and
1509 converted it into a NULL terminated array for historical reasons.
1510 * New function SD_dotload(char*) to load a DAG described in dot
1511 format. This loader and the corresponding examples require the
1512 installation of the graphviz library.
1513 * Fix a bug in the management of tasks of size 0 in the surf network
1514 models. This problem was only visible with SIMDAG and you should
1515 thus disregard results produced with earlier versions if you
1516 relied on this feature (some tasks were blocked because of this).
1517 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1518 even though your applications were doing only point-to-point
1519 communications and sequential computations. Now you can really use any
1520 model you want (of course, if you create real parallel tasks, which are
1521 not implemented in most models beside ptaskL07, this will abort).
1522 * Add an example that schedules a DAX on an heterogeneous platform
1523 using a Min-Min strategy.
1524 * New function SD_workstation_get_current_task() that returns the kind
1525 of task currently running on a workstation in the sequential access
1527 * Raise some warnings when unexecuted tasks remains at the end of the
1528 simulation. This is usually caused by cycles in the DAG.
1531 * New function: SIMIX_process_set_function() called by MSG_set_function
1532 * Change the underlying waiting queue in semaphores so that a process
1533 can wait on several of them simultaneously (as in waitany).
1534 * Fix the way to handle tokens in semaphores so that all access patterns
1535 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1536 * kill the dirty pimple SIMIX_message_sizes_output()
1537 Please use (proper) visualization instead
1540 * New data container: setset (set of sets of elements)
1541 * New module: mmalloc (mapped malloc, allowing to have several
1542 independent segments of malloc)
1543 * New function: xbt_dict_cursor_set_data()
1544 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1545 * New function: xbt_dynar_is_empty()
1546 * New function: xbt_fifo_get_last_item()
1547 * Fix xbt_dynar_shrink(): use the right element size.
1548 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1549 uninitialized areas during expand.
1550 * Fix semaphores: previous implementation was severely broken.
1551 * Use library init/fini functions for our initialization.
1552 - you can use logs and other feature as soon as you want in your
1553 code (even before the xbt_init / MSG_init)
1554 - xbt_exit is now a no-op and produce a warning when used.
1557 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1558 benefit from the latest improvement to the simulation kernel.
1559 * Kill measurement sockets for now. If you rely on them, sorry. This
1560 release is not for you. This feature will be reintroduced in the
1561 future, but we cannot delay the release any further.
1562 * New function: gras_msgtype_get_name().
1563 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1564 * Fix (at last) the pmm example: it should not randomly fail anymore.
1566 Build chain: bug fixes and overall polishing
1567 * Cmake is now stable enough. Hence, we killed the autotools.
1568 * Port to windows ( TM :)
1569 * Fix the 'make install' target.
1570 No need to use 'make install-simgrid' anymore
1571 * Introduce a 'make dist' target compiling a *source* archive
1572 'make package' compiles a binary archive
1573 * Compile java files only on need
1574 * Add --cd and --setenv command line options to tesh
1575 * Out of source builds are not fully supported yet, but we are close
1576 * Enable supernovae and optimization flags by default for our users
1579 * Add layer to set up environment directly from lua, without XML.
1580 * The effect of gras_stub_generator can be achieved through
1581 lua too (check examples/gras/console/ping_generator.lua)
1583 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1585 SimGrid (3.4.1) stable; urgency=low
1587 The "Polishing easter eggs is probably a good idea" release.
1588 This is a bug fixes release only.
1592 * Fix a bug preventing the tasks from begin garbage collected.
1595 * Fix a bug occuring when a host involved in a communication fails.
1596 This was not detected properly by the other peer involved in the
1597 communication. Now, it's reported as a network error.
1600 * Warn the user about loop dependencies in data flow of DAX files
1601 * Obey the control-flow dependencies of DAX files
1604 * Add option "enable_smpi" allowing to not compile SMPI.
1605 Probably useful for the (Mac) users experiencing a build error here
1606 * Improve the detection of lua5.1 and ruby1.8
1608 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1610 SimGrid (3.4) stable; urgency=low
1612 The "Easter in Cargese" release. Also known as (major changes):
1614 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1615 ~> bindings were greatly improved
1616 ~> new tracing infrastructure for better visualization introduced
1618 * the "Welcome to configury modernity" release.
1619 ~> we switched from autotools to cmake, and improved our cdash
1622 A more detailled list of changes follow (full detail in svn log).
1624 Java Bindings: Various Cleanups
1625 * (install java-gcj-compat-dev on debian-like to use them)
1626 * Remove put/get: no need to export deprecated interface in Java
1627 Use send/receive instead.
1628 * Cleanup the examples and add a README per directory
1629 * Remove example autoDestination (that's the only way to go now)
1630 * Remove example explicitDestination (was a plain copy of basic)
1631 * Make JniException a runtime exception, so that there is no need to
1632 declare the fact that you may encounter such a beast. I guess that
1633 nobody will ever want to survive such error.
1634 * Create specific errors for each MSG case of failure:
1635 host failure, transfer failure, timeout, task cancelled
1636 * Cleanup the exceptions that may get thrown by each function
1637 * Other internal cleanups in Java bindings. Performance still bad :/
1638 Ruby and Lua Bindings: create them
1639 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1640 * That's new and great, you should try them out.
1641 Same functionalities than Java bindings, only even less polished
1643 * Kill the useless "rate" argument of SD_task_get_execution_time()
1644 Everyone used to provide -1 as a value, it was not used, and the
1645 semantic of a possible use wasn't even clear.
1646 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1647 either as comm costs or compute costs to mean that there is no
1648 such thing for that specific task.
1649 * Add a SD_task_set_name() function
1650 * Fix SD_task_unschedule() on typed tasks
1651 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1652 * In DAX loader, accept useless 'level' attributes to <job> since
1653 LIGO DAGs have them (seem to be to ease graphical representation).
1655 * Add an example masterslave_mailbox.c using send/receive and not
1656 the deprecated put/get interface.
1657 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1658 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1660 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1661 (declare MSG_USE_DEPRECATED to still have the old name)
1662 * Add a MSG_task_set_data() function
1663 * About trace replay (see examples/msg/actions):
1665 - Allow to work with splitted trace files for each process
1666 Give the specific trace file as argument of each process,
1667 and call MSG_action_trace_run(NULL)
1668 You can still have one merged file for all processes.
1669 - Fix implementation of collective operations
1670 * Allow task_execute() on 0-sized tasks (closes #10063)
1672 * This is the first release of SimGrid where SMPI is not considered
1673 beta anymore (even if some corners should still be improved)
1674 * Port over the new SIMIX_network submodule (internal refactoring)
1675 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1676 * Implement more missing elements of the standard:
1678 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1680 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1681 MPI_Scan() MPI_Get_processor_name()
1682 - Added implementation of missing case for Alltoall (warning: it's
1683 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1684 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1685 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1686 SMPI_MPI_Allgatherv()
1687 * Bug fixes include:
1688 - MPI_Waitsome() was broken
1689 - Allow relative includes in smpicc
1690 - Command line cfg argument 'reference_speed' was ignored...
1691 - Some functions did not properly lead to auto-benching of user code
1692 - smpicc passes -O2 by default (just like openmpi one)
1694 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1695 * Bug fixes about timeouts during communications
1696 * add SIMIX_message_sizes_output() as a pimple to write to file the
1697 amount of messages per size. Use gnuplot to get histogram.
1698 Pimple because that's the only user-visible function of simix,
1699 defined directly in xbt.h (irk, sorry)
1701 - Add a SIMIX_sem_get_capacity() function
1702 - Fix interactions with processe resume/suspende
1703 - release_forever() was stupidly broken
1704 - Fix SIMIX_display_process_status() for processes in a semaphore
1705 - Make SIMIX_sem_block_onto() user-visible
1706 * Refactoring context stuff:
1707 - Use pseudo-OOP for better modularity
1708 - reimplement SIMIX_process_kill() without process_schedule() so
1709 that the latter can take as invariant that it is called from
1711 - Merge context_start into context_new for sake of simplicity
1713 * Add a Vivaldi network model, coded live during SUD'10 ;)
1714 * Rename configuration variables to start a hierarchy:
1715 o cpu_model -> cpu/model
1716 o network_model -> network/model
1717 o workstation_model -> workstation/model
1718 * New configuration variables:
1719 o network/bandwidth_factor: correction to bandwith
1720 o network/latency_factor: correction to latency
1721 o netwotk/weight_S: correction to the weight of competing streams
1722 * Add a long description to the models, that users can see with such
1723 argument on the command line: --cfg=cpu/model:help
1724 * --help-models display the long description of all known models
1726 * config: add the ability to set a default value after registration
1727 Does not override any previously set value (e.g. from cmd line)
1728 * dict: allow to have integer key and data.
1729 When so, you need to use the following functions
1730 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1731 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1732 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1733 In contrary to regular dicts, the key is not malloced before copy.
1734 Mixing scalar and regular elements in the same dict is not tested
1736 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1737 Tracing for Visualization:
1738 * SimGrid is now instrumented in order to generate a trace file for
1739 visualization analysis: to use it, need to compile SimGrid with the
1740 "tracing" option enabled, and instrument the program using SimGrid with
1741 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1742 (among other functions).
1743 * The instrumentation only traces the platform utilization for now
1744 * Documentation to use the tracing functions and how to analyze the
1745 traces with the Triva tool is written.
1746 * More information about: SimGrid FAQ (in the section Tracing Simulations
1749 * We moved to cmake as default build system. Autotools support will
1750 be dropped soon. Check the FAQ for more info about how to use it.
1751 * Greatly improved our cdash/ctest interactions
1752 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1753 * Added memory checking tests with valgrind; lot of memleak fixing.
1754 This may be the first release of simgrid with so few memory issues
1755 * Added code coverage tests.
1756 Our coverage is still improvable, but at least we see it on cdash.
1758 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1760 SimGrid (3.3.4) stable; urgency=low
1762 The "Desktop Grid needs love too" release (also called Xmas release).
1764 Models improvements:
1765 * Major speedup in the maxmin system solving by using lazy evaluation
1766 Instead of solving completely the maxmin system at each iteration,
1767 only invalidate (and recompute) the modified parts.
1768 This new feature is enabled in default models but you can try to
1769 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1770 * Cas01 IMproved as default CPU model
1771 This CPU model is the same Cas01 model, but it uses the
1772 maxmin-selective-update flag and a heap structure to manage
1773 actions on SURF kernel.
1774 It reduces the complexity to find the next action to finish and,
1775 consequently, it's faster than the old Cas01.
1776 This is the new default CPU model (Cas01).
1777 * Rename the old Cas01 model to Cas01_fullupdate
1778 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1779 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1780 * CpuTI (CPU Trace Integration)
1781 A new CPU model whose objective is simulate faster when using
1782 availability trace files.
1783 Instead of using a full featured, over engineered maxmin system for
1784 CPU modeling, this model does the pre-integration of traces files
1785 to calculate the amount of CPU power available, and so, executes
1786 faster than the old CPU models.
1787 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1788 * Use LV08 as default network model since it gives better accuracy
1789 for small messages and shouldn't change things for big ones.
1790 Use --cfg=network_model:CM02 to get the previous behavior.
1793 ******************************************
1794 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1795 ******************************************
1796 * The new CPU model may changes simulations!
1797 The point is that events occurring at the exact same timestamp
1798 are not scheduled in the same order with the old and new
1799 version. This may be enough to completely change the execution
1800 of simulations in some cases.
1801 * The new network model will change simulations!
1802 This new model is more realistic than the previous one, so you
1803 should consider redoing your old experiments with this model.
1804 Sorry for the inconvenience.
1807 * Introduce the supernovae compilation mode
1808 When compiled that way, the whole SimGrid (or almost) is put in a
1809 single compilation unit and compiled in one shoot.
1810 This is to help gcc which has difficulties to inline stuff from one
1812 The speedup seem to be above 15%, althrough more tests are needed on
1813 amd64 to confirm that gain.
1816 * Port of MSG's mailbox on top of SIMIX network
1817 The put/get mechanism was greatly simplified on the way.
1820 * New SIMIX network module. Provides:
1821 - Mailbox: rendez-vous mechanism to find with who you want to speak
1822 - Synchronous send/recv: easier and hopefully faster since the
1823 logic is handled in the maestro process directly now
1824 - Asynchronous send/recv: you dreamt of it? It's here now
1825 Too bad that nobody cared enough to propagate the change to MSG.
1826 * Add semaphores as SIMIX synchronization mechanism.
1829 * new function SD_daxload(char*) to load a DAX file
1830 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1831 * Introduce typed tasks. Specify its kind and cost at creation.
1832 At scheduling, just give where it should be placed, and the cost
1833 for each involved resource is automatically computed.
1834 Existing constructors so far (more to come of course):
1835 - SD_task_create_comm_e2e() for end-to-end communication
1836 - SD_task_create_comp_seq() for sequential computation
1837 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1838 * new function SD_task_dump() for debuging display
1839 * new function SD_task_dotty(task,FILE*) writing to file the info
1840 about the task in dotty format
1841 * SD_task_dependency_exists() can now cope with having one of its
1842 arguments NULL. If so, it tests whether the other argument has any
1844 * Add getters on list of preceding/following tasks:
1845 SD_task_get_parents(task) and SD_task_get_children(task)
1846 * Add getters on amount of workstations and list:
1847 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1848 * Add getter on task kind: SD_task_get_kind(task)
1849 * Update the start_time and finish_time of tasks on completion/failure
1850 * Bugfix: Remove task from state swags when destroyed
1853 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1856 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1857 dynar with one separate thread per value of the dynar.
1858 * Change the prototype of xbt_thread_create(), sorry.
1859 Added a boolean parameter indicating whether we want to join this
1860 thread (used in SG only for now)
1861 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1864 * GTNetS wrappers should now be usable again (and betterly tested too)
1865 * Fix a major regression from 3.2 where the timeout provided to
1866 MSG_task_put_with_timeout() was used as absolute time before which
1867 the comm should be done.
1868 * Start to fix the <cluster> tag.
1869 - Internal links should be good now (beside of the loopback, which
1870 use the private link instead)
1871 - paths to the external world is still rather broken
1872 - the <route:multi> tag is just broken. Actually that's brain-dead.
1873 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1874 to make it less stupid
1875 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1876 * Fix a source-level compatibility glitch from 3.2: after defining
1877 MSG_USE_DEPRECATED, you can use the old name
1878 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1879 * Allow to compile from the SVN with automake 1.11
1880 * Fix some problems when using the "start_time" tag in deployment XMLs.
1881 * Fix #8569: XBT/synchro.h has redundant declarations
1882 * Fix #8563: MSG return values and exceptions
1883 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1884 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1885 Thanks to Robson Peixoto.
1886 * Fix a vicious bug in dictionaries inducing that some elements were
1887 not freed on xbt_dict_free()
1889 Portability report of this version:
1890 * Main portability targets:
1891 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1892 on (amd64/i386/ia64)
1893 - mac leopard on i386
1894 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1895 but nothing critical.
1896 * Other platforms: windows, AIX and others were not tested for this release
1898 Timing report of this version:
1899 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1900 scenario parameters). From 8h to a few seconds in desktop grid settings.
1901 * Supernovae brings about 25% speedup on i386.
1903 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1905 SimGrid (3.3.3) stable; urgency=low
1907 The "Need for Speed" release.
1909 The timings done to validate the 3.3.2 were faulty.
1910 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1912 The problem was a conversion from a manually handled vector to
1913 xbt_dynar_t on the critical path.
1914 xbt_dynar_foreach calls functions, inducing stack management crap.
1916 We inlined these functions and xbt_dynar_foreach is now breath taking.
1917 We also inlined xbt_swag_belong on the way.
1919 Here are some approximate speedup measurements (on master/slaves
1920 simulations lasting between 10s and 20s each):
1921 3.3.1 -> 3.3.2: about same performance
1922 3.3.2 -> 3.3.3: 40% speedup
1923 3.3.1 -> 3.3.3: 40% speedup
1924 3.3.1 with inline patch -> 3.3.3: 30% speedup
1926 Our reading is that the refactoring which occurred in 3.3.2 made us
1927 suffer much more from the xbt_dynar_foreach low performance, but
1928 once we solved this, this refactoring proved to be very performance
1929 effective. From the 40% speedup, somehow, 10% are due to the
1930 inlining and 30% to the refactoring.
1932 That's a pitty that gcc cannot inline functions placed in other files
1933 alone. We have to choose between:
1934 - break the encapsulation (by putting private data structures and
1935 accessors in headers files to help gcc)
1936 - live with low performance
1937 - switch to a decent compiler such as icc (not quite possible).
1939 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1941 SimGrid (3.3.2) stable; urgency=low
1943 The "Simplicity does not preceed complexity, but follows it" release.
1945 The main contributors of this release were (lexical order):
1946 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1949 * Extract the routing logic into its own object.
1950 (was dupplicated in network.c and workstation_LV07.c;
1951 Allows to implement other ways of storing that info)
1952 => kill now useless network_card concept
1953 - Use dynar to represent routes (instead of void** + int*)
1954 - kill link_set (use surf_network_model->resource_set instead)
1955 - Add a command-line option to choose the routing schema to use
1956 - Add three new models:
1957 * Floyd (shortest path computed at initialization)
1958 * Dijikstra (shortest path recomputed all the time)
1959 * Cached Dijikstra (shortest path computed on need)
1960 All these models where contributed by Silas De Munck, and are
1961 described in his ICCS09 paper.
1963 * Simplify model declaration
1964 (less redirections, less function to write when defining a model)
1965 - Factorize stuff between models:
1968 surf_model_resource_set(model)
1969 surf_model_resource_by_name(model, name)
1970 - Unify the types of models in s_surf_model_t (using an union)
1971 - Embeed fields of common_public directly into s_surf_model_t
1972 - Rename model methods:
1973 action_free ~> action_unref
1974 action_change_state ~> action_state_set
1975 action_get_state ~> action_state_get
1976 - Change model methods into functions :
1977 (model)->common_public->action_use ~> surf_action_ref
1979 * Implement a generic resource; use it as ancestor to specific ones
1980 (allows to kill duplicated code in models)
1981 Drawback: timer command don't need no name nor properties;
1982 workstation_CLM03 don't need no properties
1983 (but I guess we can live with those few bytes wasted)
1985 * Improve the action object model
1986 - implement a constructor avoiding dupplicated code about field
1987 initialization in generic_action part.
1989 * Kill the SDP model: it has an external dependency, is deprecated
1990 in flavor of modern lmm models, and didn't compile since a while
1993 * Relocation of the context module from XBT to SIMIX.
1994 (the context were decoupled from the simix processes, duplicating a lot of code)
1995 => a lot of code was factorized
1996 - less overhead is introduced during scheduling
1997 - simpler API for the context factory
1998 - the logic for process creation,destruction and manipulation was simplified
1999 * Simplification of the s_smx_process_t data structure.
2000 => accesing the simix level data associated to a process is faster now,
2001 and the code is a lot more readable.
2004 * Implement some more MPI primitives:
2005 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2006 -implementation: Bcast: flat or 2-ary tree (default),
2007 Barrier: 4-ary tree,
2009 Allreduce: Reduce then Bcast
2010 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2011 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2012 Alltoallv: flat tree, like ompi
2014 * Add support for optimized collectives (Bcast is now binomial by default)
2015 * Port smpirun and smpicc to OS X
2018 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2019 Shout out if you used it.
2022 * Display the list of still queued messages in SG mode when existing
2026 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2027 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2028 * Add xbt_str_from_file(FILE*)
2029 * Add xbt_dict_get_key achieving a linear reverse search
2030 * Remove the context module
2032 Portability report of this version:
2033 * Main portability targets:
2034 - Linux(debian)/x86/context
2035 - Linux(debian)/x86/pthreads
2036 - Linux(debian)/amd64/context
2037 - Linux(debian)/amd64/pthreads
2038 On these, we still have the eratic breakages of gras/pmm and
2039 amok/saturate_sg reported in previous version. We still think
2040 that the tests are the cause of the fault, not the tested code.
2042 - Mac OSX Leopard/x86/context
2043 Still false negative in tesh autotesting.
2044 Smpi still fails, but this time because readlink does not accept -f
2045 Everything seems to work properly beside of that.
2048 - AIX version 5.3 (only tested contexts this time)
2049 Smpi still fails there because mktemp is not installed.
2050 Everything seems to work properly beside of that.
2052 I managed to compile it for the first time, but several breakages.
2053 Won't delay the release for this exotic platform.
2055 * Windows: it's still lagging behind. If you want to help, please
2058 Timing report of this version:
2059 This version seem to be more than 5% faster than 3.3.1 (on linux
2060 64bits with contextes). The gain is less than expected, we are
2061 investigating this for next release.
2063 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2065 SimGrid (3.3.1) stable; urgency=low
2068 * Implement a --cfg-help to show existing configuration variables
2069 * Build chain do not require doxygen in maintainer mode
2072 * fix a bug on struct sizeof computation, which prevented the
2073 exchange of arrays of structs in some conditions
2074 - added a regression test about this in datadesc_usage
2075 * Allow the exchange of 0-long dynamic vectors.
2076 - for that, use -1 as indicator of dynamic size instead of 0
2077 - This implied to change any size from unsigned long to long,
2078 reducing a bit communication abilities, but I guess that with
2079 64bits being quite common, this is more than enough.
2080 - This also induce a protocol change, thus bumping network protocol
2081 version from 0 to 1 (if we have external users, we have to get
2082 clean on that point too ;)
2083 - added two regression tests about this in datadesc_usage
2084 * Be more verbose when propagating local exceptions
2085 This helps debugging.
2086 * Display the status of simulated processes when receiving SIGINT in
2090 * Allow to control the simulation from a trace file.
2091 New functions MSG_action_register() and MSG_action_trace_run()
2092 The first one allows to associate a function execution to each
2093 kind of action while the second one parses a trace file and
2094 triggers the corresponding actions within the system.
2095 For now, only a toy example is provided in examples/msg/actions
2096 * Add an exemple of process migration in examples/msg/migration
2097 * Fix a bug in task exchange which broke MSG_task_get_sender()
2098 Add a teshsuite regression test for that.
2099 [Bug: if MSG_task_get_sender() is called after sender exit,
2101 * Fix a bug which prevented suspend/resume to work properly
2102 * Display the status of simulated processes when receiving SIGINT
2103 This fixes a regression of v3.3. due to the introduction of SIMIX
2104 * Bug fixing in failure management:
2105 - trace could not start by a failure at time 0
2106 - failure during communications were not working
2109 * Add SIMIX_process_set_name() to change the name of the current
2110 process in the log messages.
2111 * Store smx_hosts in a dict since we only retrieve them by name
2112 * Move the configuration infrastructure to surf
2115 * Move the configuration infrastructure to surf
2118 * Massive internal cleanups:
2119 - Store internal structures on processes instead of hosts (allows
2120 to have more than one process per host, in addition of being more
2122 - Cleanup the initialization/finalization process
2123 - Kill a whole bunch of unneeded synchronization:
2124 processes run in exclusive manner within the simulator
2125 - Move queues from global tables to process data fields
2127 - now accept -platform and -hostfile arguments
2128 - Pass the right rank value to processes according to the hostfile
2129 * Compile the examples by default, and use them as regression tests
2130 * Implement MPI_Wtime()
2131 * Change the reference speed to a command line option
2134 * TCP_gamma can now be specified as command line option using
2135 --cfg=TCP_gamma:10000000.0
2136 * Change the --surf-path cmd line option into --cfg=path:
2139 * Also include strbuff from xbt.h public header
2140 * xbt_ex_display(): do not free the exception after displaying
2141 This allows to do more with the given exception afterward.
2142 Users should call xbt_ex_free() themselves.
2146 Portability report of this version:
2147 * Main portability targets:
2148 - Linux(debian)/x86/context
2149 - Linux(debian)/x86/pthreads
2150 - Linux(debian)/amd64/context
2151 - Linux(debian)/amd64/pthreads
2152 These targets fail about 1/10 of times on gras/pmm, but we believe
2153 that this is because of the test, not because of simgrid.
2154 amok/saturate_sg fails even more rarely, and the test may not be
2157 - Mac OSX Leopard/x86/context
2158 The test suite still spits tons of errors because some obscure
2159 force prevents us from removing the temporary directories
2160 arguing that they still contain some metadata I've never heard of.
2161 Smpi fails because seq is not installed.
2162 Everything seems to work properly beside of that.
2165 - AIX version 5.3 (both contexts and pthread)
2166 Smpi still fails there because mktemp is not installed.
2167 XML inclusions seems rosty on AIX.
2169 * Windows: it's still lagging behind. If you want to help, please
2172 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2174 SimGrid (3.3) stable; urgency=high
2178 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2179 [Malek Cherier & Mt]
2181 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2184 Until now, GRAS were using MSG as an interface to SURF. It was
2185 quite difficult because both interface have several differences
2186 (MSG channels vs GRAS sockets were the most notable point).
2188 This also opens the gate to SMPI (which should occur soon) and speed
2189 up simulations by to 40% (even if it were not the main goal).
2191 **************************************
2192 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2193 **************************************
2194 The point is that events occuring at the exact same timestamp are
2195 not scheduled in the same order with the old and new version. This
2196 may be enough to completely change the execution of simulations in
2197 some cases. Sorry for the inconvenience.
2199 * Cleanup and upgrade the XML format to push further scalability
2200 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2202 * Improve the testing infrastructure with tesh. Now a very large part of
2203 the code is tested not only by being run but also by checking that the
2204 output match an expected output [Mt].
2206 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2207 is really less memory-demanding, which should allow you to use
2208 larger files in SimGrid [AL].
2210 * Inform valgrind about our contextes, so that it becomes usable
2211 with the default (and more effecient) version of SimGrid
2212 [contributed by Sékou Diakite, many thanks]
2215 * Introduce a listener thread in charge of receiving incoming
2216 messages from the network. It allows to overlap communication and
2217 computation but most notably, it removes some stupid deadlocks due
2218 to the fact that so far, a process could not send and receive at
2219 the same time. This made most non trivial communication schema
2221 * Convert the PIDs from long int to int to match the MSG ones (and
2222 linux ones too) [Mt]
2223 * New function: gras_agent_spawn() to launch a new process on
2224 current host. Only working in simulation for now. [Mt]
2225 * New function: gras_os_hostport() returning a constant form (ie,
2226 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2229 * Make the backtrace of exceptions more human readable [Mt]
2230 * New module: xbt/str [Mt]
2231 a ton of string utility functions (split, join, printf to a newly
2232 allocated buffer, trim, etc)
2233 * New module: xbt/hash [Mt]
2234 SHA1 hashing algorithm (more to come if needed)
2235 * New module: xbt/synchro [Mt]
2236 synchronization tools (mutex and conditions) working the same way
2237 in simulation and in real life (mainly useful for GRAS, but not
2239 * New module: xbt/queue [Mt]
2240 classical producer/consumer synchronization scheme
2241 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2242 (using the classical functions will get serialized) [Mt]
2243 * Make dictionary internal table dynamic. No need to specify its size
2244 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2246 * Make sure the log channels are organized as a tree under windows
2247 (because of ANSI C compatibility issue, any channel were child of
2248 root directly) [Mt].
2251 * Cleaned many thing in surf and fixed a few bugs [AL].
2252 * Add a nice command line configuration mechanism to compose models [AL].
2253 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2254 the previous one (KCCFLN05). It relies on something that looks like
2255 a max-min sharing mechanism but cannot be written as such. A new solver
2256 was thus designed [AL].
2257 * Add a new solver to lmm. Based on Lagrange optimization and
2258 gradient-based descent, it enables to efficiently maximise systems s.a
2260 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2262 This solver enables to propose two new network models for TCP Reno and
2263 TCP Vegas based on Low's work. These models still need to be fully
2264 tested though [Pedro Velho].
2267 * Bug fix in SD_simulate. Now the time bound given as argument is
2269 * Use the new parallel task model (ptask_L07) as default.
2270 * Use the SURF command line configuration mechanism.
2271 * 0-size tasks (for synchronization) should now work.
2273 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2275 SimGrid (3.2) stable; urgency=high
2279 We still experience issues on this platform, but we believe that at
2280 least MSG is usable.
2282 GRAS API BREAKAGE (for simplification purpose, sorry):
2283 * the gras_msgtype_by_name is not used anymore. Instead of
2284 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2285 you can write (and must)
2286 gras_msg_send(toserver, "request", &request);
2287 - If you still want to pass a gras_msgtype_t to the function (to cache
2288 the type and avoid the lookup time), use the gras_msg_send_() variant.
2289 - Impacted functions:
2290 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2291 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2292 * The callbacks are now expected to return 0 when everything went well
2293 (just like the main() function)
2295 GRAS new features and improvements:
2296 * New module mechanism where user code can use per process globals [Mt]
2297 This is similar to gras_userdata_*() functions, but for libraries. It
2298 factorize some code developped over and over in the examples and AMOK.
2299 It has still to be documented and used (only amok/peermanagement is
2301 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2302 loss when they were small enough to fit into the buffer and sent quickly
2303 enough so that they can all get received in one shoot.
2304 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2305 if not found. Use the *_or_null() variant for the old semantic.
2306 * In gras_msg_handle, do not discard messages without callback.
2307 They are probably messages to be explicitly awaited later (ie, proofs of
2308 mis-synchronization in userland since they are sent before being awaited)
2309 No big deal usually.
2310 * gras_socket_meas_send/recv: semantic changed!
2311 The numerical arguments used to be (1) the total amount of data to send
2312 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2313 messages. This was need for the fool willing to send more than MAXINT
2314 bytes on quite fat pipes.
2317 * Do really rename the hostmanagement module to peermanagement. [Mt]
2318 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2319 but this is rather new and this was documented in the module
2320 documentation (poor excuses, I admit)
2321 * Bandwidth measurement semantic changed! This follows the changes to
2322 gras_socket_meas_send/recv explained above.
2325 * A sequential mode has been added to the workstations. When a workstation
2326 is in sequential mode, it can execute only one task, and the other tasks
2327 are waiting in a FIFO. [Christophe Thiery]
2330 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2331 model for SIMDAG. [Christophe Thiery]
2332 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2333 equal to 0, which caused some bad behaviors in
2334 saturated_constraint_set_update. I now use a threshold mechanism like in
2338 * When running manually src/testall, you select specific units [Mt]
2339 testall is the result of our cunit mechanism, and should replace all
2340 the scripty thingy around since bash don't run easily on billware.
2342 * A mallocator system has been added. [Christophe Thiery]
2343 Mallocators allow you to recycle your unused objects instead of freeing them
2344 and allocating new ones.
2346 Documentation update:
2347 * FAQ reworking + New FAQs:
2348 - "Valgrind spits tons of errors!" [Mt]
2349 - "How to repport bugs" [Mt]
2350 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2351 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2352 - Communication time measurement within MSG [AL]
2353 - I experience weird communication times when I change the latency [AL]
2354 * GRAS tutorial [Mt]
2356 - an introduction to the framework and to the used communication model
2357 - an initiatic tour introducing the most proheminent features:
2359 . Lesson 0: Installing GRAS
2360 . Lesson 1: Setting up your own project
2361 o Part 2: Message passing
2362 . Lesson 2: Exchanging simple messages
2363 . Lesson 3: Passing arguments to the processes (in SG)
2364 . Lesson 4: Attaching callbacks to messages
2365 . Lesson 5: Using globals in processes
2366 . Lesson 6: Logging information properly
2367 . Lesson 7: Using internal timers
2368 . Lesson 8: Handling errors through exceptions
2369 . Lesson 9: Exchanging simple data
2370 . Lesson 10: Remote Procedure Calling (RPC)
2371 . Lesson 11: Explicitely waiting for messages
2372 . Recapping of message passing features in GRAS
2373 - A HOWTO section containing:
2374 o HOWTO design a GRAS application
2375 More are due, of course. They will come latter. In the meanwhile, you can
2376 check the examples which are still here.
2378 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2380 SimGrid (3.1) stable; urgency=high
2384 There was a stack corruption somewhere, visible only when optimizing
2385 with these versions. [Vince]
2388 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2389 module that enabled to play with Directed Acyclic Graphs. It is built
2390 directly on top of SURF and provides an API rather close to the old
2391 SG. Some old codes using SG are currently under rewrite to check that
2392 all needful functions are provided. [Christophe Thiery]
2395 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2396 extension of the classical CLM03 model that gracefully handles
2397 failures. This is now the default model for MSG and GRAS. It doesn't
2398 handle parallel tasks yet though. [AL]
2399 * Bug fix: Weights were not correctly set in the network part.
2400 WARNING: This may have resulted in incorrect results with simulations
2401 where there are more than one flow on a given link. [AL]
2404 * After a (long ?) discussion on simgrid-devel, we have decided that the
2405 convention we had on units was stupid. That is why it has been decided
2406 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2407 WARNING : This means that all previous platform files will not work as
2408 such with this version! A warning is issued to ask users to update
2410 A conversion script can be found in the contrib module of the CVS, under
2411 the name contrib/platform_generation/surfxml_update.pl [MQ]
2414 * Bug fix: Processes were started in reverse order, wrt deployment file.
2415 WARNING: if your code relies on this bug, please fix it. [AL]
2416 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2417 being executed on two different locations. [AL]
2418 * Bug fix: Failures are now better supported thanks to Derrick's tests
2419 (there was many failure situations I hadn't thought of and that weren't
2420 correctly handled). [AL]
2421 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2425 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2426 Exception occurring on server-side are propagated back to client (!).
2428 API CHANGE: the callback changed their prototype. Change:
2429 int my_handler(gras_socket_t expeditor, void *payload_data) {
2431 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2432 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2434 * New! function: gras_msg_handleall to deal with all messages arriving
2435 within a given period.
2436 * New! function: gras_socket_server_range to get a server socket in a
2437 range of port numbers (ease to avoid port number conflicts) [MQ]
2438 * New! gras processes display their backtrace when they get a SIGUSR1
2439 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2440 Sweet to debug RL processes [MQ]
2444 - Do not force experiment sizes to be expressed in kb, or it becomes
2445 impossible to measure the latency this way (needs one byte-long tests)
2446 WARNING: this changes the amok_bw_* function semantic. [MQ]
2447 - Implements the link saturation stuff. [MQ]
2448 * Peer management module:
2449 New! module factorizing code that we wrote over and over [MQ].
2452 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2453 - Test units are placed directly into the library code, they get extracted
2454 automatically and placed into the src/testall binary.
2455 - Convert most of the XBT tests to this system.
2456 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2457 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2458 live in the same C code. This required to change a little bit the API
2459 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2460 * New module: sparse graph structure with basic algorithms (this is work
2461 in progress and the API is not considered to be frozen yet). [AL]
2462 * Display more information on backtraces: source line & function names are
2463 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2464 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2465 * Reworked a little bit some #include statements to load only required
2466 headers. Some user code that relied on SimGrid to include stdlib or
2467 stdio may need to include it by themselves. [AL]
2468 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2470 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2471 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2472 but I guess that almost nobody used those parts. [MQ]
2474 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2476 SimGrid (3.0.1) stable; urgency=low
2479 * Unfortunately, I had missed 5 misnamed functions:
2480 xbt_fifo_item_t xbt_fifo_newitem(void);
2481 void xbt_fifo_freeitem(xbt_fifo_item_t);
2482 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2483 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2484 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2485 They're now deprecated. Please use their new versions:
2486 xbt_fifo_item_t xbt_fifo_new_item(void);
2487 void xbt_fifo_free_item(xbt_fifo_item_t);
2488 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2489 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2490 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2492 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2493 * Documentation: xbt_log module unmercifully reworked [MQ]
2494 * Bugfix: there was a problem with the ending of contexts with
2495 the pthread backend. It caused some weird deadlock or behavior
2496 depending on the pthread implementation. [AL]
2497 * Bugfix: get the exceptions raised in the simulator repport where
2498 and why they come from when they are not catched in time [AL, MQ]
2501 * Bugfix: Do repport the error when two non-connected hosts try to
2502 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2505 * Add additionnal checkings on communications. Assert that two
2506 communicating hosts are connected by a set of links... [AL]
2509 * Add additionnal checkings on channel values in communication [AL]
2510 * New: MSG_task_get_source to see on which host a task was generated [HC]
2511 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2512 the number of tasks waiting to be received on channel and sent
2514 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2515 waits for the first task coming from a given host.. [AL]
2517 GRAS new functionnalities: [MQ]
2518 * Enhance the parsing macro to allow the size of multidimentional objects
2519 to be given thru annotations.
2520 * New example (and documentation): Matrix Multiplication a la RPC
2521 (as when I was young!) and fix a bunch of bugs found on the way.
2523 GRAS performance improvements: [MQ]
2525 * Reduce the amount of cbps creation/destruction by making it static to
2526 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2529 * Change libdata to a set so that we can search for stuff by ID (and thus
2530 reduce the insane amount of dict lookups)
2533 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2534 RL communication on top of SG since it's so uneffective.
2535 It may also be used for inter-thread communication in RL, one day.
2536 * Use gras_datadesc_copy() to exchange messages on top of SG
2538 - improve message exchange performance on top of SG
2539 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2540 * Don't exchange on the network the size of the used part of buffer,
2541 instead, specify the possible buffer size to read().
2543 - reduces the amount of read/write calls (one pair per exchange)
2544 - reduces the amount of exchanged data (the size)
2545 - allows to retrieve all arrived data on receiver side, if we don't need
2546 it right now (subsequent read will peek the buffer)
2547 - allows the receiver to proceed with the begining of the stream before
2548 everything is arrived
2549 - make it possible to build an iov transport (using readv/writev)
2551 - take care of the data with non-stable storage (like stacked data),
2553 * If possible, TCP send uses vector I/O (when writev() is here)
2554 - Don't use it for receive since we send data sizes and data on the
2555 same stream, so we wouldn't be able to chain large amount of chunks
2556 before having to flush the stuff to read the size.
2557 * Rework the transport plugin mechanism to simplify it and reduce the
2558 amount of pointer dereferencement when searching for the right function
2561 * I guess that now, we do almost as few system calls as possible while
2562 doing as few data copy as possible.
2564 To improve it further, we could try to send all the sizes first and then
2565 all the data (to use iov on receiving size), but it's only a partial
2566 solution: when you have 2 dimensional data, the sizes of the second
2567 dimension is data of the first dimension, so you need 3 streams.
2569 I'm not sure the potential performance gains justify the coding burden.
2571 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2573 SimGrid (3.00) stable; urgency=high
2576 * New! Give the possibility to hijack the surf parser and thus bypass
2577 MSG_create_environment and MSG_launch_application. Have a look at
2578 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2580 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2582 SimGrid (2.96) unstable; urgency=low
2587 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2588 This deprecates the xbt_error_t mechanisms.
2589 It modifies (simplifies) all XBT and GRAS API.
2590 MSG API keeps unchanged (exceptions raised by XBT are catched from
2591 within MSG and masked with existing error handling facilities)
2594 * New! Add a FATPIPE model. [AL]
2595 * New! Add a parallel task model. [AL]
2596 * New! Add automatically a loopback interface (in the default
2597 network model) if none was precised.
2600 * Bugfix: MSG_process_resume now works with the current running process.
2602 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2603 * Modification of MSG_task_get_compute_duration. Once a task has been
2604 processed, the value returned by this function is now equal to 0. [AL]
2605 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2606 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2607 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2608 of a m_task. For now, MSG_task_cancel only works with computation
2610 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2611 of the processor (in Mflop/s) regardless of the current load on the
2613 * API Change: use proper naming convention for MSG_getClock and
2614 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2616 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2617 This function changes the priority of a computation task. This priority
2618 doesn't affect the transfer rate. A priority of 2 will make a task
2619 receive two times more cpu power than the other ones. This function
2620 has been added to suit the needs of Nguyen The Loc and hasn't been that
2621 much tested yet. So if it fails, please report it and send me your code.
2623 * API Change: removed all functions and types that were marked "deprecated"
2624 since many months. Renamed MSG_global_init_args to MSG_global_init.
2626 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2628 SimGrid (2.95) unstable; urgency=low
2631 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2632 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2633 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2636 * Bugfix: Allow absolute paths for platform description files [MQ]
2637 * Bugfix: do free the variables after use. Leads to drastic performance
2639 * Implement max_duration (ie, timeouts) on resources [AL]
2642 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2644 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2645 the needed support (timeouts) [AL]
2648 * Implement measurement sockets. You can now get the bandwidth between two
2649 hosts thanks to AMOK (see below). [MQ]
2650 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2651 dynar over the network (yeah) [MQ]
2652 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2653 * simple/static token-ring example. [Alexandre Colucci and MQ]
2654 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2655 in place before [MQ]
2658 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2659 type parsing. "short" and "long" modifiers were also ignored; other
2660 modifier (such as reference level) are still ignored. [MQ]
2661 * Embeed the buffer size within the buffer itself on SG. [MQ]
2662 That way, send() are atomic and cannot get intermixed anymore (at least
2663 the ones which are less than 100k; bigger messages still have the issue)
2664 * Array size pushed by the field, not by the field type (or each
2665 and every long int will push stuff to the cbps) [MQ]
2666 * use select() to sleep since it allows to portably sleep less than one
2669 GRAS (minor cleanups)
2670 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2673 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2674 * type category 'ignored' killed as it was never used and were difficult
2676 * whether a type can cycle or not is now a flag, leaving room for more
2677 flags (as "ignored", if we feel the need one day ;)
2678 * Rename raw sockets to measurement sockets since "raw" has another
2679 meaning in networking community.
2682 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2683 and offers features not belonging to GRAS but that most applications
2684 need. One day, it may be a set of plugins loadable at runtime.
2685 * New module: bandwidth
2686 bandwidth measurement between arbitrary nodes running this module. [MQ]
2688 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2690 SimGrid (2.94) unstable; urgency=low
2692 The first beta release of SimGrid 3 !
2696 * Update the main page and the FAQ. Adding references to gforge.
2699 * Add a gras_os_getpid function.
2702 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2703 * Extend the logs so that they also print PID, hostname, date, ... if
2705 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2706 and kill the old version which were in testsuite/
2707 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2711 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2712 for users. Instead, each programming environment (GRAS, MSG,...) use it
2713 when needed to provide such a feature to users.
2714 Don't shortcut the mechanism or you will also shortcut the virtualization
2715 you need on the simulator.
2719 * Cleanups in configury with regard to compile optimization/warning flags.
2720 Also add -fno-loop-optimize to any powerpc since it's the optimization
2721 killing gcc (< 3.4.0).
2722 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2723 needed by MSG examples complications
2724 * Borrow configury beautifications from PHP
2727 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2728 warning (thanks loris for stumbling into this one).
2729 * Bugfix: stop loading private headers (gras_config.h) from the public
2733 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2734 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2735 RL would imply a DNS resolver, which is *hard* to do in a portable way
2736 (and therefore delayed).
2737 * Implement a real timer mechanism and use it in timing macros. This allows
2738 to avoid rounding errors and get a 0.000005 sec precision in timing
2739 macros. While I was at it, various cleanups:
2740 - allow to declare more than one timed section per file (fix a stupid bug)
2741 - move some private declaration to the right place
2742 - merge conditional execution and timing macros into emulation module
2743 - document the module
2744 - make sure the module cleanups its mess on gras_exit
2745 * Documentation improvements:
2746 - (new) how to compile applications using GRAS
2747 - (new) emulation support (timing macros)
2749 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2751 SimGrid (2.93) unstable; urgency=low
2753 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2756 - Use Paje properly where used. Still to be sanitized properly.
2757 - Portability fix: Add an implementation of the contexts using pthread
2761 - Add xbt_procname(): returns the name of the current process.
2762 Use it to show the current process's name in all logging.
2764 - fix detection of older flex version and the reaction, since we do
2765 depend on modern ones (we use lex_destroy)
2766 - Better separation of SG and RL in the libs: remove all simulation code
2767 from libgras. As a result, this lib is now only 200k when stripped.
2768 Some of the xbt modules may also be duplicated (two sets and such) and
2769 should be cleaned/killed before SG3.
2770 - Insist on using xlC on AIX because of weird problems involving gcc there.
2771 - Cleanup the make remote stuff. This is now done by scripts
2772 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2773 mainly for our private use, but we're working on changing them to user
2776 - Bugfix: flush the socket on close only if there is some *output*.
2777 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2779 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2782 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2784 SimGrid (2.92) unstable; urgency=low
2786 Alpha 3 on the path to SimGrid 3
2790 - New! First try of benchmarking macros.
2791 - New! First try so that gras_stub_generator generate deployment and
2792 remote compilation helpers.
2794 - Bugfix: Initialization fix in msg_test.
2798 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2800 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2802 - API Change: gras_os_sleep to take the amount of seconds as a double.
2803 Accepting an int was error prone since it was the only location where
2804 seconds were coded as such. It leaded to damn rounding errors.
2805 - Bugfix: Hard to belive that timers ever worked before this.
2807 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2809 SimGrid (2.91) unstable; urgency=low
2811 Alpha 2 on the path to SimGrid 3
2815 - Bug fix in the lmm_solver.
2817 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2818 through the function MSG_paje_output.
2819 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2820 - It is possible to bound the rate of a communication in MSG with
2821 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2822 to write it in the changelog).
2823 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2828 - Various cleanups to the autotools stuff
2829 - Begin to move Gras examples to examples/gras/
2830 - Let make distcheck work again (yeah!)
2832 - documentation overhauled using doxygen.
2833 gtk-doc-tools is dead in SimGrid now.
2834 - Automatically extract all existing logging categories, and add the list
2835 to the documentation (long standing one, to say the less)
2837 - Cleanup the known architecture table. Reorder the entries to group what
2838 should be, and use a more consistent naming scheme.
2839 (some of the test dataset are still to be regenerated)
2840 - New! Allow library to register globals on each process just as userdata
2842 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2843 lookup time (for now).
2844 Use it in msg and trp.
2845 This cleans a lot the internals and helps enforcing privacy of the
2846 headers between the gras components.
2847 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2848 - Bugfix: gras_os_time was delirious in RL.
2849 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2851 - Reenable GRAS now that it works.
2853 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2855 SimGrid (2.90) unstable; urgency=low
2857 Alpha 1 on the path to SimGrid 3
2859 * It is a long time since the last release of SimGrid. I'm sorry about
2860 that but as I had told you, I was rewriting a lot of things. I apologize
2861 to those who had been reporting bugs to me and that I had not answered.
2862 If your bug is still in the new version, please tell me. Here is a
2863 summary of the main changes.
2865 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2866 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2867 portability, speed, and a lot more but you'll figure it out later.
2868 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2869 I wanted to release the new MSG as soon as possible and I have
2870 broken GRAS, which is the reason why, for now, only MSG is fully
2871 functional. A laconic description of these projects is available
2872 in the documentation.
2874 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2875 kernel optimized for our needs (called SURF but only the developers
2876 should use it). Hence, MSG is now roughly 30 times faster and I think
2877 that by rewriting a little bit MSG, I could event speed it up a little
2878 bit more. Beside the gain in speed, it is also much easier to encode a
2879 new platform model with SURF than it was with SG. More to come...
2881 * REVOLUTION 3: I have tried to change a little as possible the API of
2882 MSG but a few things really had to disappear. The main differences
2883 with the previous version are :
2884 1) no more m_links_t and the corresponding functions. Platforms are
2885 directly read from a XML description and cannot be hard-coded
2886 anymore. The same format is used for application deployment
2887 description. The new format is described in the documentation.
2888 Have a look in tools/platform_generation. There is a tiny script
2889 that converts from the old platform format to the new one. Concerning
2890 the application deployment format, parsing the old one is tricky.
2891 I think most of you should however be able to convert your files. If
2892 it is really an issue, I can write a C code that does the conversion.
2894 2) the toolbox tbx does not exist anymore. We now have a library
2895 with much more data-structures but without the hash-tables (we have
2896 dictionaries that are much faster).
2898 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2900 *****************************************************************************
2901 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2902 * don't feel like converting the oldies. (Mt) *
2903 *****************************************************************************
2906 Version 2.90: "the long awaited one"
2907 - Finished rewriting and debugging MSG. Rewrote the documentation.
2908 - disable GRAS for now since it needs to be ported to the newest SG
2911 - Finish the port to windows (using mingw32 for cross-compile)
2914 - Main loop and datastructures of SURF. A cpu resource object is
2915 functional. Surf can thus be used to create cpu's with variable
2916 performance on which you can execute some actions.
2918 2004-11-15 Martin Quinson
2919 - Port to ARM. Simply added the alignment and size descriptions. Should
2920 work, but the ARM machines are so slow that I didn't had the opportunity
2921 to 'make check' over there yet.
2923 2004-11-15 Arnaud Legrand
2924 - Trace manager now written. It uses a heap structure and is therefore
2925 expected to be efficient. It may however be speeded up (particularly
2926 when many events occur at the same date) by using red and black
2927 trees. One day maybe...
2928 - Max-min linear system solver written. It uses a sparse matrix
2929 structure taking advantage of its expected use. Most operations are
2930 O(1) and free/calloc are called as few as possible. The computation of
2931 the minimum could however be improved by using a red and black tree
2934 2004-11-03 Arnaud Legrand
2935 - Rename every gras_* function that was in xbt/ to its xbt_
2937 - Add a heap and a doubly-linked list to xbt
2938 - Added a dichotomy to the dictionaries. make check works as well before
2939 so I assume that the patch is correct. I do not know however if things
2940 run effectively faster than before now. :)
2942 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2943 SimGrid, and the version number is bumped to 2.x
2945 2004-10-29 Martin Quinson
2946 - Introduction of the remote errors.
2947 They are the result of a RMI/RPC on the remote machine.
2948 ErrCodes being scalar values, you can't get the host on which those
2949 errors did happen. Extending the error mechanism as in Gnome is possible.
2950 No idea yet whether it is a good idea.
2952 2004-10-28 Martin Quinson
2953 - Interface revolution: the Starred Structure Eradication.
2954 I used to do typedef struct {} toto_t; and then handle *toto_t.
2955 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2956 - ???_t is a valid type (builded with typedef)
2957 - s_toto_t is a structure (access to fields with .)
2958 - s_toto is a structure needing 'struct' keyword to be used
2959 - e_toto_t is an enum
2960 - toto_t is an 'object' (struct*)
2962 typedef struct s_toto {} s_toto_t, *toto_t;
2963 typedef enum {} e_toto_t;
2964 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2965 s_toto_t) is private.
2967 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2968 so that it can set the variable to NULL. It was so for dicts and sets,
2969 it changed for dynars.
2971 - Fix a bunch of memleaks in dict_remove
2972 - Fix a bug in sg/server_socket opening: it failed all the time.
2974 2004-10-07 Martin Quinson
2975 - Speed up dynar lookup operation a bit.
2977 gras_dynar_get is dead.
2979 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2980 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2981 (which gives you the address of the stored data).
2983 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2984 copy of the data using an affectation to do the job and not a memcpy.
2986 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2987 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2989 It does not really speedup the dynar test because they are
2990 setting elements all the time (and look them seldom). But the dict does
2991 far more lookup than setting.
2993 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2995 2004-10-05 Martin Quinson
2996 - Allow to (en/dis)able the cycle detection at run time.
2998 Whether we should check for cycle or not is now a property of each
2999 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3000 datadesc_cycle_unset allow to remove this property when previously set.
3002 Note that the cycle detection is off by default since it impacts the
3003 performance. Watch the data you feed GRAS with ;)
3005 This property is hereditary. Any element embedded in a structure having it
3006 set have it set for the time of this data exchange.
3008 You should set it both on sender and receiver side. If you don't set it on
3009 sender side, it will enter an endless loop. If you forget on receiver
3010 side, the cycles won't be recreated after communication.
3012 - Header reorganization.
3013 Kill gras_private.h, each submodule must load the headers it needs.
3015 2004-10-04 Martin Quinson
3016 - Interface revolution: do not try to survive to malloc failure.
3018 Now, gras_malloc and friends call gras_abort() on failure.
3019 As a conclusion, malloc_error is not a valid error anymore, and all
3020 functions for which it was the only gras_error_t return value are
3021 changed. They now return void, or there result directly.
3022 This simplify the API a lot.
3024 2004-09-29 Martin Quinson
3025 - Re-enable raw sockets.
3026 Created by gras_socket_{client,server}_ext;
3027 Used with gras_raw_{send,recv}
3030 It should allow to kill the last bits of gras first version soon.
3032 This is not completely satisfactory yet (duplicate code with
3033 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3036 - Simplify transport plugin (internal) interface by not passing any
3037 argument to _server and _client, but embedding them in the socket
3040 2004-09-28 Martin Quinson
3041 - Finish the port to AIX.
3042 autoconf was my problem (segfault within the malloc replacement
3043 function. No idea why)
3045 2004-09-16 Martin Quinson
3046 - Fix some size_t madness on 64bit architectures.
3048 2004-09-08 Martin Quinson
3049 - Reduce the number of system headers loaded, overload some more system
3050 calls (such as malloc to cast the result of the system one, and work
3052 - Fix and reintroduce the config support
3054 2004-09-07 Martin Quinson
3055 - Source code reorganization to allow Arnaud to surf all over there.
3056 - Allow to document the logging categories.
3057 - Remove all uppercase from logging categories and useless cleanup in names.
3059 2004-08-18 Martin Quinson
3060 Version 0.6.2 (protocol not changed; API changed)
3061 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3062 gras_error_t), and NULL when not found. Functions expecting a msgtype
3063 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3064 hopefully usefull message.
3065 - Portability to prehistoric sparcs again
3067 2004-08-17 Martin Quinson
3068 Version 0.6.1 (protocol not changed; ABI not changed)
3069 - prealloc some buffers to speed things up
3071 2004-08-11 Martin Quinson
3072 Version 0.6 (protocol not changed; ABI expended)
3073 - The parsing macro can deal with the references, provided that you add
3074 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3076 2004-08-09 Martin Quinson
3077 Version 0.5 (protocol not changed; ABI changed)
3078 - Allow to off turn the cycle detection code in data exchange at
3079 compilation time. It should be at run time, but I'm short of time (and
3080 the config stuff is still broken). That way, we keep dict out of the
3081 critical path, which is good because the performance is poor:
3082 - search not dichotomial yet
3083 - dynar give no way to access their content and memcpy everytime
3084 - In composed data description (struct, ref and so on), stop foolness of
3085 keeping the subtype's ID, but store the type itself. This keeps sets out
3086 of the critical path, which is good since they rely on dynar and
3087 dictionnaries. The only loose of that is that we cannot detect the
3088 redeclaration of a structure/union with another content (but I'm not sure
3089 the code detected well this error before anyway). We still can detect
3090 the redefinition discrepancy for the other types.
3091 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3092 it breaks the code because of type-punning used all over the place).
3093 This breaks on all non-gcc architectures (for now).
3095 All those changes (plus the buffer of last time) allow me to gain 2 order
3096 of magnitude on cruel tests consisting of 800000 array of integers on two
3097 level of a hierarchical structure (200 secondes -> 4 secondes)
3100 - the selector of reference must now return the type it points to, not
3101 the ID of this type.
3103 2004-08-06 Martin Quinson
3104 Version 0.4 (protocol changed; ABI not changed)
3105 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3106 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3107 - Implement a buffer transport, and use it by default (it relies on tcp in
3108 real life and on sg in simulation).
3109 That's a bit hackish since I had a new field to the structure to store
3110 its data without interfering with the subtype ones. Inheritance
3111 is tricky in C. And that's a kind of reverse inheritance with one class
3112 derivating two classes. Or maybe a game with java interfaces. Anyway,
3113 that's damn hard in C (at least).
3114 Moreover, I got tired while trying to ensure plugin separation and
3115 genericity in SG mode. MSG wants me to do weird things, so let's go for
3116 cruel hacks (temporarily of course ;).
3117 See comment in transport_private.h:71
3118 - do not include all the _interface headers in private but in the files
3119 which really need them (to cut the compilation time when they are
3122 2004-07-26 Martin Quinson
3123 Version 0.3 (protocol not changed; ABI changed)
3124 - Major overhault of the datadesc interface to simplify it:
3125 - shorted the function names:
3126 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3127 - add a trivial way to push/pop integers into the cbps without malloc.
3128 This allows to make really generic sub_type description, which simply
3129 pop their size of the stack.
3130 - add a function gras_datadesc_ref_pop_arr() which does what users want
3131 most of the time: Declare a dynamic array (which pops its size of the
3132 stack) and declare a reference to it. Poor name, but anyway.
3133 - kill the post-send callback, add a post-receive one
3135 2004-07-23 Martin Quinson
3136 Version 0.2 (protocol changed; ABI changed)
3137 - add some testing for cpbs in the test cases, and fix some more bugs.
3138 This invalidate again the little64 data file, since I cannot regenerate
3140 - remove an awfull optimization in the logging stuff, allowing me to:
3141 - understand it again
3142 - learn gcc how to check that the argument match the provided format
3143 - fix all errors revealed by gcc after that
3144 - internal keys of dict are not \0 terminated. Deal with it properly in
3145 loggings instead of segfaulting when the user want to see the logs :-/
3147 2004-07-22 Martin Quinson
3148 - Fix some stupid bug preventing cbps (callback postit) from working
3150 2004-07-21 Martin Quinson
3151 - Some documentation cleanups
3152 - remove the useless last argument of msgtype_declare
3153 - rename the Virtu functions to fit into the 'os' namespace
3154 - move headers src/include -> src/include/gras/ and stop fooling with
3155 gras -> . symbolic link
3156 - make distcheck is now successful
3158 2004-07-19 Martin Quinson
3160 - Build shared library also
3161 - Install html doc to the right location
3162 - stop removing maintainer files in make clean
3163 - build tests only on make check
3165 2004-07-13 Martin Quinson
3167 - No major issue in previous version => change versioning schema
3168 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3169 corresponding dataset.
3171 2004-07-11 Martin Quinson
3173 - Get it working with any kind of structure (we can compute the padding
3174 bytes remotely for all the architectures I have access to)
3175 - Implement the structure parsing macro (still not quite robust/complete)
3176 - Improvement to the remote testing toysuite
3178 2004-07-10 Martin Quinson
3179 [autoconf mechanism]
3180 - get ride of a bunch of deprecated macros
3181 - actually run the test for two-compliment, not only compile it :-/
3182 - test whether the structures get packed (and bail out if yes. Damn.
3183 Alignment is a serious matter)
3184 - test whether the structures get compacted (but respecting the alignment
3185 constraints of each types)
3186 - test whether the array fields of structures can straddle alignment boundaries
3188 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3189 (took me 2 hours to find that bug, looking at the wrong place)
3190 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3191 before sending/receiving and are used to compute the offsets of fields
3192 - Given that padding size depend even on compiler options, keep track of
3193 alignment and aligned_size only for the current architecture. Not a big
3194 deal since we send structure fields one after the other (seems
3196 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3197 program, let it work on linux/gcc/little32. portability todo.
3199 2004-07-08 Martin Quinson
3200 - import and improve remote compilation support from FAST
3201 - make sure make check works on half a dozen of machines out there
3203 2004-07-07 Martin Quinson
3204 Let's say it's version 0.0.3 ;)
3205 - Implement conversions (yuhu!)
3206 - Let it work on solaris (beside conversion, of course)
3207 - Stupid me, using rand() to generate the conversion datatests in not wise.
3209 2004-07-06 Martin Quinson
3210 - Let make dist work, since I'm gonna need it to compile on remote hosts
3211 - Let Tests/datadesc_usage write the architecture on which the file was
3212 generated as first byte.
3213 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3214 architecture descriptions.
3215 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3216 on those architectures.
3217 - Optimization: send/recv array of scalar in one shoot
3219 2004-07-05 Martin Quinson
3220 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3222 - Plug a whole bunch of memleaks
3223 - each process now have to call gras_{init,exit}. One day, their log
3224 settings will be separated
3225 - Continue the code factorisation between SG, RL and common in Transport.
3227 2004-07-04 Martin Quinson
3229 - Redistribution between SG and RL.
3230 We wanna have to accept in SG, so move accepted related parts of RL in
3231 the common part. (more precisely, the dynar of all known sockets is no
3232 more a static in transport.c, but part of the process_data)
3234 [gras_stub_generator]
3235 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3237 2004-07-03 Martin Quinson
3238 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3239 them all up in one shot)
3241 - Fix the ugly case of reference to dynamic array.
3242 - New (semi-public) function gras_datadesc_size to allow the messaging
3243 layer to malloc the needed space for the buffer.
3245 - gras_socket_close now expect the socket to close (and not its address to
3246 put NULL in it after it). This is because the socket passed to handlers
3247 is one of their argument (=> not writable).
3249 - propagate the interface cleanup from last week in datadesc, ie remove a
3250 superfluous level of indirection. User pass adress of variable
3251 containing data (both when sending and receiving), and not of a variable
3252 being a pointer to the data. Let's say that I like it better ;)
3253 The price for that is constructs like "int msg=*(int*)payload" in
3254 handlers, but it's a fine price, IMHO.
3256 - Let it work in RL (yuhu)
3258 2004-06-21 Martin Quinson
3260 - porting SG plugin and SG select to new standards (works almost).
3261 - plug memleaks and fix bugs around.
3264 - cleanup the prototype of data recv and force users to specify when they
3265 want to handle references to objects. Test case working even for cycles.
3266 - plug memleaks. Valgrind is perfectly ok with this.
3268 2004-06-12 Martin Quinson
3270 - cleanup the separation between plugin and main code in plugin creation
3272 2004-06-11 Martin Quinson
3274 - Reput hook for raw sockets, needed for BW experiments
3275 - kill a few lines of dead code
3276 [Data description] Interface cleanup
3277 - gras_datadesc_by_name returns the searched type or NULL.
3278 That way, no variable is needed to use a type desc once, which makes
3280 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3281 two parameters were strings (field name, type name), leading to
3283 [Dicos] Interface cleanup
3284 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3285 This is consistant with the dynar API.
3287 2004-04-21 Martin Quinson
3289 - Porting to new standards.
3291 - interface cleanup.
3292 There is no bag anymore, no need to take extra provision to mask the
3293 pointers behind "ID".
3294 Better splitup of functions between files create/exchange/convert.
3295 This is still a bit artificial since convert and receive are so
3296 interleaved, but anyway.
3298 - add a queued message list to procdata (the ones not matching criteria
3300 - factorize some more code between SG and RL wrt procdata
3302 - use gras_exit in example to track memleaks
3303 - get rid of gs_example now that GS is properly integrated into gras
3304 - update run_test to integrate the latest tests (datadesc)
3306 - rename WARNINGn macros to WARNn since it prooved error-prone
3308 2004-04-19 Martin Quinson
3310 - register init/exit functions within gras module mechanism
3311 - send/receive function.
3312 Convertion is not implemented, but short-cutted if not needed.
3313 struct/array elements are sent one by one (instead of block-wise), but
3314 nobody really cares (yet). Get a prototype before optimizing.
3315 - tests (using a file socket) for DD send/receive on:
3316 - base types: int, float
3317 - array: fixed size, string (ie ref to dynamic string)
3318 - structure: homogeneous, heterogeneous
3319 - chained list, graph with cycle
3320 Believe it or not, valgrind is not too unhappy with the results. The
3321 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3325 - Bugfix when using a filename explicitely (instead of '-')
3327 2004-04-09 Martin Quinson
3329 - factorize more code between RL and SG in socket creation
3330 - Complete the implementation and tests of:
3332 o file (only in RL, and mainly for debugging)
3334 I lost 3 days to design a portable address resolver, and then decided
3335 that the prototype mainly have to run on my box.
3336 Addressing portability too early may be like optimizing too early :-/
3338 - use gras_init in the Tests instead of the crappy parse_log_opt
3339 (the latter function is removed)
3340 [Conditional execution]
3341 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3342 [Code reorganisation]
3343 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3344 Build examples against the RL library, since there is no way to disable
3345 its creation for now.
3347 For information, the beginning of coding on GRAS was back in june
3348 2003. I guess that every line has been rewritten at least twice since