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 - The netzone are now available from the MSG API.
35 The old names still work, but are now deprecated.
38 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
39 - New option: smpi/keep-temps to not cleanup temp files
40 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
43 - Replay: New function xbt_replay_action_get():
44 Retrieve the function previously associated to an event type.
45 - DROPPED FUNCTION: xbt_str_varsubst()
46 - DROPPED MODULE: strbuff. We don't need it anymore.
47 - DROPPED MODULE: matrix. We don't need it anymore.
48 - DROPPED MODULE: lib. We don't need it anymore.
50 -- Release target: June 21 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
52 SimGrid (3.15) stable; urgency=low
54 The Spring Release: continuous integration servers become green
55 We fixed even the transient bugs on all target architectures:
56 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
57 Windows, FreeBSD, NetBSD.
59 - Rename NetCards to NetPoints.
60 This was intended to help NS3 users, but that's not a netcard.
61 That's a point in the routing algorithm, let's avoid wrong simplifications.
64 - New: MSG_process_yield(). Stop and yield to other processes.
65 - New: MSG_process_daemon(). Daemon processes are automatically killed
66 when the last non-daemon process terminates
67 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
68 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
69 to make room for the upcoming network energy plugin.
70 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
73 - Ensure that an actor can kill itself with Process::exit()
74 - Kill the obscure NativeException. Nobody want to survive the issues
75 it denotes, so use JniException that is a RuntimeException (not to
76 be caught explicitly).
77 - Partial bug fix in initialization. SimGrid flags on command line were
78 consumed at C level but stayed in the original Java String[] args.
79 This could mess users' args[i] if SG flags were not put at the end of
81 The SimGrid flags are now removed from the Java arguments. However,
82 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
83 if args.length is greater than the number of YOUR OWN ARGUMENTS.
84 It might be if you have --log or --cfg flags in the command line.
85 - Fix numerous memleaks all around the place. In particular, around VMs.
89 - s4u::onPlatformCreated: right before the simulation starts
90 - s4u::onSimulationEnd: right after the main simulation loop
91 - s4u::onTimeAdvance: right after a clock change (time is discrete)
92 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
93 event from the availability_file changes the avail speed.
94 - Links are now usable from s4u
95 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
96 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
97 - New examples: The conversion of MSG examples to S4U has begun
98 - Actors: create, kill, migration, and suspend
99 - Applications: master-worker and token-ring
100 - Action replay: communications and storage
101 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
104 - Backwards Compatibility breaks
105 - The SD_route_get_list and SD_route_get_size functions have been removed.
106 They are replaced by the unique
107 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
108 The route from src to dst is built in the links dynar whose size can
109 be obtained with xbt_dynar_length.
110 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
111 sg_host_route_bandwidth and sg_host_route_latency.
112 Macros ensure the backwards compatibility, but you should fix your code
115 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
116 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
117 - Bump our claim of support from MPI 1.1 to MPI 2.2.
118 We don't support 100% of it, but it should be enough. Contact us if not.
119 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
120 naive and probably inaccurate)
121 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
122 mapped onto a unique small file using some system magic.
123 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
124 - Fortran: cleanups, fixes, support of user-added operations
125 - RMA: MPI_Accumulate are applied in correct order.
126 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
127 - Better support for MPI_IN_PLACE
128 - Support for MPI_Win attrs and keyvals.
129 - Support MPI_Comm_group_create, MPI_Type_size_x
130 - MPI ops should be properly applied to all allowed types
131 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
132 algorithm by a non failing one from mpich.
135 - Kill the fifo data container: we don't use it anymore.
138 - A <cluster> can now be created with different speed values to represent pstates.
139 see examples/platforms/energy_cluster.xml for an example.
141 -- Wed Mar 22 17:50:21 CET 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
143 SimGrid (3.14.159) stable; urgency=low
145 The previous release was only a crude approximation of a Pi release;
146 this one should be a bit better.
148 - Revert a last minute change that broke on Mac OSX.
149 - Fix the changelog and NEWS files.
150 - Improve the documentation, in particular of the routing module.
151 - Integrate some patches flying around in Debian.
153 -- Wed Dec 28 17:14:49 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
155 SimGrid (3.14) stable; urgency=low
158 * The whole documentation was reworked and reorganized. There is
159 still much room for improvement here, but we're on it.
162 * We now need python3 for our testing infrastructure.
163 * Model checking needs libevent
166 * Now works on FreeBSD
169 * Privatization now works on FreeBSD
171 * Privatization is now activated by default in smpirun
172 You can turn it off with -no-privatize if something goes wrong.
174 * Call-location tracing for SMPI.
175 You can add the exact location (filename / linenumber) of an MPI call to
176 your trace files and slow down or speed up the simulation between two
177 consecutive calls by using an adjustment file (see the documentation).
179 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
180 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
181 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
182 values were ignored in some cases. The timings of these functions can now
183 be significantly different.
185 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
186 smpi/running-power is renamed to smpi/host-speed
188 * smpi/grow-injected-times option to enable or disable multiplication of the
189 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
190 make simulation less precise (but also much faster).
192 * smpirun script should be (much) faster for large deployments.
194 * SMPI tracing : fixed issue with poor matching of send/receives.
196 * Replay : Fix broken waitall
198 New functions and features
199 * MSG_parallel_task_execute_with_timeout, to timeout computations.
201 Dropped / renamed functions and features
202 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
203 - MSG_mailbox_is_empty() -> Mailbox::empty()
204 - MSG_mailbox_front() -> Mailbox::front()
205 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
206 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
207 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
208 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
209 * Don't pass the free_f parameter to property related functions:
210 - MSG_host_set_property_value()
211 - MSG_as_router_set_property_value()
212 - MSG_storage_set_property_value()
213 * VM properties. Since msg_vm_t are msg_host_t, just use
214 MSG_host_get_property_value() and friends
215 * VM I/O related things:
216 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
217 - Unimplemented save/restore methods
218 * MSG_as_router_get_property_value() was redundent with
219 MSG_environment_as_get_property_value().
220 - Removed MSG_as_router_*propert*() functions
221 - Added MSG_environment_as_set_property_value() for consistency
222 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
223 Well, they are still there for now, but deprecated with a warning.
224 Please switch to xbt_dict_new_homogeneous() before this is removed
226 * Task affinity. Its intended behavior (that was very badly tested
227 and probably not really working) was deceiving what most users
228 would have hoped here.
229 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
230 * Option network/coordinates is now useless and should be dropped.
233 * Added option storage/max_file_descriptors to allow more than 1024 files opened
236 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
237 * Added Dragonfly topology support
240 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
241 EiBps, EBps, Eibps, Ebps for bandwidth.
242 They may become useful to some lucky ones.
245 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
246 * ex/app_tokenring: new example, very similar to the MSG Token Ring
247 * ex/async_waitAll: new example, on asynchronous communications
250 * Memory usage should be decreased for simulations with a large number
251 of processes. This also helps for SMPI.
253 -- Sat Dec 24 22:34:26 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
255 SimGrid (3.13) stable; urgency=low
257 The Half Release, a.k.a. the Zealous Easter Trim.
259 - We removed half of the lines, that were mostly experimental cruft.
260 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
261 (not counting blanks and comments -- according to openhub.net)
262 - The internals are now compiled in C++ (and will soon be clean C++)
263 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
264 59 klines of C, etc. We added only 29 klines of C++ in replacement.
266 * Backwards Compatibility breaks
267 - Removed Lua simulation bindings (switch to C or Java for that).
268 Lua can still be used to describe platforms
269 - Removed Java kernel plug-ins.
270 Will be reintroduced after the ongoing major internals reorg.
272 - the following functions were removed.
273 They were too specific and should be reimplemented in a generic
274 way, with filter function.
275 - MSG_task_listen_from_host
276 - MSG_mailbox_get_count_host_waiting_tasks
277 - MSG_mailbox_put_with_timeout was removed.
278 Please use MSG_task_send_with_timeout instead.
280 - the SD_application_reinit function was removed. It has been a noop for a while.
281 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
282 replaced by a more flexible mechanism.
283 - The following functions thus do not exist anymore
284 - SD_workstation_get_access_mode
285 - SD_workstation_set_access_mode
286 - SD_workstation_get_current_task
287 - Basic estimation functions have been removed but can easily be replaced
288 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
289 - SD_workstation_get_computation_time => amount / sg_host_speed()
291 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
292 Use VM.getSpeed()*load/100 for the legacy behavior.
294 - option enable_tracing was removed. It was not doing anything for a while.
295 - In the ModelChecker:
296 - the model-checker now ptraces the model-checked process which means
297 you cannot use a debugger on the latter anymore (we might make this
298 optional in the feature);
299 - removed soft-dirty page tracking;
300 - remove model-checked side snapshot management,
301 MC_snapshot() and MC_compare_snapshot();
302 - keep the MC_cut() function as a stub (it was not really working
303 in the previous release).
306 * All options are consistently in kebab-case. Old names are kept as alias.
309 * Switch to platform v4 format.
310 - Rename from 'power' to 'speed' the attributes describing the amount of
311 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
312 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
313 - In <host> and <link>, attributes availability and state are gone.
314 It was redundent with state and availability traces, and with peak values.
315 - In <cluster>, attributes availability_file and state_file are gone.
316 It was too complex and unused.
317 - Kill <gpu>. Was not doing anything.
318 - The DOCTYPE points to the right URL:
319 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
320 (the file at this address now documents the changelog since its v1)
321 - A warning is emitted for unit-less values (they are still accepted).
322 - speed. Default: 'f' or 'flops'. Also defined:
323 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
324 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
325 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
326 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
327 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
328 - latency. Default: 's' second. Also defined:
329 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
331 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
332 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
335 * s4u::Host is now the preferred public interface to the Host features.
336 sg_host_* functions are C bindings to the exact same behavior
337 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
340 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
341 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
342 They were deprecated since a few years, and probably did not even compile anymore.
345 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
346 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
347 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
348 encouraged to update their codes. A list of the modified functions can be found at the end of
349 include/simgrid/simdag.h
352 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
353 the lambda or closure passed as a parameter will run in kernel mode.
354 All the callback functions should be rewritten to that interface at some point.
357 * Reorganizing and cleaning the internals all around the place.
360 * Remove old default barrier/bcast buggy algorithms (see #18407)
361 * Various bug fixes to handle more codes
362 * Remove the need for the --foreground option of smpirun (it is still
363 accepted for backward compatibility).
366 * Kill the setset data container: MC don't use it anymore.
367 * Kill the queue data container: it made more sense with GRAS.
368 * Kill the xbt_peer_t data type: it's useless without GRAS.
369 * Kill rm_cb feature of config sets: it was never useful.
370 * Kill graphxml parsing feature. It was not used.
371 * Kill the deprecated code protected by XBT_USE_DEPRECATED
373 - xbt_dynar_sort_strings(), when the content is char*
374 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
375 They throw exceptions on invalid input;
376 * C++ support for declaring CLI flags (simgrid::config::Flag);
377 * class for abstracting different signal backends (simgrid::xbt::signal).
378 with no external dependencies (we need very simple signals).
381 * refactoring and cleanup of the code;
382 * ongoing process to cleanly separate the model-checking algorithms
383 from the code model-checking support.
385 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
387 SimGrid (3.12) stable; urgency=low
389 The Facelift Release.
392 * Require g++ v4.7 at least to not speak prehistorical C++.
393 * Require Boost 1.48 (for signal2 component).
394 * Java must be version 7 at least when activated.
395 * Builds on Windows again (including Java bindings).
396 * Tracing is now always enabled (no way to turn it out)
397 * Remove GTNetS. It was not working anyway.
398 * Various cleanups in the cmake scripts.
399 * Move headers around to sort them out on installed systems:
400 - instr/instr.h -> simgrid/instr.h
401 - instr/jedule/* -> simgrid/jedule
402 - simdag/datatypes.h was removed
403 - simdag/simdag.h -> simgrid/simdag.h
404 - msg/datatypes.h was removed
405 - msg/msg.h -> simgrid/msg.h
408 * Interface improvement:
409 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
410 - Sanitize the interface in MSG_task_ module:
411 - Merge two functions that were close enough but misleading:
412 set_compute_duration(t) -> set_flops_amount(t)
413 get_remaining_computation(t) -> get_flops_amount(t)
414 - set_data_size(t) -> set_bytes_amount(t)
415 get_data_size(t) -> get_bytes_amount(t)
416 - Massive cleanups in the functions related to the energy
417 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
418 - New: MSG_host_get_pstate()
420 - msg/energy/onoff: switching hosts on and off
422 * Interface improvement:
423 - Rename Host.isAvail() to Host.isOn()
424 - Rename Process.currentProcess() to Process.getCurrentProcess()
425 - Rename Task.setDataSize() to Task.setBytesAmount()
426 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
428 - #18874: Actually allows the GC to reclaim tasks
432 - SIMIX_process_throw: raises an exception in a remote process
433 * Refactoring: Separate sub-modules
434 - libsmx: the public interface, as libc in a real system
435 - popping: the strange dance that converts a user request into a kernel handling
436 - smx_context_*: the virtualization mechanisms that embed the user code
437 - smx_*: the handling of each simcalls
439 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
440 * Rename smx_action_t into smx_synchro_t, making explicit that these
441 things are used to synchronize processes with their environment.
442 For example, a communication is a sort of synchronization involving
443 the communicating processes (that may block until the exchange) and
444 the platform. The same can be said from computations, etc.
446 - #18888: segfault when a process ends before its kill_time
450 - 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.
451 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
452 - MPI_Type_set_name, MPI_Type_get_name
453 - MPI_*_c2f and MPI_*_f2c functions
454 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
455 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
456 - Activate a lot of new tests from the mpich 3 testsuite
458 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
459 - InfiniBand network model added : Based on the works of Jerome Vienne
460 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
461 - When smpi/display_timing is set, also display global simulation time and application times
462 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
463 * Collective communications
464 - 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
465 - 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/).
466 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
468 - Replay now uses algorithms from wanted collective selector
469 - Replay can be used with SMP-aware algorithms
470 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
471 - 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.
473 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
474 - Correctly use loopback on fat-tree clusters
475 - Asynchronous small messages shouldn't trigger deadlocks anymore
476 * Energy/DVFS cleanup and improvement
477 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
478 - new: smpi_get_host_pstate()
481 - "Full" network optimization flag was broken since Surf++
482 - Better handling of precision flags in maxmin
483 - Fix bug causing sometimes "Impossible" errors
484 - Properly pass cluster properties to included hosts
485 * Improvement of the Energy plugin.
486 - Always update the consumption before returning that value
487 - New property: watt_off to denote the dissipation when the host is off
488 - New functions getWattMinAt and getWattMaxAt to retrieve the
489 dissipation of pstates that we are not currently at.
490 * Java: class NetworkLink renamed to Link
491 * New function: simcall_process_get_kill_time()
492 * Massive rename s/workstation/host/
493 - That's intrusive, but that's good for the project consistency. Sorry.
494 - Change config option "workstation/model" into "host/model"
497 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
498 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
499 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
500 - xbt_dirname and xbt_basename for non-POSIX systems
502 * The model checker now runs as a separate process.
503 * The model checker runs must now be launched with the new simgrid-mc program.
504 * Record/Replay: the MC can display a textual representation of a path in the
505 execution graph. It can then be replayed outside of the model checker.
507 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
509 SimGrid (3.11) stable; urgency=low
514 * Normalizing pointers addresses tool for better diff between logs
517 * Add cloud examples using new VMs
518 - examples/msg/cloud/two_tasks_vm.tesh
519 - examples/msg/cloud/simple_vm.tesh
520 - examples/java/cloud/cloud.tesh
521 - examples/java/cloud/migration/migration.tesh
522 * Add java surf examples:
523 - examples/java/surfPlugin/surf_plugin.tesh
524 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
525 - examples/java/surfCpuModel/surf_cpu_model.tesh
526 * Add SMPI+MSG example:
527 - examples/smpi/smpi_msg_masterslave/
532 - msg task destroy cancel test
533 - msg_host on/off test
534 * Move all tests in testsuite to teshsuite (adding tesh files)
535 * Restructure teshsuites
536 - one folder for each kind of test
537 * Restructure AddTests.cmake
539 - structure the order of tests (with sections)
542 * Add virtual machine
543 - creation of a VM on a PM
544 - migration of a VM from a PM to another PM
546 - MSG_process_join(msg_process_t process, double timeout)
547 - msg_bar_t MSG_barrier_init(unsigned int count)
548 - int MSG_barrier_wait(msg_bar_t barrier)
549 - void MSG_barrier_destroy(msg_bar_t barrier)
550 - msg_as_t MSG_environment_as_get_by_name(const char * name)
551 * New option "msg/debug_multiple_use" to help debugging when a task is used
554 - Improvements and finalization of MSG_storage, MSG_file APIs and their
556 - Increase code coverage in test suites
560 * Protect context stacks against stack overflow. The number of protected memory
561 pages allocated on the top of each stack (1 by default) can be configured
562 with the new command line option --cfg=contexts/guard_size:<value>.
563 * Simcalls are now generated by a python script that
564 - generates files included by SimGrid
565 - checks that all the functions exist, and proposes prototypes
567 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
568 host_set_data, host_get_data
570 - simcall_process_join(smx_process_t process, double timeout)
571 * Fix bug where sleeping processing could not be suspended.
574 * Translate surf models from C to C++
575 - Generic classes for all models: Model, Resource, Action
576 - A generic interface for each kind of model (CPU, Network, Storage
577 Workstation, WorkstationVM)
579 * Translate surf routings from C to C++
580 * Add callbacks using sigc++ or boost::signals2
581 - Add callback functions for resource creation/destruction
582 - Add callback functions for action state change
583 - Handle Energy as a plugin
584 * Replace swag by boost::intrusive
585 * Add new routing models for clusters. For documentation, see
586 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
587 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
588 parameters for cluster tag
589 - Fat trees, with topology="FAT_TREE" and
590 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
591 - see examples/platforms/torus_cluster.xml and
592 examples/platforms/fat_tree_cluster.xml
596 * Hostfiles support host:nb_processes construct to deploy several processes on
598 * Collective communication algorithms should not crash if used with
599 improper number of nodes and report the error.
600 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
601 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
602 MPI_Dims_create, MPI_Cart_sub are supported.
603 * New interface to use SMPI programmatically (still depends on MSG for
604 some parts, see examples/smpi/smpi_msg_masterslave) :
605 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
609 * Global variables privatization in MPI executables is now performed at runtime
610 with the option smpi/privatize_global_variables (default:no).
611 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
612 dynamic libraries loaded by the application are not privatized (static
613 linking with these libraries is advised in this case)
616 * Options defined in XML work correctly now.
619 * New cmake option, enable_lib_in_jar, to control whether native libraries are
620 copied into simgrid.jar or not (ON by default). Use this option if you want
621 to reduce the size of the installed simgrid.jar, *and* the native libraries
622 are kept installed elsewhere.
623 * Surf binding with SWIG (code generated in maintainer mode only):
624 - plugin to handle callbacks
625 - CPU model only for the moment
628 * Supernovae build mode is definitively removed. It was used to improve
629 inlining and inter-module optimizations. It is nowadays superseded by
630 link-time optimizations commonly available in compilers.
631 * Update ns-3 find lib. Bindings for ns-3 should work again now.
632 * Add boost dependency for surf++
633 * Add new macro for tests
634 - ADD_TESH(name <tesh_args>)
635 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
639 - xbt_bar_t XBT_barrier_init(unsigned int count)
640 - int XBT_barrier_wait(xbt_bar_t barrier)
641 - void XBT_barrier_destroy(xbt_bar_t barrier)
642 * Make the xbt_os_time module public
644 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
646 SimGrid (3.10) stable; urgency=low
648 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
651 * Reintegrate Java to the main archive as desynchronizing these
652 package is not acceptable anymore (Java is now considered stable)
653 * Add explicit synchronization facilities through semaphores
654 * Bug fix: Task.setDataSize() only changed the C world, not the value
655 cached in the Java world
658 * Dramatically change the way files are handled. API and internals changed, but
659 this part of MSG was not considered as production grade either.
660 * Add explicit synchronization facilities through semaphores
661 * Add a new function MSG_host_get_process_list()
662 * Preliminary DVFS support (see examples/msg/energy/ for details)
665 * SMPI is now included directly in the libsimgrid as the windows
666 linker doesn't force us on splitting it anymore.
667 * Improvements of the SMPI replay tool:
668 - Most of the collective communications are now rooted in the same process as
669 in the original application.
670 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
671 used until now). Multiple data types can now be used in a trace.
672 - The replay tool now supports traces produce either by TAU or a modified
674 - Bug Fix: the compute part of the reduce action is now taken into account.
675 - Gatherv collective is now supported
676 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
677 of smpirun outputs time independent traces for the current run. One file
678 is created per process. If too many processes are simulated, this behavior
679 can be changed to one file for all processes by using the
680 tracing/smpi/format/ti_one_file flag
681 * smpirun generates the host file if needed (with given host count and platform)
682 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
683 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
684 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
685 --cfg=smpi/coll_selector:(mpich/ompi)
686 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
688 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
689 happening inside a collective SMPI call.
690 * Fix the behavior of complex data types handling.
691 * Make MPI_Wtime another synchronization point to take computations into
693 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
694 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
695 * Add all missing Fortran bindings, SMPI should work with Fortran 90
696 (no privatization of global variables yet)
697 * Preliminary DVFS support (see examples/smpi/energy/ for details)
700 * Verification of liveness properties is now available for SMPI applications
701 (in addition to MSG applications)
702 * Bugged examples using SMPI in examples/smpi/mc/
703 * Add --cfg=model-check/visited option. Allows the verification of infinite
704 programs. Detection of loops in the execution thanks to the system state
705 comparison and reduction of the state space to explore. Can be combined with
706 DPOR for safety properties.
709 * Allow to change SimGrid configuration (see --help) within the code
710 thanks to SD_config() as it can be done in MSG.
711 * Add a new function SD_task_set_amount() upon user request.
714 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
715 * Remove rule based routing (no more PCRE dependency)
716 * Add a limiter_link option to cluster tag, to specify a maximum reachable
717 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
718 * Add a loopback_bw and loopback_lat options to cluster tag.
719 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
720 routing. Now peers are encapsulated in an AS and have their own private
721 router but this is transparent.
724 * Our own implementation of getline is renamed xbt_getline, and gets
725 used even if the OS provide a getline(). This should reduce the
726 configuration complexity by using the same code on all platforms.
727 * New type: xbt_cfg_elm_boolean.
728 * Allow to use yes/no for boolean configuration options in the command line.
729 * Allow to disable SimGrid cleanups at exit from command line option.
730 There are situations where one may want a simulation to end with an exit.
731 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
732 annoying when scripting around the simulator. Adding a
733 --cfg=clean_atexit:no allows to circumvent this issue.
736 * Lots of memory leaks were corrected in this release.
737 There is no known memory leaks anymore, in all of our 600+ tests.
738 * New command line option --version, to get SimGrid version information.
739 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
741 * Supernovae builds are deprecated, and expected to be removed in the next
744 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
746 SimGrid (3.9) stable; urgency=low
748 The Grasgory release: GRAS is really dead now.
750 * Complete overhaul of the internal host structures scheme.
753 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
754 was considered as experimental and badly maintained since then.
755 * Keeping it was thus a trap to our potential users, that could take
756 it instead of MSG or SMPI by mistake despite is pity state.
757 * GRAS seems to have very few users (if any), and no one volunteered
758 to maintain it further. It also induces a lot of XBT code (for
759 portability sake), that must be maintained too.
760 * For all these reasons, we killed GRAS. If someone wants to revive it
761 in the future, don't cry, our git history still remembers of GRAS.
764 * Major overhaul. Merge our documentation again as time proved that
765 splitting it was really not helping our users.
766 * Further improve the developer documentation to help newcomers
767 hacking on SimGrid itself. The user documentation (and in
768 particular, the beginner documentation) is still in a sorry state.
771 * Now works on Windows too!
772 * Much more extensive test suite, from MPICH
775 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
776 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
777 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
778 * Bug fix: let task be scheduled when the last dependency to be solved is
779 a control dependency.
780 * Remove SD_load_environment_script function.
781 Use the C sg_platf function if you want to declare a platform
785 * New function: MSG_process_get_number()
786 * Old function documented: MSG_config()
787 * Remove MSG_load_platform_script function
788 Use the C sg_platf function if you want to declare a platform
792 * Change the default value of the TCP_gamma constant (maximal size of TCP
793 congestion window) to a more realistic 4MiB value. If you notice changes in
794 your simulation results, you can fall back to the previous 20k tiny window
795 by adding --cfg=network/TCP_gamma:20000 on command line.
796 * (Hopefully) fix a bug wrt periodic availability and state traces
797 * Bug fix: use default values at start when first event in availability/state
798 trace is not at time 0.
801 * remove the "new_" part of function name sg_platf_new_trace_connect
802 (resulting in sg_platf_trace_connect), since it does not create
806 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
807 use case where it's really mandatory, and maintaining it was a pain in
809 * New: xbt_fifo_search(), search an item with a user-provided
810 comparison function instead of dumb pointer comparison.
813 * Fix the lua deployment:
814 Use `simgrid.init_application()` before deployment instead of
815 `simgrid.msg_register_application()` after.
818 * Transfer the tracing files into the corresponding modules.
820 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
822 SimGrid (3.8.1) stable; urgency=low
824 The "we are told that some people want to also *install* the simgrid
827 * Add missing file "tesh.1" to the archive.
829 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
831 SimGrid (3.8) stable; urgency=low
833 The Psssshiiiit release: SimGrid jumps into the Cloud.
836 * Add an experimental interface to manipulate VMs. They are mainly
837 process groups with very few intrinsic semantic, but they should
838 allow you to build the semantic you want easily.
839 * New function: MSG_host_set_property_value()
840 * New function: MSG_process_on_exit(). To clean memory in all cases.
841 * Bug fixes that made the host (and link) failures unusable.
842 * Add a way to auto-restart process when the host in which they are
843 executing comes back (ON_FAILURE="RESTART" on deployment file,
844 MSG_process_auto_restart_set).
845 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
846 please stop using the old ones, they are DEPRECATED.
848 * Deprecate functions MSG_global_init() / MSG_global_init_args()
849 Please use MSG_init() instead. (reducing the amount of entry
850 points in the library helps us).
851 * Make it impossible to link against the wrong version of the lib
852 * Deprecate MSG_clean(). No need to call it anymore.
853 * Function MSG_get_host_number() is not deprecated anymore.
856 * Split the doc into a user guide and a reference guide.
857 * Start a developper guide to help people hacking on SimGrid.
860 * Enable tracing by default. This modules rocks you should use it.
861 * Remove option custom_flags. Now use environment variables CFLAGS
863 * Use default cmake things to detect lua instead of home grown ones.
864 * New option "enable_mallocators" to disable mallocators, for debugging
865 purpose ("on" by default).
868 * Bug fixes around the resource failures: don't let the processes
869 survive the host they are running onto.
870 * Add an interface to auto-restart processes when the host in which they are
871 executing comes back.
872 * Ensures that SIMIX_clean is called automatically. It's not part of
873 the public interface anymore (bindings should be updated).
876 * Bug fix for when SD_Simulate is called with a positive value: be careful
877 when comparing doubles. Sometimes they are different for non significant
879 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
880 parallel task whose initial work is distributed among host according
881 to the Amdahl's law. Such tasks are created with a parameter alpha
882 that corresponds to the non-parallelizable part of the computation.
883 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
884 between two sets of workstations assuming a 1D block distribution (each
885 workstation owns a similar share of data) on both sides.
887 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
888 Data redistribution will be automatically scheduled once parent and child
889 are both scheduled. The filling of computation_amount and
890 communication_amount structures is now done seamlessly thanks to the chosen
892 * New function SD_workstation_dump to display various information
893 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
894 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
895 bandwidth on the route onto which the task is scheduled and the amount of
897 To divide the nominal bandwidth by 2, the rate then has to be :
898 rate = bandwidth/(2*amount)
899 * Compute tasks that have failed can now be rescheduled and executed again
900 (from their beginning)
901 * Increasing source code coverage (src/simdag is now covered at 95.8%
905 * Re-implement time-independent trace replay using SMPI (at the
906 smpi_smp_* level) instead of MSG. This should replace
907 examples/msg/actions/actions.c
908 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
909 hindexed and structs)
910 * Implement the exchange of non-contiguous data.
911 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
912 * Correct behavior of smpi/sender_gap and set its default value to 0
913 * Add option to asynchronously send small messages to allow better
914 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
915 specifies the size in bytes under which messages will be asynchronously sent.
916 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
917 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
920 * New C interface to define a platform: XML is now optional.
921 For more info, please check include/simgrid/platf.h.
922 * New interface to define random platforms from the C:
923 For more info, please check include/simgrid/platf_generator.h and
924 examples/msg/masterslave/masterslave_platfgen.c
925 * Export a sg_cmdline dynar containing all the arguments we got from
929 * Two new tracing options for adding comments to trace file so you
930 can track your experiments (see --help-tracing for details).
931 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
932 * Adding the SimGrid version that generated the trace file as a comment.
933 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
934 * Fix to avoid key clashes on Paje links
935 * Other minor fixes related to the Paje specification
938 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
939 and renamed to xbt_str_hash() and xbt_str_hash_ext().
940 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
941 * Greatly improve the robustness of mmalloc to user errors (such as
942 using an area after freeing it, or freeing it twice)
944 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
946 SimGrid-java (3.8.1) stable; urgency=low
948 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
949 Warning: all methods are visible, but only some of them are
950 implemented so far. Check the source (src/jtrace.c)
951 for further information.
952 * New module: org.simgrid.msg.File (SimGrid File management functions)
953 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
954 * Change the meaning of Process.restart: now restart the process from
955 the begining, like MSG_process_restart in C.
956 * Add Process.setAutoRestart: handling of process restart when failed
958 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
959 you to retrieve the properties of the processes/hosts
960 * Deprecate Msg.clean(): you can just forget about it now.
961 * New function Process.getCount(), that only works when compiling
962 with the not yet released version 3.9 of the C library.
964 * New context factory based on Coroutines. It mandates a modified JVM
965 but then, the simulations run about five times faster, and there is
966 no limit to the amount of processes (beside of the available memory).
968 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
970 SimGrid (3.7.1) stable; urgency=low
973 * Restore the prototype of MSG_process_create_with_environment() to
974 the pre-3.7 situation by removing the kill_time argument.
975 * Add a MSG_process_set_kill_time() function instead.
978 * Fix weird behaviors when dealing with parallel tasks.
981 * Simgrid is now built as a dll.
982 * Simgrid-java now works on Windows.
983 * Simgrid-Java is now included into Windows package.
986 * First pre-build package for MacOSX.
989 * Fix compilation when using MSG_USE_DEPRECATED.
990 * Fix some compilation issues on Macs and Windows.
991 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
992 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
994 We discovered that the Lua console is broken, but we are missing the
995 manpower to fix it right now. The problem existed in 3.7 too, so we
996 are not blocking the release for that. Sorry if you depended on this
997 feature, any help would be really welcome.
999 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1002 SimGrid-java (3.7.1) stable; urgency=low
1004 The "Java aint got to be bloated and slow" release
1007 * Various internal cleanups and performance improvement
1008 Simulations are expected to run up to twice faster or so
1009 * Make Process.kill(process) an instance method, not a static one
1010 * User processes are not java.lang.Thread subclasses.
1011 This breaks the compatibility (sorry), but previous API was
1012 brain-dead, making it impossible to have non-trivial
1013 initializations in the process constructor.
1014 * Require a full constructor per Process sub-class.
1015 Kinda breaks the compatibility (sorry), but this allows a much more
1016 efficient way to launch the processes at simulation startup.
1017 * Do not embeed our version of semaphores, java 1.5 can be considered
1018 as sufficiently prevalent for us to not dupplicate its features.
1022 * Add examples for almost every part of the API
1023 We spotted and fixed a lot of bugs in the process
1025 * New module: asynchronous communication API
1026 * New function: Process.sleep()
1027 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1028 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1029 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1031 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1033 SimGrid (3.7) stable; urgency=low
1035 The "spring cleanups (before the next Big Project kicks in)" release.
1038 * We can specify the SMPI latency/bandwidth factor with command line
1039 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1040 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1041 You can also use the "config tag" from platform file by adding this line
1042 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1043 (see "example/platforms/tag_config.xml" to use "config tag").
1044 Note that the command line supersedes the platform file configuration.
1045 * Change the correction factors used in LMM model, according to
1046 the latest experiments described in INRIA RR-7821.
1047 Accuracy should be improved this way.
1048 * Use the partial invalidation optimization by default for the
1049 network too. Should produce the exact same results, only faster.
1050 * Major cleanup in surf to merge models and split some optimization
1051 mechanisms from the core of the models. As a result you can now
1052 specify which model to use (e.g., --cfg=network/model:LV08
1053 --cfg=cpu/model:Cas01) and which optimization mode to use
1054 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1055 Incompatible combinations should err at initialization. See
1056 --help-models for the list of all models and optimization modes.
1057 * The CLM03 workstation model was dropped for simplicity because it
1058 used the deprecated CM02 network model. Use default instead.
1059 * Rename the TCP_gamma configuration option to network/TCP_gamma
1060 * Rename the coordinates configuration option to
1061 network/coordinates, and document it
1062 * Use now crosstraffic keyword instead of the terribly misleading
1063 fullduplex keyword. It is activated by default now in the current
1064 default model, use --cfg=network/crosstraffic:0 to turn it off.
1065 * Ongoing refactoring the model parsing to make XML files optional
1066 See include/simgrid/platf.h for details (still to be completed)
1069 * Major overhaul of the documentation. Almost instructive now :/
1070 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1071 functions and friends. This interface was considered as
1072 deprecated since over 2 years, it's time to inform our users that it is.
1073 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1074 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1075 use these (crappy) functions in your code.
1076 These functions will be removed soon. Stop using them now.
1077 * Deprecate MSG_get_host_{table,number}
1078 Implement MSG_hosts_as_dynar() instead.
1079 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1080 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1083 * Stabilize the parallel execution mode of user contexts
1084 * Introduce configuration variables to control parallel execution:
1085 - contexts/synchro: Synchronization mode to use when running
1086 contexts in parallel (either futex, posix or busy_wait)
1087 - contexts/parallel_threshold: Minimal number of user contexts
1088 that must be part of a scheduling round to switch to parallel
1089 execution mode (raw contexts only)
1090 * Fix bugs that prevented to use suspend/resume along with
1091 synchronization structures.
1092 * Fix bugs in process termination that lead to invalid memory access
1093 in very specific conditions.
1096 * Introduce a parallel mode for the models (controlled by surf/nthreads
1097 configuration item). In our tests, running the models in parallel
1098 never lead to any speedups because they are so fast that the gain
1099 of computing each model in parallel does not amortizes the
1100 synchronization costs, even when ultra fast futexes are used.
1101 This is released anyway because YMMV.
1104 * Performance boost by using a swag internally to compute the set of
1105 tasks that are finished and should constitute the return value of
1109 * Enable it by default now that it is considered rather stable.
1112 * Documentation of the tracing functions.
1113 * Performance gains when tracing categorized/uncategorized resource
1114 utilization by avoiding calls to get route when updating resource
1115 variables. LMM constraints are being used instead.
1116 * API changed to set task categories. Use MSG_task_set_category instead
1117 of TRACE_msg_set_task_category, and SD_task_set_category instead
1118 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1119 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1120 * Examples for link user variables added (see at examples/msg/tracing/)
1121 * Deprecated function TRACE_msg_set_process_category completely removed
1122 * Trace header updated according to the latest Paje file format
1123 * Tracing network lazy updates, no longer obligate users to use full updates
1124 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1125 * Experimental: let user code declare/set/push/pop application states for hosts
1126 * API changed to allow the manual creation of graph configuration files
1127 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1130 * Improve the API of Lua MSG bindings, using the Lua spirit.
1131 * Each simulated process now lives in its own Lua world (globals are
1132 automatically duplicated). It helps writing simulators. Will allow
1133 to run Splay programs within SimGrid in the future.
1134 * Add a Chord example in Lua, equivalent to the MSG one.
1137 * Start the implementation of a solution to express temporal
1138 properties, not only local assertions. This is still an
1139 experimental work in progress, stay clear from it to be safe.
1143 - Add new runtime parameters --help-logs and --help-log-categories
1144 to display information about supported logging parameters and
1146 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1147 don't exists anymore.
1148 * Mallocators: allow value NULL for the reset function.
1150 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1151 create homogeneous dictionaries, where all the elements share the
1152 same free function. Non homogeneous dictionaries will be
1153 deprecated in the next release.
1154 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1155 - Multi-level dictionaries are deprecated.
1157 - new function xbt_dynar_search_or_negative() that is useful when
1158 you have less than 2 million elements in your dynar and don't
1159 want of the extra complexity of catching exceptions when the
1160 element is not found.
1162 - Make xbt_os_thread module (for thread portability) public.
1163 Documentation is still to come, sorry.
1165 - Cleanups and simplifications to make it maintainable again.
1166 - Exotic features (such as memalign and valloc) were removed.
1167 - The metadata were extended and improved so that the
1168 model-checker becomes able to explore and inspect the heaps.
1169 - This may induce a performance drop when enable_model-checking is
1170 ON in cmake (even if it's not used in the simulation), but it is
1171 necessary at this point to get MC working.
1173 Turn model-checking OFF if simulation performance matters to you.
1174 Not enabling it at runtime is not enough, disable it in cmake.
1176 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1179 SimGrid (3.6.2) stable; urgency=low
1181 The "Not coding new stuff allows to polish old things" release.
1184 * New bindings to the NS3 packet level simulator (experimental)
1185 * Use the raw (efficient) execution contexts instead of the sysv
1186 (portable) ones when possible.
1187 * libpcre is now mandatory in any cases since not using it led to
1188 severe performance loss and possibly other issues
1189 * Update the XML platforms:
1190 - G5K: include the latest machine in Nancy
1191 - GridPP and LCG: new platforms
1192 * Documentation was partially updated, at least (more to come)
1194 Bug fixes, cosmetics and small improvements
1195 * Free terminated processes before the end of the simulation to avoid
1196 exhausting the memory of users having very dynamic amount of
1198 * Bug fix and cosmetics about canceling non-running tasks
1199 * Bug fix about the dot loader's issues when using libcgraph
1202 * Create an installer for windows with nsis (amd64 and win32)
1203 - Add an hello world project to illustrate simgrid project creation.
1204 - Embed libpcre into the Simgrid installer to avoid
1205 its compilation burden
1206 * The raw execution contexts should work on Apple now
1207 * Port to Windows 64 bits
1208 - Sysv contexts now have an implementation for this arch
1209 - GRAS communication features now support this arch
1210 * Drop support for borland compiler on windows
1211 - this code was not maintained, and we kinda depend on gcc nowadays
1212 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1213 * Fix portability issue on unstable ubuntu: linker became picky on
1216 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1218 SimGrid (3.6.1) stable; urgency=low
1220 The "Oops, we broke Macs too" release
1223 * Fixed contexts detection so that raw ones are used when possible
1224 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1225 leads to a strange error, with user code segfaulting sometimes when
1226 the generated code is not perfectly aligned (which is not
1227 controllable from the user side, depends on the amount of code)
1230 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1232 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1234 SimGrid-java (3.6) unstable; urgency=low
1237 * Split of every thing from simgrid v3.5 into a separate package.
1239 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1241 SimGrid (3.6) stable; urgency=medium
1243 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1246 * Bindings now constitute their own package, separated from the main one.
1247 Rationale: reduce our maintenance nightmare by reducing the module coupling
1248 They will soon be released on their own on gforge.
1250 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1251 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1253 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1254 * It's not quite deprecated for now because we have no replacement,
1255 but it may soon become the case.
1258 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1259 * Fortran: New user-level cache variable to store the rank of the running
1260 process. This improves performance by an order of magnitude.
1261 * C: New coccinelle script to automatically locate and modify global and
1262 local static variables.
1263 * Improved SMPI network model with a sender-side gap to account for multiple
1267 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1268 only say if a communication is finished, no matter whether it succeeded or
1269 failed. You can call MSG_comm_get_status() to know the status of a finished
1271 * New function MSG_task_dsend() to send a task and detach it. When a
1272 communication is detached, you are never notified of its success or failure
1273 and the memory is released automatically once it is finished. This function
1274 is useful when you don't care about the end nor the success of a
1276 * Change the prototypes of action replay. Sorry for inconvenience,
1277 but this is really more efficient this way (and to adapt your code,
1278 you just have to fix the initialization, that shouldn't be too long)
1279 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1280 ever managed to do anything useful with it.
1281 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1282 takes a dynar of communications. It returns immediately and gives the
1283 index of a finished communication (if any).
1284 * New example: a basic implementation of the Chord P2P algorithm.
1287 * New model for multi-core CPUs. You can now use the core attribute to
1288 precise the number of cores of a host. This is a basic model. Every
1289 process running on the host receives at most the power provided in
1290 the DTD (throughput<=power). Total throughput of process cannot exceed
1292 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1293 router linked by an up-link and a down-link (possibly asymmetrical).
1294 This kind of pattern allows to easily build last-mile model style platforms.
1295 Aggregating such patterns in a rule-based AS is thus the technique of
1296 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1297 * New model for Vivaldi routing. We transformed the Vivaldi network model
1298 into a Vivaldi routing model (based on the rule-based model). This allows to
1299 combine Vivaldi based latencies with last-mile platforms.
1302 * Added a check for NaN of IEEE754 infinite in the double entries of
1304 * Introduce a new context factory "raw", highly inspirated from the
1305 ucontext factory, but using manually crafted functions in assembly to
1306 do the work in an efficient manner.
1307 * Allow to change the used context factory at run time, not only at
1308 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1309 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1310 contexts at runtime (only with raw contexts or ucontexts).
1311 * Completely rewrote this module to allow parallel execution of user
1312 processes. Use --cfg=contexts/nthreads:N to execute user processes
1313 with N parallel threads (the default is 1, meaning no parallelism).
1314 * Allow to decide dynamically between sequential and parallel modes.
1315 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1316 processes in parallel only when their number is greater than or equal to P
1318 * Added a check for NaN of IEEE754 infinite in the double entries of
1322 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1323 won't output the state of processes when interrupted with Ctrl-C
1324 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1325 NULL-terminated array. This may solve backward compatibility issues
1326 due to the change to return type of SD_simulate. See also:
1327 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1328 * Add new macros with variable number of arguments.
1329 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1330 - in xbt/asserts.h: xbt_assert
1331 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1332 - in xbt/ex.h: THROWF and RETHROWF.
1333 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1335 * Change xbt_die() to accept a format string with arguments, just like printf.
1336 * New data structure: xbt_lib_t, like a dict but more general and with better
1340 * New configuration options
1341 Options triva/categorized and triva/uncategorized can be used to generate
1342 graph configuration files for Triva visualization tool.
1343 * Configuration option tracing/platform is renamed to tracing/categorized
1344 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1345 * New TRACE_user_link_variable interface:
1346 User provides the name of the link and the tracing variable to attach to it
1347 * the declaration of tracing categories must be done after the environment
1349 * simpler tracing interface, just one way to declare categories
1350 TRACE_category or TRACE_category_with_color, it is up to you
1351 * links in the trace file are again identified by their names
1352 * trace contains the full platform hierarchy exactly as declared using the ASes
1353 * Options tracing/msg/[task|process]:1 groups the process by hosts
1354 for both cases, tasks and processes must have names that are unique during the simulation
1355 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1356 * The experimental option tracing/msg/volume is deprecated
1357 its functionality may be reincorporated if needed
1359 The tracing generates a trace file with unordered timestamped events,
1360 because of the way the core simulator (surf) works. A script available
1361 at the tools directory (fix-paje-trace.sh) can be used to put the events
1362 in order. We have changed the tracing so it can generate ordered timestamped
1363 events in the final trace, but depending on the simulator (and how much time
1364 is simulated) that can lead to a huge memory utilization. It is deactivated
1365 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1367 Build Infrastructure
1368 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1369 - We are trying hard to keep the API stable, but it may happen that
1370 some things change (we're a research project after all, not a
1371 nuclear plant operating system). If such things should happen, you
1372 could rely on that macro to adapt.
1373 - current value: 30600 for 3.06.00, aka 3.6
1374 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1375 a number that can be compared with SIMGRID_VERSION.
1376 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1377 assertions and verbose/debug logging events are disabled at compile time.
1379 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1381 SimGrid (3.5) stable; urgency=medium
1384 * New feature to any SimGrid-based simulator: Model-Checking
1385 Check SIN#1 for more details.
1388 * New Model SMPI (three-interval linear regression for correction factors)
1389 See RR-7426, available at http://hal.inria.fr/inria-00527150
1390 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1391 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1392 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1393 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1394 * See also src/smpi/README
1399 - Tracing API changes: TRACE_start and TRACE_end should not be called
1400 by user-code. They are automatically called by simulators created
1401 with SimDAG, MSG and SMPI if the toolkit is compiled with
1402 tracing_enabled=ON. Categories declaration and utilization remain the
1403 same for MSG and SimDag.
1404 - A function was added to the tracing API to declare categories with
1406 - TRACE_category_with_color (char *category, char *color)
1407 where color must be in the following format
1408 "%f %f %f", red, green, blue
1409 and red, green, blue are float values in the interval [0, 1]
1410 - User can specify NULL as color parameter, or continue calling
1411 TRACE_category (cat)
1412 On that case, the tracing system will define random colors
1413 - The following command-line options are supported:
1414 --cfg=tracing/filename:msg.trace
1415 --cfg=tracing:1 (activate tracing, needed to use others)
1416 --cfg=tracing/platform:1 (categorized resource use)
1417 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1418 --cfg=tracing/msg/task:1 (task creation)
1419 --cfg=tracing/msg/process:1 (process creation, migration)
1420 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1421 --cfg=tracing/smpi:1 (SMPI interface tracing)
1422 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1423 - examples of examples/msg/tracing updated
1425 - DAXLoader and DOTLoader functions can generate tasks with categories
1426 - A new function to attribute a category to SD tasks:
1427 TRACE_sd_set_task_category (SD_task_t task, char *category)
1428 * Tracing the MPI interface implemented by SMPI
1429 - Collective operations are traced with states
1430 - Point-to-Point operations are traced with states/links
1431 - Tracing activated by a parameter "-trace filename" passed
1432 to smpirun during execution (considering that simgrid
1433 is compiled with tracing enabled)
1434 - To run the simulation with gdb, the simulator
1435 accepts --cfg=tracing/smpi:1 to trace SMPI
1436 - tesh files to check if smpi tracing is ok
1437 - See examples/smpi/NAS/DT-trace
1438 * GTNetS tracing re-worked
1439 - adaptation to the tracing system of GTNets to cope
1440 with modifications regarding the fullduplex mode
1441 - new tesh files to check if gtnets tracing is ok
1444 * Asynchronous communications through the functions:
1445 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1446 * New function: MSG_load_platform_script()
1447 to make possible using a lua script instead of XML files to set up platforms
1448 * New function: MSG_set_function
1449 to associate functions to processes, used when bypassing the parser
1450 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1452 Platforms: Add some more examples in examples/platforms
1453 * Grid'5000: see www.grid5000.fr
1454 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1457 * Change the XML format. This is a very important modification. SimGrid
1458 3.5 introduces a new hierarchical format based on the notion of
1459 Autonomous Systems. Compatibility with old format is ensured
1460 through the perl script provided in the install bin directory
1461 bin/simgrid_update_xml.
1462 It is now possible to build platforms with specific routing
1463 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1464 connect several platforms together. We will try to provide soon
1465 set of realistic platforms exploiting these properties (have a
1466 look at examples/platforms/ for the moment).
1467 * Take the opportunity of the XML format change to be a good XML citizen:
1468 rename link:ctn to link_ctn and similar changes (also dealt with by
1470 * Add a new routing scheme (rule-based) using regular expressions. It
1471 enables to have an extremely low memory footprint when the
1472 underlying routing is simple and can be compactly described. You
1473 need to have libpcre4-dev (perl regular expressions) installed if
1474 you want to use this routing scheme.
1475 * Revive the cluster TAG and allow to easily and efficiently (both in
1476 term of memory and speed) connect clusters together. Have a look
1477 at teshsuite/simdag/platforms/ to see how this can be done. With
1478 this tag, you can create clusters with thousands of tasks at no
1479 cost (have a look at examples/platforms/).
1480 Note: clusters are implemented as ASes, so there is no need for an
1481 enclosing AS tag if you have only one cluster in your platform.
1482 * Add new generic functions in the public interface that allows the user
1483 to call SURF 'create_resource' methods from your code (same
1484 functionality as the XML bypass mechanism but with a much lighter
1486 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1487 that uses a piecewise linear approximation to produce better
1488 results when exchanging small messages.
1489 * Add a new parameter to handle correctly full duplex link and account
1490 for interferences between uplink and downlink communications
1491 (activate with --cfg=fullduplex:1).
1494 * Rename the SD_READY (all dependencies are satisfied and task is
1495 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1496 dependencies are satisfied) state.
1497 This prevents a confusion between the notion of "ready to schedule"
1498 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1499 simulated" (SD_RUNNABLE) used by the simulation kernel.
1500 * Change the way a task is considered as ready. Instead of removing
1501 dependencies when a task is done, a counter is decreased. This way,
1502 it is always possible to reach ancestors thanks to the
1503 SD_taks_get_parents function (even after the end of the simulation.)
1504 * Change the return type of SD_Simulate from (SD_task_t*) into
1505 xbt_dynar_t. This function was in handling a dynar internally and
1506 converted it into a NULL terminated array for historical reasons.
1507 * New function SD_dotload(char*) to load a DAG described in dot
1508 format. This loader and the corresponding examples require the
1509 installation of the graphviz library.
1510 * Fix a bug in the management of tasks of size 0 in the surf network
1511 models. This problem was only visible with SIMDAG and you should
1512 thus disregard results produced with earlier versions if you
1513 relied on this feature (some tasks were blocked because of this).
1514 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1515 even though your applications were doing only point-to-point
1516 communications and sequential computations. Now you can really use any
1517 model you want (of course, if you create real parallel tasks, which are
1518 not implemented in most models beside ptaskL07, this will abort).
1519 * Add an example that schedules a DAX on an heterogeneous platform
1520 using a Min-Min strategy.
1521 * New function SD_workstation_get_current_task() that returns the kind
1522 of task currently running on a workstation in the sequential access
1524 * Raise some warnings when unexecuted tasks remains at the end of the
1525 simulation. This is usually caused by cycles in the DAG.
1528 * New function: SIMIX_process_set_function() called by MSG_set_function
1529 * Change the underlying waiting queue in semaphores so that a process
1530 can wait on several of them simultaneously (as in waitany).
1531 * Fix the way to handle tokens in semaphores so that all access patterns
1532 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1533 * kill the dirty pimple SIMIX_message_sizes_output()
1534 Please use (proper) visualization instead
1537 * New data container: setset (set of sets of elements)
1538 * New module: mmalloc (mapped malloc, allowing to have several
1539 independent segments of malloc)
1540 * New function: xbt_dict_cursor_set_data()
1541 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1542 * New function: xbt_dynar_is_empty()
1543 * New function: xbt_fifo_get_last_item()
1544 * Fix xbt_dynar_shrink(): use the right element size.
1545 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1546 uninitialized areas during expand.
1547 * Fix semaphores: previous implementation was severely broken.
1548 * Use library init/fini functions for our initialization.
1549 - you can use logs and other feature as soon as you want in your
1550 code (even before the xbt_init / MSG_init)
1551 - xbt_exit is now a no-op and produce a warning when used.
1554 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1555 benefit from the latest improvement to the simulation kernel.
1556 * Kill measurement sockets for now. If you rely on them, sorry. This
1557 release is not for you. This feature will be reintroduced in the
1558 future, but we cannot delay the release any further.
1559 * New function: gras_msgtype_get_name().
1560 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1561 * Fix (at last) the pmm example: it should not randomly fail anymore.
1563 Build chain: bug fixes and overall polishing
1564 * Cmake is now stable enough. Hence, we killed the autotools.
1565 * Port to windows ( TM :)
1566 * Fix the 'make install' target.
1567 No need to use 'make install-simgrid' anymore
1568 * Introduce a 'make dist' target compiling a *source* archive
1569 'make package' compiles a binary archive
1570 * Compile java files only on need
1571 * Add --cd and --setenv command line options to tesh
1572 * Out of source builds are not fully supported yet, but we are close
1573 * Enable supernovae and optimization flags by default for our users
1576 * Add layer to set up environment directly from lua, without XML.
1577 * The effect of gras_stub_generator can be achieved through
1578 lua too (check examples/gras/console/ping_generator.lua)
1580 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1582 SimGrid (3.4.1) stable; urgency=low
1584 The "Polishing easter eggs is probably a good idea" release.
1585 This is a bug fixes release only.
1589 * Fix a bug preventing the tasks from begin garbage collected.
1592 * Fix a bug occuring when a host involved in a communication fails.
1593 This was not detected properly by the other peer involved in the
1594 communication. Now, it's reported as a network error.
1597 * Warn the user about loop dependencies in data flow of DAX files
1598 * Obey the control-flow dependencies of DAX files
1601 * Add option "enable_smpi" allowing to not compile SMPI.
1602 Probably useful for the (Mac) users experiencing a build error here
1603 * Improve the detection of lua5.1 and ruby1.8
1605 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1607 SimGrid (3.4) stable; urgency=low
1609 The "Easter in Cargese" release. Also known as (major changes):
1611 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1612 ~> bindings were greatly improved
1613 ~> new tracing infrastructure for better visualization introduced
1615 * the "Welcome to configury modernity" release.
1616 ~> we switched from autotools to cmake, and improved our cdash
1619 A more detailled list of changes follow (full detail in svn log).
1621 Java Bindings: Various Cleanups
1622 * (install java-gcj-compat-dev on debian-like to use them)
1623 * Remove put/get: no need to export deprecated interface in Java
1624 Use send/receive instead.
1625 * Cleanup the examples and add a README per directory
1626 * Remove example autoDestination (that's the only way to go now)
1627 * Remove example explicitDestination (was a plain copy of basic)
1628 * Make JniException a runtime exception, so that there is no need to
1629 declare the fact that you may encounter such a beast. I guess that
1630 nobody will ever want to survive such error.
1631 * Create specific errors for each MSG case of failure:
1632 host failure, transfer failure, timeout, task cancelled
1633 * Cleanup the exceptions that may get thrown by each function
1634 * Other internal cleanups in Java bindings. Performance still bad :/
1635 Ruby and Lua Bindings: create them
1636 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1637 * That's new and great, you should try them out.
1638 Same functionalities than Java bindings, only even less polished
1640 * Kill the useless "rate" argument of SD_task_get_execution_time()
1641 Everyone used to provide -1 as a value, it was not used, and the
1642 semantic of a possible use wasn't even clear.
1643 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1644 either as comm costs or compute costs to mean that there is no
1645 such thing for that specific task.
1646 * Add a SD_task_set_name() function
1647 * Fix SD_task_unschedule() on typed tasks
1648 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1649 * In DAX loader, accept useless 'level' attributes to <job> since
1650 LIGO DAGs have them (seem to be to ease graphical representation).
1652 * Add an example masterslave_mailbox.c using send/receive and not
1653 the deprecated put/get interface.
1654 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1655 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1657 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1658 (declare MSG_USE_DEPRECATED to still have the old name)
1659 * Add a MSG_task_set_data() function
1660 * About trace replay (see examples/msg/actions):
1662 - Allow to work with splitted trace files for each process
1663 Give the specific trace file as argument of each process,
1664 and call MSG_action_trace_run(NULL)
1665 You can still have one merged file for all processes.
1666 - Fix implementation of collective operations
1667 * Allow task_execute() on 0-sized tasks (closes #10063)
1669 * This is the first release of SimGrid where SMPI is not considered
1670 beta anymore (even if some corners should still be improved)
1671 * Port over the new SIMIX_network submodule (internal refactoring)
1672 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1673 * Implement more missing elements of the standard:
1675 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1677 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1678 MPI_Scan() MPI_Get_processor_name()
1679 - Added implementation of missing case for Alltoall (warning: it's
1680 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1681 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1682 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1683 SMPI_MPI_Allgatherv()
1684 * Bug fixes include:
1685 - MPI_Waitsome() was broken
1686 - Allow relative includes in smpicc
1687 - Command line cfg argument 'reference_speed' was ignored...
1688 - Some functions did not properly lead to auto-benching of user code
1689 - smpicc passes -O2 by default (just like openmpi one)
1691 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1692 * Bug fixes about timeouts during communications
1693 * add SIMIX_message_sizes_output() as a pimple to write to file the
1694 amount of messages per size. Use gnuplot to get histogram.
1695 Pimple because that's the only user-visible function of simix,
1696 defined directly in xbt.h (irk, sorry)
1698 - Add a SIMIX_sem_get_capacity() function
1699 - Fix interactions with processe resume/suspende
1700 - release_forever() was stupidly broken
1701 - Fix SIMIX_display_process_status() for processes in a semaphore
1702 - Make SIMIX_sem_block_onto() user-visible
1703 * Refactoring context stuff:
1704 - Use pseudo-OOP for better modularity
1705 - reimplement SIMIX_process_kill() without process_schedule() so
1706 that the latter can take as invariant that it is called from
1708 - Merge context_start into context_new for sake of simplicity
1710 * Add a Vivaldi network model, coded live during SUD'10 ;)
1711 * Rename configuration variables to start a hierarchy:
1712 o cpu_model -> cpu/model
1713 o network_model -> network/model
1714 o workstation_model -> workstation/model
1715 * New configuration variables:
1716 o network/bandwidth_factor: correction to bandwith
1717 o network/latency_factor: correction to latency
1718 o netwotk/weight_S: correction to the weight of competing streams
1719 * Add a long description to the models, that users can see with such
1720 argument on the command line: --cfg=cpu/model:help
1721 * --help-models display the long description of all known models
1723 * config: add the ability to set a default value after registration
1724 Does not override any previously set value (e.g. from cmd line)
1725 * dict: allow to have integer key and data.
1726 When so, you need to use the following functions
1727 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1728 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1729 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1730 In contrary to regular dicts, the key is not malloced before copy.
1731 Mixing scalar and regular elements in the same dict is not tested
1733 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1734 Tracing for Visualization:
1735 * SimGrid is now instrumented in order to generate a trace file for
1736 visualization analysis: to use it, need to compile SimGrid with the
1737 "tracing" option enabled, and instrument the program using SimGrid with
1738 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1739 (among other functions).
1740 * The instrumentation only traces the platform utilization for now
1741 * Documentation to use the tracing functions and how to analyze the
1742 traces with the Triva tool is written.
1743 * More information about: SimGrid FAQ (in the section Tracing Simulations
1746 * We moved to cmake as default build system. Autotools support will
1747 be dropped soon. Check the FAQ for more info about how to use it.
1748 * Greatly improved our cdash/ctest interactions
1749 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1750 * Added memory checking tests with valgrind; lot of memleak fixing.
1751 This may be the first release of simgrid with so few memory issues
1752 * Added code coverage tests.
1753 Our coverage is still improvable, but at least we see it on cdash.
1755 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1757 SimGrid (3.3.4) stable; urgency=low
1759 The "Desktop Grid needs love too" release (also called Xmas release).
1761 Models improvements:
1762 * Major speedup in the maxmin system solving by using lazy evaluation
1763 Instead of solving completely the maxmin system at each iteration,
1764 only invalidate (and recompute) the modified parts.
1765 This new feature is enabled in default models but you can try to
1766 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1767 * Cas01 IMproved as default CPU model
1768 This CPU model is the same Cas01 model, but it uses the
1769 maxmin-selective-update flag and a heap structure to manage
1770 actions on SURF kernel.
1771 It reduces the complexity to find the next action to finish and,
1772 consequently, it's faster than the old Cas01.
1773 This is the new default CPU model (Cas01).
1774 * Rename the old Cas01 model to Cas01_fullupdate
1775 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1776 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1777 * CpuTI (CPU Trace Integration)
1778 A new CPU model whose objective is simulate faster when using
1779 availability trace files.
1780 Instead of using a full featured, over engineered maxmin system for
1781 CPU modeling, this model does the pre-integration of traces files
1782 to calculate the amount of CPU power available, and so, executes
1783 faster than the old CPU models.
1784 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1785 * Use LV08 as default network model since it gives better accuracy
1786 for small messages and shouldn't change things for big ones.
1787 Use --cfg=network_model:CM02 to get the previous behavior.
1790 ******************************************
1791 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1792 ******************************************
1793 * The new CPU model may changes simulations!
1794 The point is that events occurring at the exact same timestamp
1795 are not scheduled in the same order with the old and new
1796 version. This may be enough to completely change the execution
1797 of simulations in some cases.
1798 * The new network model will change simulations!
1799 This new model is more realistic than the previous one, so you
1800 should consider redoing your old experiments with this model.
1801 Sorry for the inconvenience.
1804 * Introduce the supernovae compilation mode
1805 When compiled that way, the whole SimGrid (or almost) is put in a
1806 single compilation unit and compiled in one shoot.
1807 This is to help gcc which has difficulties to inline stuff from one
1809 The speedup seem to be above 15%, althrough more tests are needed on
1810 amd64 to confirm that gain.
1813 * Port of MSG's mailbox on top of SIMIX network
1814 The put/get mechanism was greatly simplified on the way.
1817 * New SIMIX network module. Provides:
1818 - Mailbox: rendez-vous mechanism to find with who you want to speak
1819 - Synchronous send/recv: easier and hopefully faster since the
1820 logic is handled in the maestro process directly now
1821 - Asynchronous send/recv: you dreamt of it? It's here now
1822 Too bad that nobody cared enough to propagate the change to MSG.
1823 * Add semaphores as SIMIX synchronization mechanism.
1826 * new function SD_daxload(char*) to load a DAX file
1827 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1828 * Introduce typed tasks. Specify its kind and cost at creation.
1829 At scheduling, just give where it should be placed, and the cost
1830 for each involved resource is automatically computed.
1831 Existing constructors so far (more to come of course):
1832 - SD_task_create_comm_e2e() for end-to-end communication
1833 - SD_task_create_comp_seq() for sequential computation
1834 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1835 * new function SD_task_dump() for debuging display
1836 * new function SD_task_dotty(task,FILE*) writing to file the info
1837 about the task in dotty format
1838 * SD_task_dependency_exists() can now cope with having one of its
1839 arguments NULL. If so, it tests whether the other argument has any
1841 * Add getters on list of preceding/following tasks:
1842 SD_task_get_parents(task) and SD_task_get_children(task)
1843 * Add getters on amount of workstations and list:
1844 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1845 * Add getter on task kind: SD_task_get_kind(task)
1846 * Update the start_time and finish_time of tasks on completion/failure
1847 * Bugfix: Remove task from state swags when destroyed
1850 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1853 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1854 dynar with one separate thread per value of the dynar.
1855 * Change the prototype of xbt_thread_create(), sorry.
1856 Added a boolean parameter indicating whether we want to join this
1857 thread (used in SG only for now)
1858 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1861 * GTNetS wrappers should now be usable again (and betterly tested too)
1862 * Fix a major regression from 3.2 where the timeout provided to
1863 MSG_task_put_with_timeout() was used as absolute time before which
1864 the comm should be done.
1865 * Start to fix the <cluster> tag.
1866 - Internal links should be good now (beside of the loopback, which
1867 use the private link instead)
1868 - paths to the external world is still rather broken
1869 - the <route:multi> tag is just broken. Actually that's brain-dead.
1870 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1871 to make it less stupid
1872 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1873 * Fix a source-level compatibility glitch from 3.2: after defining
1874 MSG_USE_DEPRECATED, you can use the old name
1875 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1876 * Allow to compile from the SVN with automake 1.11
1877 * Fix some problems when using the "start_time" tag in deployment XMLs.
1878 * Fix #8569: XBT/synchro.h has redundant declarations
1879 * Fix #8563: MSG return values and exceptions
1880 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1881 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1882 Thanks to Robson Peixoto.
1883 * Fix a vicious bug in dictionaries inducing that some elements were
1884 not freed on xbt_dict_free()
1886 Portability report of this version:
1887 * Main portability targets:
1888 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1889 on (amd64/i386/ia64)
1890 - mac leopard on i386
1891 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1892 but nothing critical.
1893 * Other platforms: windows, AIX and others were not tested for this release
1895 Timing report of this version:
1896 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1897 scenario parameters). From 8h to a few seconds in desktop grid settings.
1898 * Supernovae brings about 25% speedup on i386.
1900 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1902 SimGrid (3.3.3) stable; urgency=low
1904 The "Need for Speed" release.
1906 The timings done to validate the 3.3.2 were faulty.
1907 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1909 The problem was a conversion from a manually handled vector to
1910 xbt_dynar_t on the critical path.
1911 xbt_dynar_foreach calls functions, inducing stack management crap.
1913 We inlined these functions and xbt_dynar_foreach is now breath taking.
1914 We also inlined xbt_swag_belong on the way.
1916 Here are some approximate speedup measurements (on master/slaves
1917 simulations lasting between 10s and 20s each):
1918 3.3.1 -> 3.3.2: about same performance
1919 3.3.2 -> 3.3.3: 40% speedup
1920 3.3.1 -> 3.3.3: 40% speedup
1921 3.3.1 with inline patch -> 3.3.3: 30% speedup
1923 Our reading is that the refactoring which occurred in 3.3.2 made us
1924 suffer much more from the xbt_dynar_foreach low performance, but
1925 once we solved this, this refactoring proved to be very performance
1926 effective. From the 40% speedup, somehow, 10% are due to the
1927 inlining and 30% to the refactoring.
1929 That's a pitty that gcc cannot inline functions placed in other files
1930 alone. We have to choose between:
1931 - break the encapsulation (by putting private data structures and
1932 accessors in headers files to help gcc)
1933 - live with low performance
1934 - switch to a decent compiler such as icc (not quite possible).
1936 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1938 SimGrid (3.3.2) stable; urgency=low
1940 The "Simplicity does not preceed complexity, but follows it" release.
1942 The main contributors of this release were (lexical order):
1943 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1946 * Extract the routing logic into its own object.
1947 (was dupplicated in network.c and workstation_LV07.c;
1948 Allows to implement other ways of storing that info)
1949 => kill now useless network_card concept
1950 - Use dynar to represent routes (instead of void** + int*)
1951 - kill link_set (use surf_network_model->resource_set instead)
1952 - Add a command-line option to choose the routing schema to use
1953 - Add three new models:
1954 * Floyd (shortest path computed at initialization)
1955 * Dijikstra (shortest path recomputed all the time)
1956 * Cached Dijikstra (shortest path computed on need)
1957 All these models where contributed by Silas De Munck, and are
1958 described in his ICCS09 paper.
1960 * Simplify model declaration
1961 (less redirections, less function to write when defining a model)
1962 - Factorize stuff between models:
1965 surf_model_resource_set(model)
1966 surf_model_resource_by_name(model, name)
1967 - Unify the types of models in s_surf_model_t (using an union)
1968 - Embeed fields of common_public directly into s_surf_model_t
1969 - Rename model methods:
1970 action_free ~> action_unref
1971 action_change_state ~> action_state_set
1972 action_get_state ~> action_state_get
1973 - Change model methods into functions :
1974 (model)->common_public->action_use ~> surf_action_ref
1976 * Implement a generic resource; use it as ancestor to specific ones
1977 (allows to kill duplicated code in models)
1978 Drawback: timer command don't need no name nor properties;
1979 workstation_CLM03 don't need no properties
1980 (but I guess we can live with those few bytes wasted)
1982 * Improve the action object model
1983 - implement a constructor avoiding dupplicated code about field
1984 initialization in generic_action part.
1986 * Kill the SDP model: it has an external dependency, is deprecated
1987 in flavor of modern lmm models, and didn't compile since a while
1990 * Relocation of the context module from XBT to SIMIX.
1991 (the context were decoupled from the simix processes, duplicating a lot of code)
1992 => a lot of code was factorized
1993 - less overhead is introduced during scheduling
1994 - simpler API for the context factory
1995 - the logic for process creation,destruction and manipulation was simplified
1996 * Simplification of the s_smx_process_t data structure.
1997 => accesing the simix level data associated to a process is faster now,
1998 and the code is a lot more readable.
2001 * Implement some more MPI primitives:
2002 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2003 -implementation: Bcast: flat or 2-ary tree (default),
2004 Barrier: 4-ary tree,
2006 Allreduce: Reduce then Bcast
2007 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2008 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2009 Alltoallv: flat tree, like ompi
2011 * Add support for optimized collectives (Bcast is now binomial by default)
2012 * Port smpirun and smpicc to OS X
2015 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2016 Shout out if you used it.
2019 * Display the list of still queued messages in SG mode when existing
2023 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2024 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2025 * Add xbt_str_from_file(FILE*)
2026 * Add xbt_dict_get_key achieving a linear reverse search
2027 * Remove the context module
2029 Portability report of this version:
2030 * Main portability targets:
2031 - Linux(debian)/x86/context
2032 - Linux(debian)/x86/pthreads
2033 - Linux(debian)/amd64/context
2034 - Linux(debian)/amd64/pthreads
2035 On these, we still have the eratic breakages of gras/pmm and
2036 amok/saturate_sg reported in previous version. We still think
2037 that the tests are the cause of the fault, not the tested code.
2039 - Mac OSX Leopard/x86/context
2040 Still false negative in tesh autotesting.
2041 Smpi still fails, but this time because readlink does not accept -f
2042 Everything seems to work properly beside of that.
2045 - AIX version 5.3 (only tested contexts this time)
2046 Smpi still fails there because mktemp is not installed.
2047 Everything seems to work properly beside of that.
2049 I managed to compile it for the first time, but several breakages.
2050 Won't delay the release for this exotic platform.
2052 * Windows: it's still lagging behind. If you want to help, please
2055 Timing report of this version:
2056 This version seem to be more than 5% faster than 3.3.1 (on linux
2057 64bits with contextes). The gain is less than expected, we are
2058 investigating this for next release.
2060 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2062 SimGrid (3.3.1) stable; urgency=low
2065 * Implement a --cfg-help to show existing configuration variables
2066 * Build chain do not require doxygen in maintainer mode
2069 * fix a bug on struct sizeof computation, which prevented the
2070 exchange of arrays of structs in some conditions
2071 - added a regression test about this in datadesc_usage
2072 * Allow the exchange of 0-long dynamic vectors.
2073 - for that, use -1 as indicator of dynamic size instead of 0
2074 - This implied to change any size from unsigned long to long,
2075 reducing a bit communication abilities, but I guess that with
2076 64bits being quite common, this is more than enough.
2077 - This also induce a protocol change, thus bumping network protocol
2078 version from 0 to 1 (if we have external users, we have to get
2079 clean on that point too ;)
2080 - added two regression tests about this in datadesc_usage
2081 * Be more verbose when propagating local exceptions
2082 This helps debugging.
2083 * Display the status of simulated processes when receiving SIGINT in
2087 * Allow to control the simulation from a trace file.
2088 New functions MSG_action_register() and MSG_action_trace_run()
2089 The first one allows to associate a function execution to each
2090 kind of action while the second one parses a trace file and
2091 triggers the corresponding actions within the system.
2092 For now, only a toy example is provided in examples/msg/actions
2093 * Add an exemple of process migration in examples/msg/migration
2094 * Fix a bug in task exchange which broke MSG_task_get_sender()
2095 Add a teshsuite regression test for that.
2096 [Bug: if MSG_task_get_sender() is called after sender exit,
2098 * Fix a bug which prevented suspend/resume to work properly
2099 * Display the status of simulated processes when receiving SIGINT
2100 This fixes a regression of v3.3. due to the introduction of SIMIX
2101 * Bug fixing in failure management:
2102 - trace could not start by a failure at time 0
2103 - failure during communications were not working
2106 * Add SIMIX_process_set_name() to change the name of the current
2107 process in the log messages.
2108 * Store smx_hosts in a dict since we only retrieve them by name
2109 * Move the configuration infrastructure to surf
2112 * Move the configuration infrastructure to surf
2115 * Massive internal cleanups:
2116 - Store internal structures on processes instead of hosts (allows
2117 to have more than one process per host, in addition of being more
2119 - Cleanup the initialization/finalization process
2120 - Kill a whole bunch of unneeded synchronization:
2121 processes run in exclusive manner within the simulator
2122 - Move queues from global tables to process data fields
2124 - now accept -platform and -hostfile arguments
2125 - Pass the right rank value to processes according to the hostfile
2126 * Compile the examples by default, and use them as regression tests
2127 * Implement MPI_Wtime()
2128 * Change the reference speed to a command line option
2131 * TCP_gamma can now be specified as command line option using
2132 --cfg=TCP_gamma:10000000.0
2133 * Change the --surf-path cmd line option into --cfg=path:
2136 * Also include strbuff from xbt.h public header
2137 * xbt_ex_display(): do not free the exception after displaying
2138 This allows to do more with the given exception afterward.
2139 Users should call xbt_ex_free() themselves.
2143 Portability report of this version:
2144 * Main portability targets:
2145 - Linux(debian)/x86/context
2146 - Linux(debian)/x86/pthreads
2147 - Linux(debian)/amd64/context
2148 - Linux(debian)/amd64/pthreads
2149 These targets fail about 1/10 of times on gras/pmm, but we believe
2150 that this is because of the test, not because of simgrid.
2151 amok/saturate_sg fails even more rarely, and the test may not be
2154 - Mac OSX Leopard/x86/context
2155 The test suite still spits tons of errors because some obscure
2156 force prevents us from removing the temporary directories
2157 arguing that they still contain some metadata I've never heard of.
2158 Smpi fails because seq is not installed.
2159 Everything seems to work properly beside of that.
2162 - AIX version 5.3 (both contexts and pthread)
2163 Smpi still fails there because mktemp is not installed.
2164 XML inclusions seems rosty on AIX.
2166 * Windows: it's still lagging behind. If you want to help, please
2169 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2171 SimGrid (3.3) stable; urgency=high
2175 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2176 [Malek Cherier & Mt]
2178 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2181 Until now, GRAS were using MSG as an interface to SURF. It was
2182 quite difficult because both interface have several differences
2183 (MSG channels vs GRAS sockets were the most notable point).
2185 This also opens the gate to SMPI (which should occur soon) and speed
2186 up simulations by to 40% (even if it were not the main goal).
2188 **************************************
2189 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2190 **************************************
2191 The point is that events occuring at the exact same timestamp are
2192 not scheduled in the same order with the old and new version. This
2193 may be enough to completely change the execution of simulations in
2194 some cases. Sorry for the inconvenience.
2196 * Cleanup and upgrade the XML format to push further scalability
2197 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2199 * Improve the testing infrastructure with tesh. Now a very large part of
2200 the code is tested not only by being run but also by checking that the
2201 output match an expected output [Mt].
2203 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2204 is really less memory-demanding, which should allow you to use
2205 larger files in SimGrid [AL].
2207 * Inform valgrind about our contextes, so that it becomes usable
2208 with the default (and more effecient) version of SimGrid
2209 [contributed by Sékou Diakite, many thanks]
2212 * Introduce a listener thread in charge of receiving incoming
2213 messages from the network. It allows to overlap communication and
2214 computation but most notably, it removes some stupid deadlocks due
2215 to the fact that so far, a process could not send and receive at
2216 the same time. This made most non trivial communication schema
2218 * Convert the PIDs from long int to int to match the MSG ones (and
2219 linux ones too) [Mt]
2220 * New function: gras_agent_spawn() to launch a new process on
2221 current host. Only working in simulation for now. [Mt]
2222 * New function: gras_os_hostport() returning a constant form (ie,
2223 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2226 * Make the backtrace of exceptions more human readable [Mt]
2227 * New module: xbt/str [Mt]
2228 a ton of string utility functions (split, join, printf to a newly
2229 allocated buffer, trim, etc)
2230 * New module: xbt/hash [Mt]
2231 SHA1 hashing algorithm (more to come if needed)
2232 * New module: xbt/synchro [Mt]
2233 synchronization tools (mutex and conditions) working the same way
2234 in simulation and in real life (mainly useful for GRAS, but not
2236 * New module: xbt/queue [Mt]
2237 classical producer/consumer synchronization scheme
2238 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2239 (using the classical functions will get serialized) [Mt]
2240 * Make dictionary internal table dynamic. No need to specify its size
2241 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2243 * Make sure the log channels are organized as a tree under windows
2244 (because of ANSI C compatibility issue, any channel were child of
2245 root directly) [Mt].
2248 * Cleaned many thing in surf and fixed a few bugs [AL].
2249 * Add a nice command line configuration mechanism to compose models [AL].
2250 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2251 the previous one (KCCFLN05). It relies on something that looks like
2252 a max-min sharing mechanism but cannot be written as such. A new solver
2253 was thus designed [AL].
2254 * Add a new solver to lmm. Based on Lagrange optimization and
2255 gradient-based descent, it enables to efficiently maximise systems s.a
2257 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2259 This solver enables to propose two new network models for TCP Reno and
2260 TCP Vegas based on Low's work. These models still need to be fully
2261 tested though [Pedro Velho].
2264 * Bug fix in SD_simulate. Now the time bound given as argument is
2266 * Use the new parallel task model (ptask_L07) as default.
2267 * Use the SURF command line configuration mechanism.
2268 * 0-size tasks (for synchronization) should now work.
2270 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2272 SimGrid (3.2) stable; urgency=high
2276 We still experience issues on this platform, but we believe that at
2277 least MSG is usable.
2279 GRAS API BREAKAGE (for simplification purpose, sorry):
2280 * the gras_msgtype_by_name is not used anymore. Instead of
2281 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2282 you can write (and must)
2283 gras_msg_send(toserver, "request", &request);
2284 - If you still want to pass a gras_msgtype_t to the function (to cache
2285 the type and avoid the lookup time), use the gras_msg_send_() variant.
2286 - Impacted functions:
2287 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2288 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2289 * The callbacks are now expected to return 0 when everything went well
2290 (just like the main() function)
2292 GRAS new features and improvements:
2293 * New module mechanism where user code can use per process globals [Mt]
2294 This is similar to gras_userdata_*() functions, but for libraries. It
2295 factorize some code developped over and over in the examples and AMOK.
2296 It has still to be documented and used (only amok/peermanagement is
2298 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2299 loss when they were small enough to fit into the buffer and sent quickly
2300 enough so that they can all get received in one shoot.
2301 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2302 if not found. Use the *_or_null() variant for the old semantic.
2303 * In gras_msg_handle, do not discard messages without callback.
2304 They are probably messages to be explicitly awaited later (ie, proofs of
2305 mis-synchronization in userland since they are sent before being awaited)
2306 No big deal usually.
2307 * gras_socket_meas_send/recv: semantic changed!
2308 The numerical arguments used to be (1) the total amount of data to send
2309 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2310 messages. This was need for the fool willing to send more than MAXINT
2311 bytes on quite fat pipes.
2314 * Do really rename the hostmanagement module to peermanagement. [Mt]
2315 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2316 but this is rather new and this was documented in the module
2317 documentation (poor excuses, I admit)
2318 * Bandwidth measurement semantic changed! This follows the changes to
2319 gras_socket_meas_send/recv explained above.
2322 * A sequential mode has been added to the workstations. When a workstation
2323 is in sequential mode, it can execute only one task, and the other tasks
2324 are waiting in a FIFO. [Christophe Thiery]
2327 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2328 model for SIMDAG. [Christophe Thiery]
2329 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2330 equal to 0, which caused some bad behaviors in
2331 saturated_constraint_set_update. I now use a threshold mechanism like in
2335 * When running manually src/testall, you select specific units [Mt]
2336 testall is the result of our cunit mechanism, and should replace all
2337 the scripty thingy around since bash don't run easily on billware.
2339 * A mallocator system has been added. [Christophe Thiery]
2340 Mallocators allow you to recycle your unused objects instead of freeing them
2341 and allocating new ones.
2343 Documentation update:
2344 * FAQ reworking + New FAQs:
2345 - "Valgrind spits tons of errors!" [Mt]
2346 - "How to repport bugs" [Mt]
2347 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2348 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2349 - Communication time measurement within MSG [AL]
2350 - I experience weird communication times when I change the latency [AL]
2351 * GRAS tutorial [Mt]
2353 - an introduction to the framework and to the used communication model
2354 - an initiatic tour introducing the most proheminent features:
2356 . Lesson 0: Installing GRAS
2357 . Lesson 1: Setting up your own project
2358 o Part 2: Message passing
2359 . Lesson 2: Exchanging simple messages
2360 . Lesson 3: Passing arguments to the processes (in SG)
2361 . Lesson 4: Attaching callbacks to messages
2362 . Lesson 5: Using globals in processes
2363 . Lesson 6: Logging information properly
2364 . Lesson 7: Using internal timers
2365 . Lesson 8: Handling errors through exceptions
2366 . Lesson 9: Exchanging simple data
2367 . Lesson 10: Remote Procedure Calling (RPC)
2368 . Lesson 11: Explicitely waiting for messages
2369 . Recapping of message passing features in GRAS
2370 - A HOWTO section containing:
2371 o HOWTO design a GRAS application
2372 More are due, of course. They will come latter. In the meanwhile, you can
2373 check the examples which are still here.
2375 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2377 SimGrid (3.1) stable; urgency=high
2381 There was a stack corruption somewhere, visible only when optimizing
2382 with these versions. [Vince]
2385 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2386 module that enabled to play with Directed Acyclic Graphs. It is built
2387 directly on top of SURF and provides an API rather close to the old
2388 SG. Some old codes using SG are currently under rewrite to check that
2389 all needful functions are provided. [Christophe Thiery]
2392 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2393 extension of the classical CLM03 model that gracefully handles
2394 failures. This is now the default model for MSG and GRAS. It doesn't
2395 handle parallel tasks yet though. [AL]
2396 * Bug fix: Weights were not correctly set in the network part.
2397 WARNING: This may have resulted in incorrect results with simulations
2398 where there are more than one flow on a given link. [AL]
2401 * After a (long ?) discussion on simgrid-devel, we have decided that the
2402 convention we had on units was stupid. That is why it has been decided
2403 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2404 WARNING : This means that all previous platform files will not work as
2405 such with this version! A warning is issued to ask users to update
2407 A conversion script can be found in the contrib module of the CVS, under
2408 the name contrib/platform_generation/surfxml_update.pl [MQ]
2411 * Bug fix: Processes were started in reverse order, wrt deployment file.
2412 WARNING: if your code relies on this bug, please fix it. [AL]
2413 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2414 being executed on two different locations. [AL]
2415 * Bug fix: Failures are now better supported thanks to Derrick's tests
2416 (there was many failure situations I hadn't thought of and that weren't
2417 correctly handled). [AL]
2418 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2422 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2423 Exception occurring on server-side are propagated back to client (!).
2425 API CHANGE: the callback changed their prototype. Change:
2426 int my_handler(gras_socket_t expeditor, void *payload_data) {
2428 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2429 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2431 * New! function: gras_msg_handleall to deal with all messages arriving
2432 within a given period.
2433 * New! function: gras_socket_server_range to get a server socket in a
2434 range of port numbers (ease to avoid port number conflicts) [MQ]
2435 * New! gras processes display their backtrace when they get a SIGUSR1
2436 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2437 Sweet to debug RL processes [MQ]
2441 - Do not force experiment sizes to be expressed in kb, or it becomes
2442 impossible to measure the latency this way (needs one byte-long tests)
2443 WARNING: this changes the amok_bw_* function semantic. [MQ]
2444 - Implements the link saturation stuff. [MQ]
2445 * Peer management module:
2446 New! module factorizing code that we wrote over and over [MQ].
2449 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2450 - Test units are placed directly into the library code, they get extracted
2451 automatically and placed into the src/testall binary.
2452 - Convert most of the XBT tests to this system.
2453 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2454 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2455 live in the same C code. This required to change a little bit the API
2456 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2457 * New module: sparse graph structure with basic algorithms (this is work
2458 in progress and the API is not considered to be frozen yet). [AL]
2459 * Display more information on backtraces: source line & function names are
2460 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2461 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2462 * Reworked a little bit some #include statements to load only required
2463 headers. Some user code that relied on SimGrid to include stdlib or
2464 stdio may need to include it by themselves. [AL]
2465 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2467 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2468 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2469 but I guess that almost nobody used those parts. [MQ]
2471 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2473 SimGrid (3.0.1) stable; urgency=low
2476 * Unfortunately, I had missed 5 misnamed functions:
2477 xbt_fifo_item_t xbt_fifo_newitem(void);
2478 void xbt_fifo_freeitem(xbt_fifo_item_t);
2479 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2480 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2481 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2482 They're now deprecated. Please use their new versions:
2483 xbt_fifo_item_t xbt_fifo_new_item(void);
2484 void xbt_fifo_free_item(xbt_fifo_item_t);
2485 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2486 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2487 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2489 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2490 * Documentation: xbt_log module unmercifully reworked [MQ]
2491 * Bugfix: there was a problem with the ending of contexts with
2492 the pthread backend. It caused some weird deadlock or behavior
2493 depending on the pthread implementation. [AL]
2494 * Bugfix: get the exceptions raised in the simulator repport where
2495 and why they come from when they are not catched in time [AL, MQ]
2498 * Bugfix: Do repport the error when two non-connected hosts try to
2499 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2502 * Add additionnal checkings on communications. Assert that two
2503 communicating hosts are connected by a set of links... [AL]
2506 * Add additionnal checkings on channel values in communication [AL]
2507 * New: MSG_task_get_source to see on which host a task was generated [HC]
2508 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2509 the number of tasks waiting to be received on channel and sent
2511 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2512 waits for the first task coming from a given host.. [AL]
2514 GRAS new functionnalities: [MQ]
2515 * Enhance the parsing macro to allow the size of multidimentional objects
2516 to be given thru annotations.
2517 * New example (and documentation): Matrix Multiplication a la RPC
2518 (as when I was young!) and fix a bunch of bugs found on the way.
2520 GRAS performance improvements: [MQ]
2522 * Reduce the amount of cbps creation/destruction by making it static to
2523 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2526 * Change libdata to a set so that we can search for stuff by ID (and thus
2527 reduce the insane amount of dict lookups)
2530 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2531 RL communication on top of SG since it's so uneffective.
2532 It may also be used for inter-thread communication in RL, one day.
2533 * Use gras_datadesc_copy() to exchange messages on top of SG
2535 - improve message exchange performance on top of SG
2536 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2537 * Don't exchange on the network the size of the used part of buffer,
2538 instead, specify the possible buffer size to read().
2540 - reduces the amount of read/write calls (one pair per exchange)
2541 - reduces the amount of exchanged data (the size)
2542 - allows to retrieve all arrived data on receiver side, if we don't need
2543 it right now (subsequent read will peek the buffer)
2544 - allows the receiver to proceed with the begining of the stream before
2545 everything is arrived
2546 - make it possible to build an iov transport (using readv/writev)
2548 - take care of the data with non-stable storage (like stacked data),
2550 * If possible, TCP send uses vector I/O (when writev() is here)
2551 - Don't use it for receive since we send data sizes and data on the
2552 same stream, so we wouldn't be able to chain large amount of chunks
2553 before having to flush the stuff to read the size.
2554 * Rework the transport plugin mechanism to simplify it and reduce the
2555 amount of pointer dereferencement when searching for the right function
2558 * I guess that now, we do almost as few system calls as possible while
2559 doing as few data copy as possible.
2561 To improve it further, we could try to send all the sizes first and then
2562 all the data (to use iov on receiving size), but it's only a partial
2563 solution: when you have 2 dimensional data, the sizes of the second
2564 dimension is data of the first dimension, so you need 3 streams.
2566 I'm not sure the potential performance gains justify the coding burden.
2568 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2570 SimGrid (3.00) stable; urgency=high
2573 * New! Give the possibility to hijack the surf parser and thus bypass
2574 MSG_create_environment and MSG_launch_application. Have a look at
2575 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2577 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2579 SimGrid (2.96) unstable; urgency=low
2584 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2585 This deprecates the xbt_error_t mechanisms.
2586 It modifies (simplifies) all XBT and GRAS API.
2587 MSG API keeps unchanged (exceptions raised by XBT are catched from
2588 within MSG and masked with existing error handling facilities)
2591 * New! Add a FATPIPE model. [AL]
2592 * New! Add a parallel task model. [AL]
2593 * New! Add automatically a loopback interface (in the default
2594 network model) if none was precised.
2597 * Bugfix: MSG_process_resume now works with the current running process.
2599 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2600 * Modification of MSG_task_get_compute_duration. Once a task has been
2601 processed, the value returned by this function is now equal to 0. [AL]
2602 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2603 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2604 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2605 of a m_task. For now, MSG_task_cancel only works with computation
2607 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2608 of the processor (in Mflop/s) regardless of the current load on the
2610 * API Change: use proper naming convention for MSG_getClock and
2611 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2613 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2614 This function changes the priority of a computation task. This priority
2615 doesn't affect the transfer rate. A priority of 2 will make a task
2616 receive two times more cpu power than the other ones. This function
2617 has been added to suit the needs of Nguyen The Loc and hasn't been that
2618 much tested yet. So if it fails, please report it and send me your code.
2620 * API Change: removed all functions and types that were marked "deprecated"
2621 since many months. Renamed MSG_global_init_args to MSG_global_init.
2623 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2625 SimGrid (2.95) unstable; urgency=low
2628 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2629 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2630 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2633 * Bugfix: Allow absolute paths for platform description files [MQ]
2634 * Bugfix: do free the variables after use. Leads to drastic performance
2636 * Implement max_duration (ie, timeouts) on resources [AL]
2639 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2641 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2642 the needed support (timeouts) [AL]
2645 * Implement measurement sockets. You can now get the bandwidth between two
2646 hosts thanks to AMOK (see below). [MQ]
2647 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2648 dynar over the network (yeah) [MQ]
2649 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2650 * simple/static token-ring example. [Alexandre Colucci and MQ]
2651 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2652 in place before [MQ]
2655 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2656 type parsing. "short" and "long" modifiers were also ignored; other
2657 modifier (such as reference level) are still ignored. [MQ]
2658 * Embeed the buffer size within the buffer itself on SG. [MQ]
2659 That way, send() are atomic and cannot get intermixed anymore (at least
2660 the ones which are less than 100k; bigger messages still have the issue)
2661 * Array size pushed by the field, not by the field type (or each
2662 and every long int will push stuff to the cbps) [MQ]
2663 * use select() to sleep since it allows to portably sleep less than one
2666 GRAS (minor cleanups)
2667 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2670 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2671 * type category 'ignored' killed as it was never used and were difficult
2673 * whether a type can cycle or not is now a flag, leaving room for more
2674 flags (as "ignored", if we feel the need one day ;)
2675 * Rename raw sockets to measurement sockets since "raw" has another
2676 meaning in networking community.
2679 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2680 and offers features not belonging to GRAS but that most applications
2681 need. One day, it may be a set of plugins loadable at runtime.
2682 * New module: bandwidth
2683 bandwidth measurement between arbitrary nodes running this module. [MQ]
2685 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2687 SimGrid (2.94) unstable; urgency=low
2689 The first beta release of SimGrid 3 !
2693 * Update the main page and the FAQ. Adding references to gforge.
2696 * Add a gras_os_getpid function.
2699 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2700 * Extend the logs so that they also print PID, hostname, date, ... if
2702 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2703 and kill the old version which were in testsuite/
2704 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2708 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2709 for users. Instead, each programming environment (GRAS, MSG,...) use it
2710 when needed to provide such a feature to users.
2711 Don't shortcut the mechanism or you will also shortcut the virtualization
2712 you need on the simulator.
2716 * Cleanups in configury with regard to compile optimization/warning flags.
2717 Also add -fno-loop-optimize to any powerpc since it's the optimization
2718 killing gcc (< 3.4.0).
2719 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2720 needed by MSG examples complications
2721 * Borrow configury beautifications from PHP
2724 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2725 warning (thanks loris for stumbling into this one).
2726 * Bugfix: stop loading private headers (gras_config.h) from the public
2730 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2731 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2732 RL would imply a DNS resolver, which is *hard* to do in a portable way
2733 (and therefore delayed).
2734 * Implement a real timer mechanism and use it in timing macros. This allows
2735 to avoid rounding errors and get a 0.000005 sec precision in timing
2736 macros. While I was at it, various cleanups:
2737 - allow to declare more than one timed section per file (fix a stupid bug)
2738 - move some private declaration to the right place
2739 - merge conditional execution and timing macros into emulation module
2740 - document the module
2741 - make sure the module cleanups its mess on gras_exit
2742 * Documentation improvements:
2743 - (new) how to compile applications using GRAS
2744 - (new) emulation support (timing macros)
2746 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2748 SimGrid (2.93) unstable; urgency=low
2750 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2753 - Use Paje properly where used. Still to be sanitized properly.
2754 - Portability fix: Add an implementation of the contexts using pthread
2758 - Add xbt_procname(): returns the name of the current process.
2759 Use it to show the current process's name in all logging.
2761 - fix detection of older flex version and the reaction, since we do
2762 depend on modern ones (we use lex_destroy)
2763 - Better separation of SG and RL in the libs: remove all simulation code
2764 from libgras. As a result, this lib is now only 200k when stripped.
2765 Some of the xbt modules may also be duplicated (two sets and such) and
2766 should be cleaned/killed before SG3.
2767 - Insist on using xlC on AIX because of weird problems involving gcc there.
2768 - Cleanup the make remote stuff. This is now done by scripts
2769 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2770 mainly for our private use, but we're working on changing them to user
2773 - Bugfix: flush the socket on close only if there is some *output*.
2774 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2776 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2779 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2781 SimGrid (2.92) unstable; urgency=low
2783 Alpha 3 on the path to SimGrid 3
2787 - New! First try of benchmarking macros.
2788 - New! First try so that gras_stub_generator generate deployment and
2789 remote compilation helpers.
2791 - Bugfix: Initialization fix in msg_test.
2795 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2797 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2799 - API Change: gras_os_sleep to take the amount of seconds as a double.
2800 Accepting an int was error prone since it was the only location where
2801 seconds were coded as such. It leaded to damn rounding errors.
2802 - Bugfix: Hard to belive that timers ever worked before this.
2804 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2806 SimGrid (2.91) unstable; urgency=low
2808 Alpha 2 on the path to SimGrid 3
2812 - Bug fix in the lmm_solver.
2814 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2815 through the function MSG_paje_output.
2816 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2817 - It is possible to bound the rate of a communication in MSG with
2818 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2819 to write it in the changelog).
2820 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2825 - Various cleanups to the autotools stuff
2826 - Begin to move Gras examples to examples/gras/
2827 - Let make distcheck work again (yeah!)
2829 - documentation overhauled using doxygen.
2830 gtk-doc-tools is dead in SimGrid now.
2831 - Automatically extract all existing logging categories, and add the list
2832 to the documentation (long standing one, to say the less)
2834 - Cleanup the known architecture table. Reorder the entries to group what
2835 should be, and use a more consistent naming scheme.
2836 (some of the test dataset are still to be regenerated)
2837 - New! Allow library to register globals on each process just as userdata
2839 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2840 lookup time (for now).
2841 Use it in msg and trp.
2842 This cleans a lot the internals and helps enforcing privacy of the
2843 headers between the gras components.
2844 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2845 - Bugfix: gras_os_time was delirious in RL.
2846 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2848 - Reenable GRAS now that it works.
2850 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2852 SimGrid (2.90) unstable; urgency=low
2854 Alpha 1 on the path to SimGrid 3
2856 * It is a long time since the last release of SimGrid. I'm sorry about
2857 that but as I had told you, I was rewriting a lot of things. I apologize
2858 to those who had been reporting bugs to me and that I had not answered.
2859 If your bug is still in the new version, please tell me. Here is a
2860 summary of the main changes.
2862 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2863 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2864 portability, speed, and a lot more but you'll figure it out later.
2865 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2866 I wanted to release the new MSG as soon as possible and I have
2867 broken GRAS, which is the reason why, for now, only MSG is fully
2868 functional. A laconic description of these projects is available
2869 in the documentation.
2871 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2872 kernel optimized for our needs (called SURF but only the developers
2873 should use it). Hence, MSG is now roughly 30 times faster and I think
2874 that by rewriting a little bit MSG, I could event speed it up a little
2875 bit more. Beside the gain in speed, it is also much easier to encode a
2876 new platform model with SURF than it was with SG. More to come...
2878 * REVOLUTION 3: I have tried to change a little as possible the API of
2879 MSG but a few things really had to disappear. The main differences
2880 with the previous version are :
2881 1) no more m_links_t and the corresponding functions. Platforms are
2882 directly read from a XML description and cannot be hard-coded
2883 anymore. The same format is used for application deployment
2884 description. The new format is described in the documentation.
2885 Have a look in tools/platform_generation. There is a tiny script
2886 that converts from the old platform format to the new one. Concerning
2887 the application deployment format, parsing the old one is tricky.
2888 I think most of you should however be able to convert your files. If
2889 it is really an issue, I can write a C code that does the conversion.
2891 2) the toolbox tbx does not exist anymore. We now have a library
2892 with much more data-structures but without the hash-tables (we have
2893 dictionaries that are much faster).
2895 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2897 *****************************************************************************
2898 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2899 * don't feel like converting the oldies. (Mt) *
2900 *****************************************************************************
2903 Version 2.90: "the long awaited one"
2904 - Finished rewriting and debugging MSG. Rewrote the documentation.
2905 - disable GRAS for now since it needs to be ported to the newest SG
2908 - Finish the port to windows (using mingw32 for cross-compile)
2911 - Main loop and datastructures of SURF. A cpu resource object is
2912 functional. Surf can thus be used to create cpu's with variable
2913 performance on which you can execute some actions.
2915 2004-11-15 Martin Quinson
2916 - Port to ARM. Simply added the alignment and size descriptions. Should
2917 work, but the ARM machines are so slow that I didn't had the opportunity
2918 to 'make check' over there yet.
2920 2004-11-15 Arnaud Legrand
2921 - Trace manager now written. It uses a heap structure and is therefore
2922 expected to be efficient. It may however be speeded up (particularly
2923 when many events occur at the same date) by using red and black
2924 trees. One day maybe...
2925 - Max-min linear system solver written. It uses a sparse matrix
2926 structure taking advantage of its expected use. Most operations are
2927 O(1) and free/calloc are called as few as possible. The computation of
2928 the minimum could however be improved by using a red and black tree
2931 2004-11-03 Arnaud Legrand
2932 - Rename every gras_* function that was in xbt/ to its xbt_
2934 - Add a heap and a doubly-linked list to xbt
2935 - Added a dichotomy to the dictionaries. make check works as well before
2936 so I assume that the patch is correct. I do not know however if things
2937 run effectively faster than before now. :)
2939 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2940 SimGrid, and the version number is bumped to 2.x
2942 2004-10-29 Martin Quinson
2943 - Introduction of the remote errors.
2944 They are the result of a RMI/RPC on the remote machine.
2945 ErrCodes being scalar values, you can't get the host on which those
2946 errors did happen. Extending the error mechanism as in Gnome is possible.
2947 No idea yet whether it is a good idea.
2949 2004-10-28 Martin Quinson
2950 - Interface revolution: the Starred Structure Eradication.
2951 I used to do typedef struct {} toto_t; and then handle *toto_t.
2952 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2953 - ???_t is a valid type (builded with typedef)
2954 - s_toto_t is a structure (access to fields with .)
2955 - s_toto is a structure needing 'struct' keyword to be used
2956 - e_toto_t is an enum
2957 - toto_t is an 'object' (struct*)
2959 typedef struct s_toto {} s_toto_t, *toto_t;
2960 typedef enum {} e_toto_t;
2961 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2962 s_toto_t) is private.
2964 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2965 so that it can set the variable to NULL. It was so for dicts and sets,
2966 it changed for dynars.
2968 - Fix a bunch of memleaks in dict_remove
2969 - Fix a bug in sg/server_socket opening: it failed all the time.
2971 2004-10-07 Martin Quinson
2972 - Speed up dynar lookup operation a bit.
2974 gras_dynar_get is dead.
2976 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2977 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2978 (which gives you the address of the stored data).
2980 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2981 copy of the data using an affectation to do the job and not a memcpy.
2983 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2984 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2986 It does not really speedup the dynar test because they are
2987 setting elements all the time (and look them seldom). But the dict does
2988 far more lookup than setting.
2990 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2992 2004-10-05 Martin Quinson
2993 - Allow to (en/dis)able the cycle detection at run time.
2995 Whether we should check for cycle or not is now a property of each
2996 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2997 datadesc_cycle_unset allow to remove this property when previously set.
2999 Note that the cycle detection is off by default since it impacts the
3000 performance. Watch the data you feed GRAS with ;)
3002 This property is hereditary. Any element embedded in a structure having it
3003 set have it set for the time of this data exchange.
3005 You should set it both on sender and receiver side. If you don't set it on
3006 sender side, it will enter an endless loop. If you forget on receiver
3007 side, the cycles won't be recreated after communication.
3009 - Header reorganization.
3010 Kill gras_private.h, each submodule must load the headers it needs.
3012 2004-10-04 Martin Quinson
3013 - Interface revolution: do not try to survive to malloc failure.
3015 Now, gras_malloc and friends call gras_abort() on failure.
3016 As a conclusion, malloc_error is not a valid error anymore, and all
3017 functions for which it was the only gras_error_t return value are
3018 changed. They now return void, or there result directly.
3019 This simplify the API a lot.
3021 2004-09-29 Martin Quinson
3022 - Re-enable raw sockets.
3023 Created by gras_socket_{client,server}_ext;
3024 Used with gras_raw_{send,recv}
3027 It should allow to kill the last bits of gras first version soon.
3029 This is not completely satisfactory yet (duplicate code with
3030 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3033 - Simplify transport plugin (internal) interface by not passing any
3034 argument to _server and _client, but embedding them in the socket
3037 2004-09-28 Martin Quinson
3038 - Finish the port to AIX.
3039 autoconf was my problem (segfault within the malloc replacement
3040 function. No idea why)
3042 2004-09-16 Martin Quinson
3043 - Fix some size_t madness on 64bit architectures.
3045 2004-09-08 Martin Quinson
3046 - Reduce the number of system headers loaded, overload some more system
3047 calls (such as malloc to cast the result of the system one, and work
3049 - Fix and reintroduce the config support
3051 2004-09-07 Martin Quinson
3052 - Source code reorganization to allow Arnaud to surf all over there.
3053 - Allow to document the logging categories.
3054 - Remove all uppercase from logging categories and useless cleanup in names.
3056 2004-08-18 Martin Quinson
3057 Version 0.6.2 (protocol not changed; API changed)
3058 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3059 gras_error_t), and NULL when not found. Functions expecting a msgtype
3060 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3061 hopefully usefull message.
3062 - Portability to prehistoric sparcs again
3064 2004-08-17 Martin Quinson
3065 Version 0.6.1 (protocol not changed; ABI not changed)
3066 - prealloc some buffers to speed things up
3068 2004-08-11 Martin Quinson
3069 Version 0.6 (protocol not changed; ABI expended)
3070 - The parsing macro can deal with the references, provided that you add
3071 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3073 2004-08-09 Martin Quinson
3074 Version 0.5 (protocol not changed; ABI changed)
3075 - Allow to off turn the cycle detection code in data exchange at
3076 compilation time. It should be at run time, but I'm short of time (and
3077 the config stuff is still broken). That way, we keep dict out of the
3078 critical path, which is good because the performance is poor:
3079 - search not dichotomial yet
3080 - dynar give no way to access their content and memcpy everytime
3081 - In composed data description (struct, ref and so on), stop foolness of
3082 keeping the subtype's ID, but store the type itself. This keeps sets out
3083 of the critical path, which is good since they rely on dynar and
3084 dictionnaries. The only loose of that is that we cannot detect the
3085 redeclaration of a structure/union with another content (but I'm not sure
3086 the code detected well this error before anyway). We still can detect
3087 the redefinition discrepancy for the other types.
3088 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3089 it breaks the code because of type-punning used all over the place).
3090 This breaks on all non-gcc architectures (for now).
3092 All those changes (plus the buffer of last time) allow me to gain 2 order
3093 of magnitude on cruel tests consisting of 800000 array of integers on two
3094 level of a hierarchical structure (200 secondes -> 4 secondes)
3097 - the selector of reference must now return the type it points to, not
3098 the ID of this type.
3100 2004-08-06 Martin Quinson
3101 Version 0.4 (protocol changed; ABI not changed)
3102 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3103 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3104 - Implement a buffer transport, and use it by default (it relies on tcp in
3105 real life and on sg in simulation).
3106 That's a bit hackish since I had a new field to the structure to store
3107 its data without interfering with the subtype ones. Inheritance
3108 is tricky in C. And that's a kind of reverse inheritance with one class
3109 derivating two classes. Or maybe a game with java interfaces. Anyway,
3110 that's damn hard in C (at least).
3111 Moreover, I got tired while trying to ensure plugin separation and
3112 genericity in SG mode. MSG wants me to do weird things, so let's go for
3113 cruel hacks (temporarily of course ;).
3114 See comment in transport_private.h:71
3115 - do not include all the _interface headers in private but in the files
3116 which really need them (to cut the compilation time when they are
3119 2004-07-26 Martin Quinson
3120 Version 0.3 (protocol not changed; ABI changed)
3121 - Major overhault of the datadesc interface to simplify it:
3122 - shorted the function names:
3123 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3124 - add a trivial way to push/pop integers into the cbps without malloc.
3125 This allows to make really generic sub_type description, which simply
3126 pop their size of the stack.
3127 - add a function gras_datadesc_ref_pop_arr() which does what users want
3128 most of the time: Declare a dynamic array (which pops its size of the
3129 stack) and declare a reference to it. Poor name, but anyway.
3130 - kill the post-send callback, add a post-receive one
3132 2004-07-23 Martin Quinson
3133 Version 0.2 (protocol changed; ABI changed)
3134 - add some testing for cpbs in the test cases, and fix some more bugs.
3135 This invalidate again the little64 data file, since I cannot regenerate
3137 - remove an awfull optimization in the logging stuff, allowing me to:
3138 - understand it again
3139 - learn gcc how to check that the argument match the provided format
3140 - fix all errors revealed by gcc after that
3141 - internal keys of dict are not \0 terminated. Deal with it properly in
3142 loggings instead of segfaulting when the user want to see the logs :-/
3144 2004-07-22 Martin Quinson
3145 - Fix some stupid bug preventing cbps (callback postit) from working
3147 2004-07-21 Martin Quinson
3148 - Some documentation cleanups
3149 - remove the useless last argument of msgtype_declare
3150 - rename the Virtu functions to fit into the 'os' namespace
3151 - move headers src/include -> src/include/gras/ and stop fooling with
3152 gras -> . symbolic link
3153 - make distcheck is now successful
3155 2004-07-19 Martin Quinson
3157 - Build shared library also
3158 - Install html doc to the right location
3159 - stop removing maintainer files in make clean
3160 - build tests only on make check
3162 2004-07-13 Martin Quinson
3164 - No major issue in previous version => change versioning schema
3165 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3166 corresponding dataset.
3168 2004-07-11 Martin Quinson
3170 - Get it working with any kind of structure (we can compute the padding
3171 bytes remotely for all the architectures I have access to)
3172 - Implement the structure parsing macro (still not quite robust/complete)
3173 - Improvement to the remote testing toysuite
3175 2004-07-10 Martin Quinson
3176 [autoconf mechanism]
3177 - get ride of a bunch of deprecated macros
3178 - actually run the test for two-compliment, not only compile it :-/
3179 - test whether the structures get packed (and bail out if yes. Damn.
3180 Alignment is a serious matter)
3181 - test whether the structures get compacted (but respecting the alignment
3182 constraints of each types)
3183 - test whether the array fields of structures can straddle alignment boundaries
3185 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3186 (took me 2 hours to find that bug, looking at the wrong place)
3187 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3188 before sending/receiving and are used to compute the offsets of fields
3189 - Given that padding size depend even on compiler options, keep track of
3190 alignment and aligned_size only for the current architecture. Not a big
3191 deal since we send structure fields one after the other (seems
3193 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3194 program, let it work on linux/gcc/little32. portability todo.
3196 2004-07-08 Martin Quinson
3197 - import and improve remote compilation support from FAST
3198 - make sure make check works on half a dozen of machines out there
3200 2004-07-07 Martin Quinson
3201 Let's say it's version 0.0.3 ;)
3202 - Implement conversions (yuhu!)
3203 - Let it work on solaris (beside conversion, of course)
3204 - Stupid me, using rand() to generate the conversion datatests in not wise.
3206 2004-07-06 Martin Quinson
3207 - Let make dist work, since I'm gonna need it to compile on remote hosts
3208 - Let Tests/datadesc_usage write the architecture on which the file was
3209 generated as first byte.
3210 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3211 architecture descriptions.
3212 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3213 on those architectures.
3214 - Optimization: send/recv array of scalar in one shoot
3216 2004-07-05 Martin Quinson
3217 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3219 - Plug a whole bunch of memleaks
3220 - each process now have to call gras_{init,exit}. One day, their log
3221 settings will be separated
3222 - Continue the code factorisation between SG, RL and common in Transport.
3224 2004-07-04 Martin Quinson
3226 - Redistribution between SG and RL.
3227 We wanna have to accept in SG, so move accepted related parts of RL in
3228 the common part. (more precisely, the dynar of all known sockets is no
3229 more a static in transport.c, but part of the process_data)
3231 [gras_stub_generator]
3232 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3234 2004-07-03 Martin Quinson
3235 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3236 them all up in one shot)
3238 - Fix the ugly case of reference to dynamic array.
3239 - New (semi-public) function gras_datadesc_size to allow the messaging
3240 layer to malloc the needed space for the buffer.
3242 - gras_socket_close now expect the socket to close (and not its address to
3243 put NULL in it after it). This is because the socket passed to handlers
3244 is one of their argument (=> not writable).
3246 - propagate the interface cleanup from last week in datadesc, ie remove a
3247 superfluous level of indirection. User pass adress of variable
3248 containing data (both when sending and receiving), and not of a variable
3249 being a pointer to the data. Let's say that I like it better ;)
3250 The price for that is constructs like "int msg=*(int*)payload" in
3251 handlers, but it's a fine price, IMHO.
3253 - Let it work in RL (yuhu)
3255 2004-06-21 Martin Quinson
3257 - porting SG plugin and SG select to new standards (works almost).
3258 - plug memleaks and fix bugs around.
3261 - cleanup the prototype of data recv and force users to specify when they
3262 want to handle references to objects. Test case working even for cycles.
3263 - plug memleaks. Valgrind is perfectly ok with this.
3265 2004-06-12 Martin Quinson
3267 - cleanup the separation between plugin and main code in plugin creation
3269 2004-06-11 Martin Quinson
3271 - Reput hook for raw sockets, needed for BW experiments
3272 - kill a few lines of dead code
3273 [Data description] Interface cleanup
3274 - gras_datadesc_by_name returns the searched type or NULL.
3275 That way, no variable is needed to use a type desc once, which makes
3277 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3278 two parameters were strings (field name, type name), leading to
3280 [Dicos] Interface cleanup
3281 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3282 This is consistant with the dynar API.
3284 2004-04-21 Martin Quinson
3286 - Porting to new standards.
3288 - interface cleanup.
3289 There is no bag anymore, no need to take extra provision to mask the
3290 pointers behind "ID".
3291 Better splitup of functions between files create/exchange/convert.
3292 This is still a bit artificial since convert and receive are so
3293 interleaved, but anyway.
3295 - add a queued message list to procdata (the ones not matching criteria
3297 - factorize some more code between SG and RL wrt procdata
3299 - use gras_exit in example to track memleaks
3300 - get rid of gs_example now that GS is properly integrated into gras
3301 - update run_test to integrate the latest tests (datadesc)
3303 - rename WARNINGn macros to WARNn since it prooved error-prone
3305 2004-04-19 Martin Quinson
3307 - register init/exit functions within gras module mechanism
3308 - send/receive function.
3309 Convertion is not implemented, but short-cutted if not needed.
3310 struct/array elements are sent one by one (instead of block-wise), but
3311 nobody really cares (yet). Get a prototype before optimizing.
3312 - tests (using a file socket) for DD send/receive on:
3313 - base types: int, float
3314 - array: fixed size, string (ie ref to dynamic string)
3315 - structure: homogeneous, heterogeneous
3316 - chained list, graph with cycle
3317 Believe it or not, valgrind is not too unhappy with the results. The
3318 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3322 - Bugfix when using a filename explicitely (instead of '-')
3324 2004-04-09 Martin Quinson
3326 - factorize more code between RL and SG in socket creation
3327 - Complete the implementation and tests of:
3329 o file (only in RL, and mainly for debugging)
3331 I lost 3 days to design a portable address resolver, and then decided
3332 that the prototype mainly have to run on my box.
3333 Addressing portability too early may be like optimizing too early :-/
3335 - use gras_init in the Tests instead of the crappy parse_log_opt
3336 (the latter function is removed)
3337 [Conditional execution]
3338 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3339 [Code reorganisation]
3340 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3341 Build examples against the RL library, since there is no way to disable
3342 its creation for now.
3344 For information, the beginning of coding on GRAS was back in june
3345 2003. I guess that every line has been rewritten at least twice since