1 SimGrid (3.15) stable; urgency=low
3 The Spring Release: continuous integration servers become green
4 We fixed even the transient bugs on all target architectures:
5 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
6 Windows, FreeBSD, NetBSD.
8 - Rename NetCards to NetPoints.
9 This was intended to help NS3 users, but that's not a netcard.
10 That's a point in the routing algorithm, let's avoid wrong simplifications.
13 - New: MSG_process_yield(). Stop and yield to other processes.
14 - New: MSG_process_daemon(). Daemon processes are automatically killed
15 when the last non-daemon process terminates
16 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
17 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
18 to make room for the upcoming network energy plugin.
19 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
22 - Ensure that an actor can kill itself with Process::exit()
23 - Kill the obscure NativeException. Nobody want to survive the issues
24 it denotes, so use JniException that is a RuntimeException (not to
25 be caught explicitly).
26 - Partial bug fix in initialization. SimGrid flags on command line were
27 consumed at C level but stayed in the original Java String[] args.
28 This could mess users' args[i] if SG flags were not put at the end of
30 The SimGrid flags are now removed from the Java arguments. However,
31 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
32 if args.length is greater than the number of YOUR OWN ARGUMENTS.
33 It might be if you have --log or --cfg flags in the command line.
34 - Fix numerous memleaks all around the place. In particular, around VMs.
38 - s4u::onPlatformCreated: right before the simulation starts
39 - s4u::onSimulationEnd: right after the main simulation loop
40 - s4u::onTimeAdvance: right after a clock change (time is discrete)
41 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
42 event from the availability_file changes the avail speed.
43 - Links are now usable from s4u
44 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
45 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
46 - New examples: The conversion of MSG examples to S4U has begun
47 - Actors: create, kill, migration, and suspend
48 - Applications: master-worker and token-ring
49 - Action replay: communications and storage
50 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
53 - Backwards Compatibility breaks
54 - The SD_route_get_list and SD_route_get_size functions have been removed.
55 They are replaced by the unique
56 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
57 The route from src to dst is built in the links dynar whose size can
58 be obtained with xbt_dynar_length.
59 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
60 sg_host_route_bandwidth and sg_host_route_latency.
61 Macros ensure the backwards compatibility, but you should fix your code
64 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
65 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
66 - Bump our claim of support from MPI 1.1 to MPI 2.2.
67 We don't support 100% of it, but it should be enough. Contact us if not.
68 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
69 naive and probably inaccurate)
70 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
71 mapped onto a unique small file using some system magic.
72 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
73 - Fortran: cleanups, fixes, support of user-added operations
74 - RMA: MPI_Accumulate are applied in correct order.
75 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
76 - Better support for MPI_IN_PLACE
77 - Support for MPI_Win attrs and keyvals.
78 - Support MPI_Comm_group_create, MPI_Type_size_x
79 - MPI ops should be properly applied to all allowed types
80 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
81 algorithm by a non failing one from mpich.
84 - Kill the fifo data container: we don't use it anymore.
87 - A <cluster> can now be created with different speed values to represent pstates.
88 see examples/platforms/energy_cluster.xml for an example.
90 -- Wed Mar 22 15:36:09 CET 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
92 SimGrid (3.14.159) stable; urgency=low
94 The previous release was only a crude approximation of a Pi release;
95 this one should be a bit better.
97 - Revert a last minute change that broke on Mac OSX.
98 - Fix the changelog and NEWS files.
99 - Improve the documentation, in particular of the routing module.
100 - Integrate some patches flying around in Debian.
102 -- Wed Dec 28 17:14:49 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
104 SimGrid (3.14) stable; urgency=low
107 * The whole documentation was reworked and reorganized. There is
108 still much room for improvement here, but we're on it.
111 * We now need python3 for our testing infrastructure.
112 * Model checking needs libevent
115 * Now works on FreeBSD
118 * Privatization now works on FreeBSD
120 * Privatization is now activated by default in smpirun
121 You can turn it off with -no-privatize if something goes wrong.
123 * Call-location tracing for SMPI.
124 You can add the exact location (filename / linenumber) of an MPI call to
125 your trace files and slow down or speed up the simulation between two
126 consecutive calls by using an adjustment file (see the documentation).
128 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
129 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
130 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
131 values were ignored in some cases. The timings of these functions can now
132 be significantly different.
134 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
135 smpi/running-power is renamed to smpi/host-speed
137 * smpi/grow-injected-times option to enable or disable multiplication of the
138 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
139 make simulation less precise (but also much faster).
141 * smpirun script should be (much) faster for large deployments.
143 * SMPI tracing : fixed issue with poor matching of send/receives.
145 * Replay : Fix broken waitall
147 New functions and features
148 * MSG_parallel_task_execute_with_timeout, to timeout computations.
150 Dropped / renamed functions and features
151 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
152 - MSG_mailbox_is_empty() -> Mailbox::empty()
153 - MSG_mailbox_front() -> Mailbox::front()
154 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
155 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
156 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
157 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
158 * Don't pass the free_f parameter to property related functions:
159 - MSG_host_set_property_value()
160 - MSG_as_router_set_property_value()
161 - MSG_storage_set_property_value()
162 * VM properties. Since msg_vm_t are msg_host_t, just use
163 MSG_host_get_property_value() and friends
164 * VM I/O related things:
165 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
166 - Unimplemented save/restore methods
167 * MSG_as_router_get_property_value() was redundent with
168 MSG_environment_as_get_property_value().
169 - Removed MSG_as_router_*propert*() functions
170 - Added MSG_environment_as_set_property_value() for consistency
171 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
172 Well, they are still there for now, but deprecated with a warning.
173 Please switch to xbt_dict_new_homogeneous() before this is removed
175 * Task affinity. Its intended behavior (that was very badly tested
176 and probably not really working) was deceiving what most users
177 would have hoped here.
178 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
179 * Option network/coordinates is now useless and should be dropped.
182 * Added option storage/max_file_descriptors to allow more than 1024 files opened
185 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
186 * Added Dragonfly topology support
189 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
190 EiBps, EBps, Eibps, Ebps for bandwidth.
191 They may become useful to some lucky ones.
194 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
195 * ex/app_tokenring: new example, very similar to the MSG Token Ring
196 * ex/async_waitAll: new example, on asynchronous communications
199 * Memory usage should be decreased for simulations with a large number
200 of processes. This also helps for SMPI.
202 -- Sat Dec 24 22:34:26 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
204 SimGrid (3.13) stable; urgency=low
206 The Half Release, a.k.a. the Zealous Easter Trim.
208 - We removed half of the lines, that were mostly experimental cruft.
209 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
210 (not counting blanks and comments -- according to openhub.net)
211 - The internals are now compiled in C++ (and will soon be clean C++)
212 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
213 59 klines of C, etc. We added only 29 klines of C++ in replacement.
215 * Backwards Compatibility breaks
216 - Removed Lua simulation bindings (switch to C or Java for that).
217 Lua can still be used to describe platforms
218 - Removed Java kernel plug-ins.
219 Will be reintroduced after the ongoing major internals reorg.
221 - the following functions were removed.
222 They were too specific and should be reimplemented in a generic
223 way, with filter function.
224 - MSG_task_listen_from_host
225 - MSG_mailbox_get_count_host_waiting_tasks
226 - MSG_mailbox_put_with_timeout was removed.
227 Please use MSG_task_send_with_timeout instead.
229 - the SD_application_reinit function was removed. It has been a noop for a while.
230 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
231 replaced by a more flexible mechanism.
232 - The following functions thus do not exist anymore
233 - SD_workstation_get_access_mode
234 - SD_workstation_set_access_mode
235 - SD_workstation_get_current_task
236 - Basic estimation functions have been removed but can easily be replaced
237 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
238 - SD_workstation_get_computation_time => amount / sg_host_speed()
240 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
241 Use VM.getSpeed()*load/100 for the legacy behavior.
243 - option enable_tracing was removed. It was not doing anything for a while.
244 - In the ModelChecker:
245 - the model-checker now ptraces the model-checked process which means
246 you cannot use a debugger on the latter anymore (we might make this
247 optional in the feature);
248 - removed soft-dirty page tracking;
249 - remove model-checked side snapshot management,
250 MC_snapshot() and MC_compare_snapshot();
251 - keep the MC_cut() function as a stub (it was not really working
252 in the previous release).
255 * All options are consistently in kebab-case. Old names are kept as alias.
258 * Switch to platform v4 format.
259 - Rename from 'power' to 'speed' the attributes describing the amount of
260 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
261 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
262 - In <host> and <link>, attributes availability and state are gone.
263 It was redundent with state and availability traces, and with peak values.
264 - In <cluster>, attributes availability_file and state_file are gone.
265 It was too complex and unused.
266 - Kill <gpu>. Was not doing anything.
267 - The DOCTYPE points to the right URL:
268 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
269 (the file at this address now documents the changelog since its v1)
270 - A warning is emitted for unit-less values (they are still accepted).
271 - speed. Default: 'f' or 'flops'. Also defined:
272 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
273 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
274 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
275 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
276 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
277 - latency. Default: 's' second. Also defined:
278 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
280 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
281 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
284 * s4u::Host is now the preferred public interface to the Host features.
285 sg_host_* functions are C bindings to the exact same behavior
286 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
289 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
290 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
291 They were deprecated since a few years, and probably did not even compile anymore.
294 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
295 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
296 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
297 encouraged to update their codes. A list of the modified functions can be found at the end of
298 include/simgrid/simdag.h
301 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
302 the lambda or closure passed as a parameter will run in kernel mode.
303 All the callback functions should be rewritten to that interface at some point.
306 * Reorganizing and cleaning the internals all around the place.
309 * Remove old default barrier/bcast buggy algorithms (see #18407)
310 * Various bug fixes to handle more codes
311 * Remove the need for the --foreground option of smpirun (it is still
312 accepted for backward compatibility).
315 * Kill the setset data container: MC don't use it anymore.
316 * Kill the queue data container: it made more sense with GRAS.
317 * Kill the xbt_peer_t data type: it's useless without GRAS.
318 * Kill rm_cb feature of config sets: it was never useful.
319 * Kill graphxml parsing feature. It was not used.
320 * Kill the deprecated code protected by XBT_USE_DEPRECATED
322 - xbt_dynar_sort_strings(), when the content is char*
323 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
324 They throw exceptions on invalid input;
325 * C++ support for declaring CLI flags (simgrid::config::Flag);
326 * class for abstracting different signal backends (simgrid::xbt::signal).
327 with no external dependencies (we need very simple signals).
330 * refactoring and cleanup of the code;
331 * ongoing process to cleanly separate the model-checking algorithms
332 from the code model-checking support.
334 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
336 SimGrid (3.12) stable; urgency=low
338 The Facelift Release.
341 * Require g++ v4.7 at least to not speak prehistorical C++.
342 * Require Boost 1.48 (for signal2 component).
343 * Java must be version 7 at least when activated.
344 * Builds on Windows again (including Java bindings).
345 * Tracing is now always enabled (no way to turn it out)
346 * Remove GTNetS. It was not working anyway.
347 * Various cleanups in the cmake scripts.
348 * Move headers around to sort them out on installed systems:
349 - instr/instr.h -> simgrid/instr.h
350 - instr/jedule/* -> simgrid/jedule
351 - simdag/datatypes.h was removed
352 - simdag/simdag.h -> simgrid/simdag.h
353 - msg/datatypes.h was removed
354 - msg/msg.h -> simgrid/msg.h
357 * Interface improvement:
358 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
359 - Sanitize the interface in MSG_task_ module:
360 - Merge two functions that were close enough but misleading:
361 set_compute_duration(t) -> set_flops_amount(t)
362 get_remaining_computation(t) -> get_flops_amount(t)
363 - set_data_size(t) -> set_bytes_amount(t)
364 get_data_size(t) -> get_bytes_amount(t)
365 - Massive cleanups in the functions related to the energy
366 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
367 - New: MSG_host_get_pstate()
369 - msg/energy/onoff: switching hosts on and off
371 * Interface improvement:
372 - Rename Host.isAvail() to Host.isOn()
373 - Rename Process.currentProcess() to Process.getCurrentProcess()
374 - Rename Task.setDataSize() to Task.setBytesAmount()
375 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
377 - #18874: Actually allows the GC to reclaim tasks
381 - SIMIX_process_throw: raises an exception in a remote process
382 * Refactoring: Separate sub-modules
383 - libsmx: the public interface, as libc in a real system
384 - popping: the strange dance that converts a user request into a kernel handling
385 - smx_context_*: the virtualization mechanisms that embed the user code
386 - smx_*: the handling of each simcalls
388 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
389 * Rename smx_action_t into smx_synchro_t, making explicit that these
390 things are used to synchronize processes with their environment.
391 For example, a communication is a sort of synchronization involving
392 the communicating processes (that may block until the exchange) and
393 the platform. The same can be said from computations, etc.
395 - #18888: segfault when a process ends before its kill_time
399 - 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.
400 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
401 - MPI_Type_set_name, MPI_Type_get_name
402 - MPI_*_c2f and MPI_*_f2c functions
403 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
404 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
405 - Activate a lot of new tests from the mpich 3 testsuite
407 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
408 - InfiniBand network model added : Based on the works of Jerome Vienne
409 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
410 - When smpi/display_timing is set, also display global simulation time and application times
411 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
412 * Collective communications
413 - 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
414 - 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/).
415 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
417 - Replay now uses algorithms from wanted collective selector
418 - Replay can be used with SMP-aware algorithms
419 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
420 - 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.
422 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
423 - Correctly use loopback on fat-tree clusters
424 - Asynchronous small messages shouldn't trigger deadlocks anymore
425 * Energy/DVFS cleanup and improvement
426 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
427 - new: smpi_get_host_pstate()
430 - "Full" network optimization flag was broken since Surf++
431 - Better handling of precision flags in maxmin
432 - Fix bug causing sometimes "Impossible" errors
433 - Properly pass cluster properties to included hosts
434 * Improvement of the Energy plugin.
435 - Always update the consumption before returning that value
436 - New property: watt_off to denote the dissipation when the host is off
437 - New functions getWattMinAt and getWattMaxAt to retrieve the
438 dissipation of pstates that we are not currently at.
439 * Java: class NetworkLink renamed to Link
440 * New function: simcall_process_get_kill_time()
441 * Massive rename s/workstation/host/
442 - That's intrusive, but that's good for the project consistency. Sorry.
443 - Change config option "workstation/model" into "host/model"
446 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
447 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
448 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
449 - xbt_dirname and xbt_basename for non-POSIX systems
451 * The model checker now runs as a separate process.
452 * The model checker runs must now be launched with the new simgrid-mc program.
453 * Record/Replay: the MC can display a textual representation of a path in the
454 execution graph. It can then be replayed outside of the model checker.
456 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
458 SimGrid (3.11) stable; urgency=low
463 * Normalizing pointers addresses tool for better diff between logs
466 * Add cloud examples using new VMs
467 - examples/msg/cloud/two_tasks_vm.tesh
468 - examples/msg/cloud/simple_vm.tesh
469 - examples/java/cloud/cloud.tesh
470 - examples/java/cloud/migration/migration.tesh
471 * Add java surf examples:
472 - examples/java/surfPlugin/surf_plugin.tesh
473 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
474 - examples/java/surfCpuModel/surf_cpu_model.tesh
475 * Add SMPI+MSG example:
476 - examples/smpi/smpi_msg_masterslave/
481 - msg task destroy cancel test
482 - msg_host on/off test
483 * Move all tests in testsuite to teshsuite (adding tesh files)
484 * Restructure teshsuites
485 - one folder for each kind of test
486 * Restructure AddTests.cmake
488 - structure the order of tests (with sections)
491 * Add virtual machine
492 - creation of a VM on a PM
493 - migration of a VM from a PM to another PM
495 - MSG_process_join(msg_process_t process, double timeout)
496 - msg_bar_t MSG_barrier_init(unsigned int count)
497 - int MSG_barrier_wait(msg_bar_t barrier)
498 - void MSG_barrier_destroy(msg_bar_t barrier)
499 - msg_as_t MSG_environment_as_get_by_name(const char * name)
500 * New option "msg/debug_multiple_use" to help debugging when a task is used
503 - Improvements and finalization of MSG_storage, MSG_file APIs and their
505 - Increase code coverage in test suites
509 * Protect context stacks against stack overflow. The number of protected memory
510 pages allocated on the top of each stack (1 by default) can be configured
511 with the new command line option --cfg=contexts/guard_size:<value>.
512 * Simcalls are now generated by a python script that
513 - generates files included by SimGrid
514 - checks that all the functions exist, and proposes prototypes
516 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
517 host_set_data, host_get_data
519 - simcall_process_join(smx_process_t process, double timeout)
520 * Fix bug where sleeping processing could not be suspended.
523 * Translate surf models from C to C++
524 - Generic classes for all models: Model, Resource, Action
525 - A generic interface for each kind of model (CPU, Network, Storage
526 Workstation, WorkstationVM)
528 * Translate surf routings from C to C++
529 * Add callbacks using sigc++ or boost::signals2
530 - Add callback functions for resource creation/destruction
531 - Add callback functions for action state change
532 - Handle Energy as a plugin
533 * Replace swag by boost::intrusive
534 * Add new routing models for clusters. For documentation, see
535 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
536 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
537 parameters for cluster tag
538 - Fat trees, with topology="FAT_TREE" and
539 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
540 - see examples/platforms/torus_cluster.xml and
541 examples/platforms/fat_tree_cluster.xml
545 * Hostfiles support host:nb_processes construct to deploy several processes on
547 * Collective communication algorithms should not crash if used with
548 improper number of nodes and report the error.
549 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
550 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
551 MPI_Dims_create, MPI_Cart_sub are supported.
552 * New interface to use SMPI programmatically (still depends on MSG for
553 some parts, see examples/smpi/smpi_msg_masterslave) :
554 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
558 * Global variables privatization in MPI executables is now performed at runtime
559 with the option smpi/privatize_global_variables (default:no).
560 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
561 dynamic libraries loaded by the application are not privatized (static
562 linking with these libraries is advised in this case)
565 * Options defined in XML work correctly now.
568 * New cmake option, enable_lib_in_jar, to control whether native libraries are
569 copied into simgrid.jar or not (ON by default). Use this option if you want
570 to reduce the size of the installed simgrid.jar, *and* the native libraries
571 are kept installed elsewhere.
572 * Surf binding with SWIG (code generated in maintainer mode only):
573 - plugin to handle callbacks
574 - CPU model only for the moment
577 * Supernovae build mode is definitively removed. It was used to improve
578 inlining and inter-module optimizations. It is nowadays superseded by
579 link-time optimizations commonly available in compilers.
580 * Update ns-3 find lib. Bindings for ns-3 should work again now.
581 * Add boost dependency for surf++
582 * Add new macro for tests
583 - ADD_TESH(name <tesh_args>)
584 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
588 - xbt_bar_t XBT_barrier_init(unsigned int count)
589 - int XBT_barrier_wait(xbt_bar_t barrier)
590 - void XBT_barrier_destroy(xbt_bar_t barrier)
591 * Make the xbt_os_time module public
593 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
595 SimGrid (3.10) stable; urgency=low
597 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
600 * Reintegrate Java to the main archive as desynchronizing these
601 package is not acceptable anymore (Java is now considered stable)
602 * Add explicit synchronization facilities through semaphores
603 * Bug fix: Task.setDataSize() only changed the C world, not the value
604 cached in the Java world
607 * Dramatically change the way files are handled. API and internals changed, but
608 this part of MSG was not considered as production grade either.
609 * Add explicit synchronization facilities through semaphores
610 * Add a new function MSG_host_get_process_list()
611 * Preliminary DVFS support (see examples/msg/energy/ for details)
614 * SMPI is now included directly in the libsimgrid as the windows
615 linker doesn't force us on splitting it anymore.
616 * Improvements of the SMPI replay tool:
617 - Most of the collective communications are now rooted in the same process as
618 in the original application.
619 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
620 used until now). Multiple data types can now be used in a trace.
621 - The replay tool now supports traces produce either by TAU or a modified
623 - Bug Fix: the compute part of the reduce action is now taken into account.
624 - Gatherv collective is now supported
625 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
626 of smpirun outputs time independent traces for the current run. One file
627 is created per process. If too many processes are simulated, this behavior
628 can be changed to one file for all processes by using the
629 tracing/smpi/format/ti_one_file flag
630 * smpirun generates the host file if needed (with given host count and platform)
631 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
632 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
633 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
634 --cfg=smpi/coll_selector:(mpich/ompi)
635 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
637 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
638 happening inside a collective SMPI call.
639 * Fix the behavior of complex data types handling.
640 * Make MPI_Wtime another synchronization point to take computations into
642 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
643 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
644 * Add all missing Fortran bindings, SMPI should work with Fortran 90
645 (no privatization of global variables yet)
646 * Preliminary DVFS support (see examples/smpi/energy/ for details)
649 * Verification of liveness properties is now available for SMPI applications
650 (in addition to MSG applications)
651 * Bugged examples using SMPI in examples/smpi/mc/
652 * Add --cfg=model-check/visited option. Allows the verification of infinite
653 programs. Detection of loops in the execution thanks to the system state
654 comparison and reduction of the state space to explore. Can be combined with
655 DPOR for safety properties.
658 * Allow to change SimGrid configuration (see --help) within the code
659 thanks to SD_config() as it can be done in MSG.
660 * Add a new function SD_task_set_amount() upon user request.
663 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
664 * Remove rule based routing (no more PCRE dependency)
665 * Add a limiter_link option to cluster tag, to specify a maximum reachable
666 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
667 * Add a loopback_bw and loopback_lat options to cluster tag.
668 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
669 routing. Now peers are encapsulated in an AS and have their own private
670 router but this is transparent.
673 * Our own implementation of getline is renamed xbt_getline, and gets
674 used even if the OS provide a getline(). This should reduce the
675 configuration complexity by using the same code on all platforms.
676 * New type: xbt_cfg_elm_boolean.
677 * Allow to use yes/no for boolean configuration options in the command line.
678 * Allow to disable SimGrid cleanups at exit from command line option.
679 There are situations where one may want a simulation to end with an exit.
680 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
681 annoying when scripting around the simulator. Adding a
682 --cfg=clean_atexit:no allows to circumvent this issue.
685 * Lots of memory leaks were corrected in this release.
686 There is no known memory leaks anymore, in all of our 600+ tests.
687 * New command line option --version, to get SimGrid version information.
688 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
690 * Supernovae builds are deprecated, and expected to be removed in the next
693 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
695 SimGrid (3.9) stable; urgency=low
697 The Grasgory release: GRAS is really dead now.
699 * Complete overhaul of the internal host structures scheme.
702 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
703 was considered as experimental and badly maintained since then.
704 * Keeping it was thus a trap to our potential users, that could take
705 it instead of MSG or SMPI by mistake despite is pity state.
706 * GRAS seems to have very few users (if any), and no one volunteered
707 to maintain it further. It also induces a lot of XBT code (for
708 portability sake), that must be maintained too.
709 * For all these reasons, we killed GRAS. If someone wants to revive it
710 in the future, don't cry, our git history still remembers of GRAS.
713 * Major overhaul. Merge our documentation again as time proved that
714 splitting it was really not helping our users.
715 * Further improve the developer documentation to help newcomers
716 hacking on SimGrid itself. The user documentation (and in
717 particular, the beginner documentation) is still in a sorry state.
720 * Now works on Windows too!
721 * Much more extensive test suite, from MPICH
724 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
725 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
726 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
727 * Bug fix: let task be scheduled when the last dependency to be solved is
728 a control dependency.
729 * Remove SD_load_environment_script function.
730 Use the C sg_platf function if you want to declare a platform
734 * New function: MSG_process_get_number()
735 * Old function documented: MSG_config()
736 * Remove MSG_load_platform_script function
737 Use the C sg_platf function if you want to declare a platform
741 * Change the default value of the TCP_gamma constant (maximal size of TCP
742 congestion window) to a more realistic 4MiB value. If you notice changes in
743 your simulation results, you can fall back to the previous 20k tiny window
744 by adding --cfg=network/TCP_gamma:20000 on command line.
745 * (Hopefully) fix a bug wrt periodic availability and state traces
746 * Bug fix: use default values at start when first event in availability/state
747 trace is not at time 0.
750 * remove the "new_" part of function name sg_platf_new_trace_connect
751 (resulting in sg_platf_trace_connect), since it does not create
755 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
756 use case where it's really mandatory, and maintaining it was a pain in
758 * New: xbt_fifo_search(), search an item with a user-provided
759 comparison function instead of dumb pointer comparison.
762 * Fix the lua deployment:
763 Use `simgrid.init_application()` before deployment instead of
764 `simgrid.msg_register_application()` after.
767 * Transfer the tracing files into the corresponding modules.
769 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
771 SimGrid (3.8.1) stable; urgency=low
773 The "we are told that some people want to also *install* the simgrid
776 * Add missing file "tesh.1" to the archive.
778 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
780 SimGrid (3.8) stable; urgency=low
782 The Psssshiiiit release: SimGrid jumps into the Cloud.
785 * Add an experimental interface to manipulate VMs. They are mainly
786 process groups with very few intrinsic semantic, but they should
787 allow you to build the semantic you want easily.
788 * New function: MSG_host_set_property_value()
789 * New function: MSG_process_on_exit(). To clean memory in all cases.
790 * Bug fixes that made the host (and link) failures unusable.
791 * Add a way to auto-restart process when the host in which they are
792 executing comes back (ON_FAILURE="RESTART" on deployment file,
793 MSG_process_auto_restart_set).
794 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
795 please stop using the old ones, they are DEPRECATED.
797 * Deprecate functions MSG_global_init() / MSG_global_init_args()
798 Please use MSG_init() instead. (reducing the amount of entry
799 points in the library helps us).
800 * Make it impossible to link against the wrong version of the lib
801 * Deprecate MSG_clean(). No need to call it anymore.
802 * Function MSG_get_host_number() is not deprecated anymore.
805 * Split the doc into a user guide and a reference guide.
806 * Start a developper guide to help people hacking on SimGrid.
809 * Enable tracing by default. This modules rocks you should use it.
810 * Remove option custom_flags. Now use environment variables CFLAGS
812 * Use default cmake things to detect lua instead of home grown ones.
813 * New option "enable_mallocators" to disable mallocators, for debugging
814 purpose ("on" by default).
817 * Bug fixes around the resource failures: don't let the processes
818 survive the host they are running onto.
819 * Add an interface to auto-restart processes when the host in which they are
820 executing comes back.
821 * Ensures that SIMIX_clean is called automatically. It's not part of
822 the public interface anymore (bindings should be updated).
825 * Bug fix for when SD_Simulate is called with a positive value: be careful
826 when comparing doubles. Sometimes they are different for non significant
828 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
829 parallel task whose initial work is distributed among host according
830 to the Amdahl's law. Such tasks are created with a parameter alpha
831 that corresponds to the non-parallelizable part of the computation.
832 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
833 between two sets of workstations assuming a 1D block distribution (each
834 workstation owns a similar share of data) on both sides.
836 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
837 Data redistribution will be automatically scheduled once parent and child
838 are both scheduled. The filling of computation_amount and
839 communication_amount structures is now done seamlessly thanks to the chosen
841 * New function SD_workstation_dump to display various information
842 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
843 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
844 bandwidth on the route onto which the task is scheduled and the amount of
846 To divide the nominal bandwidth by 2, the rate then has to be :
847 rate = bandwidth/(2*amount)
848 * Compute tasks that have failed can now be rescheduled and executed again
849 (from their beginning)
850 * Increasing source code coverage (src/simdag is now covered at 95.8%
854 * Re-implement time-independent trace replay using SMPI (at the
855 smpi_smp_* level) instead of MSG. This should replace
856 examples/msg/actions/actions.c
857 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
858 hindexed and structs)
859 * Implement the exchange of non-contiguous data.
860 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
861 * Correct behavior of smpi/sender_gap and set its default value to 0
862 * Add option to asynchronously send small messages to allow better
863 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
864 specifies the size in bytes under which messages will be asynchronously sent.
865 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
866 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
869 * New C interface to define a platform: XML is now optional.
870 For more info, please check include/simgrid/platf.h.
871 * New interface to define random platforms from the C:
872 For more info, please check include/simgrid/platf_generator.h and
873 examples/msg/masterslave/masterslave_platfgen.c
874 * Export a sg_cmdline dynar containing all the arguments we got from
878 * Two new tracing options for adding comments to trace file so you
879 can track your experiments (see --help-tracing for details).
880 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
881 * Adding the SimGrid version that generated the trace file as a comment.
882 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
883 * Fix to avoid key clashes on Paje links
884 * Other minor fixes related to the Paje specification
887 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
888 and renamed to xbt_str_hash() and xbt_str_hash_ext().
889 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
890 * Greatly improve the robustness of mmalloc to user errors (such as
891 using an area after freeing it, or freeing it twice)
893 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
895 SimGrid-java (3.8.1) stable; urgency=low
897 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
898 Warning: all methods are visible, but only some of them are
899 implemented so far. Check the source (src/jtrace.c)
900 for further information.
901 * New module: org.simgrid.msg.File (SimGrid File management functions)
902 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
903 * Change the meaning of Process.restart: now restart the process from
904 the begining, like MSG_process_restart in C.
905 * Add Process.setAutoRestart: handling of process restart when failed
907 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
908 you to retrieve the properties of the processes/hosts
909 * Deprecate Msg.clean(): you can just forget about it now.
910 * New function Process.getCount(), that only works when compiling
911 with the not yet released version 3.9 of the C library.
913 * New context factory based on Coroutines. It mandates a modified JVM
914 but then, the simulations run about five times faster, and there is
915 no limit to the amount of processes (beside of the available memory).
917 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
919 SimGrid (3.7.1) stable; urgency=low
922 * Restore the prototype of MSG_process_create_with_environment() to
923 the pre-3.7 situation by removing the kill_time argument.
924 * Add a MSG_process_set_kill_time() function instead.
927 * Fix weird behaviors when dealing with parallel tasks.
930 * Simgrid is now built as a dll.
931 * Simgrid-java now works on Windows.
932 * Simgrid-Java is now included into Windows package.
935 * First pre-build package for MacOSX.
938 * Fix compilation when using MSG_USE_DEPRECATED.
939 * Fix some compilation issues on Macs and Windows.
940 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
941 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
943 We discovered that the Lua console is broken, but we are missing the
944 manpower to fix it right now. The problem existed in 3.7 too, so we
945 are not blocking the release for that. Sorry if you depended on this
946 feature, any help would be really welcome.
948 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
951 SimGrid-java (3.7.1) stable; urgency=low
953 The "Java aint got to be bloated and slow" release
956 * Various internal cleanups and performance improvement
957 Simulations are expected to run up to twice faster or so
958 * Make Process.kill(process) an instance method, not a static one
959 * User processes are not java.lang.Thread subclasses.
960 This breaks the compatibility (sorry), but previous API was
961 brain-dead, making it impossible to have non-trivial
962 initializations in the process constructor.
963 * Require a full constructor per Process sub-class.
964 Kinda breaks the compatibility (sorry), but this allows a much more
965 efficient way to launch the processes at simulation startup.
966 * Do not embeed our version of semaphores, java 1.5 can be considered
967 as sufficiently prevalent for us to not dupplicate its features.
971 * Add examples for almost every part of the API
972 We spotted and fixed a lot of bugs in the process
974 * New module: asynchronous communication API
975 * New function: Process.sleep()
976 It takes milliseconds as argument, just as java.lang.Thread.sleep()
977 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
978 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
980 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
982 SimGrid (3.7) stable; urgency=low
984 The "spring cleanups (before the next Big Project kicks in)" release.
987 * We can specify the SMPI latency/bandwidth factor with command line
988 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
989 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
990 You can also use the "config tag" from platform file by adding this line
991 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
992 (see "example/platforms/tag_config.xml" to use "config tag").
993 Note that the command line supersedes the platform file configuration.
994 * Change the correction factors used in LMM model, according to
995 the latest experiments described in INRIA RR-7821.
996 Accuracy should be improved this way.
997 * Use the partial invalidation optimization by default for the
998 network too. Should produce the exact same results, only faster.
999 * Major cleanup in surf to merge models and split some optimization
1000 mechanisms from the core of the models. As a result you can now
1001 specify which model to use (e.g., --cfg=network/model:LV08
1002 --cfg=cpu/model:Cas01) and which optimization mode to use
1003 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1004 Incompatible combinations should err at initialization. See
1005 --help-models for the list of all models and optimization modes.
1006 * The CLM03 workstation model was dropped for simplicity because it
1007 used the deprecated CM02 network model. Use default instead.
1008 * Rename the TCP_gamma configuration option to network/TCP_gamma
1009 * Rename the coordinates configuration option to
1010 network/coordinates, and document it
1011 * Use now crosstraffic keyword instead of the terribly misleading
1012 fullduplex keyword. It is activated by default now in the current
1013 default model, use --cfg=network/crosstraffic:0 to turn it off.
1014 * Ongoing refactoring the model parsing to make XML files optional
1015 See include/simgrid/platf.h for details (still to be completed)
1018 * Major overhaul of the documentation. Almost instructive now :/
1019 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1020 functions and friends. This interface was considered as
1021 deprecated since over 2 years, it's time to inform our users that it is.
1022 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1023 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1024 use these (crappy) functions in your code.
1025 These functions will be removed soon. Stop using them now.
1026 * Deprecate MSG_get_host_{table,number}
1027 Implement MSG_hosts_as_dynar() instead.
1028 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1029 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1032 * Stabilize the parallel execution mode of user contexts
1033 * Introduce configuration variables to control parallel execution:
1034 - contexts/synchro: Synchronization mode to use when running
1035 contexts in parallel (either futex, posix or busy_wait)
1036 - contexts/parallel_threshold: Minimal number of user contexts
1037 that must be part of a scheduling round to switch to parallel
1038 execution mode (raw contexts only)
1039 * Fix bugs that prevented to use suspend/resume along with
1040 synchronization structures.
1041 * Fix bugs in process termination that lead to invalid memory access
1042 in very specific conditions.
1045 * Introduce a parallel mode for the models (controlled by surf/nthreads
1046 configuration item). In our tests, running the models in parallel
1047 never lead to any speedups because they are so fast that the gain
1048 of computing each model in parallel does not amortizes the
1049 synchronization costs, even when ultra fast futexes are used.
1050 This is released anyway because YMMV.
1053 * Performance boost by using a swag internally to compute the set of
1054 tasks that are finished and should constitute the return value of
1058 * Enable it by default now that it is considered rather stable.
1061 * Documentation of the tracing functions.
1062 * Performance gains when tracing categorized/uncategorized resource
1063 utilization by avoiding calls to get route when updating resource
1064 variables. LMM constraints are being used instead.
1065 * API changed to set task categories. Use MSG_task_set_category instead
1066 of TRACE_msg_set_task_category, and SD_task_set_category instead
1067 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1068 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1069 * Examples for link user variables added (see at examples/msg/tracing/)
1070 * Deprecated function TRACE_msg_set_process_category completely removed
1071 * Trace header updated according to the latest Paje file format
1072 * Tracing network lazy updates, no longer obligate users to use full updates
1073 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1074 * Experimental: let user code declare/set/push/pop application states for hosts
1075 * API changed to allow the manual creation of graph configuration files
1076 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1079 * Improve the API of Lua MSG bindings, using the Lua spirit.
1080 * Each simulated process now lives in its own Lua world (globals are
1081 automatically duplicated). It helps writing simulators. Will allow
1082 to run Splay programs within SimGrid in the future.
1083 * Add a Chord example in Lua, equivalent to the MSG one.
1086 * Start the implementation of a solution to express temporal
1087 properties, not only local assertions. This is still an
1088 experimental work in progress, stay clear from it to be safe.
1092 - Add new runtime parameters --help-logs and --help-log-categories
1093 to display information about supported logging parameters and
1095 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1096 don't exists anymore.
1097 * Mallocators: allow value NULL for the reset function.
1099 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1100 create homogeneous dictionaries, where all the elements share the
1101 same free function. Non homogeneous dictionaries will be
1102 deprecated in the next release.
1103 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1104 - Multi-level dictionaries are deprecated.
1106 - new function xbt_dynar_search_or_negative() that is useful when
1107 you have less than 2 million elements in your dynar and don't
1108 want of the extra complexity of catching exceptions when the
1109 element is not found.
1111 - Make xbt_os_thread module (for thread portability) public.
1112 Documentation is still to come, sorry.
1114 - Cleanups and simplifications to make it maintainable again.
1115 - Exotic features (such as memalign and valloc) were removed.
1116 - The metadata were extended and improved so that the
1117 model-checker becomes able to explore and inspect the heaps.
1118 - This may induce a performance drop when enable_model-checking is
1119 ON in cmake (even if it's not used in the simulation), but it is
1120 necessary at this point to get MC working.
1122 Turn model-checking OFF if simulation performance matters to you.
1123 Not enabling it at runtime is not enough, disable it in cmake.
1125 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1128 SimGrid (3.6.2) stable; urgency=low
1130 The "Not coding new stuff allows to polish old things" release.
1133 * New bindings to the NS3 packet level simulator (experimental)
1134 * Use the raw (efficient) execution contexts instead of the sysv
1135 (portable) ones when possible.
1136 * libpcre is now mandatory in any cases since not using it led to
1137 severe performance loss and possibly other issues
1138 * Update the XML platforms:
1139 - G5K: include the latest machine in Nancy
1140 - GridPP and LCG: new platforms
1141 * Documentation was partially updated, at least (more to come)
1143 Bug fixes, cosmetics and small improvements
1144 * Free terminated processes before the end of the simulation to avoid
1145 exhausting the memory of users having very dynamic amount of
1147 * Bug fix and cosmetics about canceling non-running tasks
1148 * Bug fix about the dot loader's issues when using libcgraph
1151 * Create an installer for windows with nsis (amd64 and win32)
1152 - Add an hello world project to illustrate simgrid project creation.
1153 - Embed libpcre into the Simgrid installer to avoid
1154 its compilation burden
1155 * The raw execution contexts should work on Apple now
1156 * Port to Windows 64 bits
1157 - Sysv contexts now have an implementation for this arch
1158 - GRAS communication features now support this arch
1159 * Drop support for borland compiler on windows
1160 - this code was not maintained, and we kinda depend on gcc nowadays
1161 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1162 * Fix portability issue on unstable ubuntu: linker became picky on
1165 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1167 SimGrid (3.6.1) stable; urgency=low
1169 The "Oops, we broke Macs too" release
1172 * Fixed contexts detection so that raw ones are used when possible
1173 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1174 leads to a strange error, with user code segfaulting sometimes when
1175 the generated code is not perfectly aligned (which is not
1176 controllable from the user side, depends on the amount of code)
1179 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1181 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1183 SimGrid-java (3.6) unstable; urgency=low
1186 * Split of every thing from simgrid v3.5 into a separate package.
1188 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1190 SimGrid (3.6) stable; urgency=medium
1192 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1195 * Bindings now constitute their own package, separated from the main one.
1196 Rationale: reduce our maintenance nightmare by reducing the module coupling
1197 They will soon be released on their own on gforge.
1199 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1200 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1202 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1203 * It's not quite deprecated for now because we have no replacement,
1204 but it may soon become the case.
1207 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1208 * Fortran: New user-level cache variable to store the rank of the running
1209 process. This improves performance by an order of magnitude.
1210 * C: New coccinelle script to automatically locate and modify global and
1211 local static variables.
1212 * Improved SMPI network model with a sender-side gap to account for multiple
1216 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1217 only say if a communication is finished, no matter whether it succeeded or
1218 failed. You can call MSG_comm_get_status() to know the status of a finished
1220 * New function MSG_task_dsend() to send a task and detach it. When a
1221 communication is detached, you are never notified of its success or failure
1222 and the memory is released automatically once it is finished. This function
1223 is useful when you don't care about the end nor the success of a
1225 * Change the prototypes of action replay. Sorry for inconvenience,
1226 but this is really more efficient this way (and to adapt your code,
1227 you just have to fix the initialization, that shouldn't be too long)
1228 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1229 ever managed to do anything useful with it.
1230 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1231 takes a dynar of communications. It returns immediately and gives the
1232 index of a finished communication (if any).
1233 * New example: a basic implementation of the Chord P2P algorithm.
1236 * New model for multi-core CPUs. You can now use the core attribute to
1237 precise the number of cores of a host. This is a basic model. Every
1238 process running on the host receives at most the power provided in
1239 the DTD (throughput<=power). Total throughput of process cannot exceed
1241 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1242 router linked by an up-link and a down-link (possibly asymmetrical).
1243 This kind of pattern allows to easily build last-mile model style platforms.
1244 Aggregating such patterns in a rule-based AS is thus the technique of
1245 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1246 * New model for Vivaldi routing. We transformed the Vivaldi network model
1247 into a Vivaldi routing model (based on the rule-based model). This allows to
1248 combine Vivaldi based latencies with last-mile platforms.
1251 * Added a check for NaN of IEEE754 infinite in the double entries of
1253 * Introduce a new context factory "raw", highly inspirated from the
1254 ucontext factory, but using manually crafted functions in assembly to
1255 do the work in an efficient manner.
1256 * Allow to change the used context factory at run time, not only at
1257 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1258 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1259 contexts at runtime (only with raw contexts or ucontexts).
1260 * Completely rewrote this module to allow parallel execution of user
1261 processes. Use --cfg=contexts/nthreads:N to execute user processes
1262 with N parallel threads (the default is 1, meaning no parallelism).
1263 * Allow to decide dynamically between sequential and parallel modes.
1264 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1265 processes in parallel only when their number is greater than or equal to P
1267 * Added a check for NaN of IEEE754 infinite in the double entries of
1271 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1272 won't output the state of processes when interrupted with Ctrl-C
1273 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1274 NULL-terminated array. This may solve backward compatibility issues
1275 due to the change to return type of SD_simulate. See also:
1276 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1277 * Add new macros with variable number of arguments.
1278 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1279 - in xbt/asserts.h: xbt_assert
1280 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1281 - in xbt/ex.h: THROWF and RETHROWF.
1282 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1284 * Change xbt_die() to accept a format string with arguments, just like printf.
1285 * New data structure: xbt_lib_t, like a dict but more general and with better
1289 * New configuration options
1290 Options triva/categorized and triva/uncategorized can be used to generate
1291 graph configuration files for Triva visualization tool.
1292 * Configuration option tracing/platform is renamed to tracing/categorized
1293 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1294 * New TRACE_user_link_variable interface:
1295 User provides the name of the link and the tracing variable to attach to it
1296 * the declaration of tracing categories must be done after the environment
1298 * simpler tracing interface, just one way to declare categories
1299 TRACE_category or TRACE_category_with_color, it is up to you
1300 * links in the trace file are again identified by their names
1301 * trace contains the full platform hierarchy exactly as declared using the ASes
1302 * Options tracing/msg/[task|process]:1 groups the process by hosts
1303 for both cases, tasks and processes must have names that are unique during the simulation
1304 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1305 * The experimental option tracing/msg/volume is deprecated
1306 its functionality may be reincorporated if needed
1308 The tracing generates a trace file with unordered timestamped events,
1309 because of the way the core simulator (surf) works. A script available
1310 at the tools directory (fix-paje-trace.sh) can be used to put the events
1311 in order. We have changed the tracing so it can generate ordered timestamped
1312 events in the final trace, but depending on the simulator (and how much time
1313 is simulated) that can lead to a huge memory utilization. It is deactivated
1314 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1316 Build Infrastructure
1317 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1318 - We are trying hard to keep the API stable, but it may happen that
1319 some things change (we're a research project after all, not a
1320 nuclear plant operating system). If such things should happen, you
1321 could rely on that macro to adapt.
1322 - current value: 30600 for 3.06.00, aka 3.6
1323 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1324 a number that can be compared with SIMGRID_VERSION.
1325 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1326 assertions and verbose/debug logging events are disabled at compile time.
1328 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1330 SimGrid (3.5) stable; urgency=medium
1333 * New feature to any SimGrid-based simulator: Model-Checking
1334 Check SIN#1 for more details.
1337 * New Model SMPI (three-interval linear regression for correction factors)
1338 See RR-7426, available at http://hal.inria.fr/inria-00527150
1339 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1340 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1341 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1342 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1343 * See also src/smpi/README
1348 - Tracing API changes: TRACE_start and TRACE_end should not be called
1349 by user-code. They are automatically called by simulators created
1350 with SimDAG, MSG and SMPI if the toolkit is compiled with
1351 tracing_enabled=ON. Categories declaration and utilization remain the
1352 same for MSG and SimDag.
1353 - A function was added to the tracing API to declare categories with
1355 - TRACE_category_with_color (char *category, char *color)
1356 where color must be in the following format
1357 "%f %f %f", red, green, blue
1358 and red, green, blue are float values in the interval [0, 1]
1359 - User can specify NULL as color parameter, or continue calling
1360 TRACE_category (cat)
1361 On that case, the tracing system will define random colors
1362 - The following command-line options are supported:
1363 --cfg=tracing/filename:msg.trace
1364 --cfg=tracing:1 (activate tracing, needed to use others)
1365 --cfg=tracing/platform:1 (categorized resource use)
1366 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1367 --cfg=tracing/msg/task:1 (task creation)
1368 --cfg=tracing/msg/process:1 (process creation, migration)
1369 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1370 --cfg=tracing/smpi:1 (SMPI interface tracing)
1371 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1372 - examples of examples/msg/tracing updated
1374 - DAXLoader and DOTLoader functions can generate tasks with categories
1375 - A new function to attribute a category to SD tasks:
1376 TRACE_sd_set_task_category (SD_task_t task, char *category)
1377 * Tracing the MPI interface implemented by SMPI
1378 - Collective operations are traced with states
1379 - Point-to-Point operations are traced with states/links
1380 - Tracing activated by a parameter "-trace filename" passed
1381 to smpirun during execution (considering that simgrid
1382 is compiled with tracing enabled)
1383 - To run the simulation with gdb, the simulator
1384 accepts --cfg=tracing/smpi:1 to trace SMPI
1385 - tesh files to check if smpi tracing is ok
1386 - See examples/smpi/NAS/DT-trace
1387 * GTNetS tracing re-worked
1388 - adaptation to the tracing system of GTNets to cope
1389 with modifications regarding the fullduplex mode
1390 - new tesh files to check if gtnets tracing is ok
1393 * Asynchronous communications through the functions:
1394 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1395 * New function: MSG_load_platform_script()
1396 to make possible using a lua script instead of XML files to set up platforms
1397 * New function: MSG_set_function
1398 to associate functions to processes, used when bypassing the parser
1399 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1401 Platforms: Add some more examples in examples/platforms
1402 * Grid'5000: see www.grid5000.fr
1403 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1406 * Change the XML format. This is a very important modification. SimGrid
1407 3.5 introduces a new hierarchical format based on the notion of
1408 Autonomous Systems. Compatibility with old format is ensured
1409 through the perl script provided in the install bin directory
1410 bin/simgrid_update_xml.
1411 It is now possible to build platforms with specific routing
1412 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1413 connect several platforms together. We will try to provide soon
1414 set of realistic platforms exploiting these properties (have a
1415 look at examples/platforms/ for the moment).
1416 * Take the opportunity of the XML format change to be a good XML citizen:
1417 rename link:ctn to link_ctn and similar changes (also dealt with by
1419 * Add a new routing scheme (rule-based) using regular expressions. It
1420 enables to have an extremely low memory footprint when the
1421 underlying routing is simple and can be compactly described. You
1422 need to have libpcre4-dev (perl regular expressions) installed if
1423 you want to use this routing scheme.
1424 * Revive the cluster TAG and allow to easily and efficiently (both in
1425 term of memory and speed) connect clusters together. Have a look
1426 at teshsuite/simdag/platforms/ to see how this can be done. With
1427 this tag, you can create clusters with thousands of tasks at no
1428 cost (have a look at examples/platforms/).
1429 Note: clusters are implemented as ASes, so there is no need for an
1430 enclosing AS tag if you have only one cluster in your platform.
1431 * Add new generic functions in the public interface that allows the user
1432 to call SURF 'create_resource' methods from your code (same
1433 functionality as the XML bypass mechanism but with a much lighter
1435 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1436 that uses a piecewise linear approximation to produce better
1437 results when exchanging small messages.
1438 * Add a new parameter to handle correctly full duplex link and account
1439 for interferences between uplink and downlink communications
1440 (activate with --cfg=fullduplex:1).
1443 * Rename the SD_READY (all dependencies are satisfied and task is
1444 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1445 dependencies are satisfied) state.
1446 This prevents a confusion between the notion of "ready to schedule"
1447 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1448 simulated" (SD_RUNNABLE) used by the simulation kernel.
1449 * Change the way a task is considered as ready. Instead of removing
1450 dependencies when a task is done, a counter is decreased. This way,
1451 it is always possible to reach ancestors thanks to the
1452 SD_taks_get_parents function (even after the end of the simulation.)
1453 * Change the return type of SD_Simulate from (SD_task_t*) into
1454 xbt_dynar_t. This function was in handling a dynar internally and
1455 converted it into a NULL terminated array for historical reasons.
1456 * New function SD_dotload(char*) to load a DAG described in dot
1457 format. This loader and the corresponding examples require the
1458 installation of the graphviz library.
1459 * Fix a bug in the management of tasks of size 0 in the surf network
1460 models. This problem was only visible with SIMDAG and you should
1461 thus disregard results produced with earlier versions if you
1462 relied on this feature (some tasks were blocked because of this).
1463 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1464 even though your applications were doing only point-to-point
1465 communications and sequential computations. Now you can really use any
1466 model you want (of course, if you create real parallel tasks, which are
1467 not implemented in most models beside ptaskL07, this will abort).
1468 * Add an example that schedules a DAX on an heterogeneous platform
1469 using a Min-Min strategy.
1470 * New function SD_workstation_get_current_task() that returns the kind
1471 of task currently running on a workstation in the sequential access
1473 * Raise some warnings when unexecuted tasks remains at the end of the
1474 simulation. This is usually caused by cycles in the DAG.
1477 * New function: SIMIX_process_set_function() called by MSG_set_function
1478 * Change the underlying waiting queue in semaphores so that a process
1479 can wait on several of them simultaneously (as in waitany).
1480 * Fix the way to handle tokens in semaphores so that all access patterns
1481 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1482 * kill the dirty pimple SIMIX_message_sizes_output()
1483 Please use (proper) visualization instead
1486 * New data container: setset (set of sets of elements)
1487 * New module: mmalloc (mapped malloc, allowing to have several
1488 independent segments of malloc)
1489 * New function: xbt_dict_cursor_set_data()
1490 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1491 * New function: xbt_dynar_is_empty()
1492 * New function: xbt_fifo_get_last_item()
1493 * Fix xbt_dynar_shrink(): use the right element size.
1494 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1495 uninitialized areas during expand.
1496 * Fix semaphores: previous implementation was severely broken.
1497 * Use library init/fini functions for our initialization.
1498 - you can use logs and other feature as soon as you want in your
1499 code (even before the xbt_init / MSG_init)
1500 - xbt_exit is now a no-op and produce a warning when used.
1503 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1504 benefit from the latest improvement to the simulation kernel.
1505 * Kill measurement sockets for now. If you rely on them, sorry. This
1506 release is not for you. This feature will be reintroduced in the
1507 future, but we cannot delay the release any further.
1508 * New function: gras_msgtype_get_name().
1509 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1510 * Fix (at last) the pmm example: it should not randomly fail anymore.
1512 Build chain: bug fixes and overall polishing
1513 * Cmake is now stable enough. Hence, we killed the autotools.
1514 * Port to windows ( TM :)
1515 * Fix the 'make install' target.
1516 No need to use 'make install-simgrid' anymore
1517 * Introduce a 'make dist' target compiling a *source* archive
1518 'make package' compiles a binary archive
1519 * Compile java files only on need
1520 * Add --cd and --setenv command line options to tesh
1521 * Out of source builds are not fully supported yet, but we are close
1522 * Enable supernovae and optimization flags by default for our users
1525 * Add layer to set up environment directly from lua, without XML.
1526 * The effect of gras_stub_generator can be achieved through
1527 lua too (check examples/gras/console/ping_generator.lua)
1529 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1531 SimGrid (3.4.1) stable; urgency=low
1533 The "Polishing easter eggs is probably a good idea" release.
1534 This is a bug fixes release only.
1538 * Fix a bug preventing the tasks from begin garbage collected.
1541 * Fix a bug occuring when a host involved in a communication fails.
1542 This was not detected properly by the other peer involved in the
1543 communication. Now, it's reported as a network error.
1546 * Warn the user about loop dependencies in data flow of DAX files
1547 * Obey the control-flow dependencies of DAX files
1550 * Add option "enable_smpi" allowing to not compile SMPI.
1551 Probably useful for the (Mac) users experiencing a build error here
1552 * Improve the detection of lua5.1 and ruby1.8
1554 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1556 SimGrid (3.4) stable; urgency=low
1558 The "Easter in Cargese" release. Also known as (major changes):
1560 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1561 ~> bindings were greatly improved
1562 ~> new tracing infrastructure for better visualization introduced
1564 * the "Welcome to configury modernity" release.
1565 ~> we switched from autotools to cmake, and improved our cdash
1568 A more detailled list of changes follow (full detail in svn log).
1570 Java Bindings: Various Cleanups
1571 * (install java-gcj-compat-dev on debian-like to use them)
1572 * Remove put/get: no need to export deprecated interface in Java
1573 Use send/receive instead.
1574 * Cleanup the examples and add a README per directory
1575 * Remove example autoDestination (that's the only way to go now)
1576 * Remove example explicitDestination (was a plain copy of basic)
1577 * Make JniException a runtime exception, so that there is no need to
1578 declare the fact that you may encounter such a beast. I guess that
1579 nobody will ever want to survive such error.
1580 * Create specific errors for each MSG case of failure:
1581 host failure, transfer failure, timeout, task cancelled
1582 * Cleanup the exceptions that may get thrown by each function
1583 * Other internal cleanups in Java bindings. Performance still bad :/
1584 Ruby and Lua Bindings: create them
1585 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1586 * That's new and great, you should try them out.
1587 Same functionalities than Java bindings, only even less polished
1589 * Kill the useless "rate" argument of SD_task_get_execution_time()
1590 Everyone used to provide -1 as a value, it was not used, and the
1591 semantic of a possible use wasn't even clear.
1592 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1593 either as comm costs or compute costs to mean that there is no
1594 such thing for that specific task.
1595 * Add a SD_task_set_name() function
1596 * Fix SD_task_unschedule() on typed tasks
1597 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1598 * In DAX loader, accept useless 'level' attributes to <job> since
1599 LIGO DAGs have them (seem to be to ease graphical representation).
1601 * Add an example masterslave_mailbox.c using send/receive and not
1602 the deprecated put/get interface.
1603 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1604 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1606 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1607 (declare MSG_USE_DEPRECATED to still have the old name)
1608 * Add a MSG_task_set_data() function
1609 * About trace replay (see examples/msg/actions):
1611 - Allow to work with splitted trace files for each process
1612 Give the specific trace file as argument of each process,
1613 and call MSG_action_trace_run(NULL)
1614 You can still have one merged file for all processes.
1615 - Fix implementation of collective operations
1616 * Allow task_execute() on 0-sized tasks (closes #10063)
1618 * This is the first release of SimGrid where SMPI is not considered
1619 beta anymore (even if some corners should still be improved)
1620 * Port over the new SIMIX_network submodule (internal refactoring)
1621 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1622 * Implement more missing elements of the standard:
1624 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1626 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1627 MPI_Scan() MPI_Get_processor_name()
1628 - Added implementation of missing case for Alltoall (warning: it's
1629 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1630 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1631 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1632 SMPI_MPI_Allgatherv()
1633 * Bug fixes include:
1634 - MPI_Waitsome() was broken
1635 - Allow relative includes in smpicc
1636 - Command line cfg argument 'reference_speed' was ignored...
1637 - Some functions did not properly lead to auto-benching of user code
1638 - smpicc passes -O2 by default (just like openmpi one)
1640 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1641 * Bug fixes about timeouts during communications
1642 * add SIMIX_message_sizes_output() as a pimple to write to file the
1643 amount of messages per size. Use gnuplot to get histogram.
1644 Pimple because that's the only user-visible function of simix,
1645 defined directly in xbt.h (irk, sorry)
1647 - Add a SIMIX_sem_get_capacity() function
1648 - Fix interactions with processe resume/suspende
1649 - release_forever() was stupidly broken
1650 - Fix SIMIX_display_process_status() for processes in a semaphore
1651 - Make SIMIX_sem_block_onto() user-visible
1652 * Refactoring context stuff:
1653 - Use pseudo-OOP for better modularity
1654 - reimplement SIMIX_process_kill() without process_schedule() so
1655 that the latter can take as invariant that it is called from
1657 - Merge context_start into context_new for sake of simplicity
1659 * Add a Vivaldi network model, coded live during SUD'10 ;)
1660 * Rename configuration variables to start a hierarchy:
1661 o cpu_model -> cpu/model
1662 o network_model -> network/model
1663 o workstation_model -> workstation/model
1664 * New configuration variables:
1665 o network/bandwidth_factor: correction to bandwith
1666 o network/latency_factor: correction to latency
1667 o netwotk/weight_S: correction to the weight of competing streams
1668 * Add a long description to the models, that users can see with such
1669 argument on the command line: --cfg=cpu/model:help
1670 * --help-models display the long description of all known models
1672 * config: add the ability to set a default value after registration
1673 Does not override any previously set value (e.g. from cmd line)
1674 * dict: allow to have integer key and data.
1675 When so, you need to use the following functions
1676 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1677 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1678 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1679 In contrary to regular dicts, the key is not malloced before copy.
1680 Mixing scalar and regular elements in the same dict is not tested
1682 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1683 Tracing for Visualization:
1684 * SimGrid is now instrumented in order to generate a trace file for
1685 visualization analysis: to use it, need to compile SimGrid with the
1686 "tracing" option enabled, and instrument the program using SimGrid with
1687 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1688 (among other functions).
1689 * The instrumentation only traces the platform utilization for now
1690 * Documentation to use the tracing functions and how to analyze the
1691 traces with the Triva tool is written.
1692 * More information about: SimGrid FAQ (in the section Tracing Simulations
1695 * We moved to cmake as default build system. Autotools support will
1696 be dropped soon. Check the FAQ for more info about how to use it.
1697 * Greatly improved our cdash/ctest interactions
1698 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1699 * Added memory checking tests with valgrind; lot of memleak fixing.
1700 This may be the first release of simgrid with so few memory issues
1701 * Added code coverage tests.
1702 Our coverage is still improvable, but at least we see it on cdash.
1704 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1706 SimGrid (3.3.4) stable; urgency=low
1708 The "Desktop Grid needs love too" release (also called Xmas release).
1710 Models improvements:
1711 * Major speedup in the maxmin system solving by using lazy evaluation
1712 Instead of solving completely the maxmin system at each iteration,
1713 only invalidate (and recompute) the modified parts.
1714 This new feature is enabled in default models but you can try to
1715 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1716 * Cas01 IMproved as default CPU model
1717 This CPU model is the same Cas01 model, but it uses the
1718 maxmin-selective-update flag and a heap structure to manage
1719 actions on SURF kernel.
1720 It reduces the complexity to find the next action to finish and,
1721 consequently, it's faster than the old Cas01.
1722 This is the new default CPU model (Cas01).
1723 * Rename the old Cas01 model to Cas01_fullupdate
1724 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1725 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1726 * CpuTI (CPU Trace Integration)
1727 A new CPU model whose objective is simulate faster when using
1728 availability trace files.
1729 Instead of using a full featured, over engineered maxmin system for
1730 CPU modeling, this model does the pre-integration of traces files
1731 to calculate the amount of CPU power available, and so, executes
1732 faster than the old CPU models.
1733 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1734 * Use LV08 as default network model since it gives better accuracy
1735 for small messages and shouldn't change things for big ones.
1736 Use --cfg=network_model:CM02 to get the previous behavior.
1739 ******************************************
1740 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1741 ******************************************
1742 * The new CPU model may changes simulations!
1743 The point is that events occurring at the exact same timestamp
1744 are not scheduled in the same order with the old and new
1745 version. This may be enough to completely change the execution
1746 of simulations in some cases.
1747 * The new network model will change simulations!
1748 This new model is more realistic than the previous one, so you
1749 should consider redoing your old experiments with this model.
1750 Sorry for the inconvenience.
1753 * Introduce the supernovae compilation mode
1754 When compiled that way, the whole SimGrid (or almost) is put in a
1755 single compilation unit and compiled in one shoot.
1756 This is to help gcc which has difficulties to inline stuff from one
1758 The speedup seem to be above 15%, althrough more tests are needed on
1759 amd64 to confirm that gain.
1762 * Port of MSG's mailbox on top of SIMIX network
1763 The put/get mechanism was greatly simplified on the way.
1766 * New SIMIX network module. Provides:
1767 - Mailbox: rendez-vous mechanism to find with who you want to speak
1768 - Synchronous send/recv: easier and hopefully faster since the
1769 logic is handled in the maestro process directly now
1770 - Asynchronous send/recv: you dreamt of it? It's here now
1771 Too bad that nobody cared enough to propagate the change to MSG.
1772 * Add semaphores as SIMIX synchronization mechanism.
1775 * new function SD_daxload(char*) to load a DAX file
1776 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1777 * Introduce typed tasks. Specify its kind and cost at creation.
1778 At scheduling, just give where it should be placed, and the cost
1779 for each involved resource is automatically computed.
1780 Existing constructors so far (more to come of course):
1781 - SD_task_create_comm_e2e() for end-to-end communication
1782 - SD_task_create_comp_seq() for sequential computation
1783 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1784 * new function SD_task_dump() for debuging display
1785 * new function SD_task_dotty(task,FILE*) writing to file the info
1786 about the task in dotty format
1787 * SD_task_dependency_exists() can now cope with having one of its
1788 arguments NULL. If so, it tests whether the other argument has any
1790 * Add getters on list of preceding/following tasks:
1791 SD_task_get_parents(task) and SD_task_get_children(task)
1792 * Add getters on amount of workstations and list:
1793 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1794 * Add getter on task kind: SD_task_get_kind(task)
1795 * Update the start_time and finish_time of tasks on completion/failure
1796 * Bugfix: Remove task from state swags when destroyed
1799 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1802 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1803 dynar with one separate thread per value of the dynar.
1804 * Change the prototype of xbt_thread_create(), sorry.
1805 Added a boolean parameter indicating whether we want to join this
1806 thread (used in SG only for now)
1807 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1810 * GTNetS wrappers should now be usable again (and betterly tested too)
1811 * Fix a major regression from 3.2 where the timeout provided to
1812 MSG_task_put_with_timeout() was used as absolute time before which
1813 the comm should be done.
1814 * Start to fix the <cluster> tag.
1815 - Internal links should be good now (beside of the loopback, which
1816 use the private link instead)
1817 - paths to the external world is still rather broken
1818 - the <route:multi> tag is just broken. Actually that's brain-dead.
1819 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1820 to make it less stupid
1821 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1822 * Fix a source-level compatibility glitch from 3.2: after defining
1823 MSG_USE_DEPRECATED, you can use the old name
1824 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1825 * Allow to compile from the SVN with automake 1.11
1826 * Fix some problems when using the "start_time" tag in deployment XMLs.
1827 * Fix #8569: XBT/synchro.h has redundant declarations
1828 * Fix #8563: MSG return values and exceptions
1829 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1830 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1831 Thanks to Robson Peixoto.
1832 * Fix a vicious bug in dictionaries inducing that some elements were
1833 not freed on xbt_dict_free()
1835 Portability report of this version:
1836 * Main portability targets:
1837 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1838 on (amd64/i386/ia64)
1839 - mac leopard on i386
1840 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1841 but nothing critical.
1842 * Other platforms: windows, AIX and others were not tested for this release
1844 Timing report of this version:
1845 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1846 scenario parameters). From 8h to a few seconds in desktop grid settings.
1847 * Supernovae brings about 25% speedup on i386.
1849 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1851 SimGrid (3.3.3) stable; urgency=low
1853 The "Need for Speed" release.
1855 The timings done to validate the 3.3.2 were faulty.
1856 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1858 The problem was a conversion from a manually handled vector to
1859 xbt_dynar_t on the critical path.
1860 xbt_dynar_foreach calls functions, inducing stack management crap.
1862 We inlined these functions and xbt_dynar_foreach is now breath taking.
1863 We also inlined xbt_swag_belong on the way.
1865 Here are some approximate speedup measurements (on master/slaves
1866 simulations lasting between 10s and 20s each):
1867 3.3.1 -> 3.3.2: about same performance
1868 3.3.2 -> 3.3.3: 40% speedup
1869 3.3.1 -> 3.3.3: 40% speedup
1870 3.3.1 with inline patch -> 3.3.3: 30% speedup
1872 Our reading is that the refactoring which occurred in 3.3.2 made us
1873 suffer much more from the xbt_dynar_foreach low performance, but
1874 once we solved this, this refactoring proved to be very performance
1875 effective. From the 40% speedup, somehow, 10% are due to the
1876 inlining and 30% to the refactoring.
1878 That's a pitty that gcc cannot inline functions placed in other files
1879 alone. We have to choose between:
1880 - break the encapsulation (by putting private data structures and
1881 accessors in headers files to help gcc)
1882 - live with low performance
1883 - switch to a decent compiler such as icc (not quite possible).
1885 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1887 SimGrid (3.3.2) stable; urgency=low
1889 The "Simplicity does not preceed complexity, but follows it" release.
1891 The main contributors of this release were (lexical order):
1892 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1895 * Extract the routing logic into its own object.
1896 (was dupplicated in network.c and workstation_LV07.c;
1897 Allows to implement other ways of storing that info)
1898 => kill now useless network_card concept
1899 - Use dynar to represent routes (instead of void** + int*)
1900 - kill link_set (use surf_network_model->resource_set instead)
1901 - Add a command-line option to choose the routing schema to use
1902 - Add three new models:
1903 * Floyd (shortest path computed at initialization)
1904 * Dijikstra (shortest path recomputed all the time)
1905 * Cached Dijikstra (shortest path computed on need)
1906 All these models where contributed by Silas De Munck, and are
1907 described in his ICCS09 paper.
1909 * Simplify model declaration
1910 (less redirections, less function to write when defining a model)
1911 - Factorize stuff between models:
1914 surf_model_resource_set(model)
1915 surf_model_resource_by_name(model, name)
1916 - Unify the types of models in s_surf_model_t (using an union)
1917 - Embeed fields of common_public directly into s_surf_model_t
1918 - Rename model methods:
1919 action_free ~> action_unref
1920 action_change_state ~> action_state_set
1921 action_get_state ~> action_state_get
1922 - Change model methods into functions :
1923 (model)->common_public->action_use ~> surf_action_ref
1925 * Implement a generic resource; use it as ancestor to specific ones
1926 (allows to kill duplicated code in models)
1927 Drawback: timer command don't need no name nor properties;
1928 workstation_CLM03 don't need no properties
1929 (but I guess we can live with those few bytes wasted)
1931 * Improve the action object model
1932 - implement a constructor avoiding dupplicated code about field
1933 initialization in generic_action part.
1935 * Kill the SDP model: it has an external dependency, is deprecated
1936 in flavor of modern lmm models, and didn't compile since a while
1939 * Relocation of the context module from XBT to SIMIX.
1940 (the context were decoupled from the simix processes, duplicating a lot of code)
1941 => a lot of code was factorized
1942 - less overhead is introduced during scheduling
1943 - simpler API for the context factory
1944 - the logic for process creation,destruction and manipulation was simplified
1945 * Simplification of the s_smx_process_t data structure.
1946 => accesing the simix level data associated to a process is faster now,
1947 and the code is a lot more readable.
1950 * Implement some more MPI primitives:
1951 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1952 -implementation: Bcast: flat or 2-ary tree (default),
1953 Barrier: 4-ary tree,
1955 Allreduce: Reduce then Bcast
1956 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1957 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1958 Alltoallv: flat tree, like ompi
1960 * Add support for optimized collectives (Bcast is now binomial by default)
1961 * Port smpirun and smpicc to OS X
1964 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1965 Shout out if you used it.
1968 * Display the list of still queued messages in SG mode when existing
1972 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1973 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1974 * Add xbt_str_from_file(FILE*)
1975 * Add xbt_dict_get_key achieving a linear reverse search
1976 * Remove the context module
1978 Portability report of this version:
1979 * Main portability targets:
1980 - Linux(debian)/x86/context
1981 - Linux(debian)/x86/pthreads
1982 - Linux(debian)/amd64/context
1983 - Linux(debian)/amd64/pthreads
1984 On these, we still have the eratic breakages of gras/pmm and
1985 amok/saturate_sg reported in previous version. We still think
1986 that the tests are the cause of the fault, not the tested code.
1988 - Mac OSX Leopard/x86/context
1989 Still false negative in tesh autotesting.
1990 Smpi still fails, but this time because readlink does not accept -f
1991 Everything seems to work properly beside of that.
1994 - AIX version 5.3 (only tested contexts this time)
1995 Smpi still fails there because mktemp is not installed.
1996 Everything seems to work properly beside of that.
1998 I managed to compile it for the first time, but several breakages.
1999 Won't delay the release for this exotic platform.
2001 * Windows: it's still lagging behind. If you want to help, please
2004 Timing report of this version:
2005 This version seem to be more than 5% faster than 3.3.1 (on linux
2006 64bits with contextes). The gain is less than expected, we are
2007 investigating this for next release.
2009 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2011 SimGrid (3.3.1) stable; urgency=low
2014 * Implement a --cfg-help to show existing configuration variables
2015 * Build chain do not require doxygen in maintainer mode
2018 * fix a bug on struct sizeof computation, which prevented the
2019 exchange of arrays of structs in some conditions
2020 - added a regression test about this in datadesc_usage
2021 * Allow the exchange of 0-long dynamic vectors.
2022 - for that, use -1 as indicator of dynamic size instead of 0
2023 - This implied to change any size from unsigned long to long,
2024 reducing a bit communication abilities, but I guess that with
2025 64bits being quite common, this is more than enough.
2026 - This also induce a protocol change, thus bumping network protocol
2027 version from 0 to 1 (if we have external users, we have to get
2028 clean on that point too ;)
2029 - added two regression tests about this in datadesc_usage
2030 * Be more verbose when propagating local exceptions
2031 This helps debugging.
2032 * Display the status of simulated processes when receiving SIGINT in
2036 * Allow to control the simulation from a trace file.
2037 New functions MSG_action_register() and MSG_action_trace_run()
2038 The first one allows to associate a function execution to each
2039 kind of action while the second one parses a trace file and
2040 triggers the corresponding actions within the system.
2041 For now, only a toy example is provided in examples/msg/actions
2042 * Add an exemple of process migration in examples/msg/migration
2043 * Fix a bug in task exchange which broke MSG_task_get_sender()
2044 Add a teshsuite regression test for that.
2045 [Bug: if MSG_task_get_sender() is called after sender exit,
2047 * Fix a bug which prevented suspend/resume to work properly
2048 * Display the status of simulated processes when receiving SIGINT
2049 This fixes a regression of v3.3. due to the introduction of SIMIX
2050 * Bug fixing in failure management:
2051 - trace could not start by a failure at time 0
2052 - failure during communications were not working
2055 * Add SIMIX_process_set_name() to change the name of the current
2056 process in the log messages.
2057 * Store smx_hosts in a dict since we only retrieve them by name
2058 * Move the configuration infrastructure to surf
2061 * Move the configuration infrastructure to surf
2064 * Massive internal cleanups:
2065 - Store internal structures on processes instead of hosts (allows
2066 to have more than one process per host, in addition of being more
2068 - Cleanup the initialization/finalization process
2069 - Kill a whole bunch of unneeded synchronization:
2070 processes run in exclusive manner within the simulator
2071 - Move queues from global tables to process data fields
2073 - now accept -platform and -hostfile arguments
2074 - Pass the right rank value to processes according to the hostfile
2075 * Compile the examples by default, and use them as regression tests
2076 * Implement MPI_Wtime()
2077 * Change the reference speed to a command line option
2080 * TCP_gamma can now be specified as command line option using
2081 --cfg=TCP_gamma:10000000.0
2082 * Change the --surf-path cmd line option into --cfg=path:
2085 * Also include strbuff from xbt.h public header
2086 * xbt_ex_display(): do not free the exception after displaying
2087 This allows to do more with the given exception afterward.
2088 Users should call xbt_ex_free() themselves.
2092 Portability report of this version:
2093 * Main portability targets:
2094 - Linux(debian)/x86/context
2095 - Linux(debian)/x86/pthreads
2096 - Linux(debian)/amd64/context
2097 - Linux(debian)/amd64/pthreads
2098 These targets fail about 1/10 of times on gras/pmm, but we believe
2099 that this is because of the test, not because of simgrid.
2100 amok/saturate_sg fails even more rarely, and the test may not be
2103 - Mac OSX Leopard/x86/context
2104 The test suite still spits tons of errors because some obscure
2105 force prevents us from removing the temporary directories
2106 arguing that they still contain some metadata I've never heard of.
2107 Smpi fails because seq is not installed.
2108 Everything seems to work properly beside of that.
2111 - AIX version 5.3 (both contexts and pthread)
2112 Smpi still fails there because mktemp is not installed.
2113 XML inclusions seems rosty on AIX.
2115 * Windows: it's still lagging behind. If you want to help, please
2118 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2120 SimGrid (3.3) stable; urgency=high
2124 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2125 [Malek Cherier & Mt]
2127 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2130 Until now, GRAS were using MSG as an interface to SURF. It was
2131 quite difficult because both interface have several differences
2132 (MSG channels vs GRAS sockets were the most notable point).
2134 This also opens the gate to SMPI (which should occur soon) and speed
2135 up simulations by to 40% (even if it were not the main goal).
2137 **************************************
2138 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2139 **************************************
2140 The point is that events occuring at the exact same timestamp are
2141 not scheduled in the same order with the old and new version. This
2142 may be enough to completely change the execution of simulations in
2143 some cases. Sorry for the inconvenience.
2145 * Cleanup and upgrade the XML format to push further scalability
2146 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2148 * Improve the testing infrastructure with tesh. Now a very large part of
2149 the code is tested not only by being run but also by checking that the
2150 output match an expected output [Mt].
2152 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2153 is really less memory-demanding, which should allow you to use
2154 larger files in SimGrid [AL].
2156 * Inform valgrind about our contextes, so that it becomes usable
2157 with the default (and more effecient) version of SimGrid
2158 [contributed by Sékou Diakite, many thanks]
2161 * Introduce a listener thread in charge of receiving incoming
2162 messages from the network. It allows to overlap communication and
2163 computation but most notably, it removes some stupid deadlocks due
2164 to the fact that so far, a process could not send and receive at
2165 the same time. This made most non trivial communication schema
2167 * Convert the PIDs from long int to int to match the MSG ones (and
2168 linux ones too) [Mt]
2169 * New function: gras_agent_spawn() to launch a new process on
2170 current host. Only working in simulation for now. [Mt]
2171 * New function: gras_os_hostport() returning a constant form (ie,
2172 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2175 * Make the backtrace of exceptions more human readable [Mt]
2176 * New module: xbt/str [Mt]
2177 a ton of string utility functions (split, join, printf to a newly
2178 allocated buffer, trim, etc)
2179 * New module: xbt/hash [Mt]
2180 SHA1 hashing algorithm (more to come if needed)
2181 * New module: xbt/synchro [Mt]
2182 synchronization tools (mutex and conditions) working the same way
2183 in simulation and in real life (mainly useful for GRAS, but not
2185 * New module: xbt/queue [Mt]
2186 classical producer/consumer synchronization scheme
2187 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2188 (using the classical functions will get serialized) [Mt]
2189 * Make dictionary internal table dynamic. No need to specify its size
2190 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2192 * Make sure the log channels are organized as a tree under windows
2193 (because of ANSI C compatibility issue, any channel were child of
2194 root directly) [Mt].
2197 * Cleaned many thing in surf and fixed a few bugs [AL].
2198 * Add a nice command line configuration mechanism to compose models [AL].
2199 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2200 the previous one (KCCFLN05). It relies on something that looks like
2201 a max-min sharing mechanism but cannot be written as such. A new solver
2202 was thus designed [AL].
2203 * Add a new solver to lmm. Based on Lagrange optimization and
2204 gradient-based descent, it enables to efficiently maximise systems s.a
2206 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2208 This solver enables to propose two new network models for TCP Reno and
2209 TCP Vegas based on Low's work. These models still need to be fully
2210 tested though [Pedro Velho].
2213 * Bug fix in SD_simulate. Now the time bound given as argument is
2215 * Use the new parallel task model (ptask_L07) as default.
2216 * Use the SURF command line configuration mechanism.
2217 * 0-size tasks (for synchronization) should now work.
2219 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2221 SimGrid (3.2) stable; urgency=high
2225 We still experience issues on this platform, but we believe that at
2226 least MSG is usable.
2228 GRAS API BREAKAGE (for simplification purpose, sorry):
2229 * the gras_msgtype_by_name is not used anymore. Instead of
2230 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2231 you can write (and must)
2232 gras_msg_send(toserver, "request", &request);
2233 - If you still want to pass a gras_msgtype_t to the function (to cache
2234 the type and avoid the lookup time), use the gras_msg_send_() variant.
2235 - Impacted functions:
2236 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2237 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2238 * The callbacks are now expected to return 0 when everything went well
2239 (just like the main() function)
2241 GRAS new features and improvements:
2242 * New module mechanism where user code can use per process globals [Mt]
2243 This is similar to gras_userdata_*() functions, but for libraries. It
2244 factorize some code developped over and over in the examples and AMOK.
2245 It has still to be documented and used (only amok/peermanagement is
2247 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2248 loss when they were small enough to fit into the buffer and sent quickly
2249 enough so that they can all get received in one shoot.
2250 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2251 if not found. Use the *_or_null() variant for the old semantic.
2252 * In gras_msg_handle, do not discard messages without callback.
2253 They are probably messages to be explicitly awaited later (ie, proofs of
2254 mis-synchronization in userland since they are sent before being awaited)
2255 No big deal usually.
2256 * gras_socket_meas_send/recv: semantic changed!
2257 The numerical arguments used to be (1) the total amount of data to send
2258 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2259 messages. This was need for the fool willing to send more than MAXINT
2260 bytes on quite fat pipes.
2263 * Do really rename the hostmanagement module to peermanagement. [Mt]
2264 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2265 but this is rather new and this was documented in the module
2266 documentation (poor excuses, I admit)
2267 * Bandwidth measurement semantic changed! This follows the changes to
2268 gras_socket_meas_send/recv explained above.
2271 * A sequential mode has been added to the workstations. When a workstation
2272 is in sequential mode, it can execute only one task, and the other tasks
2273 are waiting in a FIFO. [Christophe Thiery]
2276 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2277 model for SIMDAG. [Christophe Thiery]
2278 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2279 equal to 0, which caused some bad behaviors in
2280 saturated_constraint_set_update. I now use a threshold mechanism like in
2284 * When running manually src/testall, you select specific units [Mt]
2285 testall is the result of our cunit mechanism, and should replace all
2286 the scripty thingy around since bash don't run easily on billware.
2288 * A mallocator system has been added. [Christophe Thiery]
2289 Mallocators allow you to recycle your unused objects instead of freeing them
2290 and allocating new ones.
2292 Documentation update:
2293 * FAQ reworking + New FAQs:
2294 - "Valgrind spits tons of errors!" [Mt]
2295 - "How to repport bugs" [Mt]
2296 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2297 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2298 - Communication time measurement within MSG [AL]
2299 - I experience weird communication times when I change the latency [AL]
2300 * GRAS tutorial [Mt]
2302 - an introduction to the framework and to the used communication model
2303 - an initiatic tour introducing the most proheminent features:
2305 . Lesson 0: Installing GRAS
2306 . Lesson 1: Setting up your own project
2307 o Part 2: Message passing
2308 . Lesson 2: Exchanging simple messages
2309 . Lesson 3: Passing arguments to the processes (in SG)
2310 . Lesson 4: Attaching callbacks to messages
2311 . Lesson 5: Using globals in processes
2312 . Lesson 6: Logging information properly
2313 . Lesson 7: Using internal timers
2314 . Lesson 8: Handling errors through exceptions
2315 . Lesson 9: Exchanging simple data
2316 . Lesson 10: Remote Procedure Calling (RPC)
2317 . Lesson 11: Explicitely waiting for messages
2318 . Recapping of message passing features in GRAS
2319 - A HOWTO section containing:
2320 o HOWTO design a GRAS application
2321 More are due, of course. They will come latter. In the meanwhile, you can
2322 check the examples which are still here.
2324 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2326 SimGrid (3.1) stable; urgency=high
2330 There was a stack corruption somewhere, visible only when optimizing
2331 with these versions. [Vince]
2334 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2335 module that enabled to play with Directed Acyclic Graphs. It is built
2336 directly on top of SURF and provides an API rather close to the old
2337 SG. Some old codes using SG are currently under rewrite to check that
2338 all needful functions are provided. [Christophe Thiery]
2341 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2342 extension of the classical CLM03 model that gracefully handles
2343 failures. This is now the default model for MSG and GRAS. It doesn't
2344 handle parallel tasks yet though. [AL]
2345 * Bug fix: Weights were not correctly set in the network part.
2346 WARNING: This may have resulted in incorrect results with simulations
2347 where there are more than one flow on a given link. [AL]
2350 * After a (long ?) discussion on simgrid-devel, we have decided that the
2351 convention we had on units was stupid. That is why it has been decided
2352 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2353 WARNING : This means that all previous platform files will not work as
2354 such with this version! A warning is issued to ask users to update
2356 A conversion script can be found in the contrib module of the CVS, under
2357 the name contrib/platform_generation/surfxml_update.pl [MQ]
2360 * Bug fix: Processes were started in reverse order, wrt deployment file.
2361 WARNING: if your code relies on this bug, please fix it. [AL]
2362 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2363 being executed on two different locations. [AL]
2364 * Bug fix: Failures are now better supported thanks to Derrick's tests
2365 (there was many failure situations I hadn't thought of and that weren't
2366 correctly handled). [AL]
2367 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2371 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2372 Exception occurring on server-side are propagated back to client (!).
2374 API CHANGE: the callback changed their prototype. Change:
2375 int my_handler(gras_socket_t expeditor, void *payload_data) {
2377 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2378 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2380 * New! function: gras_msg_handleall to deal with all messages arriving
2381 within a given period.
2382 * New! function: gras_socket_server_range to get a server socket in a
2383 range of port numbers (ease to avoid port number conflicts) [MQ]
2384 * New! gras processes display their backtrace when they get a SIGUSR1
2385 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2386 Sweet to debug RL processes [MQ]
2390 - Do not force experiment sizes to be expressed in kb, or it becomes
2391 impossible to measure the latency this way (needs one byte-long tests)
2392 WARNING: this changes the amok_bw_* function semantic. [MQ]
2393 - Implements the link saturation stuff. [MQ]
2394 * Peer management module:
2395 New! module factorizing code that we wrote over and over [MQ].
2398 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2399 - Test units are placed directly into the library code, they get extracted
2400 automatically and placed into the src/testall binary.
2401 - Convert most of the XBT tests to this system.
2402 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2403 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2404 live in the same C code. This required to change a little bit the API
2405 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2406 * New module: sparse graph structure with basic algorithms (this is work
2407 in progress and the API is not considered to be frozen yet). [AL]
2408 * Display more information on backtraces: source line & function names are
2409 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2410 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2411 * Reworked a little bit some #include statements to load only required
2412 headers. Some user code that relied on SimGrid to include stdlib or
2413 stdio may need to include it by themselves. [AL]
2414 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2416 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2417 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2418 but I guess that almost nobody used those parts. [MQ]
2420 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2422 SimGrid (3.0.1) stable; urgency=low
2425 * Unfortunately, I had missed 5 misnamed functions:
2426 xbt_fifo_item_t xbt_fifo_newitem(void);
2427 void xbt_fifo_freeitem(xbt_fifo_item_t);
2428 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2429 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2430 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2431 They're now deprecated. Please use their new versions:
2432 xbt_fifo_item_t xbt_fifo_new_item(void);
2433 void xbt_fifo_free_item(xbt_fifo_item_t);
2434 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2435 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2436 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2438 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2439 * Documentation: xbt_log module unmercifully reworked [MQ]
2440 * Bugfix: there was a problem with the ending of contexts with
2441 the pthread backend. It caused some weird deadlock or behavior
2442 depending on the pthread implementation. [AL]
2443 * Bugfix: get the exceptions raised in the simulator repport where
2444 and why they come from when they are not catched in time [AL, MQ]
2447 * Bugfix: Do repport the error when two non-connected hosts try to
2448 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2451 * Add additionnal checkings on communications. Assert that two
2452 communicating hosts are connected by a set of links... [AL]
2455 * Add additionnal checkings on channel values in communication [AL]
2456 * New: MSG_task_get_source to see on which host a task was generated [HC]
2457 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2458 the number of tasks waiting to be received on channel and sent
2460 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2461 waits for the first task coming from a given host.. [AL]
2463 GRAS new functionnalities: [MQ]
2464 * Enhance the parsing macro to allow the size of multidimentional objects
2465 to be given thru annotations.
2466 * New example (and documentation): Matrix Multiplication a la RPC
2467 (as when I was young!) and fix a bunch of bugs found on the way.
2469 GRAS performance improvements: [MQ]
2471 * Reduce the amount of cbps creation/destruction by making it static to
2472 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2475 * Change libdata to a set so that we can search for stuff by ID (and thus
2476 reduce the insane amount of dict lookups)
2479 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2480 RL communication on top of SG since it's so uneffective.
2481 It may also be used for inter-thread communication in RL, one day.
2482 * Use gras_datadesc_copy() to exchange messages on top of SG
2484 - improve message exchange performance on top of SG
2485 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2486 * Don't exchange on the network the size of the used part of buffer,
2487 instead, specify the possible buffer size to read().
2489 - reduces the amount of read/write calls (one pair per exchange)
2490 - reduces the amount of exchanged data (the size)
2491 - allows to retrieve all arrived data on receiver side, if we don't need
2492 it right now (subsequent read will peek the buffer)
2493 - allows the receiver to proceed with the begining of the stream before
2494 everything is arrived
2495 - make it possible to build an iov transport (using readv/writev)
2497 - take care of the data with non-stable storage (like stacked data),
2499 * If possible, TCP send uses vector I/O (when writev() is here)
2500 - Don't use it for receive since we send data sizes and data on the
2501 same stream, so we wouldn't be able to chain large amount of chunks
2502 before having to flush the stuff to read the size.
2503 * Rework the transport plugin mechanism to simplify it and reduce the
2504 amount of pointer dereferencement when searching for the right function
2507 * I guess that now, we do almost as few system calls as possible while
2508 doing as few data copy as possible.
2510 To improve it further, we could try to send all the sizes first and then
2511 all the data (to use iov on receiving size), but it's only a partial
2512 solution: when you have 2 dimensional data, the sizes of the second
2513 dimension is data of the first dimension, so you need 3 streams.
2515 I'm not sure the potential performance gains justify the coding burden.
2517 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2519 SimGrid (3.00) stable; urgency=high
2522 * New! Give the possibility to hijack the surf parser and thus bypass
2523 MSG_create_environment and MSG_launch_application. Have a look at
2524 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2526 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2528 SimGrid (2.96) unstable; urgency=low
2533 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2534 This deprecates the xbt_error_t mechanisms.
2535 It modifies (simplifies) all XBT and GRAS API.
2536 MSG API keeps unchanged (exceptions raised by XBT are catched from
2537 within MSG and masked with existing error handling facilities)
2540 * New! Add a FATPIPE model. [AL]
2541 * New! Add a parallel task model. [AL]
2542 * New! Add automatically a loopback interface (in the default
2543 network model) if none was precised.
2546 * Bugfix: MSG_process_resume now works with the current running process.
2548 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2549 * Modification of MSG_task_get_compute_duration. Once a task has been
2550 processed, the value returned by this function is now equal to 0. [AL]
2551 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2552 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2553 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2554 of a m_task. For now, MSG_task_cancel only works with computation
2556 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2557 of the processor (in Mflop/s) regardless of the current load on the
2559 * API Change: use proper naming convention for MSG_getClock and
2560 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2562 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2563 This function changes the priority of a computation task. This priority
2564 doesn't affect the transfer rate. A priority of 2 will make a task
2565 receive two times more cpu power than the other ones. This function
2566 has been added to suit the needs of Nguyen The Loc and hasn't been that
2567 much tested yet. So if it fails, please report it and send me your code.
2569 * API Change: removed all functions and types that were marked "deprecated"
2570 since many months. Renamed MSG_global_init_args to MSG_global_init.
2572 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2574 SimGrid (2.95) unstable; urgency=low
2577 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2578 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2579 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2582 * Bugfix: Allow absolute paths for platform description files [MQ]
2583 * Bugfix: do free the variables after use. Leads to drastic performance
2585 * Implement max_duration (ie, timeouts) on resources [AL]
2588 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2590 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2591 the needed support (timeouts) [AL]
2594 * Implement measurement sockets. You can now get the bandwidth between two
2595 hosts thanks to AMOK (see below). [MQ]
2596 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2597 dynar over the network (yeah) [MQ]
2598 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2599 * simple/static token-ring example. [Alexandre Colucci and MQ]
2600 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2601 in place before [MQ]
2604 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2605 type parsing. "short" and "long" modifiers were also ignored; other
2606 modifier (such as reference level) are still ignored. [MQ]
2607 * Embeed the buffer size within the buffer itself on SG. [MQ]
2608 That way, send() are atomic and cannot get intermixed anymore (at least
2609 the ones which are less than 100k; bigger messages still have the issue)
2610 * Array size pushed by the field, not by the field type (or each
2611 and every long int will push stuff to the cbps) [MQ]
2612 * use select() to sleep since it allows to portably sleep less than one
2615 GRAS (minor cleanups)
2616 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2619 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2620 * type category 'ignored' killed as it was never used and were difficult
2622 * whether a type can cycle or not is now a flag, leaving room for more
2623 flags (as "ignored", if we feel the need one day ;)
2624 * Rename raw sockets to measurement sockets since "raw" has another
2625 meaning in networking community.
2628 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2629 and offers features not belonging to GRAS but that most applications
2630 need. One day, it may be a set of plugins loadable at runtime.
2631 * New module: bandwidth
2632 bandwidth measurement between arbitrary nodes running this module. [MQ]
2634 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2636 SimGrid (2.94) unstable; urgency=low
2638 The first beta release of SimGrid 3 !
2642 * Update the main page and the FAQ. Adding references to gforge.
2645 * Add a gras_os_getpid function.
2648 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2649 * Extend the logs so that they also print PID, hostname, date, ... if
2651 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2652 and kill the old version which were in testsuite/
2653 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2657 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2658 for users. Instead, each programming environment (GRAS, MSG,...) use it
2659 when needed to provide such a feature to users.
2660 Don't shortcut the mechanism or you will also shortcut the virtualization
2661 you need on the simulator.
2665 * Cleanups in configury with regard to compile optimization/warning flags.
2666 Also add -fno-loop-optimize to any powerpc since it's the optimization
2667 killing gcc (< 3.4.0).
2668 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2669 needed by MSG examples complications
2670 * Borrow configury beautifications from PHP
2673 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2674 warning (thanks loris for stumbling into this one).
2675 * Bugfix: stop loading private headers (gras_config.h) from the public
2679 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2680 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2681 RL would imply a DNS resolver, which is *hard* to do in a portable way
2682 (and therefore delayed).
2683 * Implement a real timer mechanism and use it in timing macros. This allows
2684 to avoid rounding errors and get a 0.000005 sec precision in timing
2685 macros. While I was at it, various cleanups:
2686 - allow to declare more than one timed section per file (fix a stupid bug)
2687 - move some private declaration to the right place
2688 - merge conditional execution and timing macros into emulation module
2689 - document the module
2690 - make sure the module cleanups its mess on gras_exit
2691 * Documentation improvements:
2692 - (new) how to compile applications using GRAS
2693 - (new) emulation support (timing macros)
2695 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2697 SimGrid (2.93) unstable; urgency=low
2699 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2702 - Use Paje properly where used. Still to be sanitized properly.
2703 - Portability fix: Add an implementation of the contexts using pthread
2707 - Add xbt_procname(): returns the name of the current process.
2708 Use it to show the current process's name in all logging.
2710 - fix detection of older flex version and the reaction, since we do
2711 depend on modern ones (we use lex_destroy)
2712 - Better separation of SG and RL in the libs: remove all simulation code
2713 from libgras. As a result, this lib is now only 200k when stripped.
2714 Some of the xbt modules may also be duplicated (two sets and such) and
2715 should be cleaned/killed before SG3.
2716 - Insist on using xlC on AIX because of weird problems involving gcc there.
2717 - Cleanup the make remote stuff. This is now done by scripts
2718 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2719 mainly for our private use, but we're working on changing them to user
2722 - Bugfix: flush the socket on close only if there is some *output*.
2723 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2725 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2728 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2730 SimGrid (2.92) unstable; urgency=low
2732 Alpha 3 on the path to SimGrid 3
2736 - New! First try of benchmarking macros.
2737 - New! First try so that gras_stub_generator generate deployment and
2738 remote compilation helpers.
2740 - Bugfix: Initialization fix in msg_test.
2744 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2746 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2748 - API Change: gras_os_sleep to take the amount of seconds as a double.
2749 Accepting an int was error prone since it was the only location where
2750 seconds were coded as such. It leaded to damn rounding errors.
2751 - Bugfix: Hard to belive that timers ever worked before this.
2753 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2755 SimGrid (2.91) unstable; urgency=low
2757 Alpha 2 on the path to SimGrid 3
2761 - Bug fix in the lmm_solver.
2763 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2764 through the function MSG_paje_output.
2765 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2766 - It is possible to bound the rate of a communication in MSG with
2767 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2768 to write it in the changelog).
2769 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2774 - Various cleanups to the autotools stuff
2775 - Begin to move Gras examples to examples/gras/
2776 - Let make distcheck work again (yeah!)
2778 - documentation overhauled using doxygen.
2779 gtk-doc-tools is dead in SimGrid now.
2780 - Automatically extract all existing logging categories, and add the list
2781 to the documentation (long standing one, to say the less)
2783 - Cleanup the known architecture table. Reorder the entries to group what
2784 should be, and use a more consistent naming scheme.
2785 (some of the test dataset are still to be regenerated)
2786 - New! Allow library to register globals on each process just as userdata
2788 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2789 lookup time (for now).
2790 Use it in msg and trp.
2791 This cleans a lot the internals and helps enforcing privacy of the
2792 headers between the gras components.
2793 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2794 - Bugfix: gras_os_time was delirious in RL.
2795 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2797 - Reenable GRAS now that it works.
2799 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2801 SimGrid (2.90) unstable; urgency=low
2803 Alpha 1 on the path to SimGrid 3
2805 * It is a long time since the last release of SimGrid. I'm sorry about
2806 that but as I had told you, I was rewriting a lot of things. I apologize
2807 to those who had been reporting bugs to me and that I had not answered.
2808 If your bug is still in the new version, please tell me. Here is a
2809 summary of the main changes.
2811 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2812 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2813 portability, speed, and a lot more but you'll figure it out later.
2814 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2815 I wanted to release the new MSG as soon as possible and I have
2816 broken GRAS, which is the reason why, for now, only MSG is fully
2817 functional. A laconic description of these projects is available
2818 in the documentation.
2820 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2821 kernel optimized for our needs (called SURF but only the developers
2822 should use it). Hence, MSG is now roughly 30 times faster and I think
2823 that by rewriting a little bit MSG, I could event speed it up a little
2824 bit more. Beside the gain in speed, it is also much easier to encode a
2825 new platform model with SURF than it was with SG. More to come...
2827 * REVOLUTION 3: I have tried to change a little as possible the API of
2828 MSG but a few things really had to disappear. The main differences
2829 with the previous version are :
2830 1) no more m_links_t and the corresponding functions. Platforms are
2831 directly read from a XML description and cannot be hard-coded
2832 anymore. The same format is used for application deployment
2833 description. The new format is described in the documentation.
2834 Have a look in tools/platform_generation. There is a tiny script
2835 that converts from the old platform format to the new one. Concerning
2836 the application deployment format, parsing the old one is tricky.
2837 I think most of you should however be able to convert your files. If
2838 it is really an issue, I can write a C code that does the conversion.
2840 2) the toolbox tbx does not exist anymore. We now have a library
2841 with much more data-structures but without the hash-tables (we have
2842 dictionaries that are much faster).
2844 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2846 *****************************************************************************
2847 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2848 * don't feel like converting the oldies. (Mt) *
2849 *****************************************************************************
2852 Version 2.90: "the long awaited one"
2853 - Finished rewriting and debugging MSG. Rewrote the documentation.
2854 - disable GRAS for now since it needs to be ported to the newest SG
2857 - Finish the port to windows (using mingw32 for cross-compile)
2860 - Main loop and datastructures of SURF. A cpu resource object is
2861 functional. Surf can thus be used to create cpu's with variable
2862 performance on which you can execute some actions.
2864 2004-11-15 Martin Quinson
2865 - Port to ARM. Simply added the alignment and size descriptions. Should
2866 work, but the ARM machines are so slow that I didn't had the opportunity
2867 to 'make check' over there yet.
2869 2004-11-15 Arnaud Legrand
2870 - Trace manager now written. It uses a heap structure and is therefore
2871 expected to be efficient. It may however be speeded up (particularly
2872 when many events occur at the same date) by using red and black
2873 trees. One day maybe...
2874 - Max-min linear system solver written. It uses a sparse matrix
2875 structure taking advantage of its expected use. Most operations are
2876 O(1) and free/calloc are called as few as possible. The computation of
2877 the minimum could however be improved by using a red and black tree
2880 2004-11-03 Arnaud Legrand
2881 - Rename every gras_* function that was in xbt/ to its xbt_
2883 - Add a heap and a doubly-linked list to xbt
2884 - Added a dichotomy to the dictionaries. make check works as well before
2885 so I assume that the patch is correct. I do not know however if things
2886 run effectively faster than before now. :)
2888 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2889 SimGrid, and the version number is bumped to 2.x
2891 2004-10-29 Martin Quinson
2892 - Introduction of the remote errors.
2893 They are the result of a RMI/RPC on the remote machine.
2894 ErrCodes being scalar values, you can't get the host on which those
2895 errors did happen. Extending the error mechanism as in Gnome is possible.
2896 No idea yet whether it is a good idea.
2898 2004-10-28 Martin Quinson
2899 - Interface revolution: the Starred Structure Eradication.
2900 I used to do typedef struct {} toto_t; and then handle *toto_t.
2901 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2902 - ???_t is a valid type (builded with typedef)
2903 - s_toto_t is a structure (access to fields with .)
2904 - s_toto is a structure needing 'struct' keyword to be used
2905 - e_toto_t is an enum
2906 - toto_t is an 'object' (struct*)
2908 typedef struct s_toto {} s_toto_t, *toto_t;
2909 typedef enum {} e_toto_t;
2910 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2911 s_toto_t) is private.
2913 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2914 so that it can set the variable to NULL. It was so for dicts and sets,
2915 it changed for dynars.
2917 - Fix a bunch of memleaks in dict_remove
2918 - Fix a bug in sg/server_socket opening: it failed all the time.
2920 2004-10-07 Martin Quinson
2921 - Speed up dynar lookup operation a bit.
2923 gras_dynar_get is dead.
2925 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2926 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2927 (which gives you the address of the stored data).
2929 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2930 copy of the data using an affectation to do the job and not a memcpy.
2932 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2933 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2935 It does not really speedup the dynar test because they are
2936 setting elements all the time (and look them seldom). But the dict does
2937 far more lookup than setting.
2939 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2941 2004-10-05 Martin Quinson
2942 - Allow to (en/dis)able the cycle detection at run time.
2944 Whether we should check for cycle or not is now a property of each
2945 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2946 datadesc_cycle_unset allow to remove this property when previously set.
2948 Note that the cycle detection is off by default since it impacts the
2949 performance. Watch the data you feed GRAS with ;)
2951 This property is hereditary. Any element embedded in a structure having it
2952 set have it set for the time of this data exchange.
2954 You should set it both on sender and receiver side. If you don't set it on
2955 sender side, it will enter an endless loop. If you forget on receiver
2956 side, the cycles won't be recreated after communication.
2958 - Header reorganization.
2959 Kill gras_private.h, each submodule must load the headers it needs.
2961 2004-10-04 Martin Quinson
2962 - Interface revolution: do not try to survive to malloc failure.
2964 Now, gras_malloc and friends call gras_abort() on failure.
2965 As a conclusion, malloc_error is not a valid error anymore, and all
2966 functions for which it was the only gras_error_t return value are
2967 changed. They now return void, or there result directly.
2968 This simplify the API a lot.
2970 2004-09-29 Martin Quinson
2971 - Re-enable raw sockets.
2972 Created by gras_socket_{client,server}_ext;
2973 Used with gras_raw_{send,recv}
2976 It should allow to kill the last bits of gras first version soon.
2978 This is not completely satisfactory yet (duplicate code with
2979 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2982 - Simplify transport plugin (internal) interface by not passing any
2983 argument to _server and _client, but embedding them in the socket
2986 2004-09-28 Martin Quinson
2987 - Finish the port to AIX.
2988 autoconf was my problem (segfault within the malloc replacement
2989 function. No idea why)
2991 2004-09-16 Martin Quinson
2992 - Fix some size_t madness on 64bit architectures.
2994 2004-09-08 Martin Quinson
2995 - Reduce the number of system headers loaded, overload some more system
2996 calls (such as malloc to cast the result of the system one, and work
2998 - Fix and reintroduce the config support
3000 2004-09-07 Martin Quinson
3001 - Source code reorganization to allow Arnaud to surf all over there.
3002 - Allow to document the logging categories.
3003 - Remove all uppercase from logging categories and useless cleanup in names.
3005 2004-08-18 Martin Quinson
3006 Version 0.6.2 (protocol not changed; API changed)
3007 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3008 gras_error_t), and NULL when not found. Functions expecting a msgtype
3009 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3010 hopefully usefull message.
3011 - Portability to prehistoric sparcs again
3013 2004-08-17 Martin Quinson
3014 Version 0.6.1 (protocol not changed; ABI not changed)
3015 - prealloc some buffers to speed things up
3017 2004-08-11 Martin Quinson
3018 Version 0.6 (protocol not changed; ABI expended)
3019 - The parsing macro can deal with the references, provided that you add
3020 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3022 2004-08-09 Martin Quinson
3023 Version 0.5 (protocol not changed; ABI changed)
3024 - Allow to off turn the cycle detection code in data exchange at
3025 compilation time. It should be at run time, but I'm short of time (and
3026 the config stuff is still broken). That way, we keep dict out of the
3027 critical path, which is good because the performance is poor:
3028 - search not dichotomial yet
3029 - dynar give no way to access their content and memcpy everytime
3030 - In composed data description (struct, ref and so on), stop foolness of
3031 keeping the subtype's ID, but store the type itself. This keeps sets out
3032 of the critical path, which is good since they rely on dynar and
3033 dictionnaries. The only loose of that is that we cannot detect the
3034 redeclaration of a structure/union with another content (but I'm not sure
3035 the code detected well this error before anyway). We still can detect
3036 the redefinition discrepancy for the other types.
3037 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3038 it breaks the code because of type-punning used all over the place).
3039 This breaks on all non-gcc architectures (for now).
3041 All those changes (plus the buffer of last time) allow me to gain 2 order
3042 of magnitude on cruel tests consisting of 800000 array of integers on two
3043 level of a hierarchical structure (200 secondes -> 4 secondes)
3046 - the selector of reference must now return the type it points to, not
3047 the ID of this type.
3049 2004-08-06 Martin Quinson
3050 Version 0.4 (protocol changed; ABI not changed)
3051 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3052 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3053 - Implement a buffer transport, and use it by default (it relies on tcp in
3054 real life and on sg in simulation).
3055 That's a bit hackish since I had a new field to the structure to store
3056 its data without interfering with the subtype ones. Inheritance
3057 is tricky in C. And that's a kind of reverse inheritance with one class
3058 derivating two classes. Or maybe a game with java interfaces. Anyway,
3059 that's damn hard in C (at least).
3060 Moreover, I got tired while trying to ensure plugin separation and
3061 genericity in SG mode. MSG wants me to do weird things, so let's go for
3062 cruel hacks (temporarily of course ;).
3063 See comment in transport_private.h:71
3064 - do not include all the _interface headers in private but in the files
3065 which really need them (to cut the compilation time when they are
3068 2004-07-26 Martin Quinson
3069 Version 0.3 (protocol not changed; ABI changed)
3070 - Major overhault of the datadesc interface to simplify it:
3071 - shorted the function names:
3072 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3073 - add a trivial way to push/pop integers into the cbps without malloc.
3074 This allows to make really generic sub_type description, which simply
3075 pop their size of the stack.
3076 - add a function gras_datadesc_ref_pop_arr() which does what users want
3077 most of the time: Declare a dynamic array (which pops its size of the
3078 stack) and declare a reference to it. Poor name, but anyway.
3079 - kill the post-send callback, add a post-receive one
3081 2004-07-23 Martin Quinson
3082 Version 0.2 (protocol changed; ABI changed)
3083 - add some testing for cpbs in the test cases, and fix some more bugs.
3084 This invalidate again the little64 data file, since I cannot regenerate
3086 - remove an awfull optimization in the logging stuff, allowing me to:
3087 - understand it again
3088 - learn gcc how to check that the argument match the provided format
3089 - fix all errors revealed by gcc after that
3090 - internal keys of dict are not \0 terminated. Deal with it properly in
3091 loggings instead of segfaulting when the user want to see the logs :-/
3093 2004-07-22 Martin Quinson
3094 - Fix some stupid bug preventing cbps (callback postit) from working
3096 2004-07-21 Martin Quinson
3097 - Some documentation cleanups
3098 - remove the useless last argument of msgtype_declare
3099 - rename the Virtu functions to fit into the 'os' namespace
3100 - move headers src/include -> src/include/gras/ and stop fooling with
3101 gras -> . symbolic link
3102 - make distcheck is now successful
3104 2004-07-19 Martin Quinson
3106 - Build shared library also
3107 - Install html doc to the right location
3108 - stop removing maintainer files in make clean
3109 - build tests only on make check
3111 2004-07-13 Martin Quinson
3113 - No major issue in previous version => change versioning schema
3114 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3115 corresponding dataset.
3117 2004-07-11 Martin Quinson
3119 - Get it working with any kind of structure (we can compute the padding
3120 bytes remotely for all the architectures I have access to)
3121 - Implement the structure parsing macro (still not quite robust/complete)
3122 - Improvement to the remote testing toysuite
3124 2004-07-10 Martin Quinson
3125 [autoconf mechanism]
3126 - get ride of a bunch of deprecated macros
3127 - actually run the test for two-compliment, not only compile it :-/
3128 - test whether the structures get packed (and bail out if yes. Damn.
3129 Alignment is a serious matter)
3130 - test whether the structures get compacted (but respecting the alignment
3131 constraints of each types)
3132 - test whether the array fields of structures can straddle alignment boundaries
3134 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3135 (took me 2 hours to find that bug, looking at the wrong place)
3136 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3137 before sending/receiving and are used to compute the offsets of fields
3138 - Given that padding size depend even on compiler options, keep track of
3139 alignment and aligned_size only for the current architecture. Not a big
3140 deal since we send structure fields one after the other (seems
3142 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3143 program, let it work on linux/gcc/little32. portability todo.
3145 2004-07-08 Martin Quinson
3146 - import and improve remote compilation support from FAST
3147 - make sure make check works on half a dozen of machines out there
3149 2004-07-07 Martin Quinson
3150 Let's say it's version 0.0.3 ;)
3151 - Implement conversions (yuhu!)
3152 - Let it work on solaris (beside conversion, of course)
3153 - Stupid me, using rand() to generate the conversion datatests in not wise.
3155 2004-07-06 Martin Quinson
3156 - Let make dist work, since I'm gonna need it to compile on remote hosts
3157 - Let Tests/datadesc_usage write the architecture on which the file was
3158 generated as first byte.
3159 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3160 architecture descriptions.
3161 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3162 on those architectures.
3163 - Optimization: send/recv array of scalar in one shoot
3165 2004-07-05 Martin Quinson
3166 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3168 - Plug a whole bunch of memleaks
3169 - each process now have to call gras_{init,exit}. One day, their log
3170 settings will be separated
3171 - Continue the code factorisation between SG, RL and common in Transport.
3173 2004-07-04 Martin Quinson
3175 - Redistribution between SG and RL.
3176 We wanna have to accept in SG, so move accepted related parts of RL in
3177 the common part. (more precisely, the dynar of all known sockets is no
3178 more a static in transport.c, but part of the process_data)
3180 [gras_stub_generator]
3181 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3183 2004-07-03 Martin Quinson
3184 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3185 them all up in one shot)
3187 - Fix the ugly case of reference to dynamic array.
3188 - New (semi-public) function gras_datadesc_size to allow the messaging
3189 layer to malloc the needed space for the buffer.
3191 - gras_socket_close now expect the socket to close (and not its address to
3192 put NULL in it after it). This is because the socket passed to handlers
3193 is one of their argument (=> not writable).
3195 - propagate the interface cleanup from last week in datadesc, ie remove a
3196 superfluous level of indirection. User pass adress of variable
3197 containing data (both when sending and receiving), and not of a variable
3198 being a pointer to the data. Let's say that I like it better ;)
3199 The price for that is constructs like "int msg=*(int*)payload" in
3200 handlers, but it's a fine price, IMHO.
3202 - Let it work in RL (yuhu)
3204 2004-06-21 Martin Quinson
3206 - porting SG plugin and SG select to new standards (works almost).
3207 - plug memleaks and fix bugs around.
3210 - cleanup the prototype of data recv and force users to specify when they
3211 want to handle references to objects. Test case working even for cycles.
3212 - plug memleaks. Valgrind is perfectly ok with this.
3214 2004-06-12 Martin Quinson
3216 - cleanup the separation between plugin and main code in plugin creation
3218 2004-06-11 Martin Quinson
3220 - Reput hook for raw sockets, needed for BW experiments
3221 - kill a few lines of dead code
3222 [Data description] Interface cleanup
3223 - gras_datadesc_by_name returns the searched type or NULL.
3224 That way, no variable is needed to use a type desc once, which makes
3226 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3227 two parameters were strings (field name, type name), leading to
3229 [Dicos] Interface cleanup
3230 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3231 This is consistant with the dynar API.
3233 2004-04-21 Martin Quinson
3235 - Porting to new standards.
3237 - interface cleanup.
3238 There is no bag anymore, no need to take extra provision to mask the
3239 pointers behind "ID".
3240 Better splitup of functions between files create/exchange/convert.
3241 This is still a bit artificial since convert and receive are so
3242 interleaved, but anyway.
3244 - add a queued message list to procdata (the ones not matching criteria
3246 - factorize some more code between SG and RL wrt procdata
3248 - use gras_exit in example to track memleaks
3249 - get rid of gs_example now that GS is properly integrated into gras
3250 - update run_test to integrate the latest tests (datadesc)
3252 - rename WARNINGn macros to WARNn since it prooved error-prone
3254 2004-04-19 Martin Quinson
3256 - register init/exit functions within gras module mechanism
3257 - send/receive function.
3258 Convertion is not implemented, but short-cutted if not needed.
3259 struct/array elements are sent one by one (instead of block-wise), but
3260 nobody really cares (yet). Get a prototype before optimizing.
3261 - tests (using a file socket) for DD send/receive on:
3262 - base types: int, float
3263 - array: fixed size, string (ie ref to dynamic string)
3264 - structure: homogeneous, heterogeneous
3265 - chained list, graph with cycle
3266 Believe it or not, valgrind is not too unhappy with the results. The
3267 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3271 - Bugfix when using a filename explicitely (instead of '-')
3273 2004-04-09 Martin Quinson
3275 - factorize more code between RL and SG in socket creation
3276 - Complete the implementation and tests of:
3278 o file (only in RL, and mainly for debugging)
3280 I lost 3 days to design a portable address resolver, and then decided
3281 that the prototype mainly have to run on my box.
3282 Addressing portability too early may be like optimizing too early :-/
3284 - use gras_init in the Tests instead of the crappy parse_log_opt
3285 (the latter function is removed)
3286 [Conditional execution]
3287 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3288 [Code reorganisation]
3289 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3290 Build examples against the RL library, since there is no way to disable
3291 its creation for now.
3293 For information, the beginning of coding on GRAS was back in june
3294 2003. I guess that every line has been rewritten at least twice since