1 SimGrid (3.15) UNRELEASED; urgency=low
3 The Ever Green Release.
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: dupplicates 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 catched explicitely).
29 - s4u::onPlatformCreated: right before the simulation starts
30 - s4u::onSimulationEnd: right after the main simulation loop
31 - s4u::onTimeAdvance: right after a clock change (time is discrete)
32 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
33 event from the availability_file changes the avail speed.
34 - Links are now usable from s4u
35 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
36 - Drop Host::getPstateSpeedCurrent() which dupplicates Host::speed()
39 - Backwards Compatibility breaks
40 - The SD_route_get_list and SD_route_get_size functions have been removed.
41 They are replaced by the unique
42 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
43 The route from src to dst is built in the links dynar whose size can
44 be obtained with xbt_dynar_length.
45 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
46 sg_host_route_bandwidth and sg_host_route_latency.
47 Macros ensure the backwards compatibility, but you should fix your code
50 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++). This can break codes that were using internals of SMPI (from private.h instead of the public smpi.h).
51 - Bump our claim of support from MPI 1.1 to MPI 2.2. We don't support 100% of
52 it, but it should be enough. Contact us if not.
53 - MPI_Win_lock/unlock support added (as for all RMA, implementation is naive and probably inaccurate)
54 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
55 mapped onto a unique file.
56 - Bugfix : smpirun was sometimes misusing hostfile when wrongly terminated
57 - Fortran : cleanups, fixes, support of user-added operations
58 - RMA : MPI_Accumulate are applied in correct order, and
59 MPI_Win_complete or MPI_Win_post shouldn't miss messages anymore.
60 - Better support for MPI_IN_PLACE
61 - Support for MPI_Win attrs and keyvals.
62 - Support MPI_Comm_group_create, MPI_Type_size_x
63 - MPI ops should be properly applied to all allowed types
64 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
65 algorithm by a non failing one from mpich.
68 - Kill the fifo data container: we don't use it anymore.
71 - A <cluster> can now be created with different speed values to represent pstates.
72 see examples/platforms/energy_cluster.xml for an example.
74 -- target_date=March 20 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
76 SimGrid (3.14.159) stable; urgency=low
78 The previous release was only a crude approximation of a Pi release;
79 this one should be a bit better.
81 - Revert a last minute change that broke on Mac OSX.
82 - Fix the changelog and NEWS files.
83 - Improve the documentation, in particular of the routing module.
84 - Integrate some patches flying around in Debian.
86 -- Wed Dec 28 17:14:49 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
88 SimGrid (3.14) stable; urgency=low
91 * The whole documentation was reworked and reorganized. There is
92 still much room for improvement here, but we're on it.
95 * We now need python3 for our testing infrastructure.
96 * Model checking needs libevent
99 * Now works on FreeBSD
102 * Privatization now works on FreeBSD
104 * Privatization is now activated by default in smpirun
105 You can turn it off with -no-privatize if something goes wrong.
107 * Call-location tracing for SMPI.
108 You can add the exact location (filename / linenumber) of an MPI call to
109 your trace files and slow down or speed up the simulation between two
110 consecutive calls by using an adjustment file (see the documentation).
112 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
113 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
114 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
115 values were ignored in some cases. The timings of these functions can now
116 be significantly different.
118 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
119 smpi/running-power is renamed to smpi/host-speed
121 * smpi/grow-injected-times option to enable or disable multiplication of the
122 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
123 make simulation less precise (but also much faster).
125 * smpirun script should be (much) faster for large deployments.
127 * SMPI tracing : fixed issue with poor matching of send/receives.
129 * Replay : Fix broken waitall
131 New functions and features
132 * MSG_parallel_task_execute_with_timeout, to timeout computations.
134 Dropped / renamed functions and features
135 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
136 - MSG_mailbox_is_empty() -> Mailbox::empty()
137 - MSG_mailbox_front() -> Mailbox::front()
138 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
139 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
140 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
141 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
142 * Don't pass the free_f parameter to property related functions:
143 - MSG_host_set_property_value()
144 - MSG_as_router_set_property_value()
145 - MSG_storage_set_property_value()
146 * VM properties. Since msg_vm_t are msg_host_t, just use
147 MSG_host_get_property_value() and friends
148 * VM I/O related things:
149 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
150 - Unimplemented save/restore methods
151 * MSG_as_router_get_property_value() was redundent with
152 MSG_environment_as_get_property_value().
153 - Removed MSG_as_router_*propert*() functions
154 - Added MSG_environment_as_set_property_value() for consistency
155 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
156 Well, they are still there for now, but deprecated with a warning.
157 Please switch to xbt_dict_new_homogeneous() before this is removed
159 * Task affinity. Its intended behavior (that was very badly tested
160 and probably not really working) was deceiving what most users
161 would have hoped here.
162 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
163 * Option network/coordinates is now useless and should be dropped.
166 * Added option storage/max_file_descriptors to allow more than 1024 files opened
169 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
170 * Added Dragonfly topology support
173 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
174 EiBps, EBps, Eibps, Ebps for bandwidth.
175 They may become useful to some lucky ones.
178 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
179 * ex/app_tokenring: new example, very similar to the MSG Token Ring
180 * ex/async_waitAll: new example, on asynchronous communications
183 * Memory usage should be decreased for simulations with a large number
184 of processes. This also helps for SMPI.
186 -- Sat Dec 24 22:34:26 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
188 SimGrid (3.13) stable; urgency=low
190 The Half Release, a.k.a. the Zealous Easter Trim.
192 - We removed half of the lines, that were mostly experimental cruft.
193 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
194 (not counting blanks and comments -- according to openhub.net)
195 - The internals are now compiled in C++ (and will soon be clean C++)
196 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
197 59 klines of C, etc. We added only 29 klines of C++ in replacement.
199 * Backwards Compatibility breaks
200 - Removed Lua simulation bindings (switch to C or Java for that).
201 Lua can still be used to describe platforms
202 - Removed Java kernel plug-ins.
203 Will be reintroduced after the ongoing major internals reorg.
205 - the following functions were removed.
206 They were too specific and should be reimplemented in a generic
207 way, with filter function.
208 - MSG_task_listen_from_host
209 - MSG_mailbox_get_count_host_waiting_tasks
210 - MSG_mailbox_put_with_timeout was removed.
211 Please use MSG_task_send_with_timeout instead.
213 - the SD_application_reinit function was removed. It has been a noop for a while.
214 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
215 replaced by a more flexible mechanism.
216 - The following functions thus do not exist anymore
217 - SD_workstation_get_access_mode
218 - SD_workstation_set_access_mode
219 - SD_workstation_get_current_task
220 - Basic estimation functions have been removed but can easily be replaced
221 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
222 - SD_workstation_get_computation_time => amount / sg_host_speed()
224 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
225 Use VM.getSpeed()*load/100 for the legacy behavior.
227 - option enable_tracing was removed. It was not doing anything for a while.
228 - In the ModelChecker:
229 - the model-checker now ptraces the model-checked process which means
230 you cannot use a debugger on the latter anymore (we might make this
231 optional in the feature);
232 - removed soft-dirty page tracking;
233 - remove model-checked side snapshot management,
234 MC_snapshot() and MC_compare_snapshot();
235 - keep the MC_cut() function as a stub (it was not really working
236 in the previous release).
239 * All options are consistently in kebab-case. Old names are kept as alias.
242 * Switch to platform v4 format.
243 - Rename from 'power' to 'speed' the attributes describing the amount of
244 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
245 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
246 - In <host> and <link>, attributes availability and state are gone.
247 It was redundent with state and availability traces, and with peak values.
248 - In <cluster>, attributes availability_file and state_file are gone.
249 It was too complex and unused.
250 - Kill <gpu>. Was not doing anything.
251 - The DOCTYPE points to the right URL:
252 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
253 (the file at this address now documents the changelog since its v1)
254 - A warning is emitted for unit-less values (they are still accepted).
255 - speed. Default: 'f' or 'flops'. Also defined:
256 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
257 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
258 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
259 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
260 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
261 - latency. Default: 's' second. Also defined:
262 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
264 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
265 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
268 * s4u::Host is now the preferred public interface to the Host features.
269 sg_host_* functions are C bindings to the exact same behavior
270 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
273 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
274 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
275 They were deprecated since a few years, and probably did not even compile anymore.
278 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
279 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
280 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
281 encouraged to update their codes. A list of the modified functions can be found at the end of
282 include/simgrid/simdag.h
285 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
286 the lambda or closure passed as a parameter will run in kernel mode.
287 All the callback functions should be rewritten to that interface at some point.
290 * Reorganizing and cleaning the internals all around the place.
293 * Remove old default barrier/bcast buggy algorithms (see #18407)
294 * Various bug fixes to handle more codes
295 * Remove the need for the --foreground option of smpirun (it is still
296 accepted for backward compatibility).
299 * Kill the setset data container: MC don't use it anymore.
300 * Kill the queue data container: it made more sense with GRAS.
301 * Kill the xbt_peer_t data type: it's useless without GRAS.
302 * Kill rm_cb feature of config sets: it was never useful.
303 * Kill graphxml parsing feature. It was not used.
304 * Kill the deprecated code protected by XBT_USE_DEPRECATED
306 - xbt_dynar_sort_strings(), when the content is char*
307 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
308 They throw exceptions on invalid input;
309 * C++ support for declaring CLI flags (simgrid::config::Flag);
310 * class for abstracting different signal backends (simgrid::xbt::signal).
311 with no external dependencies (we need very simple signals).
314 * refactoring and cleanup of the code;
315 * ongoing process to cleanly separate the model-checking algorithms
316 from the code model-checking support.
318 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
320 SimGrid (3.12) stable; urgency=low
322 The Facelift Release.
325 * Require g++ v4.7 at least to not speak prehistorical C++.
326 * Require Boost 1.48 (for signal2 component).
327 * Java must be version 7 at least when activated.
328 * Builds on Windows again (including Java bindings).
329 * Tracing is now always enabled (no way to turn it out)
330 * Remove GTNetS. It was not working anyway.
331 * Various cleanups in the cmake scripts.
332 * Move headers around to sort them out on installed systems:
333 - instr/instr.h -> simgrid/instr.h
334 - instr/jedule/* -> simgrid/jedule
335 - simdag/datatypes.h was removed
336 - simdag/simdag.h -> simgrid/simdag.h
337 - msg/datatypes.h was removed
338 - msg/msg.h -> simgrid/msg.h
341 * Interface improvement:
342 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
343 - Sanitize the interface in MSG_task_ module:
344 - Merge two functions that were close enough but misleading:
345 set_compute_duration(t) -> set_flops_amount(t)
346 get_remaining_computation(t) -> get_flops_amount(t)
347 - set_data_size(t) -> set_bytes_amount(t)
348 get_data_size(t) -> get_bytes_amount(t)
349 - Massive cleanups in the functions related to the energy
350 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
351 - New: MSG_host_get_pstate()
353 - msg/energy/onoff: switching hosts on and off
355 * Interface improvement:
356 - Rename Host.isAvail() to Host.isOn()
357 - Rename Process.currentProcess() to Process.getCurrentProcess()
358 - Rename Task.setDataSize() to Task.setBytesAmount()
359 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
361 - #18874: Actually allows the GC to reclaim tasks
365 - SIMIX_process_throw: raises an exception in a remote process
366 * Refactoring: Separate sub-modules
367 - libsmx: the public interface, as libc in a real system
368 - popping: the strange dance that converts a user request into a kernel handling
369 - smx_context_*: the virtualization mechanisms that embed the user code
370 - smx_*: the handling of each simcalls
372 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
373 * Rename smx_action_t into smx_synchro_t, making explicit that these
374 things are used to synchronize processes with their environment.
375 For example, a communication is a sort of synchronization involving
376 the communicating processes (that may block until the exchange) and
377 the platform. The same can be said from computations, etc.
379 - #18888: segfault when a process ends before its kill_time
383 - 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.
384 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
385 - MPI_Type_set_name, MPI_Type_get_name
386 - MPI_*_c2f and MPI_*_f2c functions
387 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
388 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
389 - Activate a lot of new tests from the mpich 3 testsuite
391 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
392 - InfiniBand network model added : Based on the works of Jerome Vienne
393 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
394 - When smpi/display_timing is set, also display global simulation time and application times
395 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
396 * Collective communications
397 - 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
398 - 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/).
399 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
401 - Replay now uses algorithms from wanted collective selector
402 - Replay can be used with SMP-aware algorithms
403 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
404 - 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.
406 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
407 - Correctly use loopback on fat-tree clusters
408 - Asynchronous small messages shouldn't trigger deadlocks anymore
409 * Energy/DVFS cleanup and improvement
410 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
411 - new: smpi_get_host_pstate()
414 - "Full" network optimization flag was broken since Surf++
415 - Better handling of precision flags in maxmin
416 - Fix bug causing sometimes "Impossible" errors
417 - Properly pass cluster properties to included hosts
418 * Improvement of the Energy plugin.
419 - Always update the consumption before returning that value
420 - New property: watt_off to denote the dissipation when the host is off
421 - New functions getWattMinAt and getWattMaxAt to retrieve the
422 dissipation of pstates that we are not currently at.
423 * Java: class NetworkLink renamed to Link
424 * New function: simcall_process_get_kill_time()
425 * Massive rename s/workstation/host/
426 - That's intrusive, but that's good for the project consistency. Sorry.
427 - Change config option "workstation/model" into "host/model"
430 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
431 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
432 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
433 - xbt_dirname and xbt_basename for non-POSIX systems
435 * The model checker now runs as a separate process.
436 * The model checker runs must now be launched with the new simgrid-mc program.
437 * Record/Replay: the MC can display a textual representation of a path in the
438 execution graph. It can then be replayed outside of the model checker.
440 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
442 SimGrid (3.11) stable; urgency=low
447 * Normalizing pointers addresses tool for better diff between logs
450 * Add cloud examples using new VMs
451 - examples/msg/cloud/two_tasks_vm.tesh
452 - examples/msg/cloud/simple_vm.tesh
453 - examples/java/cloud/cloud.tesh
454 - examples/java/cloud/migration/migration.tesh
455 * Add java surf examples:
456 - examples/java/surfPlugin/surf_plugin.tesh
457 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
458 - examples/java/surfCpuModel/surf_cpu_model.tesh
459 * Add SMPI+MSG example:
460 - examples/smpi/smpi_msg_masterslave/
465 - msg task destroy cancel test
466 - msg_host on/off test
467 * Move all tests in testsuite to teshsuite (adding tesh files)
468 * Restructure teshsuites
469 - one folder for each kind of test
470 * Restructure AddTests.cmake
472 - structure the order of tests (with sections)
475 * Add virtual machine
476 - creation of a VM on a PM
477 - migration of a VM from a PM to another PM
479 - MSG_process_join(msg_process_t process, double timeout)
480 - msg_bar_t MSG_barrier_init(unsigned int count)
481 - int MSG_barrier_wait(msg_bar_t barrier)
482 - void MSG_barrier_destroy(msg_bar_t barrier)
483 - msg_as_t MSG_environment_as_get_by_name(const char * name)
484 * New option "msg/debug_multiple_use" to help debugging when a task is used
487 - Improvements and finalization of MSG_storage, MSG_file APIs and their
489 - Increase code coverage in test suites
493 * Protect context stacks against stack overflow. The number of protected memory
494 pages allocated on the top of each stack (1 by default) can be configured
495 with the new command line option --cfg=contexts/guard_size:<value>.
496 * Simcalls are now generated by a python script that
497 - generates files included by SimGrid
498 - checks that all the functions exist, and proposes prototypes
500 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
501 host_set_data, host_get_data
503 - simcall_process_join(smx_process_t process, double timeout)
504 * Fix bug where sleeping processing could not be suspended.
507 * Translate surf models from C to C++
508 - Generic classes for all models: Model, Resource, Action
509 - A generic interface for each kind of model (CPU, Network, Storage
510 Workstation, WorkstationVM)
512 * Translate surf routings from C to C++
513 * Add callbacks using sigc++ or boost::signals2
514 - Add callback functions for resource creation/destruction
515 - Add callback functions for action state change
516 - Handle Energy as a plugin
517 * Replace swag by boost::intrusive
518 * Add new routing models for clusters. For documentation, see
519 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
520 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
521 parameters for cluster tag
522 - Fat trees, with topology="FAT_TREE" and
523 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
524 - see examples/platforms/torus_cluster.xml and
525 examples/platforms/fat_tree_cluster.xml
529 * Hostfiles support host:nb_processes construct to deploy several processes on
531 * Collective communication algorithms should not crash if used with
532 improper number of nodes and report the error.
533 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
534 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
535 MPI_Dims_create, MPI_Cart_sub are supported.
536 * New interface to use SMPI programmatically (still depends on MSG for
537 some parts, see examples/smpi/smpi_msg_masterslave) :
538 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
542 * Global variables privatization in MPI executables is now performed at runtime
543 with the option smpi/privatize_global_variables (default:no).
544 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
545 dynamic libraries loaded by the application are not privatized (static
546 linking with these libraries is advised in this case)
549 * Options defined in XML work correctly now.
552 * New cmake option, enable_lib_in_jar, to control whether native libraries are
553 copied into simgrid.jar or not (ON by default). Use this option if you want
554 to reduce the size of the installed simgrid.jar, *and* the native libraries
555 are kept installed elsewhere.
556 * Surf binding with SWIG (code generated in maintainer mode only):
557 - plugin to handle callbacks
558 - CPU model only for the moment
561 * Supernovae build mode is definitively removed. It was used to improve
562 inlining and inter-module optimizations. It is nowadays superseded by
563 link-time optimizations commonly available in compilers.
564 * Update ns-3 find lib. Bindings for ns-3 should work again now.
565 * Add boost dependency for surf++
566 * Add new macro for tests
567 - ADD_TESH(name <tesh_args>)
568 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
572 - xbt_bar_t XBT_barrier_init(unsigned int count)
573 - int XBT_barrier_wait(xbt_bar_t barrier)
574 - void XBT_barrier_destroy(xbt_bar_t barrier)
575 * Make the xbt_os_time module public
577 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
579 SimGrid (3.10) stable; urgency=low
581 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
584 * Reintegrate Java to the main archive as desynchronizing these
585 package is not acceptable anymore (Java is now considered stable)
586 * Add explicit synchronization facilities through semaphores
587 * Bug fix: Task.setDataSize() only changed the C world, not the value
588 cached in the Java world
591 * Dramatically change the way files are handled. API and internals changed, but
592 this part of MSG was not considered as production grade either.
593 * Add explicit synchronization facilities through semaphores
594 * Add a new function MSG_host_get_process_list()
595 * Preliminary DVFS support (see examples/msg/energy/ for details)
598 * SMPI is now included directly in the libsimgrid as the windows
599 linker doesn't force us on splitting it anymore.
600 * Improvements of the SMPI replay tool:
601 - Most of the collective communications are now rooted in the same process as
602 in the original application.
603 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
604 used until now). Multiple data types can now be used in a trace.
605 - The replay tool now supports traces produce either by TAU or a modified
607 - Bug Fix: the compute part of the reduce action is now taken into account.
608 - Gatherv collective is now supported
609 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
610 of smpirun outputs time independent traces for the current run. One file
611 is created per process. If too many processes are simulated, this behavior
612 can be changed to one file for all processes by using the
613 tracing/smpi/format/ti_one_file flag
614 * smpirun generates the host file if needed (with given host count and platform)
615 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
616 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
617 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
618 --cfg=smpi/coll_selector:(mpich/ompi)
619 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
621 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
622 happening inside a collective SMPI call.
623 * Fix the behavior of complex data types handling.
624 * Make MPI_Wtime another synchronization point to take computations into
626 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
627 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
628 * Add all missing Fortran bindings, SMPI should work with Fortran 90
629 (no privatization of global variables yet)
630 * Preliminary DVFS support (see examples/smpi/energy/ for details)
633 * Verification of liveness properties is now available for SMPI applications
634 (in addition to MSG applications)
635 * Bugged examples using SMPI in examples/smpi/mc/
636 * Add --cfg=model-check/visited option. Allows the verification of infinite
637 programs. Detection of loops in the execution thanks to the system state
638 comparison and reduction of the state space to explore. Can be combined with
639 DPOR for safety properties.
642 * Allow to change SimGrid configuration (see --help) within the code
643 thanks to SD_config() as it can be done in MSG.
644 * Add a new function SD_task_set_amount() upon user request.
647 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
648 * Remove rule based routing (no more PCRE dependency)
649 * Add a limiter_link option to cluster tag, to specify a maximum reachable
650 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
651 * Add a loopback_bw and loopback_lat options to cluster tag.
652 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
653 routing. Now peers are encapsulated in an AS and have their own private
654 router but this is transparent.
657 * Our own implementation of getline is renamed xbt_getline, and gets
658 used even if the OS provide a getline(). This should reduce the
659 configuration complexity by using the same code on all platforms.
660 * New type: xbt_cfg_elm_boolean.
661 * Allow to use yes/no for boolean configuration options in the command line.
662 * Allow to disable SimGrid cleanups at exit from command line option.
663 There are situations where one may want a simulation to end with an exit.
664 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
665 annoying when scripting around the simulator. Adding a
666 --cfg=clean_atexit:no allows to circumvent this issue.
669 * Lots of memory leaks were corrected in this release.
670 There is no known memory leaks anymore, in all of our 600+ tests.
671 * New command line option --version, to get SimGrid version information.
672 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
674 * Supernovae builds are deprecated, and expected to be removed in the next
677 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
679 SimGrid (3.9) stable; urgency=low
681 The Grasgory release: GRAS is really dead now.
683 * Complete overhaul of the internal host structures scheme.
686 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
687 was considered as experimental and badly maintained since then.
688 * Keeping it was thus a trap to our potential users, that could take
689 it instead of MSG or SMPI by mistake despite is pity state.
690 * GRAS seems to have very few users (if any), and no one volunteered
691 to maintain it further. It also induces a lot of XBT code (for
692 portability sake), that must be maintained too.
693 * For all these reasons, we killed GRAS. If someone wants to revive it
694 in the future, don't cry, our git history still remembers of GRAS.
697 * Major overhaul. Merge our documentation again as time proved that
698 splitting it was really not helping our users.
699 * Further improve the developer documentation to help newcomers
700 hacking on SimGrid itself. The user documentation (and in
701 particular, the beginner documentation) is still in a sorry state.
704 * Now works on Windows too!
705 * Much more extensive test suite, from MPICH
708 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
709 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
710 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
711 * Bug fix: let task be scheduled when the last dependency to be solved is
712 a control dependency.
713 * Remove SD_load_environment_script function.
714 Use the C sg_platf function if you want to declare a platform
718 * New function: MSG_process_get_number()
719 * Old function documented: MSG_config()
720 * Remove MSG_load_platform_script function
721 Use the C sg_platf function if you want to declare a platform
725 * Change the default value of the TCP_gamma constant (maximal size of TCP
726 congestion window) to a more realistic 4MiB value. If you notice changes in
727 your simulation results, you can fall back to the previous 20k tiny window
728 by adding --cfg=network/TCP_gamma:20000 on command line.
729 * (Hopefully) fix a bug wrt periodic availability and state traces
730 * Bug fix: use default values at start when first event in availability/state
731 trace is not at time 0.
734 * remove the "new_" part of function name sg_platf_new_trace_connect
735 (resulting in sg_platf_trace_connect), since it does not create
739 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
740 use case where it's really mandatory, and maintaining it was a pain in
742 * New: xbt_fifo_search(), search an item with a user-provided
743 comparison function instead of dumb pointer comparison.
746 * Fix the lua deployment:
747 Use `simgrid.init_application()` before deployment instead of
748 `simgrid.msg_register_application()` after.
751 * Transfer the tracing files into the corresponding modules.
753 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
755 SimGrid (3.8.1) stable; urgency=low
757 The "we are told that some people want to also *install* the simgrid
760 * Add missing file "tesh.1" to the archive.
762 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
764 SimGrid (3.8) stable; urgency=low
766 The Psssshiiiit release: SimGrid jumps into the Cloud.
769 * Add an experimental interface to manipulate VMs. They are mainly
770 process groups with very few intrinsic semantic, but they should
771 allow you to build the semantic you want easily.
772 * New function: MSG_host_set_property_value()
773 * New function: MSG_process_on_exit(). To clean memory in all cases.
774 * Bug fixes that made the host (and link) failures unusable.
775 * Add a way to auto-restart process when the host in which they are
776 executing comes back (ON_FAILURE="RESTART" on deployment file,
777 MSG_process_auto_restart_set).
778 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
779 please stop using the old ones, they are DEPRECATED.
781 * Deprecate functions MSG_global_init() / MSG_global_init_args()
782 Please use MSG_init() instead. (reducing the amount of entry
783 points in the library helps us).
784 * Make it impossible to link against the wrong version of the lib
785 * Deprecate MSG_clean(). No need to call it anymore.
786 * Function MSG_get_host_number() is not deprecated anymore.
789 * Split the doc into a user guide and a reference guide.
790 * Start a developper guide to help people hacking on SimGrid.
793 * Enable tracing by default. This modules rocks you should use it.
794 * Remove option custom_flags. Now use environment variables CFLAGS
796 * Use default cmake things to detect lua instead of home grown ones.
797 * New option "enable_mallocators" to disable mallocators, for debugging
798 purpose ("on" by default).
801 * Bug fixes around the resource failures: don't let the processes
802 survive the host they are running onto.
803 * Add an interface to auto-restart processes when the host in which they are
804 executing comes back.
805 * Ensures that SIMIX_clean is called automatically. It's not part of
806 the public interface anymore (bindings should be updated).
809 * Bug fix for when SD_Simulate is called with a positive value: be careful
810 when comparing doubles. Sometimes they are different for non significant
812 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
813 parallel task whose initial work is distributed among host according
814 to the Amdahl's law. Such tasks are created with a parameter alpha
815 that corresponds to the non-parallelizable part of the computation.
816 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
817 between two sets of workstations assuming a 1D block distribution (each
818 workstation owns a similar share of data) on both sides.
820 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
821 Data redistribution will be automatically scheduled once parent and child
822 are both scheduled. The filling of computation_amount and
823 communication_amount structures is now done seamlessly thanks to the chosen
825 * New function SD_workstation_dump to display various information
826 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
827 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
828 bandwidth on the route onto which the task is scheduled and the amount of
830 To divide the nominal bandwidth by 2, the rate then has to be :
831 rate = bandwidth/(2*amount)
832 * Compute tasks that have failed can now be rescheduled and executed again
833 (from their beginning)
834 * Increasing source code coverage (src/simdag is now covered at 95.8%
838 * Re-implement time-independent trace replay using SMPI (at the
839 smpi_smp_* level) instead of MSG. This should replace
840 examples/msg/actions/actions.c
841 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
842 hindexed and structs)
843 * Implement the exchange of non-contiguous data.
844 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
845 * Correct behavior of smpi/sender_gap and set its default value to 0
846 * Add option to asynchronously send small messages to allow better
847 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
848 specifies the size in bytes under which messages will be asynchronously sent.
849 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
850 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
853 * New C interface to define a platform: XML is now optional.
854 For more info, please check include/simgrid/platf.h.
855 * New interface to define random platforms from the C:
856 For more info, please check include/simgrid/platf_generator.h and
857 examples/msg/masterslave/masterslave_platfgen.c
858 * Export a sg_cmdline dynar containing all the arguments we got from
862 * Two new tracing options for adding comments to trace file so you
863 can track your experiments (see --help-tracing for details).
864 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
865 * Adding the SimGrid version that generated the trace file as a comment.
866 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
867 * Fix to avoid key clashes on Paje links
868 * Other minor fixes related to the Paje specification
871 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
872 and renamed to xbt_str_hash() and xbt_str_hash_ext().
873 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
874 * Greatly improve the robustness of mmalloc to user errors (such as
875 using an area after freeing it, or freeing it twice)
877 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
879 SimGrid-java (3.8.1) stable; urgency=low
881 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
882 Warning: all methods are visible, but only some of them are
883 implemented so far. Check the source (src/jtrace.c)
884 for further information.
885 * New module: org.simgrid.msg.File (SimGrid File management functions)
886 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
887 * Change the meaning of Process.restart: now restart the process from
888 the begining, like MSG_process_restart in C.
889 * Add Process.setAutoRestart: handling of process restart when failed
891 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
892 you to retrieve the properties of the processes/hosts
893 * Deprecate Msg.clean(): you can just forget about it now.
894 * New function Process.getCount(), that only works when compiling
895 with the not yet released version 3.9 of the C library.
897 * New context factory based on Coroutines. It mandates a modified JVM
898 but then, the simulations run about five times faster, and there is
899 no limit to the amount of processes (beside of the available memory).
901 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
903 SimGrid (3.7.1) stable; urgency=low
906 * Restore the prototype of MSG_process_create_with_environment() to
907 the pre-3.7 situation by removing the kill_time argument.
908 * Add a MSG_process_set_kill_time() function instead.
911 * Fix weird behaviors when dealing with parallel tasks.
914 * Simgrid is now built as a dll.
915 * Simgrid-java now works on Windows.
916 * Simgrid-Java is now included into Windows package.
919 * First pre-build package for MacOSX.
922 * Fix compilation when using MSG_USE_DEPRECATED.
923 * Fix some compilation issues on Macs and Windows.
924 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
925 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
927 We discovered that the Lua console is broken, but we are missing the
928 manpower to fix it right now. The problem existed in 3.7 too, so we
929 are not blocking the release for that. Sorry if you depended on this
930 feature, any help would be really welcome.
932 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
935 SimGrid-java (3.7.1) stable; urgency=low
937 The "Java aint got to be bloated and slow" release
940 * Various internal cleanups and performance improvement
941 Simulations are expected to run up to twice faster or so
942 * Make Process.kill(process) an instance method, not a static one
943 * User processes are not java.lang.Thread subclasses.
944 This breaks the compatibility (sorry), but previous API was
945 brain-dead, making it impossible to have non-trivial
946 initializations in the process constructor.
947 * Require a full constructor per Process sub-class.
948 Kinda breaks the compatibility (sorry), but this allows a much more
949 efficient way to launch the processes at simulation startup.
950 * Do not embeed our version of semaphores, java 1.5 can be considered
951 as sufficiently prevalent for us to not dupplicate its features.
955 * Add examples for almost every part of the API
956 We spotted and fixed a lot of bugs in the process
958 * New module: asynchronous communication API
959 * New function: Process.sleep()
960 It takes milliseconds as argument, just as java.lang.Thread.sleep()
961 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
962 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
964 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
966 SimGrid (3.7) stable; urgency=low
968 The "spring cleanups (before the next Big Project kicks in)" release.
971 * We can specify the SMPI latency/bandwidth factor with command line
972 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
973 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
974 You can also use the "config tag" from platform file by adding this line
975 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
976 (see "example/platforms/tag_config.xml" to use "config tag").
977 Note that the command line supersedes the platform file configuration.
978 * Change the correction factors used in LMM model, according to
979 the latest experiments described in INRIA RR-7821.
980 Accuracy should be improved this way.
981 * Use the partial invalidation optimization by default for the
982 network too. Should produce the exact same results, only faster.
983 * Major cleanup in surf to merge models and split some optimization
984 mechanisms from the core of the models. As a result you can now
985 specify which model to use (e.g., --cfg=network/model:LV08
986 --cfg=cpu/model:Cas01) and which optimization mode to use
987 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
988 Incompatible combinations should err at initialization. See
989 --help-models for the list of all models and optimization modes.
990 * The CLM03 workstation model was dropped for simplicity because it
991 used the deprecated CM02 network model. Use default instead.
992 * Rename the TCP_gamma configuration option to network/TCP_gamma
993 * Rename the coordinates configuration option to
994 network/coordinates, and document it
995 * Use now crosstraffic keyword instead of the terribly misleading
996 fullduplex keyword. It is activated by default now in the current
997 default model, use --cfg=network/crosstraffic:0 to turn it off.
998 * Ongoing refactoring the model parsing to make XML files optional
999 See include/simgrid/platf.h for details (still to be completed)
1002 * Major overhaul of the documentation. Almost instructive now :/
1003 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1004 functions and friends. This interface was considered as
1005 deprecated since over 2 years, it's time to inform our users that it is.
1006 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1007 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1008 use these (crappy) functions in your code.
1009 These functions will be removed soon. Stop using them now.
1010 * Deprecate MSG_get_host_{table,number}
1011 Implement MSG_hosts_as_dynar() instead.
1012 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1013 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1016 * Stabilize the parallel execution mode of user contexts
1017 * Introduce configuration variables to control parallel execution:
1018 - contexts/synchro: Synchronization mode to use when running
1019 contexts in parallel (either futex, posix or busy_wait)
1020 - contexts/parallel_threshold: Minimal number of user contexts
1021 that must be part of a scheduling round to switch to parallel
1022 execution mode (raw contexts only)
1023 * Fix bugs that prevented to use suspend/resume along with
1024 synchronization structures.
1025 * Fix bugs in process termination that lead to invalid memory access
1026 in very specific conditions.
1029 * Introduce a parallel mode for the models (controlled by surf/nthreads
1030 configuration item). In our tests, running the models in parallel
1031 never lead to any speedups because they are so fast that the gain
1032 of computing each model in parallel does not amortizes the
1033 synchronization costs, even when ultra fast futexes are used.
1034 This is released anyway because YMMV.
1037 * Performance boost by using a swag internally to compute the set of
1038 tasks that are finished and should constitute the return value of
1042 * Enable it by default now that it is considered rather stable.
1045 * Documentation of the tracing functions.
1046 * Performance gains when tracing categorized/uncategorized resource
1047 utilization by avoiding calls to get route when updating resource
1048 variables. LMM constraints are being used instead.
1049 * API changed to set task categories. Use MSG_task_set_category instead
1050 of TRACE_msg_set_task_category, and SD_task_set_category instead
1051 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1052 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1053 * Examples for link user variables added (see at examples/msg/tracing/)
1054 * Deprecated function TRACE_msg_set_process_category completely removed
1055 * Trace header updated according to the latest Paje file format
1056 * Tracing network lazy updates, no longer obligate users to use full updates
1057 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1058 * Experimental: let user code declare/set/push/pop application states for hosts
1059 * API changed to allow the manual creation of graph configuration files
1060 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1063 * Improve the API of Lua MSG bindings, using the Lua spirit.
1064 * Each simulated process now lives in its own Lua world (globals are
1065 automatically duplicated). It helps writing simulators. Will allow
1066 to run Splay programs within SimGrid in the future.
1067 * Add a Chord example in Lua, equivalent to the MSG one.
1070 * Start the implementation of a solution to express temporal
1071 properties, not only local assertions. This is still an
1072 experimental work in progress, stay clear from it to be safe.
1076 - Add new runtime parameters --help-logs and --help-log-categories
1077 to display information about supported logging parameters and
1079 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1080 don't exists anymore.
1081 * Mallocators: allow value NULL for the reset function.
1083 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1084 create homogeneous dictionaries, where all the elements share the
1085 same free function. Non homogeneous dictionaries will be
1086 deprecated in the next release.
1087 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1088 - Multi-level dictionaries are deprecated.
1090 - new function xbt_dynar_search_or_negative() that is useful when
1091 you have less than 2 million elements in your dynar and don't
1092 want of the extra complexity of catching exceptions when the
1093 element is not found.
1095 - Make xbt_os_thread module (for thread portability) public.
1096 Documentation is still to come, sorry.
1098 - Cleanups and simplifications to make it maintainable again.
1099 - Exotic features (such as memalign and valloc) were removed.
1100 - The metadata were extended and improved so that the
1101 model-checker becomes able to explore and inspect the heaps.
1102 - This may induce a performance drop when enable_model-checking is
1103 ON in cmake (even if it's not used in the simulation), but it is
1104 necessary at this point to get MC working.
1106 Turn model-checking OFF if simulation performance matters to you.
1107 Not enabling it at runtime is not enough, disable it in cmake.
1109 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1112 SimGrid (3.6.2) stable; urgency=low
1114 The "Not coding new stuff allows to polish old things" release.
1117 * New bindings to the NS3 packet level simulator (experimental)
1118 * Use the raw (efficient) execution contexts instead of the sysv
1119 (portable) ones when possible.
1120 * libpcre is now mandatory in any cases since not using it led to
1121 severe performance loss and possibly other issues
1122 * Update the XML platforms:
1123 - G5K: include the latest machine in Nancy
1124 - GridPP and LCG: new platforms
1125 * Documentation was partially updated, at least (more to come)
1127 Bug fixes, cosmetics and small improvements
1128 * Free terminated processes before the end of the simulation to avoid
1129 exhausting the memory of users having very dynamic amount of
1131 * Bug fix and cosmetics about canceling non-running tasks
1132 * Bug fix about the dot loader's issues when using libcgraph
1135 * Create an installer for windows with nsis (amd64 and win32)
1136 - Add an hello world project to illustrate simgrid project creation.
1137 - Embed libpcre into the Simgrid installer to avoid
1138 its compilation burden
1139 * The raw execution contexts should work on Apple now
1140 * Port to Windows 64 bits
1141 - Sysv contexts now have an implementation for this arch
1142 - GRAS communication features now support this arch
1143 * Drop support for borland compiler on windows
1144 - this code was not maintained, and we kinda depend on gcc nowadays
1145 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1146 * Fix portability issue on unstable ubuntu: linker became picky on
1149 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1151 SimGrid (3.6.1) stable; urgency=low
1153 The "Oops, we broke Macs too" release
1156 * Fixed contexts detection so that raw ones are used when possible
1157 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1158 leads to a strange error, with user code segfaulting sometimes when
1159 the generated code is not perfectly aligned (which is not
1160 controllable from the user side, depends on the amount of code)
1163 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1165 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1167 SimGrid-java (3.6) unstable; urgency=low
1170 * Split of every thing from simgrid v3.5 into a separate package.
1172 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1174 SimGrid (3.6) stable; urgency=medium
1176 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1179 * Bindings now constitute their own package, separated from the main one.
1180 Rationale: reduce our maintenance nightmare by reducing the module coupling
1181 They will soon be released on their own on gforge.
1183 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1184 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1186 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1187 * It's not quite deprecated for now because we have no replacement,
1188 but it may soon become the case.
1191 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1192 * Fortran: New user-level cache variable to store the rank of the running
1193 process. This improves performance by an order of magnitude.
1194 * C: New coccinelle script to automatically locate and modify global and
1195 local static variables.
1196 * Improved SMPI network model with a sender-side gap to account for multiple
1200 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1201 only say if a communication is finished, no matter whether it succeeded or
1202 failed. You can call MSG_comm_get_status() to know the status of a finished
1204 * New function MSG_task_dsend() to send a task and detach it. When a
1205 communication is detached, you are never notified of its success or failure
1206 and the memory is released automatically once it is finished. This function
1207 is useful when you don't care about the end nor the success of a
1209 * Change the prototypes of action replay. Sorry for inconvenience,
1210 but this is really more efficient this way (and to adapt your code,
1211 you just have to fix the initialization, that shouldn't be too long)
1212 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1213 ever managed to do anything useful with it.
1214 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1215 takes a dynar of communications. It returns immediately and gives the
1216 index of a finished communication (if any).
1217 * New example: a basic implementation of the Chord P2P algorithm.
1220 * New model for multi-core CPUs. You can now use the core attribute to
1221 precise the number of cores of a host. This is a basic model. Every
1222 process running on the host receives at most the power provided in
1223 the DTD (throughput<=power). Total throughput of process cannot exceed
1225 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1226 router linked by an up-link and a down-link (possibly asymmetrical).
1227 This kind of pattern allows to easily build last-mile model style platforms.
1228 Aggregating such patterns in a rule-based AS is thus the technique of
1229 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1230 * New model for Vivaldi routing. We transformed the Vivaldi network model
1231 into a Vivaldi routing model (based on the rule-based model). This allows to
1232 combine Vivaldi based latencies with last-mile platforms.
1235 * Added a check for NaN of IEEE754 infinite in the double entries of
1237 * Introduce a new context factory "raw", highly inspirated from the
1238 ucontext factory, but using manually crafted functions in assembly to
1239 do the work in an efficient manner.
1240 * Allow to change the used context factory at run time, not only at
1241 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1242 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1243 contexts at runtime (only with raw contexts or ucontexts).
1244 * Completely rewrote this module to allow parallel execution of user
1245 processes. Use --cfg=contexts/nthreads:N to execute user processes
1246 with N parallel threads (the default is 1, meaning no parallelism).
1247 * Allow to decide dynamically between sequential and parallel modes.
1248 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1249 processes in parallel only when their number is greater than or equal to P
1251 * Added a check for NaN of IEEE754 infinite in the double entries of
1255 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1256 won't output the state of processes when interrupted with Ctrl-C
1257 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1258 NULL-terminated array. This may solve backward compatibility issues
1259 due to the change to return type of SD_simulate. See also:
1260 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1261 * Add new macros with variable number of arguments.
1262 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1263 - in xbt/asserts.h: xbt_assert
1264 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1265 - in xbt/ex.h: THROWF and RETHROWF.
1266 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1268 * Change xbt_die() to accept a format string with arguments, just like printf.
1269 * New data structure: xbt_lib_t, like a dict but more general and with better
1273 * New configuration options
1274 Options triva/categorized and triva/uncategorized can be used to generate
1275 graph configuration files for Triva visualization tool.
1276 * Configuration option tracing/platform is renamed to tracing/categorized
1277 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1278 * New TRACE_user_link_variable interface:
1279 User provides the name of the link and the tracing variable to attach to it
1280 * the declaration of tracing categories must be done after the environment
1282 * simpler tracing interface, just one way to declare categories
1283 TRACE_category or TRACE_category_with_color, it is up to you
1284 * links in the trace file are again identified by their names
1285 * trace contains the full platform hierarchy exactly as declared using the ASes
1286 * Options tracing/msg/[task|process]:1 groups the process by hosts
1287 for both cases, tasks and processes must have names that are unique during the simulation
1288 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1289 * The experimental option tracing/msg/volume is deprecated
1290 its functionality may be reincorporated if needed
1292 The tracing generates a trace file with unordered timestamped events,
1293 because of the way the core simulator (surf) works. A script available
1294 at the tools directory (fix-paje-trace.sh) can be used to put the events
1295 in order. We have changed the tracing so it can generate ordered timestamped
1296 events in the final trace, but depending on the simulator (and how much time
1297 is simulated) that can lead to a huge memory utilization. It is deactivated
1298 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1300 Build Infrastructure
1301 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1302 - We are trying hard to keep the API stable, but it may happen that
1303 some things change (we're a research project after all, not a
1304 nuclear plant operating system). If such things should happen, you
1305 could rely on that macro to adapt.
1306 - current value: 30600 for 3.06.00, aka 3.6
1307 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1308 a number that can be compared with SIMGRID_VERSION.
1309 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1310 assertions and verbose/debug logging events are disabled at compile time.
1312 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1314 SimGrid (3.5) stable; urgency=medium
1317 * New feature to any SimGrid-based simulator: Model-Checking
1318 Check SIN#1 for more details.
1321 * New Model SMPI (three-interval linear regression for correction factors)
1322 See RR-7426, available at http://hal.inria.fr/inria-00527150
1323 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1324 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1325 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1326 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1327 * See also src/smpi/README
1332 - Tracing API changes: TRACE_start and TRACE_end should not be called
1333 by user-code. They are automatically called by simulators created
1334 with SimDAG, MSG and SMPI if the toolkit is compiled with
1335 tracing_enabled=ON. Categories declaration and utilization remain the
1336 same for MSG and SimDag.
1337 - A function was added to the tracing API to declare categories with
1339 - TRACE_category_with_color (char *category, char *color)
1340 where color must be in the following format
1341 "%f %f %f", red, green, blue
1342 and red, green, blue are float values in the interval [0, 1]
1343 - User can specify NULL as color parameter, or continue calling
1344 TRACE_category (cat)
1345 On that case, the tracing system will define random colors
1346 - The following command-line options are supported:
1347 --cfg=tracing/filename:msg.trace
1348 --cfg=tracing:1 (activate tracing, needed to use others)
1349 --cfg=tracing/platform:1 (categorized resource use)
1350 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1351 --cfg=tracing/msg/task:1 (task creation)
1352 --cfg=tracing/msg/process:1 (process creation, migration)
1353 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1354 --cfg=tracing/smpi:1 (SMPI interface tracing)
1355 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1356 - examples of examples/msg/tracing updated
1358 - DAXLoader and DOTLoader functions can generate tasks with categories
1359 - A new function to attribute a category to SD tasks:
1360 TRACE_sd_set_task_category (SD_task_t task, char *category)
1361 * Tracing the MPI interface implemented by SMPI
1362 - Collective operations are traced with states
1363 - Point-to-Point operations are traced with states/links
1364 - Tracing activated by a parameter "-trace filename" passed
1365 to smpirun during execution (considering that simgrid
1366 is compiled with tracing enabled)
1367 - To run the simulation with gdb, the simulator
1368 accepts --cfg=tracing/smpi:1 to trace SMPI
1369 - tesh files to check if smpi tracing is ok
1370 - See examples/smpi/NAS/DT-trace
1371 * GTNetS tracing re-worked
1372 - adaptation to the tracing system of GTNets to cope
1373 with modifications regarding the fullduplex mode
1374 - new tesh files to check if gtnets tracing is ok
1377 * Asynchronous communications through the functions:
1378 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1379 * New function: MSG_load_platform_script()
1380 to make possible using a lua script instead of XML files to set up platforms
1381 * New function: MSG_set_function
1382 to associate functions to processes, used when bypassing the parser
1383 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1385 Platforms: Add some more examples in examples/platforms
1386 * Grid'5000: see www.grid5000.fr
1387 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1390 * Change the XML format. This is a very important modification. SimGrid
1391 3.5 introduces a new hierarchical format based on the notion of
1392 Autonomous Systems. Compatibility with old format is ensured
1393 through the perl script provided in the install bin directory
1394 bin/simgrid_update_xml.
1395 It is now possible to build platforms with specific routing
1396 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1397 connect several platforms together. We will try to provide soon
1398 set of realistic platforms exploiting these properties (have a
1399 look at examples/platforms/ for the moment).
1400 * Take the opportunity of the XML format change to be a good XML citizen:
1401 rename link:ctn to link_ctn and similar changes (also dealt with by
1403 * Add a new routing scheme (rule-based) using regular expressions. It
1404 enables to have an extremely low memory footprint when the
1405 underlying routing is simple and can be compactly described. You
1406 need to have libpcre4-dev (perl regular expressions) installed if
1407 you want to use this routing scheme.
1408 * Revive the cluster TAG and allow to easily and efficiently (both in
1409 term of memory and speed) connect clusters together. Have a look
1410 at teshsuite/simdag/platforms/ to see how this can be done. With
1411 this tag, you can create clusters with thousands of tasks at no
1412 cost (have a look at examples/platforms/).
1413 Note: clusters are implemented as ASes, so there is no need for an
1414 enclosing AS tag if you have only one cluster in your platform.
1415 * Add new generic functions in the public interface that allows the user
1416 to call SURF 'create_resource' methods from your code (same
1417 functionality as the XML bypass mechanism but with a much lighter
1419 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1420 that uses a piecewise linear approximation to produce better
1421 results when exchanging small messages.
1422 * Add a new parameter to handle correctly full duplex link and account
1423 for interferences between uplink and downlink communications
1424 (activate with --cfg=fullduplex:1).
1427 * Rename the SD_READY (all dependencies are satisfied and task is
1428 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1429 dependencies are satisfied) state.
1430 This prevents a confusion between the notion of "ready to schedule"
1431 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1432 simulated" (SD_RUNNABLE) used by the simulation kernel.
1433 * Change the way a task is considered as ready. Instead of removing
1434 dependencies when a task is done, a counter is decreased. This way,
1435 it is always possible to reach ancestors thanks to the
1436 SD_taks_get_parents function (even after the end of the simulation.)
1437 * Change the return type of SD_Simulate from (SD_task_t*) into
1438 xbt_dynar_t. This function was in handling a dynar internally and
1439 converted it into a NULL terminated array for historical reasons.
1440 * New function SD_dotload(char*) to load a DAG described in dot
1441 format. This loader and the corresponding examples require the
1442 installation of the graphviz library.
1443 * Fix a bug in the management of tasks of size 0 in the surf network
1444 models. This problem was only visible with SIMDAG and you should
1445 thus disregard results produced with earlier versions if you
1446 relied on this feature (some tasks were blocked because of this).
1447 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1448 even though your applications were doing only point-to-point
1449 communications and sequential computations. Now you can really use any
1450 model you want (of course, if you create real parallel tasks, which are
1451 not implemented in most models beside ptaskL07, this will abort).
1452 * Add an example that schedules a DAX on an heterogeneous platform
1453 using a Min-Min strategy.
1454 * New function SD_workstation_get_current_task() that returns the kind
1455 of task currently running on a workstation in the sequential access
1457 * Raise some warnings when unexecuted tasks remains at the end of the
1458 simulation. This is usually caused by cycles in the DAG.
1461 * New function: SIMIX_process_set_function() called by MSG_set_function
1462 * Change the underlying waiting queue in semaphores so that a process
1463 can wait on several of them simultaneously (as in waitany).
1464 * Fix the way to handle tokens in semaphores so that all access patterns
1465 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1466 * kill the dirty pimple SIMIX_message_sizes_output()
1467 Please use (proper) visualization instead
1470 * New data container: setset (set of sets of elements)
1471 * New module: mmalloc (mapped malloc, allowing to have several
1472 independent segments of malloc)
1473 * New function: xbt_dict_cursor_set_data()
1474 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1475 * New function: xbt_dynar_is_empty()
1476 * New function: xbt_fifo_get_last_item()
1477 * Fix xbt_dynar_shrink(): use the right element size.
1478 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1479 uninitialized areas during expand.
1480 * Fix semaphores: previous implementation was severely broken.
1481 * Use library init/fini functions for our initialization.
1482 - you can use logs and other feature as soon as you want in your
1483 code (even before the xbt_init / MSG_init)
1484 - xbt_exit is now a no-op and produce a warning when used.
1487 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1488 benefit from the latest improvement to the simulation kernel.
1489 * Kill measurement sockets for now. If you rely on them, sorry. This
1490 release is not for you. This feature will be reintroduced in the
1491 future, but we cannot delay the release any further.
1492 * New function: gras_msgtype_get_name().
1493 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1494 * Fix (at last) the pmm example: it should not randomly fail anymore.
1496 Build chain: bug fixes and overall polishing
1497 * Cmake is now stable enough. Hence, we killed the autotools.
1498 * Port to windows ( TM :)
1499 * Fix the 'make install' target.
1500 No need to use 'make install-simgrid' anymore
1501 * Introduce a 'make dist' target compiling a *source* archive
1502 'make package' compiles a binary archive
1503 * Compile java files only on need
1504 * Add --cd and --setenv command line options to tesh
1505 * Out of source builds are not fully supported yet, but we are close
1506 * Enable supernovae and optimization flags by default for our users
1509 * Add layer to set up environment directly from lua, without XML.
1510 * The effect of gras_stub_generator can be achieved through
1511 lua too (check examples/gras/console/ping_generator.lua)
1513 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1515 SimGrid (3.4.1) stable; urgency=low
1517 The "Polishing easter eggs is probably a good idea" release.
1518 This is a bug fixes release only.
1522 * Fix a bug preventing the tasks from begin garbage collected.
1525 * Fix a bug occuring when a host involved in a communication fails.
1526 This was not detected properly by the other peer involved in the
1527 communication. Now, it's reported as a network error.
1530 * Warn the user about loop dependencies in data flow of DAX files
1531 * Obey the control-flow dependencies of DAX files
1534 * Add option "enable_smpi" allowing to not compile SMPI.
1535 Probably useful for the (Mac) users experiencing a build error here
1536 * Improve the detection of lua5.1 and ruby1.8
1538 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1540 SimGrid (3.4) stable; urgency=low
1542 The "Easter in Cargese" release. Also known as (major changes):
1544 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1545 ~> bindings were greatly improved
1546 ~> new tracing infrastructure for better visualization introduced
1548 * the "Welcome to configury modernity" release.
1549 ~> we switched from autotools to cmake, and improved our cdash
1552 A more detailled list of changes follow (full detail in svn log).
1554 Java Bindings: Various Cleanups
1555 * (install java-gcj-compat-dev on debian-like to use them)
1556 * Remove put/get: no need to export deprecated interface in Java
1557 Use send/receive instead.
1558 * Cleanup the examples and add a README per directory
1559 * Remove example autoDestination (that's the only way to go now)
1560 * Remove example explicitDestination (was a plain copy of basic)
1561 * Make JniException a runtime exception, so that there is no need to
1562 declare the fact that you may encounter such a beast. I guess that
1563 nobody will ever want to survive such error.
1564 * Create specific errors for each MSG case of failure:
1565 host failure, transfer failure, timeout, task cancelled
1566 * Cleanup the exceptions that may get thrown by each function
1567 * Other internal cleanups in Java bindings. Performance still bad :/
1568 Ruby and Lua Bindings: create them
1569 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1570 * That's new and great, you should try them out.
1571 Same functionalities than Java bindings, only even less polished
1573 * Kill the useless "rate" argument of SD_task_get_execution_time()
1574 Everyone used to provide -1 as a value, it was not used, and the
1575 semantic of a possible use wasn't even clear.
1576 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1577 either as comm costs or compute costs to mean that there is no
1578 such thing for that specific task.
1579 * Add a SD_task_set_name() function
1580 * Fix SD_task_unschedule() on typed tasks
1581 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1582 * In DAX loader, accept useless 'level' attributes to <job> since
1583 LIGO DAGs have them (seem to be to ease graphical representation).
1585 * Add an example masterslave_mailbox.c using send/receive and not
1586 the deprecated put/get interface.
1587 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1588 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1590 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1591 (declare MSG_USE_DEPRECATED to still have the old name)
1592 * Add a MSG_task_set_data() function
1593 * About trace replay (see examples/msg/actions):
1595 - Allow to work with splitted trace files for each process
1596 Give the specific trace file as argument of each process,
1597 and call MSG_action_trace_run(NULL)
1598 You can still have one merged file for all processes.
1599 - Fix implementation of collective operations
1600 * Allow task_execute() on 0-sized tasks (closes #10063)
1602 * This is the first release of SimGrid where SMPI is not considered
1603 beta anymore (even if some corners should still be improved)
1604 * Port over the new SIMIX_network submodule (internal refactoring)
1605 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1606 * Implement more missing elements of the standard:
1608 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1610 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1611 MPI_Scan() MPI_Get_processor_name()
1612 - Added implementation of missing case for Alltoall (warning: it's
1613 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1614 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1615 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1616 SMPI_MPI_Allgatherv()
1617 * Bug fixes include:
1618 - MPI_Waitsome() was broken
1619 - Allow relative includes in smpicc
1620 - Command line cfg argument 'reference_speed' was ignored...
1621 - Some functions did not properly lead to auto-benching of user code
1622 - smpicc passes -O2 by default (just like openmpi one)
1624 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1625 * Bug fixes about timeouts during communications
1626 * add SIMIX_message_sizes_output() as a pimple to write to file the
1627 amount of messages per size. Use gnuplot to get histogram.
1628 Pimple because that's the only user-visible function of simix,
1629 defined directly in xbt.h (irk, sorry)
1631 - Add a SIMIX_sem_get_capacity() function
1632 - Fix interactions with processe resume/suspende
1633 - release_forever() was stupidly broken
1634 - Fix SIMIX_display_process_status() for processes in a semaphore
1635 - Make SIMIX_sem_block_onto() user-visible
1636 * Refactoring context stuff:
1637 - Use pseudo-OOP for better modularity
1638 - reimplement SIMIX_process_kill() without process_schedule() so
1639 that the latter can take as invariant that it is called from
1641 - Merge context_start into context_new for sake of simplicity
1643 * Add a Vivaldi network model, coded live during SUD'10 ;)
1644 * Rename configuration variables to start a hierarchy:
1645 o cpu_model -> cpu/model
1646 o network_model -> network/model
1647 o workstation_model -> workstation/model
1648 * New configuration variables:
1649 o network/bandwidth_factor: correction to bandwith
1650 o network/latency_factor: correction to latency
1651 o netwotk/weight_S: correction to the weight of competing streams
1652 * Add a long description to the models, that users can see with such
1653 argument on the command line: --cfg=cpu/model:help
1654 * --help-models display the long description of all known models
1656 * config: add the ability to set a default value after registration
1657 Does not override any previously set value (e.g. from cmd line)
1658 * dict: allow to have integer key and data.
1659 When so, you need to use the following functions
1660 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1661 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1662 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1663 In contrary to regular dicts, the key is not malloced before copy.
1664 Mixing scalar and regular elements in the same dict is not tested
1666 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1667 Tracing for Visualization:
1668 * SimGrid is now instrumented in order to generate a trace file for
1669 visualization analysis: to use it, need to compile SimGrid with the
1670 "tracing" option enabled, and instrument the program using SimGrid with
1671 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1672 (among other functions).
1673 * The instrumentation only traces the platform utilization for now
1674 * Documentation to use the tracing functions and how to analyze the
1675 traces with the Triva tool is written.
1676 * More information about: SimGrid FAQ (in the section Tracing Simulations
1679 * We moved to cmake as default build system. Autotools support will
1680 be dropped soon. Check the FAQ for more info about how to use it.
1681 * Greatly improved our cdash/ctest interactions
1682 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1683 * Added memory checking tests with valgrind; lot of memleak fixing.
1684 This may be the first release of simgrid with so few memory issues
1685 * Added code coverage tests.
1686 Our coverage is still improvable, but at least we see it on cdash.
1688 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1690 SimGrid (3.3.4) stable; urgency=low
1692 The "Desktop Grid needs love too" release (also called Xmas release).
1694 Models improvements:
1695 * Major speedup in the maxmin system solving by using lazy evaluation
1696 Instead of solving completely the maxmin system at each iteration,
1697 only invalidate (and recompute) the modified parts.
1698 This new feature is enabled in default models but you can try to
1699 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1700 * Cas01 IMproved as default CPU model
1701 This CPU model is the same Cas01 model, but it uses the
1702 maxmin-selective-update flag and a heap structure to manage
1703 actions on SURF kernel.
1704 It reduces the complexity to find the next action to finish and,
1705 consequently, it's faster than the old Cas01.
1706 This is the new default CPU model (Cas01).
1707 * Rename the old Cas01 model to Cas01_fullupdate
1708 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1709 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1710 * CpuTI (CPU Trace Integration)
1711 A new CPU model whose objective is simulate faster when using
1712 availability trace files.
1713 Instead of using a full featured, over engineered maxmin system for
1714 CPU modeling, this model does the pre-integration of traces files
1715 to calculate the amount of CPU power available, and so, executes
1716 faster than the old CPU models.
1717 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1718 * Use LV08 as default network model since it gives better accuracy
1719 for small messages and shouldn't change things for big ones.
1720 Use --cfg=network_model:CM02 to get the previous behavior.
1723 ******************************************
1724 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1725 ******************************************
1726 * The new CPU model may changes simulations!
1727 The point is that events occurring at the exact same timestamp
1728 are not scheduled in the same order with the old and new
1729 version. This may be enough to completely change the execution
1730 of simulations in some cases.
1731 * The new network model will change simulations!
1732 This new model is more realistic than the previous one, so you
1733 should consider redoing your old experiments with this model.
1734 Sorry for the inconvenience.
1737 * Introduce the supernovae compilation mode
1738 When compiled that way, the whole SimGrid (or almost) is put in a
1739 single compilation unit and compiled in one shoot.
1740 This is to help gcc which has difficulties to inline stuff from one
1742 The speedup seem to be above 15%, althrough more tests are needed on
1743 amd64 to confirm that gain.
1746 * Port of MSG's mailbox on top of SIMIX network
1747 The put/get mechanism was greatly simplified on the way.
1750 * New SIMIX network module. Provides:
1751 - Mailbox: rendez-vous mechanism to find with who you want to speak
1752 - Synchronous send/recv: easier and hopefully faster since the
1753 logic is handled in the maestro process directly now
1754 - Asynchronous send/recv: you dreamt of it? It's here now
1755 Too bad that nobody cared enough to propagate the change to MSG.
1756 * Add semaphores as SIMIX synchronization mechanism.
1759 * new function SD_daxload(char*) to load a DAX file
1760 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1761 * Introduce typed tasks. Specify its kind and cost at creation.
1762 At scheduling, just give where it should be placed, and the cost
1763 for each involved resource is automatically computed.
1764 Existing constructors so far (more to come of course):
1765 - SD_task_create_comm_e2e() for end-to-end communication
1766 - SD_task_create_comp_seq() for sequential computation
1767 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1768 * new function SD_task_dump() for debuging display
1769 * new function SD_task_dotty(task,FILE*) writing to file the info
1770 about the task in dotty format
1771 * SD_task_dependency_exists() can now cope with having one of its
1772 arguments NULL. If so, it tests whether the other argument has any
1774 * Add getters on list of preceding/following tasks:
1775 SD_task_get_parents(task) and SD_task_get_children(task)
1776 * Add getters on amount of workstations and list:
1777 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1778 * Add getter on task kind: SD_task_get_kind(task)
1779 * Update the start_time and finish_time of tasks on completion/failure
1780 * Bugfix: Remove task from state swags when destroyed
1783 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1786 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1787 dynar with one separate thread per value of the dynar.
1788 * Change the prototype of xbt_thread_create(), sorry.
1789 Added a boolean parameter indicating whether we want to join this
1790 thread (used in SG only for now)
1791 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1794 * GTNetS wrappers should now be usable again (and betterly tested too)
1795 * Fix a major regression from 3.2 where the timeout provided to
1796 MSG_task_put_with_timeout() was used as absolute time before which
1797 the comm should be done.
1798 * Start to fix the <cluster> tag.
1799 - Internal links should be good now (beside of the loopback, which
1800 use the private link instead)
1801 - paths to the external world is still rather broken
1802 - the <route:multi> tag is just broken. Actually that's brain-dead.
1803 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1804 to make it less stupid
1805 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1806 * Fix a source-level compatibility glitch from 3.2: after defining
1807 MSG_USE_DEPRECATED, you can use the old name
1808 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1809 * Allow to compile from the SVN with automake 1.11
1810 * Fix some problems when using the "start_time" tag in deployment XMLs.
1811 * Fix #8569: XBT/synchro.h has redundant declarations
1812 * Fix #8563: MSG return values and exceptions
1813 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1814 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1815 Thanks to Robson Peixoto.
1816 * Fix a vicious bug in dictionaries inducing that some elements were
1817 not freed on xbt_dict_free()
1819 Portability report of this version:
1820 * Main portability targets:
1821 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1822 on (amd64/i386/ia64)
1823 - mac leopard on i386
1824 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1825 but nothing critical.
1826 * Other platforms: windows, AIX and others were not tested for this release
1828 Timing report of this version:
1829 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1830 scenario parameters). From 8h to a few seconds in desktop grid settings.
1831 * Supernovae brings about 25% speedup on i386.
1833 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1835 SimGrid (3.3.3) stable; urgency=low
1837 The "Need for Speed" release.
1839 The timings done to validate the 3.3.2 were faulty.
1840 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1842 The problem was a conversion from a manually handled vector to
1843 xbt_dynar_t on the critical path.
1844 xbt_dynar_foreach calls functions, inducing stack management crap.
1846 We inlined these functions and xbt_dynar_foreach is now breath taking.
1847 We also inlined xbt_swag_belong on the way.
1849 Here are some approximate speedup measurements (on master/slaves
1850 simulations lasting between 10s and 20s each):
1851 3.3.1 -> 3.3.2: about same performance
1852 3.3.2 -> 3.3.3: 40% speedup
1853 3.3.1 -> 3.3.3: 40% speedup
1854 3.3.1 with inline patch -> 3.3.3: 30% speedup
1856 Our reading is that the refactoring which occurred in 3.3.2 made us
1857 suffer much more from the xbt_dynar_foreach low performance, but
1858 once we solved this, this refactoring proved to be very performance
1859 effective. From the 40% speedup, somehow, 10% are due to the
1860 inlining and 30% to the refactoring.
1862 That's a pitty that gcc cannot inline functions placed in other files
1863 alone. We have to choose between:
1864 - break the encapsulation (by putting private data structures and
1865 accessors in headers files to help gcc)
1866 - live with low performance
1867 - switch to a decent compiler such as icc (not quite possible).
1869 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1871 SimGrid (3.3.2) stable; urgency=low
1873 The "Simplicity does not preceed complexity, but follows it" release.
1875 The main contributors of this release were (lexical order):
1876 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1879 * Extract the routing logic into its own object.
1880 (was dupplicated in network.c and workstation_LV07.c;
1881 Allows to implement other ways of storing that info)
1882 => kill now useless network_card concept
1883 - Use dynar to represent routes (instead of void** + int*)
1884 - kill link_set (use surf_network_model->resource_set instead)
1885 - Add a command-line option to choose the routing schema to use
1886 - Add three new models:
1887 * Floyd (shortest path computed at initialization)
1888 * Dijikstra (shortest path recomputed all the time)
1889 * Cached Dijikstra (shortest path computed on need)
1890 All these models where contributed by Silas De Munck, and are
1891 described in his ICCS09 paper.
1893 * Simplify model declaration
1894 (less redirections, less function to write when defining a model)
1895 - Factorize stuff between models:
1898 surf_model_resource_set(model)
1899 surf_model_resource_by_name(model, name)
1900 - Unify the types of models in s_surf_model_t (using an union)
1901 - Embeed fields of common_public directly into s_surf_model_t
1902 - Rename model methods:
1903 action_free ~> action_unref
1904 action_change_state ~> action_state_set
1905 action_get_state ~> action_state_get
1906 - Change model methods into functions :
1907 (model)->common_public->action_use ~> surf_action_ref
1909 * Implement a generic resource; use it as ancestor to specific ones
1910 (allows to kill duplicated code in models)
1911 Drawback: timer command don't need no name nor properties;
1912 workstation_CLM03 don't need no properties
1913 (but I guess we can live with those few bytes wasted)
1915 * Improve the action object model
1916 - implement a constructor avoiding dupplicated code about field
1917 initialization in generic_action part.
1919 * Kill the SDP model: it has an external dependency, is deprecated
1920 in flavor of modern lmm models, and didn't compile since a while
1923 * Relocation of the context module from XBT to SIMIX.
1924 (the context were decoupled from the simix processes, duplicating a lot of code)
1925 => a lot of code was factorized
1926 - less overhead is introduced during scheduling
1927 - simpler API for the context factory
1928 - the logic for process creation,destruction and manipulation was simplified
1929 * Simplification of the s_smx_process_t data structure.
1930 => accesing the simix level data associated to a process is faster now,
1931 and the code is a lot more readable.
1934 * Implement some more MPI primitives:
1935 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1936 -implementation: Bcast: flat or 2-ary tree (default),
1937 Barrier: 4-ary tree,
1939 Allreduce: Reduce then Bcast
1940 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1941 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1942 Alltoallv: flat tree, like ompi
1944 * Add support for optimized collectives (Bcast is now binomial by default)
1945 * Port smpirun and smpicc to OS X
1948 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1949 Shout out if you used it.
1952 * Display the list of still queued messages in SG mode when existing
1956 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1957 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1958 * Add xbt_str_from_file(FILE*)
1959 * Add xbt_dict_get_key achieving a linear reverse search
1960 * Remove the context module
1962 Portability report of this version:
1963 * Main portability targets:
1964 - Linux(debian)/x86/context
1965 - Linux(debian)/x86/pthreads
1966 - Linux(debian)/amd64/context
1967 - Linux(debian)/amd64/pthreads
1968 On these, we still have the eratic breakages of gras/pmm and
1969 amok/saturate_sg reported in previous version. We still think
1970 that the tests are the cause of the fault, not the tested code.
1972 - Mac OSX Leopard/x86/context
1973 Still false negative in tesh autotesting.
1974 Smpi still fails, but this time because readlink does not accept -f
1975 Everything seems to work properly beside of that.
1978 - AIX version 5.3 (only tested contexts this time)
1979 Smpi still fails there because mktemp is not installed.
1980 Everything seems to work properly beside of that.
1982 I managed to compile it for the first time, but several breakages.
1983 Won't delay the release for this exotic platform.
1985 * Windows: it's still lagging behind. If you want to help, please
1988 Timing report of this version:
1989 This version seem to be more than 5% faster than 3.3.1 (on linux
1990 64bits with contextes). The gain is less than expected, we are
1991 investigating this for next release.
1993 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1995 SimGrid (3.3.1) stable; urgency=low
1998 * Implement a --cfg-help to show existing configuration variables
1999 * Build chain do not require doxygen in maintainer mode
2002 * fix a bug on struct sizeof computation, which prevented the
2003 exchange of arrays of structs in some conditions
2004 - added a regression test about this in datadesc_usage
2005 * Allow the exchange of 0-long dynamic vectors.
2006 - for that, use -1 as indicator of dynamic size instead of 0
2007 - This implied to change any size from unsigned long to long,
2008 reducing a bit communication abilities, but I guess that with
2009 64bits being quite common, this is more than enough.
2010 - This also induce a protocol change, thus bumping network protocol
2011 version from 0 to 1 (if we have external users, we have to get
2012 clean on that point too ;)
2013 - added two regression tests about this in datadesc_usage
2014 * Be more verbose when propagating local exceptions
2015 This helps debugging.
2016 * Display the status of simulated processes when receiving SIGINT in
2020 * Allow to control the simulation from a trace file.
2021 New functions MSG_action_register() and MSG_action_trace_run()
2022 The first one allows to associate a function execution to each
2023 kind of action while the second one parses a trace file and
2024 triggers the corresponding actions within the system.
2025 For now, only a toy example is provided in examples/msg/actions
2026 * Add an exemple of process migration in examples/msg/migration
2027 * Fix a bug in task exchange which broke MSG_task_get_sender()
2028 Add a teshsuite regression test for that.
2029 [Bug: if MSG_task_get_sender() is called after sender exit,
2031 * Fix a bug which prevented suspend/resume to work properly
2032 * Display the status of simulated processes when receiving SIGINT
2033 This fixes a regression of v3.3. due to the introduction of SIMIX
2034 * Bug fixing in failure management:
2035 - trace could not start by a failure at time 0
2036 - failure during communications were not working
2039 * Add SIMIX_process_set_name() to change the name of the current
2040 process in the log messages.
2041 * Store smx_hosts in a dict since we only retrieve them by name
2042 * Move the configuration infrastructure to surf
2045 * Move the configuration infrastructure to surf
2048 * Massive internal cleanups:
2049 - Store internal structures on processes instead of hosts (allows
2050 to have more than one process per host, in addition of being more
2052 - Cleanup the initialization/finalization process
2053 - Kill a whole bunch of unneeded synchronization:
2054 processes run in exclusive manner within the simulator
2055 - Move queues from global tables to process data fields
2057 - now accept -platform and -hostfile arguments
2058 - Pass the right rank value to processes according to the hostfile
2059 * Compile the examples by default, and use them as regression tests
2060 * Implement MPI_Wtime()
2061 * Change the reference speed to a command line option
2064 * TCP_gamma can now be specified as command line option using
2065 --cfg=TCP_gamma:10000000.0
2066 * Change the --surf-path cmd line option into --cfg=path:
2069 * Also include strbuff from xbt.h public header
2070 * xbt_ex_display(): do not free the exception after displaying
2071 This allows to do more with the given exception afterward.
2072 Users should call xbt_ex_free() themselves.
2076 Portability report of this version:
2077 * Main portability targets:
2078 - Linux(debian)/x86/context
2079 - Linux(debian)/x86/pthreads
2080 - Linux(debian)/amd64/context
2081 - Linux(debian)/amd64/pthreads
2082 These targets fail about 1/10 of times on gras/pmm, but we believe
2083 that this is because of the test, not because of simgrid.
2084 amok/saturate_sg fails even more rarely, and the test may not be
2087 - Mac OSX Leopard/x86/context
2088 The test suite still spits tons of errors because some obscure
2089 force prevents us from removing the temporary directories
2090 arguing that they still contain some metadata I've never heard of.
2091 Smpi fails because seq is not installed.
2092 Everything seems to work properly beside of that.
2095 - AIX version 5.3 (both contexts and pthread)
2096 Smpi still fails there because mktemp is not installed.
2097 XML inclusions seems rosty on AIX.
2099 * Windows: it's still lagging behind. If you want to help, please
2102 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2104 SimGrid (3.3) stable; urgency=high
2108 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2109 [Malek Cherier & Mt]
2111 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2114 Until now, GRAS were using MSG as an interface to SURF. It was
2115 quite difficult because both interface have several differences
2116 (MSG channels vs GRAS sockets were the most notable point).
2118 This also opens the gate to SMPI (which should occur soon) and speed
2119 up simulations by to 40% (even if it were not the main goal).
2121 **************************************
2122 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2123 **************************************
2124 The point is that events occuring at the exact same timestamp are
2125 not scheduled in the same order with the old and new version. This
2126 may be enough to completely change the execution of simulations in
2127 some cases. Sorry for the inconvenience.
2129 * Cleanup and upgrade the XML format to push further scalability
2130 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2132 * Improve the testing infrastructure with tesh. Now a very large part of
2133 the code is tested not only by being run but also by checking that the
2134 output match an expected output [Mt].
2136 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2137 is really less memory-demanding, which should allow you to use
2138 larger files in SimGrid [AL].
2140 * Inform valgrind about our contextes, so that it becomes usable
2141 with the default (and more effecient) version of SimGrid
2142 [contributed by Sékou Diakite, many thanks]
2145 * Introduce a listener thread in charge of receiving incoming
2146 messages from the network. It allows to overlap communication and
2147 computation but most notably, it removes some stupid deadlocks due
2148 to the fact that so far, a process could not send and receive at
2149 the same time. This made most non trivial communication schema
2151 * Convert the PIDs from long int to int to match the MSG ones (and
2152 linux ones too) [Mt]
2153 * New function: gras_agent_spawn() to launch a new process on
2154 current host. Only working in simulation for now. [Mt]
2155 * New function: gras_os_hostport() returning a constant form (ie,
2156 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2159 * Make the backtrace of exceptions more human readable [Mt]
2160 * New module: xbt/str [Mt]
2161 a ton of string utility functions (split, join, printf to a newly
2162 allocated buffer, trim, etc)
2163 * New module: xbt/hash [Mt]
2164 SHA1 hashing algorithm (more to come if needed)
2165 * New module: xbt/synchro [Mt]
2166 synchronization tools (mutex and conditions) working the same way
2167 in simulation and in real life (mainly useful for GRAS, but not
2169 * New module: xbt/queue [Mt]
2170 classical producer/consumer synchronization scheme
2171 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2172 (using the classical functions will get serialized) [Mt]
2173 * Make dictionary internal table dynamic. No need to specify its size
2174 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2176 * Make sure the log channels are organized as a tree under windows
2177 (because of ANSI C compatibility issue, any channel were child of
2178 root directly) [Mt].
2181 * Cleaned many thing in surf and fixed a few bugs [AL].
2182 * Add a nice command line configuration mechanism to compose models [AL].
2183 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2184 the previous one (KCCFLN05). It relies on something that looks like
2185 a max-min sharing mechanism but cannot be written as such. A new solver
2186 was thus designed [AL].
2187 * Add a new solver to lmm. Based on Lagrange optimization and
2188 gradient-based descent, it enables to efficiently maximise systems s.a
2190 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2192 This solver enables to propose two new network models for TCP Reno and
2193 TCP Vegas based on Low's work. These models still need to be fully
2194 tested though [Pedro Velho].
2197 * Bug fix in SD_simulate. Now the time bound given as argument is
2199 * Use the new parallel task model (ptask_L07) as default.
2200 * Use the SURF command line configuration mechanism.
2201 * 0-size tasks (for synchronization) should now work.
2203 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2205 SimGrid (3.2) stable; urgency=high
2209 We still experience issues on this platform, but we believe that at
2210 least MSG is usable.
2212 GRAS API BREAKAGE (for simplification purpose, sorry):
2213 * the gras_msgtype_by_name is not used anymore. Instead of
2214 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2215 you can write (and must)
2216 gras_msg_send(toserver, "request", &request);
2217 - If you still want to pass a gras_msgtype_t to the function (to cache
2218 the type and avoid the lookup time), use the gras_msg_send_() variant.
2219 - Impacted functions:
2220 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2221 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2222 * The callbacks are now expected to return 0 when everything went well
2223 (just like the main() function)
2225 GRAS new features and improvements:
2226 * New module mechanism where user code can use per process globals [Mt]
2227 This is similar to gras_userdata_*() functions, but for libraries. It
2228 factorize some code developped over and over in the examples and AMOK.
2229 It has still to be documented and used (only amok/peermanagement is
2231 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2232 loss when they were small enough to fit into the buffer and sent quickly
2233 enough so that they can all get received in one shoot.
2234 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2235 if not found. Use the *_or_null() variant for the old semantic.
2236 * In gras_msg_handle, do not discard messages without callback.
2237 They are probably messages to be explicitly awaited later (ie, proofs of
2238 mis-synchronization in userland since they are sent before being awaited)
2239 No big deal usually.
2240 * gras_socket_meas_send/recv: semantic changed!
2241 The numerical arguments used to be (1) the total amount of data to send
2242 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2243 messages. This was need for the fool willing to send more than MAXINT
2244 bytes on quite fat pipes.
2247 * Do really rename the hostmanagement module to peermanagement. [Mt]
2248 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2249 but this is rather new and this was documented in the module
2250 documentation (poor excuses, I admit)
2251 * Bandwidth measurement semantic changed! This follows the changes to
2252 gras_socket_meas_send/recv explained above.
2255 * A sequential mode has been added to the workstations. When a workstation
2256 is in sequential mode, it can execute only one task, and the other tasks
2257 are waiting in a FIFO. [Christophe Thiery]
2260 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2261 model for SIMDAG. [Christophe Thiery]
2262 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2263 equal to 0, which caused some bad behaviors in
2264 saturated_constraint_set_update. I now use a threshold mechanism like in
2268 * When running manually src/testall, you select specific units [Mt]
2269 testall is the result of our cunit mechanism, and should replace all
2270 the scripty thingy around since bash don't run easily on billware.
2272 * A mallocator system has been added. [Christophe Thiery]
2273 Mallocators allow you to recycle your unused objects instead of freeing them
2274 and allocating new ones.
2276 Documentation update:
2277 * FAQ reworking + New FAQs:
2278 - "Valgrind spits tons of errors!" [Mt]
2279 - "How to repport bugs" [Mt]
2280 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2281 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2282 - Communication time measurement within MSG [AL]
2283 - I experience weird communication times when I change the latency [AL]
2284 * GRAS tutorial [Mt]
2286 - an introduction to the framework and to the used communication model
2287 - an initiatic tour introducing the most proheminent features:
2289 . Lesson 0: Installing GRAS
2290 . Lesson 1: Setting up your own project
2291 o Part 2: Message passing
2292 . Lesson 2: Exchanging simple messages
2293 . Lesson 3: Passing arguments to the processes (in SG)
2294 . Lesson 4: Attaching callbacks to messages
2295 . Lesson 5: Using globals in processes
2296 . Lesson 6: Logging information properly
2297 . Lesson 7: Using internal timers
2298 . Lesson 8: Handling errors through exceptions
2299 . Lesson 9: Exchanging simple data
2300 . Lesson 10: Remote Procedure Calling (RPC)
2301 . Lesson 11: Explicitely waiting for messages
2302 . Recapping of message passing features in GRAS
2303 - A HOWTO section containing:
2304 o HOWTO design a GRAS application
2305 More are due, of course. They will come latter. In the meanwhile, you can
2306 check the examples which are still here.
2308 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2310 SimGrid (3.1) stable; urgency=high
2314 There was a stack corruption somewhere, visible only when optimizing
2315 with these versions. [Vince]
2318 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2319 module that enabled to play with Directed Acyclic Graphs. It is built
2320 directly on top of SURF and provides an API rather close to the old
2321 SG. Some old codes using SG are currently under rewrite to check that
2322 all needful functions are provided. [Christophe Thiery]
2325 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2326 extension of the classical CLM03 model that gracefully handles
2327 failures. This is now the default model for MSG and GRAS. It doesn't
2328 handle parallel tasks yet though. [AL]
2329 * Bug fix: Weights were not correctly set in the network part.
2330 WARNING: This may have resulted in incorrect results with simulations
2331 where there are more than one flow on a given link. [AL]
2334 * After a (long ?) discussion on simgrid-devel, we have decided that the
2335 convention we had on units was stupid. That is why it has been decided
2336 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2337 WARNING : This means that all previous platform files will not work as
2338 such with this version! A warning is issued to ask users to update
2340 A conversion script can be found in the contrib module of the CVS, under
2341 the name contrib/platform_generation/surfxml_update.pl [MQ]
2344 * Bug fix: Processes were started in reverse order, wrt deployment file.
2345 WARNING: if your code relies on this bug, please fix it. [AL]
2346 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2347 being executed on two different locations. [AL]
2348 * Bug fix: Failures are now better supported thanks to Derrick's tests
2349 (there was many failure situations I hadn't thought of and that weren't
2350 correctly handled). [AL]
2351 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2355 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2356 Exception occurring on server-side are propagated back to client (!).
2358 API CHANGE: the callback changed their prototype. Change:
2359 int my_handler(gras_socket_t expeditor, void *payload_data) {
2361 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2362 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2364 * New! function: gras_msg_handleall to deal with all messages arriving
2365 within a given period.
2366 * New! function: gras_socket_server_range to get a server socket in a
2367 range of port numbers (ease to avoid port number conflicts) [MQ]
2368 * New! gras processes display their backtrace when they get a SIGUSR1
2369 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2370 Sweet to debug RL processes [MQ]
2374 - Do not force experiment sizes to be expressed in kb, or it becomes
2375 impossible to measure the latency this way (needs one byte-long tests)
2376 WARNING: this changes the amok_bw_* function semantic. [MQ]
2377 - Implements the link saturation stuff. [MQ]
2378 * Peer management module:
2379 New! module factorizing code that we wrote over and over [MQ].
2382 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2383 - Test units are placed directly into the library code, they get extracted
2384 automatically and placed into the src/testall binary.
2385 - Convert most of the XBT tests to this system.
2386 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2387 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2388 live in the same C code. This required to change a little bit the API
2389 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2390 * New module: sparse graph structure with basic algorithms (this is work
2391 in progress and the API is not considered to be frozen yet). [AL]
2392 * Display more information on backtraces: source line & function names are
2393 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2394 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2395 * Reworked a little bit some #include statements to load only required
2396 headers. Some user code that relied on SimGrid to include stdlib or
2397 stdio may need to include it by themselves. [AL]
2398 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2400 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2401 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2402 but I guess that almost nobody used those parts. [MQ]
2404 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2406 SimGrid (3.0.1) stable; urgency=low
2409 * Unfortunately, I had missed 5 misnamed functions:
2410 xbt_fifo_item_t xbt_fifo_newitem(void);
2411 void xbt_fifo_freeitem(xbt_fifo_item_t);
2412 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2413 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2414 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2415 They're now deprecated. Please use their new versions:
2416 xbt_fifo_item_t xbt_fifo_new_item(void);
2417 void xbt_fifo_free_item(xbt_fifo_item_t);
2418 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2419 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2420 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2422 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2423 * Documentation: xbt_log module unmercifully reworked [MQ]
2424 * Bugfix: there was a problem with the ending of contexts with
2425 the pthread backend. It caused some weird deadlock or behavior
2426 depending on the pthread implementation. [AL]
2427 * Bugfix: get the exceptions raised in the simulator repport where
2428 and why they come from when they are not catched in time [AL, MQ]
2431 * Bugfix: Do repport the error when two non-connected hosts try to
2432 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2435 * Add additionnal checkings on communications. Assert that two
2436 communicating hosts are connected by a set of links... [AL]
2439 * Add additionnal checkings on channel values in communication [AL]
2440 * New: MSG_task_get_source to see on which host a task was generated [HC]
2441 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2442 the number of tasks waiting to be received on channel and sent
2444 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2445 waits for the first task coming from a given host.. [AL]
2447 GRAS new functionnalities: [MQ]
2448 * Enhance the parsing macro to allow the size of multidimentional objects
2449 to be given thru annotations.
2450 * New example (and documentation): Matrix Multiplication a la RPC
2451 (as when I was young!) and fix a bunch of bugs found on the way.
2453 GRAS performance improvements: [MQ]
2455 * Reduce the amount of cbps creation/destruction by making it static to
2456 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2459 * Change libdata to a set so that we can search for stuff by ID (and thus
2460 reduce the insane amount of dict lookups)
2463 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2464 RL communication on top of SG since it's so uneffective.
2465 It may also be used for inter-thread communication in RL, one day.
2466 * Use gras_datadesc_copy() to exchange messages on top of SG
2468 - improve message exchange performance on top of SG
2469 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2470 * Don't exchange on the network the size of the used part of buffer,
2471 instead, specify the possible buffer size to read().
2473 - reduces the amount of read/write calls (one pair per exchange)
2474 - reduces the amount of exchanged data (the size)
2475 - allows to retrieve all arrived data on receiver side, if we don't need
2476 it right now (subsequent read will peek the buffer)
2477 - allows the receiver to proceed with the begining of the stream before
2478 everything is arrived
2479 - make it possible to build an iov transport (using readv/writev)
2481 - take care of the data with non-stable storage (like stacked data),
2483 * If possible, TCP send uses vector I/O (when writev() is here)
2484 - Don't use it for receive since we send data sizes and data on the
2485 same stream, so we wouldn't be able to chain large amount of chunks
2486 before having to flush the stuff to read the size.
2487 * Rework the transport plugin mechanism to simplify it and reduce the
2488 amount of pointer dereferencement when searching for the right function
2491 * I guess that now, we do almost as few system calls as possible while
2492 doing as few data copy as possible.
2494 To improve it further, we could try to send all the sizes first and then
2495 all the data (to use iov on receiving size), but it's only a partial
2496 solution: when you have 2 dimensional data, the sizes of the second
2497 dimension is data of the first dimension, so you need 3 streams.
2499 I'm not sure the potential performance gains justify the coding burden.
2501 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2503 SimGrid (3.00) stable; urgency=high
2506 * New! Give the possibility to hijack the surf parser and thus bypass
2507 MSG_create_environment and MSG_launch_application. Have a look at
2508 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2510 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2512 SimGrid (2.96) unstable; urgency=low
2517 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2518 This deprecates the xbt_error_t mechanisms.
2519 It modifies (simplifies) all XBT and GRAS API.
2520 MSG API keeps unchanged (exceptions raised by XBT are catched from
2521 within MSG and masked with existing error handling facilities)
2524 * New! Add a FATPIPE model. [AL]
2525 * New! Add a parallel task model. [AL]
2526 * New! Add automatically a loopback interface (in the default
2527 network model) if none was precised.
2530 * Bugfix: MSG_process_resume now works with the current running process.
2532 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2533 * Modification of MSG_task_get_compute_duration. Once a task has been
2534 processed, the value returned by this function is now equal to 0. [AL]
2535 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2536 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2537 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2538 of a m_task. For now, MSG_task_cancel only works with computation
2540 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2541 of the processor (in Mflop/s) regardless of the current load on the
2543 * API Change: use proper naming convention for MSG_getClock and
2544 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2546 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2547 This function changes the priority of a computation task. This priority
2548 doesn't affect the transfer rate. A priority of 2 will make a task
2549 receive two times more cpu power than the other ones. This function
2550 has been added to suit the needs of Nguyen The Loc and hasn't been that
2551 much tested yet. So if it fails, please report it and send me your code.
2553 * API Change: removed all functions and types that were marked "deprecated"
2554 since many months. Renamed MSG_global_init_args to MSG_global_init.
2556 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2558 SimGrid (2.95) unstable; urgency=low
2561 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2562 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2563 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2566 * Bugfix: Allow absolute paths for platform description files [MQ]
2567 * Bugfix: do free the variables after use. Leads to drastic performance
2569 * Implement max_duration (ie, timeouts) on resources [AL]
2572 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2574 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2575 the needed support (timeouts) [AL]
2578 * Implement measurement sockets. You can now get the bandwidth between two
2579 hosts thanks to AMOK (see below). [MQ]
2580 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2581 dynar over the network (yeah) [MQ]
2582 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2583 * simple/static token-ring example. [Alexandre Colucci and MQ]
2584 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2585 in place before [MQ]
2588 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2589 type parsing. "short" and "long" modifiers were also ignored; other
2590 modifier (such as reference level) are still ignored. [MQ]
2591 * Embeed the buffer size within the buffer itself on SG. [MQ]
2592 That way, send() are atomic and cannot get intermixed anymore (at least
2593 the ones which are less than 100k; bigger messages still have the issue)
2594 * Array size pushed by the field, not by the field type (or each
2595 and every long int will push stuff to the cbps) [MQ]
2596 * use select() to sleep since it allows to portably sleep less than one
2599 GRAS (minor cleanups)
2600 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2603 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2604 * type category 'ignored' killed as it was never used and were difficult
2606 * whether a type can cycle or not is now a flag, leaving room for more
2607 flags (as "ignored", if we feel the need one day ;)
2608 * Rename raw sockets to measurement sockets since "raw" has another
2609 meaning in networking community.
2612 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2613 and offers features not belonging to GRAS but that most applications
2614 need. One day, it may be a set of plugins loadable at runtime.
2615 * New module: bandwidth
2616 bandwidth measurement between arbitrary nodes running this module. [MQ]
2618 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2620 SimGrid (2.94) unstable; urgency=low
2622 The first beta release of SimGrid 3 !
2626 * Update the main page and the FAQ. Adding references to gforge.
2629 * Add a gras_os_getpid function.
2632 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2633 * Extend the logs so that they also print PID, hostname, date, ... if
2635 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2636 and kill the old version which were in testsuite/
2637 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2641 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2642 for users. Instead, each programming environment (GRAS, MSG,...) use it
2643 when needed to provide such a feature to users.
2644 Don't shortcut the mechanism or you will also shortcut the virtualization
2645 you need on the simulator.
2649 * Cleanups in configury with regard to compile optimization/warning flags.
2650 Also add -fno-loop-optimize to any powerpc since it's the optimization
2651 killing gcc (< 3.4.0).
2652 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2653 needed by MSG examples complications
2654 * Borrow configury beautifications from PHP
2657 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2658 warning (thanks loris for stumbling into this one).
2659 * Bugfix: stop loading private headers (gras_config.h) from the public
2663 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2664 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2665 RL would imply a DNS resolver, which is *hard* to do in a portable way
2666 (and therefore delayed).
2667 * Implement a real timer mechanism and use it in timing macros. This allows
2668 to avoid rounding errors and get a 0.000005 sec precision in timing
2669 macros. While I was at it, various cleanups:
2670 - allow to declare more than one timed section per file (fix a stupid bug)
2671 - move some private declaration to the right place
2672 - merge conditional execution and timing macros into emulation module
2673 - document the module
2674 - make sure the module cleanups its mess on gras_exit
2675 * Documentation improvements:
2676 - (new) how to compile applications using GRAS
2677 - (new) emulation support (timing macros)
2679 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2681 SimGrid (2.93) unstable; urgency=low
2683 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2686 - Use Paje properly where used. Still to be sanitized properly.
2687 - Portability fix: Add an implementation of the contexts using pthread
2691 - Add xbt_procname(): returns the name of the current process.
2692 Use it to show the current process's name in all logging.
2694 - fix detection of older flex version and the reaction, since we do
2695 depend on modern ones (we use lex_destroy)
2696 - Better separation of SG and RL in the libs: remove all simulation code
2697 from libgras. As a result, this lib is now only 200k when stripped.
2698 Some of the xbt modules may also be duplicated (two sets and such) and
2699 should be cleaned/killed before SG3.
2700 - Insist on using xlC on AIX because of weird problems involving gcc there.
2701 - Cleanup the make remote stuff. This is now done by scripts
2702 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2703 mainly for our private use, but we're working on changing them to user
2706 - Bugfix: flush the socket on close only if there is some *output*.
2707 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2709 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2712 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2714 SimGrid (2.92) unstable; urgency=low
2716 Alpha 3 on the path to SimGrid 3
2720 - New! First try of benchmarking macros.
2721 - New! First try so that gras_stub_generator generate deployment and
2722 remote compilation helpers.
2724 - Bugfix: Initialization fix in msg_test.
2728 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2730 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2732 - API Change: gras_os_sleep to take the amount of seconds as a double.
2733 Accepting an int was error prone since it was the only location where
2734 seconds were coded as such. It leaded to damn rounding errors.
2735 - Bugfix: Hard to belive that timers ever worked before this.
2737 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2739 SimGrid (2.91) unstable; urgency=low
2741 Alpha 2 on the path to SimGrid 3
2745 - Bug fix in the lmm_solver.
2747 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2748 through the function MSG_paje_output.
2749 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2750 - It is possible to bound the rate of a communication in MSG with
2751 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2752 to write it in the changelog).
2753 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2758 - Various cleanups to the autotools stuff
2759 - Begin to move Gras examples to examples/gras/
2760 - Let make distcheck work again (yeah!)
2762 - documentation overhauled using doxygen.
2763 gtk-doc-tools is dead in SimGrid now.
2764 - Automatically extract all existing logging categories, and add the list
2765 to the documentation (long standing one, to say the less)
2767 - Cleanup the known architecture table. Reorder the entries to group what
2768 should be, and use a more consistent naming scheme.
2769 (some of the test dataset are still to be regenerated)
2770 - New! Allow library to register globals on each process just as userdata
2772 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2773 lookup time (for now).
2774 Use it in msg and trp.
2775 This cleans a lot the internals and helps enforcing privacy of the
2776 headers between the gras components.
2777 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2778 - Bugfix: gras_os_time was delirious in RL.
2779 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2781 - Reenable GRAS now that it works.
2783 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2785 SimGrid (2.90) unstable; urgency=low
2787 Alpha 1 on the path to SimGrid 3
2789 * It is a long time since the last release of SimGrid. I'm sorry about
2790 that but as I had told you, I was rewriting a lot of things. I apologize
2791 to those who had been reporting bugs to me and that I had not answered.
2792 If your bug is still in the new version, please tell me. Here is a
2793 summary of the main changes.
2795 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2796 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2797 portability, speed, and a lot more but you'll figure it out later.
2798 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2799 I wanted to release the new MSG as soon as possible and I have
2800 broken GRAS, which is the reason why, for now, only MSG is fully
2801 functional. A laconic description of these projects is available
2802 in the documentation.
2804 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2805 kernel optimized for our needs (called SURF but only the developers
2806 should use it). Hence, MSG is now roughly 30 times faster and I think
2807 that by rewriting a little bit MSG, I could event speed it up a little
2808 bit more. Beside the gain in speed, it is also much easier to encode a
2809 new platform model with SURF than it was with SG. More to come...
2811 * REVOLUTION 3: I have tried to change a little as possible the API of
2812 MSG but a few things really had to disappear. The main differences
2813 with the previous version are :
2814 1) no more m_links_t and the corresponding functions. Platforms are
2815 directly read from a XML description and cannot be hard-coded
2816 anymore. The same format is used for application deployment
2817 description. The new format is described in the documentation.
2818 Have a look in tools/platform_generation. There is a tiny script
2819 that converts from the old platform format to the new one. Concerning
2820 the application deployment format, parsing the old one is tricky.
2821 I think most of you should however be able to convert your files. If
2822 it is really an issue, I can write a C code that does the conversion.
2824 2) the toolbox tbx does not exist anymore. We now have a library
2825 with much more data-structures but without the hash-tables (we have
2826 dictionaries that are much faster).
2828 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2830 *****************************************************************************
2831 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2832 * don't feel like converting the oldies. (Mt) *
2833 *****************************************************************************
2836 Version 2.90: "the long awaited one"
2837 - Finished rewriting and debugging MSG. Rewrote the documentation.
2838 - disable GRAS for now since it needs to be ported to the newest SG
2841 - Finish the port to windows (using mingw32 for cross-compile)
2844 - Main loop and datastructures of SURF. A cpu resource object is
2845 functional. Surf can thus be used to create cpu's with variable
2846 performance on which you can execute some actions.
2848 2004-11-15 Martin Quinson
2849 - Port to ARM. Simply added the alignment and size descriptions. Should
2850 work, but the ARM machines are so slow that I didn't had the opportunity
2851 to 'make check' over there yet.
2853 2004-11-15 Arnaud Legrand
2854 - Trace manager now written. It uses a heap structure and is therefore
2855 expected to be efficient. It may however be speeded up (particularly
2856 when many events occur at the same date) by using red and black
2857 trees. One day maybe...
2858 - Max-min linear system solver written. It uses a sparse matrix
2859 structure taking advantage of its expected use. Most operations are
2860 O(1) and free/calloc are called as few as possible. The computation of
2861 the minimum could however be improved by using a red and black tree
2864 2004-11-03 Arnaud Legrand
2865 - Rename every gras_* function that was in xbt/ to its xbt_
2867 - Add a heap and a doubly-linked list to xbt
2868 - Added a dichotomy to the dictionaries. make check works as well before
2869 so I assume that the patch is correct. I do not know however if things
2870 run effectively faster than before now. :)
2872 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2873 SimGrid, and the version number is bumped to 2.x
2875 2004-10-29 Martin Quinson
2876 - Introduction of the remote errors.
2877 They are the result of a RMI/RPC on the remote machine.
2878 ErrCodes being scalar values, you can't get the host on which those
2879 errors did happen. Extending the error mechanism as in Gnome is possible.
2880 No idea yet whether it is a good idea.
2882 2004-10-28 Martin Quinson
2883 - Interface revolution: the Starred Structure Eradication.
2884 I used to do typedef struct {} toto_t; and then handle *toto_t.
2885 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2886 - ???_t is a valid type (builded with typedef)
2887 - s_toto_t is a structure (access to fields with .)
2888 - s_toto is a structure needing 'struct' keyword to be used
2889 - e_toto_t is an enum
2890 - toto_t is an 'object' (struct*)
2892 typedef struct s_toto {} s_toto_t, *toto_t;
2893 typedef enum {} e_toto_t;
2894 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2895 s_toto_t) is private.
2897 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2898 so that it can set the variable to NULL. It was so for dicts and sets,
2899 it changed for dynars.
2901 - Fix a bunch of memleaks in dict_remove
2902 - Fix a bug in sg/server_socket opening: it failed all the time.
2904 2004-10-07 Martin Quinson
2905 - Speed up dynar lookup operation a bit.
2907 gras_dynar_get is dead.
2909 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2910 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2911 (which gives you the address of the stored data).
2913 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2914 copy of the data using an affectation to do the job and not a memcpy.
2916 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2917 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2919 It does not really speedup the dynar test because they are
2920 setting elements all the time (and look them seldom). But the dict does
2921 far more lookup than setting.
2923 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2925 2004-10-05 Martin Quinson
2926 - Allow to (en/dis)able the cycle detection at run time.
2928 Whether we should check for cycle or not is now a property of each
2929 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2930 datadesc_cycle_unset allow to remove this property when previously set.
2932 Note that the cycle detection is off by default since it impacts the
2933 performance. Watch the data you feed GRAS with ;)
2935 This property is hereditary. Any element embedded in a structure having it
2936 set have it set for the time of this data exchange.
2938 You should set it both on sender and receiver side. If you don't set it on
2939 sender side, it will enter an endless loop. If you forget on receiver
2940 side, the cycles won't be recreated after communication.
2942 - Header reorganization.
2943 Kill gras_private.h, each submodule must load the headers it needs.
2945 2004-10-04 Martin Quinson
2946 - Interface revolution: do not try to survive to malloc failure.
2948 Now, gras_malloc and friends call gras_abort() on failure.
2949 As a conclusion, malloc_error is not a valid error anymore, and all
2950 functions for which it was the only gras_error_t return value are
2951 changed. They now return void, or there result directly.
2952 This simplify the API a lot.
2954 2004-09-29 Martin Quinson
2955 - Re-enable raw sockets.
2956 Created by gras_socket_{client,server}_ext;
2957 Used with gras_raw_{send,recv}
2960 It should allow to kill the last bits of gras first version soon.
2962 This is not completely satisfactory yet (duplicate code with
2963 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2966 - Simplify transport plugin (internal) interface by not passing any
2967 argument to _server and _client, but embedding them in the socket
2970 2004-09-28 Martin Quinson
2971 - Finish the port to AIX.
2972 autoconf was my problem (segfault within the malloc replacement
2973 function. No idea why)
2975 2004-09-16 Martin Quinson
2976 - Fix some size_t madness on 64bit architectures.
2978 2004-09-08 Martin Quinson
2979 - Reduce the number of system headers loaded, overload some more system
2980 calls (such as malloc to cast the result of the system one, and work
2982 - Fix and reintroduce the config support
2984 2004-09-07 Martin Quinson
2985 - Source code reorganization to allow Arnaud to surf all over there.
2986 - Allow to document the logging categories.
2987 - Remove all uppercase from logging categories and useless cleanup in names.
2989 2004-08-18 Martin Quinson
2990 Version 0.6.2 (protocol not changed; API changed)
2991 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2992 gras_error_t), and NULL when not found. Functions expecting a msgtype
2993 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2994 hopefully usefull message.
2995 - Portability to prehistoric sparcs again
2997 2004-08-17 Martin Quinson
2998 Version 0.6.1 (protocol not changed; ABI not changed)
2999 - prealloc some buffers to speed things up
3001 2004-08-11 Martin Quinson
3002 Version 0.6 (protocol not changed; ABI expended)
3003 - The parsing macro can deal with the references, provided that you add
3004 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3006 2004-08-09 Martin Quinson
3007 Version 0.5 (protocol not changed; ABI changed)
3008 - Allow to off turn the cycle detection code in data exchange at
3009 compilation time. It should be at run time, but I'm short of time (and
3010 the config stuff is still broken). That way, we keep dict out of the
3011 critical path, which is good because the performance is poor:
3012 - search not dichotomial yet
3013 - dynar give no way to access their content and memcpy everytime
3014 - In composed data description (struct, ref and so on), stop foolness of
3015 keeping the subtype's ID, but store the type itself. This keeps sets out
3016 of the critical path, which is good since they rely on dynar and
3017 dictionnaries. The only loose of that is that we cannot detect the
3018 redeclaration of a structure/union with another content (but I'm not sure
3019 the code detected well this error before anyway). We still can detect
3020 the redefinition discrepancy for the other types.
3021 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3022 it breaks the code because of type-punning used all over the place).
3023 This breaks on all non-gcc architectures (for now).
3025 All those changes (plus the buffer of last time) allow me to gain 2 order
3026 of magnitude on cruel tests consisting of 800000 array of integers on two
3027 level of a hierarchical structure (200 secondes -> 4 secondes)
3030 - the selector of reference must now return the type it points to, not
3031 the ID of this type.
3033 2004-08-06 Martin Quinson
3034 Version 0.4 (protocol changed; ABI not changed)
3035 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3036 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3037 - Implement a buffer transport, and use it by default (it relies on tcp in
3038 real life and on sg in simulation).
3039 That's a bit hackish since I had a new field to the structure to store
3040 its data without interfering with the subtype ones. Inheritance
3041 is tricky in C. And that's a kind of reverse inheritance with one class
3042 derivating two classes. Or maybe a game with java interfaces. Anyway,
3043 that's damn hard in C (at least).
3044 Moreover, I got tired while trying to ensure plugin separation and
3045 genericity in SG mode. MSG wants me to do weird things, so let's go for
3046 cruel hacks (temporarily of course ;).
3047 See comment in transport_private.h:71
3048 - do not include all the _interface headers in private but in the files
3049 which really need them (to cut the compilation time when they are
3052 2004-07-26 Martin Quinson
3053 Version 0.3 (protocol not changed; ABI changed)
3054 - Major overhault of the datadesc interface to simplify it:
3055 - shorted the function names:
3056 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3057 - add a trivial way to push/pop integers into the cbps without malloc.
3058 This allows to make really generic sub_type description, which simply
3059 pop their size of the stack.
3060 - add a function gras_datadesc_ref_pop_arr() which does what users want
3061 most of the time: Declare a dynamic array (which pops its size of the
3062 stack) and declare a reference to it. Poor name, but anyway.
3063 - kill the post-send callback, add a post-receive one
3065 2004-07-23 Martin Quinson
3066 Version 0.2 (protocol changed; ABI changed)
3067 - add some testing for cpbs in the test cases, and fix some more bugs.
3068 This invalidate again the little64 data file, since I cannot regenerate
3070 - remove an awfull optimization in the logging stuff, allowing me to:
3071 - understand it again
3072 - learn gcc how to check that the argument match the provided format
3073 - fix all errors revealed by gcc after that
3074 - internal keys of dict are not \0 terminated. Deal with it properly in
3075 loggings instead of segfaulting when the user want to see the logs :-/
3077 2004-07-22 Martin Quinson
3078 - Fix some stupid bug preventing cbps (callback postit) from working
3080 2004-07-21 Martin Quinson
3081 - Some documentation cleanups
3082 - remove the useless last argument of msgtype_declare
3083 - rename the Virtu functions to fit into the 'os' namespace
3084 - move headers src/include -> src/include/gras/ and stop fooling with
3085 gras -> . symbolic link
3086 - make distcheck is now successful
3088 2004-07-19 Martin Quinson
3090 - Build shared library also
3091 - Install html doc to the right location
3092 - stop removing maintainer files in make clean
3093 - build tests only on make check
3095 2004-07-13 Martin Quinson
3097 - No major issue in previous version => change versioning schema
3098 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3099 corresponding dataset.
3101 2004-07-11 Martin Quinson
3103 - Get it working with any kind of structure (we can compute the padding
3104 bytes remotely for all the architectures I have access to)
3105 - Implement the structure parsing macro (still not quite robust/complete)
3106 - Improvement to the remote testing toysuite
3108 2004-07-10 Martin Quinson
3109 [autoconf mechanism]
3110 - get ride of a bunch of deprecated macros
3111 - actually run the test for two-compliment, not only compile it :-/
3112 - test whether the structures get packed (and bail out if yes. Damn.
3113 Alignment is a serious matter)
3114 - test whether the structures get compacted (but respecting the alignment
3115 constraints of each types)
3116 - test whether the array fields of structures can straddle alignment boundaries
3118 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3119 (took me 2 hours to find that bug, looking at the wrong place)
3120 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3121 before sending/receiving and are used to compute the offsets of fields
3122 - Given that padding size depend even on compiler options, keep track of
3123 alignment and aligned_size only for the current architecture. Not a big
3124 deal since we send structure fields one after the other (seems
3126 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3127 program, let it work on linux/gcc/little32. portability todo.
3129 2004-07-08 Martin Quinson
3130 - import and improve remote compilation support from FAST
3131 - make sure make check works on half a dozen of machines out there
3133 2004-07-07 Martin Quinson
3134 Let's say it's version 0.0.3 ;)
3135 - Implement conversions (yuhu!)
3136 - Let it work on solaris (beside conversion, of course)
3137 - Stupid me, using rand() to generate the conversion datatests in not wise.
3139 2004-07-06 Martin Quinson
3140 - Let make dist work, since I'm gonna need it to compile on remote hosts
3141 - Let Tests/datadesc_usage write the architecture on which the file was
3142 generated as first byte.
3143 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3144 architecture descriptions.
3145 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3146 on those architectures.
3147 - Optimization: send/recv array of scalar in one shoot
3149 2004-07-05 Martin Quinson
3150 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3152 - Plug a whole bunch of memleaks
3153 - each process now have to call gras_{init,exit}. One day, their log
3154 settings will be separated
3155 - Continue the code factorisation between SG, RL and common in Transport.
3157 2004-07-04 Martin Quinson
3159 - Redistribution between SG and RL.
3160 We wanna have to accept in SG, so move accepted related parts of RL in
3161 the common part. (more precisely, the dynar of all known sockets is no
3162 more a static in transport.c, but part of the process_data)
3164 [gras_stub_generator]
3165 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3167 2004-07-03 Martin Quinson
3168 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3169 them all up in one shot)
3171 - Fix the ugly case of reference to dynamic array.
3172 - New (semi-public) function gras_datadesc_size to allow the messaging
3173 layer to malloc the needed space for the buffer.
3175 - gras_socket_close now expect the socket to close (and not its address to
3176 put NULL in it after it). This is because the socket passed to handlers
3177 is one of their argument (=> not writable).
3179 - propagate the interface cleanup from last week in datadesc, ie remove a
3180 superfluous level of indirection. User pass adress of variable
3181 containing data (both when sending and receiving), and not of a variable
3182 being a pointer to the data. Let's say that I like it better ;)
3183 The price for that is constructs like "int msg=*(int*)payload" in
3184 handlers, but it's a fine price, IMHO.
3186 - Let it work in RL (yuhu)
3188 2004-06-21 Martin Quinson
3190 - porting SG plugin and SG select to new standards (works almost).
3191 - plug memleaks and fix bugs around.
3194 - cleanup the prototype of data recv and force users to specify when they
3195 want to handle references to objects. Test case working even for cycles.
3196 - plug memleaks. Valgrind is perfectly ok with this.
3198 2004-06-12 Martin Quinson
3200 - cleanup the separation between plugin and main code in plugin creation
3202 2004-06-11 Martin Quinson
3204 - Reput hook for raw sockets, needed for BW experiments
3205 - kill a few lines of dead code
3206 [Data description] Interface cleanup
3207 - gras_datadesc_by_name returns the searched type or NULL.
3208 That way, no variable is needed to use a type desc once, which makes
3210 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3211 two parameters were strings (field name, type name), leading to
3213 [Dicos] Interface cleanup
3214 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3215 This is consistant with the dynar API.
3217 2004-04-21 Martin Quinson
3219 - Porting to new standards.
3221 - interface cleanup.
3222 There is no bag anymore, no need to take extra provision to mask the
3223 pointers behind "ID".
3224 Better splitup of functions between files create/exchange/convert.
3225 This is still a bit artificial since convert and receive are so
3226 interleaved, but anyway.
3228 - add a queued message list to procdata (the ones not matching criteria
3230 - factorize some more code between SG and RL wrt procdata
3232 - use gras_exit in example to track memleaks
3233 - get rid of gs_example now that GS is properly integrated into gras
3234 - update run_test to integrate the latest tests (datadesc)
3236 - rename WARNINGn macros to WARNn since it prooved error-prone
3238 2004-04-19 Martin Quinson
3240 - register init/exit functions within gras module mechanism
3241 - send/receive function.
3242 Convertion is not implemented, but short-cutted if not needed.
3243 struct/array elements are sent one by one (instead of block-wise), but
3244 nobody really cares (yet). Get a prototype before optimizing.
3245 - tests (using a file socket) for DD send/receive on:
3246 - base types: int, float
3247 - array: fixed size, string (ie ref to dynamic string)
3248 - structure: homogeneous, heterogeneous
3249 - chained list, graph with cycle
3250 Believe it or not, valgrind is not too unhappy with the results. The
3251 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3255 - Bugfix when using a filename explicitely (instead of '-')
3257 2004-04-09 Martin Quinson
3259 - factorize more code between RL and SG in socket creation
3260 - Complete the implementation and tests of:
3262 o file (only in RL, and mainly for debugging)
3264 I lost 3 days to design a portable address resolver, and then decided
3265 that the prototype mainly have to run on my box.
3266 Addressing portability too early may be like optimizing too early :-/
3268 - use gras_init in the Tests instead of the crappy parse_log_opt
3269 (the latter function is removed)
3270 [Conditional execution]
3271 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3272 [Code reorganisation]
3273 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3274 Build examples against the RL library, since there is no way to disable
3275 its creation for now.
3277 For information, the beginning of coding on GRAS was back in june
3278 2003. I guess that every line has been rewritten at least twice since