1 SimGrid (3.15) UNRELEASED; urgency=low
3 - Rename NetCards to NetPoints.
4 This was intended to help NS3 users, but that's not a netcard.
5 That's a point in the routing algorithm, let's avoid wrong simplifications.
8 - New function: MSG_process_yield()
9 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
10 to make room for the upcoming network energy plugin.
13 - Ensure that an actor can kill itself with Process::exit()
14 - Kill the obscure NativeException. Nobody want to survive the issues
15 it denotes, so use JniException that is a RuntimeException (not to
16 be catched explicitely).
20 - s4u::onPlatformCreated: right before the simulation starts
21 - s4u::onSimulationEnd: right after the main simulation loop
22 - s4u::onTimeAdvance: right after a clock change (time is discrete)
23 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
24 event from the availability_file changes the avail speed.
25 - Links are now usable from s4u
28 - Kill the fifo data container: we don't use it anymore.
30 -- target_date=March 20 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
32 SimGrid (3.14.159) stable; urgency=low
34 The previous release was only a crude approximation of a Pi release;
35 this one should be a bit better.
37 - Revert a last minute change that broke on Mac OSX.
38 - Fix the changelog and NEWS files.
39 - Improve the documentation, in particular of the routing module.
40 - Integrate some patches flying around in Debian.
42 -- Wed Dec 28 17:14:49 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
44 SimGrid (3.14) stable; urgency=low
47 * The whole documentation was reworked and reorganized. There is
48 still much room for improvement here, but we're on it.
51 * We now need python3 for our testing infrastructure.
52 * Model checking needs libevent
55 * Privatization is now activated by default in smpirun
56 You can turn it off with -no-privatize if something goes wrong.
58 * Call-location tracing for SMPI.
59 You can add the exact location (filename / linenumber) of an MPI call to
60 your trace files and slow down or speed up the simulation between two
61 consecutive calls by using an adjustment file (see the documentation).
63 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
64 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
65 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
66 values were ignored in some cases. The timings of these functions can now
67 be significantly different.
69 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
70 smpi/running-power is renamed to smpi/host-speed
72 * smpi/grow-injected-times option to enable or disable multiplication of the
73 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
74 make simulation less precise (but also much faster).
76 * smpirun script should be (much) faster for large deployments.
78 * SMPI tracing : fixed issue with poor matching of send/receives.
80 * Replay : Fix broken waitall
82 New functions and features
83 * MSG_parallel_task_execute_with_timeout, to timeout computations.
85 Dropped / renamed functions and features
86 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
87 - MSG_mailbox_is_empty() -> Mailbox::empty()
88 - MSG_mailbox_front() -> Mailbox::front()
89 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
90 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
91 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
92 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
93 * Don't pass the free_f parameter to property related functions:
94 - MSG_host_set_property_value()
95 - MSG_as_router_set_property_value()
96 - MSG_storage_set_property_value()
97 * VM properties. Since msg_vm_t are msg_host_t, just use
98 MSG_host_get_property_value() and friends
99 * VM I/O related things:
100 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
101 - Unimplemented save/restore methods
102 * MSG_as_router_get_property_value() was redundent with
103 MSG_environment_as_get_property_value().
104 - Removed MSG_as_router_*propert*() functions
105 - Added MSG_environment_as_set_property_value() for consistency
106 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
107 Well, they are still there for now, but deprecated with a warning.
108 Please switch to xbt_dict_new_homogeneous() before this is removed
110 * Task affinity. Its intended behavior (that was very badly tested
111 and probably not really working) was deceiving what most users
112 would have hoped here.
113 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
114 * Option network/coordinates is now useless and should be dropped.
117 * Added option storage/max_file_descriptors to allow more than 1024 files opened
120 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
121 * Added Dragonfly topology support
124 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
125 EiBps, EBps, Eibps, Ebps for bandwidth.
126 They may become useful to some lucky ones.
129 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
130 * ex/app_tokenring: new example, very similar to the MSG Token Ring
131 * ex/async_waitAll: new example, on asynchronous communications
134 * Memory usage should be decreased for simulations with a large number
135 of processes. This also helps for SMPI.
137 -- Sat Dec 24 22:34:26 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
139 SimGrid (3.13) stable; urgency=low
141 The Half Release, a.k.a. the Zealous Easter Trim.
143 - We removed half of the lines, that were mostly experimental cruft.
144 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
145 (not counting blanks and comments -- according to openhub.net)
146 - The internals are now compiled in C++ (and will soon be clean C++)
147 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
148 59 klines of C, etc. We added only 29 klines of C++ in replacement.
150 * Backwards Compatibility breaks
151 - Removed Lua simulation bindings (switch to C or Java for that).
152 Lua can still be used to describe platforms
153 - Removed Java kernel plug-ins.
154 Will be reintroduced after the ongoing major internals reorg.
156 - the following functions were removed.
157 They were too specific and should be reimplemented in a generic
158 way, with filter function.
159 - MSG_task_listen_from_host
160 - MSG_mailbox_get_count_host_waiting_tasks
161 - MSG_mailbox_put_with_timeout was removed.
162 Please use MSG_task_send_with_timeout instead.
164 - the SD_application_reinit function was removed. It has been a noop for a while.
165 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
166 replaced by a more flexible mechanism.
167 - The following functions thus do not exist anymore
168 - SD_workstation_get_access_mode
169 - SD_workstation_set_access_mode
170 - SD_workstation_get_current_task
171 - Basic estimation functions have been removed but can easily be replaced
172 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
173 - SD_workstation_get_computation_time => amount / sg_host_speed()
175 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
176 Use VM.getSpeed()*load/100 for the legacy behavior.
178 - option enable_tracing was removed. It was not doing anything for a while.
179 - In the ModelChecker:
180 - the model-checker now ptraces the model-checked process which means
181 you cannot use a debugger on the latter anymore (we might make this
182 optional in the feature);
183 - removed soft-dirty page tracking;
184 - remove model-checked side snapshot management,
185 MC_snapshot() and MC_compare_snapshot();
186 - keep the MC_cut() function as a stub (it was not really working
187 in the previous release).
190 * All options are consistently in kebab-case. Old names are kept as alias.
193 * Switch to platform v4 format.
194 - Rename from 'power' to 'speed' the attributes describing the amount of
195 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
196 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
197 - In <host> and <link>, attributes availability and state are gone.
198 It was redundent with state and availability traces, and with peak values.
199 - In <cluster>, attributes availability_file and state_file are gone.
200 It was too complex and unused.
201 - Kill <gpu>. Was not doing anything.
202 - The DOCTYPE points to the right URL:
203 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
204 (the file at this address now documents the changelog since its v1)
205 - A warning is emitted for unit-less values (they are still accepted).
206 - speed. Default: 'f' or 'flops'. Also defined:
207 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
208 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
209 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
210 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
211 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
212 - latency. Default: 's' second. Also defined:
213 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
215 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
216 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
219 * s4u::Host is now the preferred public interface to the Host features.
220 sg_host_* functions are C bindings to the exact same behavior
221 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
224 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
225 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
226 They were deprecated since a few years, and probably did not even compile anymore.
229 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
230 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
231 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
232 encouraged to update their codes. A list of the modified functions can be found at the end of
233 include/simgrid/simdag.h
236 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
237 the lambda or closure passed as a parameter will run in kernel mode.
238 All the callback functions should be rewritten to that interface at some point.
241 * Reorganizing and cleaning the internals all around the place.
244 * Remove old default barrier/bcast buggy algorithms (see #18407)
245 * Various bug fixes to handle more codes
246 * Remove the need for the --foreground option of smpirun (it is still
247 accepted for backward compatibility).
250 * Kill the setset data container: MC don't use it anymore.
251 * Kill the queue data container: it made more sense with GRAS.
252 * Kill the xbt_peer_t data type: it's useless without GRAS.
253 * Kill rm_cb feature of config sets: it was never useful.
254 * Kill graphxml parsing feature. It was not used.
255 * Kill the deprecated code protected by XBT_USE_DEPRECATED
257 - xbt_dynar_sort_strings(), when the content is char*
258 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
259 They throw exceptions on invalid input;
260 * C++ support for declaring CLI flags (simgrid::config::Flag);
261 * class for abstracting different signal backends (simgrid::xbt::signal).
262 with no external dependencies (we need very simple signals).
265 * refactoring and cleanup of the code;
266 * ongoing process to cleanly separate the model-checking algorithms
267 from the code model-checking support.
269 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
271 SimGrid (3.12) stable; urgency=low
273 The Facelift Release.
276 * Require g++ v4.7 at least to not speak prehistorical C++.
277 * Require Boost 1.48 (for signal2 component).
278 * Java must be version 7 at least when activated.
279 * Builds on Windows again (including Java bindings).
280 * Tracing is now always enabled (no way to turn it out)
281 * Remove GTNetS. It was not working anyway.
282 * Various cleanups in the cmake scripts.
283 * Move headers around to sort them out on installed systems:
284 - instr/instr.h -> simgrid/instr.h
285 - instr/jedule/* -> simgrid/jedule
286 - simdag/datatypes.h was removed
287 - simdag/simdag.h -> simgrid/simdag.h
288 - msg/datatypes.h was removed
289 - msg/msg.h -> simgrid/msg.h
292 * Interface improvement:
293 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
294 - Sanitize the interface in MSG_task_ module:
295 - Merge two functions that were close enough but misleading:
296 set_compute_duration(t) -> set_flops_amount(t)
297 get_remaining_computation(t) -> get_flops_amount(t)
298 - set_data_size(t) -> set_bytes_amount(t)
299 get_data_size(t) -> get_bytes_amount(t)
300 - Massive cleanups in the functions related to the energy
301 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
302 - New: MSG_host_get_pstate()
304 - msg/energy/onoff: switching hosts on and off
306 * Interface improvement:
307 - Rename Host.isAvail() to Host.isOn()
308 - Rename Process.currentProcess() to Process.getCurrentProcess()
309 - Rename Task.setDataSize() to Task.setBytesAmount()
310 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
312 - #18874: Actually allows the GC to reclaim tasks
316 - SIMIX_process_throw: raises an exception in a remote process
317 * Refactoring: Separate sub-modules
318 - libsmx: the public interface, as libc in a real system
319 - popping: the strange dance that converts a user request into a kernel handling
320 - smx_context_*: the virtualization mechanisms that embed the user code
321 - smx_*: the handling of each simcalls
323 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
324 * Rename smx_action_t into smx_synchro_t, making explicit that these
325 things are used to synchronize processes with their environment.
326 For example, a communication is a sort of synchronization involving
327 the communicating processes (that may block until the exchange) and
328 the platform. The same can be said from computations, etc.
330 - #18888: segfault when a process ends before its kill_time
334 - 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.
335 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
336 - MPI_Type_set_name, MPI_Type_get_name
337 - MPI_*_c2f and MPI_*_f2c functions
338 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
339 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
340 - Activate a lot of new tests from the mpich 3 testsuite
342 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
343 - InfiniBand network model added : Based on the works of Jerome Vienne
344 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
345 - When smpi/display_timing is set, also display global simulation time and application times
346 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
347 * Collective communications
348 - 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
349 - 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/).
350 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
352 - Replay now uses algorithms from wanted collective selector
353 - Replay can be used with SMP-aware algorithms
354 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
355 - 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.
357 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
358 - Correctly use loopback on fat-tree clusters
359 - Asynchronous small messages shouldn't trigger deadlocks anymore
360 * Energy/DVFS cleanup and improvement
361 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
362 - new: smpi_get_host_pstate()
365 - "Full" network optimization flag was broken since Surf++
366 - Better handling of precision flags in maxmin
367 - Fix bug causing sometimes "Impossible" errors
368 - Properly pass cluster properties to included hosts
369 * Improvement of the Energy plugin.
370 - Always update the consumption before returning that value
371 - New property: watt_off to denote the dissipation when the host is off
372 - New functions getWattMinAt and getWattMaxAt to retrieve the
373 dissipation of pstates that we are not currently at.
374 * Java: class NetworkLink renamed to Link
375 * New function: simcall_process_get_kill_time()
376 * Massive rename s/workstation/host/
377 - That's intrusive, but that's good for the project consistency. Sorry.
378 - Change config option "workstation/model" into "host/model"
381 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
382 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
383 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
384 - xbt_dirname and xbt_basename for non-POSIX systems
386 * The model checker now runs as a separate process.
387 * The model checker runs must now be launched with the new simgrid-mc program.
388 * Record/Replay: the MC can display a textual representation of a path in the
389 execution graph. It can then be replayed outside of the model checker.
391 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
393 SimGrid (3.11) stable; urgency=low
398 * Normalizing pointers addresses tool for better diff between logs
401 * Add cloud examples using new VMs
402 - examples/msg/cloud/two_tasks_vm.tesh
403 - examples/msg/cloud/simple_vm.tesh
404 - examples/java/cloud/cloud.tesh
405 - examples/java/cloud/migration/migration.tesh
406 * Add java surf examples:
407 - examples/java/surfPlugin/surf_plugin.tesh
408 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
409 - examples/java/surfCpuModel/surf_cpu_model.tesh
410 * Add SMPI+MSG example:
411 - examples/smpi/smpi_msg_masterslave/
416 - msg task destroy cancel test
417 - msg_host on/off test
418 * Move all tests in testsuite to teshsuite (adding tesh files)
419 * Restructure teshsuites
420 - one folder for each kind of test
421 * Restructure AddTests.cmake
423 - structure the order of tests (with sections)
426 * Add virtual machine
427 - creation of a VM on a PM
428 - migration of a VM from a PM to another PM
430 - MSG_process_join(msg_process_t process, double timeout)
431 - msg_bar_t MSG_barrier_init(unsigned int count)
432 - int MSG_barrier_wait(msg_bar_t barrier)
433 - void MSG_barrier_destroy(msg_bar_t barrier)
434 - msg_as_t MSG_environment_as_get_by_name(const char * name)
435 * New option "msg/debug_multiple_use" to help debugging when a task is used
438 - Improvements and finalization of MSG_storage, MSG_file APIs and their
440 - Increase code coverage in test suites
444 * Protect context stacks against stack overflow. The number of protected memory
445 pages allocated on the top of each stack (1 by default) can be configured
446 with the new command line option --cfg=contexts/guard_size:<value>.
447 * Simcalls are now generated by a python script that
448 - generates files included by SimGrid
449 - checks that all the functions exist, and proposes prototypes
451 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
452 host_set_data, host_get_data
454 - simcall_process_join(smx_process_t process, double timeout)
455 * Fix bug where sleeping processing could not be suspended.
458 * Translate surf models from C to C++
459 - Generic classes for all models: Model, Resource, Action
460 - A generic interface for each kind of model (CPU, Network, Storage
461 Workstation, WorkstationVM)
463 * Translate surf routings from C to C++
464 * Add callbacks using sigc++ or boost::signals2
465 - Add callback functions for resource creation/destruction
466 - Add callback functions for action state change
467 - Handle Energy as a plugin
468 * Replace swag by boost::intrusive
469 * Add new routing models for clusters. For documentation, see
470 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
471 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
472 parameters for cluster tag
473 - Fat trees, with topology="FAT_TREE" and
474 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
475 - see examples/platforms/torus_cluster.xml and
476 examples/platforms/fat_tree_cluster.xml
480 * Hostfiles support host:nb_processes construct to deploy several processes on
482 * Collective communication algorithms should not crash if used with
483 improper number of nodes and report the error.
484 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
485 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
486 MPI_Dims_create, MPI_Cart_sub are supported.
487 * New interface to use SMPI programmatically (still depends on MSG for
488 some parts, see examples/smpi/smpi_msg_masterslave) :
489 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
493 * Global variables privatization in MPI executables is now performed at runtime
494 with the option smpi/privatize_global_variables (default:no).
495 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
496 dynamic libraries loaded by the application are not privatized (static
497 linking with these libraries is advised in this case)
500 * Options defined in XML work correctly now.
503 * New cmake option, enable_lib_in_jar, to control whether native libraries are
504 copied into simgrid.jar or not (ON by default). Use this option if you want
505 to reduce the size of the installed simgrid.jar, *and* the native libraries
506 are kept installed elsewhere.
507 * Surf binding with SWIG (code generated in maintainer mode only):
508 - plugin to handle callbacks
509 - CPU model only for the moment
512 * Supernovae build mode is definitively removed. It was used to improve
513 inlining and inter-module optimizations. It is nowadays superseded by
514 link-time optimizations commonly available in compilers.
515 * Update ns-3 find lib. Bindings for ns-3 should work again now.
516 * Add boost dependency for surf++
517 * Add new macro for tests
518 - ADD_TESH(name <tesh_args>)
519 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
523 - xbt_bar_t XBT_barrier_init(unsigned int count)
524 - int XBT_barrier_wait(xbt_bar_t barrier)
525 - void XBT_barrier_destroy(xbt_bar_t barrier)
526 * Make the xbt_os_time module public
528 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
530 SimGrid (3.10) stable; urgency=low
532 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
535 * Reintegrate Java to the main archive as desynchronizing these
536 package is not acceptable anymore (Java is now considered stable)
537 * Add explicit synchronization facilities through semaphores
538 * Bug fix: Task.setDataSize() only changed the C world, not the value
539 cached in the Java world
542 * Dramatically change the way files are handled. API and internals changed, but
543 this part of MSG was not considered as production grade either.
544 * Add explicit synchronization facilities through semaphores
545 * Add a new function MSG_host_get_process_list()
546 * Preliminary DVFS support (see examples/msg/energy/ for details)
549 * SMPI is now included directly in the libsimgrid as the windows
550 linker doesn't force us on splitting it anymore.
551 * Improvements of the SMPI replay tool:
552 - Most of the collective communications are now rooted in the same process as
553 in the original application.
554 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
555 used until now). Multiple data types can now be used in a trace.
556 - The replay tool now supports traces produce either by TAU or a modified
558 - Bug Fix: the compute part of the reduce action is now taken into account.
559 - Gatherv collective is now supported
560 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
561 of smpirun outputs time independent traces for the current run. One file
562 is created per process. If too many processes are simulated, this behavior
563 can be changed to one file for all processes by using the
564 tracing/smpi/format/ti_one_file flag
565 * smpirun generates the host file if needed (with given host count and platform)
566 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
567 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
568 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
569 --cfg=smpi/coll_selector:(mpich/ompi)
570 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
572 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
573 happening inside a collective SMPI call.
574 * Fix the behavior of complex data types handling.
575 * Make MPI_Wtime another synchronization point to take computations into
577 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
578 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
579 * Add all missing Fortran bindings, SMPI should work with Fortran 90
580 (no privatization of global variables yet)
581 * Preliminary DVFS support (see examples/smpi/energy/ for details)
584 * Verification of liveness properties is now available for SMPI applications
585 (in addition to MSG applications)
586 * Bugged examples using SMPI in examples/smpi/mc/
587 * Add --cfg=model-check/visited option. Allows the verification of infinite
588 programs. Detection of loops in the execution thanks to the system state
589 comparison and reduction of the state space to explore. Can be combined with
590 DPOR for safety properties.
593 * Allow to change SimGrid configuration (see --help) within the code
594 thanks to SD_config() as it can be done in MSG.
595 * Add a new function SD_task_set_amount() upon user request.
598 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
599 * Remove rule based routing (no more PCRE dependency)
600 * Add a limiter_link option to cluster tag, to specify a maximum reachable
601 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
602 * Add a loopback_bw and loopback_lat options to cluster tag.
603 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
604 routing. Now peers are encapsulated in an AS and have their own private
605 router but this is transparent.
608 * Our own implementation of getline is renamed xbt_getline, and gets
609 used even if the OS provide a getline(). This should reduce the
610 configuration complexity by using the same code on all platforms.
611 * New type: xbt_cfg_elm_boolean.
612 * Allow to use yes/no for boolean configuration options in the command line.
613 * Allow to disable SimGrid cleanups at exit from command line option.
614 There are situations where one may want a simulation to end with an exit.
615 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
616 annoying when scripting around the simulator. Adding a
617 --cfg=clean_atexit:no allows to circumvent this issue.
620 * Lots of memory leaks were corrected in this release.
621 There is no known memory leaks anymore, in all of our 600+ tests.
622 * New command line option --version, to get SimGrid version information.
623 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
625 * Supernovae builds are deprecated, and expected to be removed in the next
628 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
630 SimGrid (3.9) stable; urgency=low
632 The Grasgory release: GRAS is really dead now.
634 * Complete overhaul of the internal host structures scheme.
637 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
638 was considered as experimental and badly maintained since then.
639 * Keeping it was thus a trap to our potential users, that could take
640 it instead of MSG or SMPI by mistake despite is pity state.
641 * GRAS seems to have very few users (if any), and no one volunteered
642 to maintain it further. It also induces a lot of XBT code (for
643 portability sake), that must be maintained too.
644 * For all these reasons, we killed GRAS. If someone wants to revive it
645 in the future, don't cry, our git history still remembers of GRAS.
648 * Major overhaul. Merge our documentation again as time proved that
649 splitting it was really not helping our users.
650 * Further improve the developer documentation to help newcomers
651 hacking on SimGrid itself. The user documentation (and in
652 particular, the beginner documentation) is still in a sorry state.
655 * Now works on Windows too!
656 * Much more extensive test suite, from MPICH
659 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
660 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
661 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
662 * Bug fix: let task be scheduled when the last dependency to be solved is
663 a control dependency.
664 * Remove SD_load_environment_script function.
665 Use the C sg_platf function if you want to declare a platform
669 * New function: MSG_process_get_number()
670 * Old function documented: MSG_config()
671 * Remove MSG_load_platform_script function
672 Use the C sg_platf function if you want to declare a platform
676 * Change the default value of the TCP_gamma constant (maximal size of TCP
677 congestion window) to a more realistic 4MiB value. If you notice changes in
678 your simulation results, you can fall back to the previous 20k tiny window
679 by adding --cfg=network/TCP_gamma:20000 on command line.
680 * (Hopefully) fix a bug wrt periodic availability and state traces
681 * Bug fix: use default values at start when first event in availability/state
682 trace is not at time 0.
685 * remove the "new_" part of function name sg_platf_new_trace_connect
686 (resulting in sg_platf_trace_connect), since it does not create
690 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
691 use case where it's really mandatory, and maintaining it was a pain in
693 * New: xbt_fifo_search(), search an item with a user-provided
694 comparison function instead of dumb pointer comparison.
697 * Fix the lua deployment:
698 Use `simgrid.init_application()` before deployment instead of
699 `simgrid.msg_register_application()` after.
702 * Transfer the tracing files into the corresponding modules.
704 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
706 SimGrid (3.8.1) stable; urgency=low
708 The "we are told that some people want to also *install* the simgrid
711 * Add missing file "tesh.1" to the archive.
713 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
715 SimGrid (3.8) stable; urgency=low
717 The Psssshiiiit release: SimGrid jumps into the Cloud.
720 * Add an experimental interface to manipulate VMs. They are mainly
721 process groups with very few intrinsic semantic, but they should
722 allow you to build the semantic you want easily.
723 * New function: MSG_host_set_property_value()
724 * New function: MSG_process_on_exit(). To clean memory in all cases.
725 * Bug fixes that made the host (and link) failures unusable.
726 * Add a way to auto-restart process when the host in which they are
727 executing comes back (ON_FAILURE="RESTART" on deployment file,
728 MSG_process_auto_restart_set).
729 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
730 please stop using the old ones, they are DEPRECATED.
732 * Deprecate functions MSG_global_init() / MSG_global_init_args()
733 Please use MSG_init() instead. (reducing the amount of entry
734 points in the library helps us).
735 * Make it impossible to link against the wrong version of the lib
736 * Deprecate MSG_clean(). No need to call it anymore.
737 * Function MSG_get_host_number() is not deprecated anymore.
740 * Split the doc into a user guide and a reference guide.
741 * Start a developper guide to help people hacking on SimGrid.
744 * Enable tracing by default. This modules rocks you should use it.
745 * Remove option custom_flags. Now use environment variables CFLAGS
747 * Use default cmake things to detect lua instead of home grown ones.
748 * New option "enable_mallocators" to disable mallocators, for debugging
749 purpose ("on" by default).
752 * Bug fixes around the resource failures: don't let the processes
753 survive the host they are running onto.
754 * Add an interface to auto-restart processes when the host in which they are
755 executing comes back.
756 * Ensures that SIMIX_clean is called automatically. It's not part of
757 the public interface anymore (bindings should be updated).
760 * Bug fix for when SD_Simulate is called with a positive value: be careful
761 when comparing doubles. Sometimes they are different for non significant
763 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
764 parallel task whose initial work is distributed among host according
765 to the Amdahl's law. Such tasks are created with a parameter alpha
766 that corresponds to the non-parallelizable part of the computation.
767 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
768 between two sets of workstations assuming a 1D block distribution (each
769 workstation owns a similar share of data) on both sides.
771 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
772 Data redistribution will be automatically scheduled once parent and child
773 are both scheduled. The filling of computation_amount and
774 communication_amount structures is now done seamlessly thanks to the chosen
776 * New function SD_workstation_dump to display various information
777 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
778 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
779 bandwidth on the route onto which the task is scheduled and the amount of
781 To divide the nominal bandwidth by 2, the rate then has to be :
782 rate = bandwidth/(2*amount)
783 * Compute tasks that have failed can now be rescheduled and executed again
784 (from their beginning)
785 * Increasing source code coverage (src/simdag is now covered at 95.8%
789 * Re-implement time-independent trace replay using SMPI (at the
790 smpi_smp_* level) instead of MSG. This should replace
791 examples/msg/actions/actions.c
792 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
793 hindexed and structs)
794 * Implement the exchange of non-contiguous data.
795 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
796 * Correct behavior of smpi/sender_gap and set its default value to 0
797 * Add option to asynchronously send small messages to allow better
798 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
799 specifies the size in bytes under which messages will be asynchronously sent.
800 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
801 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
804 * New C interface to define a platform: XML is now optional.
805 For more info, please check include/simgrid/platf.h.
806 * New interface to define random platforms from the C:
807 For more info, please check include/simgrid/platf_generator.h and
808 examples/msg/masterslave/masterslave_platfgen.c
809 * Export a sg_cmdline dynar containing all the arguments we got from
813 * Two new tracing options for adding comments to trace file so you
814 can track your experiments (see --help-tracing for details).
815 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
816 * Adding the SimGrid version that generated the trace file as a comment.
817 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
818 * Fix to avoid key clashes on Paje links
819 * Other minor fixes related to the Paje specification
822 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
823 and renamed to xbt_str_hash() and xbt_str_hash_ext().
824 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
825 * Greatly improve the robustness of mmalloc to user errors (such as
826 using an area after freeing it, or freeing it twice)
828 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
830 SimGrid-java (3.8.1) stable; urgency=low
832 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
833 Warning: all methods are visible, but only some of them are
834 implemented so far. Check the source (src/jtrace.c)
835 for further information.
836 * New module: org.simgrid.msg.File (SimGrid File management functions)
837 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
838 * Change the meaning of Process.restart: now restart the process from
839 the begining, like MSG_process_restart in C.
840 * Add Process.setAutoRestart: handling of process restart when failed
842 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
843 you to retrieve the properties of the processes/hosts
844 * Deprecate Msg.clean(): you can just forget about it now.
845 * New function Process.getCount(), that only works when compiling
846 with the not yet released version 3.9 of the C library.
848 * New context factory based on Coroutines. It mandates a modified JVM
849 but then, the simulations run about five times faster, and there is
850 no limit to the amount of processes (beside of the available memory).
852 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
854 SimGrid (3.7.1) stable; urgency=low
857 * Restore the prototype of MSG_process_create_with_environment() to
858 the pre-3.7 situation by removing the kill_time argument.
859 * Add a MSG_process_set_kill_time() function instead.
862 * Fix weird behaviors when dealing with parallel tasks.
865 * Simgrid is now built as a dll.
866 * Simgrid-java now works on Windows.
867 * Simgrid-Java is now included into Windows package.
870 * First pre-build package for MacOSX.
873 * Fix compilation when using MSG_USE_DEPRECATED.
874 * Fix some compilation issues on Macs and Windows.
875 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
876 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
878 We discovered that the Lua console is broken, but we are missing the
879 manpower to fix it right now. The problem existed in 3.7 too, so we
880 are not blocking the release for that. Sorry if you depended on this
881 feature, any help would be really welcome.
883 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
886 SimGrid-java (3.7.1) stable; urgency=low
888 The "Java aint got to be bloated and slow" release
891 * Various internal cleanups and performance improvement
892 Simulations are expected to run up to twice faster or so
893 * Make Process.kill(process) an instance method, not a static one
894 * User processes are not java.lang.Thread subclasses.
895 This breaks the compatibility (sorry), but previous API was
896 brain-dead, making it impossible to have non-trivial
897 initializations in the process constructor.
898 * Require a full constructor per Process sub-class.
899 Kinda breaks the compatibility (sorry), but this allows a much more
900 efficient way to launch the processes at simulation startup.
901 * Do not embeed our version of semaphores, java 1.5 can be considered
902 as sufficiently prevalent for us to not dupplicate its features.
906 * Add examples for almost every part of the API
907 We spotted and fixed a lot of bugs in the process
909 * New module: asynchronous communication API
910 * New function: Process.sleep()
911 It takes milliseconds as argument, just as java.lang.Thread.sleep()
912 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
913 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
915 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
917 SimGrid (3.7) stable; urgency=low
919 The "spring cleanups (before the next Big Project kicks in)" release.
922 * We can specify the SMPI latency/bandwidth factor with command line
923 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
924 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
925 You can also use the "config tag" from platform file by adding this line
926 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
927 (see "example/platforms/tag_config.xml" to use "config tag").
928 Note that the command line supersedes the platform file configuration.
929 * Change the correction factors used in LMM model, according to
930 the latest experiments described in INRIA RR-7821.
931 Accuracy should be improved this way.
932 * Use the partial invalidation optimization by default for the
933 network too. Should produce the exact same results, only faster.
934 * Major cleanup in surf to merge models and split some optimization
935 mechanisms from the core of the models. As a result you can now
936 specify which model to use (e.g., --cfg=network/model:LV08
937 --cfg=cpu/model:Cas01) and which optimization mode to use
938 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
939 Incompatible combinations should err at initialization. See
940 --help-models for the list of all models and optimization modes.
941 * The CLM03 workstation model was dropped for simplicity because it
942 used the deprecated CM02 network model. Use default instead.
943 * Rename the TCP_gamma configuration option to network/TCP_gamma
944 * Rename the coordinates configuration option to
945 network/coordinates, and document it
946 * Use now crosstraffic keyword instead of the terribly misleading
947 fullduplex keyword. It is activated by default now in the current
948 default model, use --cfg=network/crosstraffic:0 to turn it off.
949 * Ongoing refactoring the model parsing to make XML files optional
950 See include/simgrid/platf.h for details (still to be completed)
953 * Major overhaul of the documentation. Almost instructive now :/
954 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
955 functions and friends. This interface was considered as
956 deprecated since over 2 years, it's time to inform our users that it is.
957 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
958 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
959 use these (crappy) functions in your code.
960 These functions will be removed soon. Stop using them now.
961 * Deprecate MSG_get_host_{table,number}
962 Implement MSG_hosts_as_dynar() instead.
963 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
964 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
967 * Stabilize the parallel execution mode of user contexts
968 * Introduce configuration variables to control parallel execution:
969 - contexts/synchro: Synchronization mode to use when running
970 contexts in parallel (either futex, posix or busy_wait)
971 - contexts/parallel_threshold: Minimal number of user contexts
972 that must be part of a scheduling round to switch to parallel
973 execution mode (raw contexts only)
974 * Fix bugs that prevented to use suspend/resume along with
975 synchronization structures.
976 * Fix bugs in process termination that lead to invalid memory access
977 in very specific conditions.
980 * Introduce a parallel mode for the models (controlled by surf/nthreads
981 configuration item). In our tests, running the models in parallel
982 never lead to any speedups because they are so fast that the gain
983 of computing each model in parallel does not amortizes the
984 synchronization costs, even when ultra fast futexes are used.
985 This is released anyway because YMMV.
988 * Performance boost by using a swag internally to compute the set of
989 tasks that are finished and should constitute the return value of
993 * Enable it by default now that it is considered rather stable.
996 * Documentation of the tracing functions.
997 * Performance gains when tracing categorized/uncategorized resource
998 utilization by avoiding calls to get route when updating resource
999 variables. LMM constraints are being used instead.
1000 * API changed to set task categories. Use MSG_task_set_category instead
1001 of TRACE_msg_set_task_category, and SD_task_set_category instead
1002 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1003 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1004 * Examples for link user variables added (see at examples/msg/tracing/)
1005 * Deprecated function TRACE_msg_set_process_category completely removed
1006 * Trace header updated according to the latest Paje file format
1007 * Tracing network lazy updates, no longer obligate users to use full updates
1008 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1009 * Experimental: let user code declare/set/push/pop application states for hosts
1010 * API changed to allow the manual creation of graph configuration files
1011 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1014 * Improve the API of Lua MSG bindings, using the Lua spirit.
1015 * Each simulated process now lives in its own Lua world (globals are
1016 automatically duplicated). It helps writing simulators. Will allow
1017 to run Splay programs within SimGrid in the future.
1018 * Add a Chord example in Lua, equivalent to the MSG one.
1021 * Start the implementation of a solution to express temporal
1022 properties, not only local assertions. This is still an
1023 experimental work in progress, stay clear from it to be safe.
1027 - Add new runtime parameters --help-logs and --help-log-categories
1028 to display information about supported logging parameters and
1030 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1031 don't exists anymore.
1032 * Mallocators: allow value NULL for the reset function.
1034 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1035 create homogeneous dictionaries, where all the elements share the
1036 same free function. Non homogeneous dictionaries will be
1037 deprecated in the next release.
1038 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1039 - Multi-level dictionaries are deprecated.
1041 - new function xbt_dynar_search_or_negative() that is useful when
1042 you have less than 2 million elements in your dynar and don't
1043 want of the extra complexity of catching exceptions when the
1044 element is not found.
1046 - Make xbt_os_thread module (for thread portability) public.
1047 Documentation is still to come, sorry.
1049 - Cleanups and simplifications to make it maintainable again.
1050 - Exotic features (such as memalign and valloc) were removed.
1051 - The metadata were extended and improved so that the
1052 model-checker becomes able to explore and inspect the heaps.
1053 - This may induce a performance drop when enable_model-checking is
1054 ON in cmake (even if it's not used in the simulation), but it is
1055 necessary at this point to get MC working.
1057 Turn model-checking OFF if simulation performance matters to you.
1058 Not enabling it at runtime is not enough, disable it in cmake.
1060 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1063 SimGrid (3.6.2) stable; urgency=low
1065 The "Not coding new stuff allows to polish old things" release.
1068 * New bindings to the NS3 packet level simulator (experimental)
1069 * Use the raw (efficient) execution contexts instead of the sysv
1070 (portable) ones when possible.
1071 * libpcre is now mandatory in any cases since not using it led to
1072 severe performance loss and possibly other issues
1073 * Update the XML platforms:
1074 - G5K: include the latest machine in Nancy
1075 - GridPP and LCG: new platforms
1076 * Documentation was partially updated, at least (more to come)
1078 Bug fixes, cosmetics and small improvements
1079 * Free terminated processes before the end of the simulation to avoid
1080 exhausting the memory of users having very dynamic amount of
1082 * Bug fix and cosmetics about canceling non-running tasks
1083 * Bug fix about the dot loader's issues when using libcgraph
1086 * Create an installer for windows with nsis (amd64 and win32)
1087 - Add an hello world project to illustrate simgrid project creation.
1088 - Embed libpcre into the Simgrid installer to avoid
1089 its compilation burden
1090 * The raw execution contexts should work on Apple now
1091 * Port to Windows 64 bits
1092 - Sysv contexts now have an implementation for this arch
1093 - GRAS communication features now support this arch
1094 * Drop support for borland compiler on windows
1095 - this code was not maintained, and we kinda depend on gcc nowadays
1096 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1097 * Fix portability issue on unstable ubuntu: linker became picky on
1100 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1102 SimGrid (3.6.1) stable; urgency=low
1104 The "Oops, we broke Macs too" release
1107 * Fixed contexts detection so that raw ones are used when possible
1108 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1109 leads to a strange error, with user code segfaulting sometimes when
1110 the generated code is not perfectly aligned (which is not
1111 controllable from the user side, depends on the amount of code)
1114 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1116 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1118 SimGrid-java (3.6) unstable; urgency=low
1121 * Split of every thing from simgrid v3.5 into a separate package.
1123 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1125 SimGrid (3.6) stable; urgency=medium
1127 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1130 * Bindings now constitute their own package, separated from the main one.
1131 Rationale: reduce our maintenance nightmare by reducing the module coupling
1132 They will soon be released on their own on gforge.
1134 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1135 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1137 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1138 * It's not quite deprecated for now because we have no replacement,
1139 but it may soon become the case.
1142 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1143 * Fortran: New user-level cache variable to store the rank of the running
1144 process. This improves performance by an order of magnitude.
1145 * C: New coccinelle script to automatically locate and modify global and
1146 local static variables.
1147 * Improved SMPI network model with a sender-side gap to account for multiple
1151 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1152 only say if a communication is finished, no matter whether it succeeded or
1153 failed. You can call MSG_comm_get_status() to know the status of a finished
1155 * New function MSG_task_dsend() to send a task and detach it. When a
1156 communication is detached, you are never notified of its success or failure
1157 and the memory is released automatically once it is finished. This function
1158 is useful when you don't care about the end nor the success of a
1160 * Change the prototypes of action replay. Sorry for inconvenience,
1161 but this is really more efficient this way (and to adapt your code,
1162 you just have to fix the initialization, that shouldn't be too long)
1163 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1164 ever managed to do anything useful with it.
1165 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1166 takes a dynar of communications. It returns immediately and gives the
1167 index of a finished communication (if any).
1168 * New example: a basic implementation of the Chord P2P algorithm.
1171 * New model for multi-core CPUs. You can now use the core attribute to
1172 precise the number of cores of a host. This is a basic model. Every
1173 process running on the host receives at most the power provided in
1174 the DTD (throughput<=power). Total throughput of process cannot exceed
1176 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1177 router linked by an up-link and a down-link (possibly asymmetrical).
1178 This kind of pattern allows to easily build last-mile model style platforms.
1179 Aggregating such patterns in a rule-based AS is thus the technique of
1180 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1181 * New model for Vivaldi routing. We transformed the Vivaldi network model
1182 into a Vivaldi routing model (based on the rule-based model). This allows to
1183 combine Vivaldi based latencies with last-mile platforms.
1186 * Added a check for NaN of IEEE754 infinite in the double entries of
1188 * Introduce a new context factory "raw", highly inspirated from the
1189 ucontext factory, but using manually crafted functions in assembly to
1190 do the work in an efficient manner.
1191 * Allow to change the used context factory at run time, not only at
1192 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1193 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1194 contexts at runtime (only with raw contexts or ucontexts).
1195 * Completely rewrote this module to allow parallel execution of user
1196 processes. Use --cfg=contexts/nthreads:N to execute user processes
1197 with N parallel threads (the default is 1, meaning no parallelism).
1198 * Allow to decide dynamically between sequential and parallel modes.
1199 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1200 processes in parallel only when their number is greater than or equal to P
1202 * Added a check for NaN of IEEE754 infinite in the double entries of
1206 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1207 won't output the state of processes when interrupted with Ctrl-C
1208 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1209 NULL-terminated array. This may solve backward compatibility issues
1210 due to the change to return type of SD_simulate. See also:
1211 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1212 * Add new macros with variable number of arguments.
1213 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1214 - in xbt/asserts.h: xbt_assert
1215 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1216 - in xbt/ex.h: THROWF and RETHROWF.
1217 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1219 * Change xbt_die() to accept a format string with arguments, just like printf.
1220 * New data structure: xbt_lib_t, like a dict but more general and with better
1224 * New configuration options
1225 Options triva/categorized and triva/uncategorized can be used to generate
1226 graph configuration files for Triva visualization tool.
1227 * Configuration option tracing/platform is renamed to tracing/categorized
1228 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1229 * New TRACE_user_link_variable interface:
1230 User provides the name of the link and the tracing variable to attach to it
1231 * the declaration of tracing categories must be done after the environment
1233 * simpler tracing interface, just one way to declare categories
1234 TRACE_category or TRACE_category_with_color, it is up to you
1235 * links in the trace file are again identified by their names
1236 * trace contains the full platform hierarchy exactly as declared using the ASes
1237 * Options tracing/msg/[task|process]:1 groups the process by hosts
1238 for both cases, tasks and processes must have names that are unique during the simulation
1239 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1240 * The experimental option tracing/msg/volume is deprecated
1241 its functionality may be reincorporated if needed
1243 The tracing generates a trace file with unordered timestamped events,
1244 because of the way the core simulator (surf) works. A script available
1245 at the tools directory (fix-paje-trace.sh) can be used to put the events
1246 in order. We have changed the tracing so it can generate ordered timestamped
1247 events in the final trace, but depending on the simulator (and how much time
1248 is simulated) that can lead to a huge memory utilization. It is deactivated
1249 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1251 Build Infrastructure
1252 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1253 - We are trying hard to keep the API stable, but it may happen that
1254 some things change (we're a research project after all, not a
1255 nuclear plant operating system). If such things should happen, you
1256 could rely on that macro to adapt.
1257 - current value: 30600 for 3.06.00, aka 3.6
1258 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1259 a number that can be compared with SIMGRID_VERSION.
1260 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1261 assertions and verbose/debug logging events are disabled at compile time.
1263 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1265 SimGrid (3.5) stable; urgency=medium
1268 * New feature to any SimGrid-based simulator: Model-Checking
1269 Check SIN#1 for more details.
1272 * New Model SMPI (three-interval linear regression for correction factors)
1273 See RR-7426, available at http://hal.inria.fr/inria-00527150
1274 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1275 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1276 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1277 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1278 * See also src/smpi/README
1283 - Tracing API changes: TRACE_start and TRACE_end should not be called
1284 by user-code. They are automatically called by simulators created
1285 with SimDAG, MSG and SMPI if the toolkit is compiled with
1286 tracing_enabled=ON. Categories declaration and utilization remain the
1287 same for MSG and SimDag.
1288 - A function was added to the tracing API to declare categories with
1290 - TRACE_category_with_color (char *category, char *color)
1291 where color must be in the following format
1292 "%f %f %f", red, green, blue
1293 and red, green, blue are float values in the interval [0, 1]
1294 - User can specify NULL as color parameter, or continue calling
1295 TRACE_category (cat)
1296 On that case, the tracing system will define random colors
1297 - The following command-line options are supported:
1298 --cfg=tracing/filename:msg.trace
1299 --cfg=tracing:1 (activate tracing, needed to use others)
1300 --cfg=tracing/platform:1 (categorized resource use)
1301 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1302 --cfg=tracing/msg/task:1 (task creation)
1303 --cfg=tracing/msg/process:1 (process creation, migration)
1304 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1305 --cfg=tracing/smpi:1 (SMPI interface tracing)
1306 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1307 - examples of examples/msg/tracing updated
1309 - DAXLoader and DOTLoader functions can generate tasks with categories
1310 - A new function to attribute a category to SD tasks:
1311 TRACE_sd_set_task_category (SD_task_t task, char *category)
1312 * Tracing the MPI interface implemented by SMPI
1313 - Collective operations are traced with states
1314 - Point-to-Point operations are traced with states/links
1315 - Tracing activated by a parameter "-trace filename" passed
1316 to smpirun during execution (considering that simgrid
1317 is compiled with tracing enabled)
1318 - To run the simulation with gdb, the simulator
1319 accepts --cfg=tracing/smpi:1 to trace SMPI
1320 - tesh files to check if smpi tracing is ok
1321 - See examples/smpi/NAS/DT-trace
1322 * GTNetS tracing re-worked
1323 - adaptation to the tracing system of GTNets to cope
1324 with modifications regarding the fullduplex mode
1325 - new tesh files to check if gtnets tracing is ok
1328 * Asynchronous communications through the functions:
1329 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1330 * New function: MSG_load_platform_script()
1331 to make possible using a lua script instead of XML files to set up platforms
1332 * New function: MSG_set_function
1333 to associate functions to processes, used when bypassing the parser
1334 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1336 Platforms: Add some more examples in examples/platforms
1337 * Grid'5000: see www.grid5000.fr
1338 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1341 * Change the XML format. This is a very important modification. SimGrid
1342 3.5 introduces a new hierarchical format based on the notion of
1343 Autonomous Systems. Compatibility with old format is ensured
1344 through the perl script provided in the install bin directory
1345 bin/simgrid_update_xml.
1346 It is now possible to build platforms with specific routing
1347 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1348 connect several platforms together. We will try to provide soon
1349 set of realistic platforms exploiting these properties (have a
1350 look at examples/platforms/ for the moment).
1351 * Take the opportunity of the XML format change to be a good XML citizen:
1352 rename link:ctn to link_ctn and similar changes (also dealt with by
1354 * Add a new routing scheme (rule-based) using regular expressions. It
1355 enables to have an extremely low memory footprint when the
1356 underlying routing is simple and can be compactly described. You
1357 need to have libpcre4-dev (perl regular expressions) installed if
1358 you want to use this routing scheme.
1359 * Revive the cluster TAG and allow to easily and efficiently (both in
1360 term of memory and speed) connect clusters together. Have a look
1361 at teshsuite/simdag/platforms/ to see how this can be done. With
1362 this tag, you can create clusters with thousands of tasks at no
1363 cost (have a look at examples/platforms/).
1364 Note: clusters are implemented as ASes, so there is no need for an
1365 enclosing AS tag if you have only one cluster in your platform.
1366 * Add new generic functions in the public interface that allows the user
1367 to call SURF 'create_resource' methods from your code (same
1368 functionality as the XML bypass mechanism but with a much lighter
1370 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1371 that uses a piecewise linear approximation to produce better
1372 results when exchanging small messages.
1373 * Add a new parameter to handle correctly full duplex link and account
1374 for interferences between uplink and downlink communications
1375 (activate with --cfg=fullduplex:1).
1378 * Rename the SD_READY (all dependencies are satisfied and task is
1379 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1380 dependencies are satisfied) state.
1381 This prevents a confusion between the notion of "ready to schedule"
1382 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1383 simulated" (SD_RUNNABLE) used by the simulation kernel.
1384 * Change the way a task is considered as ready. Instead of removing
1385 dependencies when a task is done, a counter is decreased. This way,
1386 it is always possible to reach ancestors thanks to the
1387 SD_taks_get_parents function (even after the end of the simulation.)
1388 * Change the return type of SD_Simulate from (SD_task_t*) into
1389 xbt_dynar_t. This function was in handling a dynar internally and
1390 converted it into a NULL terminated array for historical reasons.
1391 * New function SD_dotload(char*) to load a DAG described in dot
1392 format. This loader and the corresponding examples require the
1393 installation of the graphviz library.
1394 * Fix a bug in the management of tasks of size 0 in the surf network
1395 models. This problem was only visible with SIMDAG and you should
1396 thus disregard results produced with earlier versions if you
1397 relied on this feature (some tasks were blocked because of this).
1398 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1399 even though your applications were doing only point-to-point
1400 communications and sequential computations. Now you can really use any
1401 model you want (of course, if you create real parallel tasks, which are
1402 not implemented in most models beside ptaskL07, this will abort).
1403 * Add an example that schedules a DAX on an heterogeneous platform
1404 using a Min-Min strategy.
1405 * New function SD_workstation_get_current_task() that returns the kind
1406 of task currently running on a workstation in the sequential access
1408 * Raise some warnings when unexecuted tasks remains at the end of the
1409 simulation. This is usually caused by cycles in the DAG.
1412 * New function: SIMIX_process_set_function() called by MSG_set_function
1413 * Change the underlying waiting queue in semaphores so that a process
1414 can wait on several of them simultaneously (as in waitany).
1415 * Fix the way to handle tokens in semaphores so that all access patterns
1416 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1417 * kill the dirty pimple SIMIX_message_sizes_output()
1418 Please use (proper) visualization instead
1421 * New data container: setset (set of sets of elements)
1422 * New module: mmalloc (mapped malloc, allowing to have several
1423 independent segments of malloc)
1424 * New function: xbt_dict_cursor_set_data()
1425 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1426 * New function: xbt_dynar_is_empty()
1427 * New function: xbt_fifo_get_last_item()
1428 * Fix xbt_dynar_shrink(): use the right element size.
1429 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1430 uninitialized areas during expand.
1431 * Fix semaphores: previous implementation was severely broken.
1432 * Use library init/fini functions for our initialization.
1433 - you can use logs and other feature as soon as you want in your
1434 code (even before the xbt_init / MSG_init)
1435 - xbt_exit is now a no-op and produce a warning when used.
1438 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1439 benefit from the latest improvement to the simulation kernel.
1440 * Kill measurement sockets for now. If you rely on them, sorry. This
1441 release is not for you. This feature will be reintroduced in the
1442 future, but we cannot delay the release any further.
1443 * New function: gras_msgtype_get_name().
1444 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1445 * Fix (at last) the pmm example: it should not randomly fail anymore.
1447 Build chain: bug fixes and overall polishing
1448 * Cmake is now stable enough. Hence, we killed the autotools.
1449 * Port to windows ( TM :)
1450 * Fix the 'make install' target.
1451 No need to use 'make install-simgrid' anymore
1452 * Introduce a 'make dist' target compiling a *source* archive
1453 'make package' compiles a binary archive
1454 * Compile java files only on need
1455 * Add --cd and --setenv command line options to tesh
1456 * Out of source builds are not fully supported yet, but we are close
1457 * Enable supernovae and optimization flags by default for our users
1460 * Add layer to set up environment directly from lua, without XML.
1461 * The effect of gras_stub_generator can be achieved through
1462 lua too (check examples/gras/console/ping_generator.lua)
1464 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1466 SimGrid (3.4.1) stable; urgency=low
1468 The "Polishing easter eggs is probably a good idea" release.
1469 This is a bug fixes release only.
1473 * Fix a bug preventing the tasks from begin garbage collected.
1476 * Fix a bug occuring when a host involved in a communication fails.
1477 This was not detected properly by the other peer involved in the
1478 communication. Now, it's reported as a network error.
1481 * Warn the user about loop dependencies in data flow of DAX files
1482 * Obey the control-flow dependencies of DAX files
1485 * Add option "enable_smpi" allowing to not compile SMPI.
1486 Probably useful for the (Mac) users experiencing a build error here
1487 * Improve the detection of lua5.1 and ruby1.8
1489 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1491 SimGrid (3.4) stable; urgency=low
1493 The "Easter in Cargese" release. Also known as (major changes):
1495 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1496 ~> bindings were greatly improved
1497 ~> new tracing infrastructure for better visualization introduced
1499 * the "Welcome to configury modernity" release.
1500 ~> we switched from autotools to cmake, and improved our cdash
1503 A more detailled list of changes follow (full detail in svn log).
1505 Java Bindings: Various Cleanups
1506 * (install java-gcj-compat-dev on debian-like to use them)
1507 * Remove put/get: no need to export deprecated interface in Java
1508 Use send/receive instead.
1509 * Cleanup the examples and add a README per directory
1510 * Remove example autoDestination (that's the only way to go now)
1511 * Remove example explicitDestination (was a plain copy of basic)
1512 * Make JniException a runtime exception, so that there is no need to
1513 declare the fact that you may encounter such a beast. I guess that
1514 nobody will ever want to survive such error.
1515 * Create specific errors for each MSG case of failure:
1516 host failure, transfer failure, timeout, task cancelled
1517 * Cleanup the exceptions that may get thrown by each function
1518 * Other internal cleanups in Java bindings. Performance still bad :/
1519 Ruby and Lua Bindings: create them
1520 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1521 * That's new and great, you should try them out.
1522 Same functionalities than Java bindings, only even less polished
1524 * Kill the useless "rate" argument of SD_task_get_execution_time()
1525 Everyone used to provide -1 as a value, it was not used, and the
1526 semantic of a possible use wasn't even clear.
1527 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1528 either as comm costs or compute costs to mean that there is no
1529 such thing for that specific task.
1530 * Add a SD_task_set_name() function
1531 * Fix SD_task_unschedule() on typed tasks
1532 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1533 * In DAX loader, accept useless 'level' attributes to <job> since
1534 LIGO DAGs have them (seem to be to ease graphical representation).
1536 * Add an example masterslave_mailbox.c using send/receive and not
1537 the deprecated put/get interface.
1538 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1539 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1541 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1542 (declare MSG_USE_DEPRECATED to still have the old name)
1543 * Add a MSG_task_set_data() function
1544 * About trace replay (see examples/msg/actions):
1546 - Allow to work with splitted trace files for each process
1547 Give the specific trace file as argument of each process,
1548 and call MSG_action_trace_run(NULL)
1549 You can still have one merged file for all processes.
1550 - Fix implementation of collective operations
1551 * Allow task_execute() on 0-sized tasks (closes #10063)
1553 * This is the first release of SimGrid where SMPI is not considered
1554 beta anymore (even if some corners should still be improved)
1555 * Port over the new SIMIX_network submodule (internal refactoring)
1556 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1557 * Implement more missing elements of the standard:
1559 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1561 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1562 MPI_Scan() MPI_Get_processor_name()
1563 - Added implementation of missing case for Alltoall (warning: it's
1564 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1565 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1566 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1567 SMPI_MPI_Allgatherv()
1568 * Bug fixes include:
1569 - MPI_Waitsome() was broken
1570 - Allow relative includes in smpicc
1571 - Command line cfg argument 'reference_speed' was ignored...
1572 - Some functions did not properly lead to auto-benching of user code
1573 - smpicc passes -O2 by default (just like openmpi one)
1575 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1576 * Bug fixes about timeouts during communications
1577 * add SIMIX_message_sizes_output() as a pimple to write to file the
1578 amount of messages per size. Use gnuplot to get histogram.
1579 Pimple because that's the only user-visible function of simix,
1580 defined directly in xbt.h (irk, sorry)
1582 - Add a SIMIX_sem_get_capacity() function
1583 - Fix interactions with processe resume/suspende
1584 - release_forever() was stupidly broken
1585 - Fix SIMIX_display_process_status() for processes in a semaphore
1586 - Make SIMIX_sem_block_onto() user-visible
1587 * Refactoring context stuff:
1588 - Use pseudo-OOP for better modularity
1589 - reimplement SIMIX_process_kill() without process_schedule() so
1590 that the latter can take as invariant that it is called from
1592 - Merge context_start into context_new for sake of simplicity
1594 * Add a Vivaldi network model, coded live during SUD'10 ;)
1595 * Rename configuration variables to start a hierarchy:
1596 o cpu_model -> cpu/model
1597 o network_model -> network/model
1598 o workstation_model -> workstation/model
1599 * New configuration variables:
1600 o network/bandwidth_factor: correction to bandwith
1601 o network/latency_factor: correction to latency
1602 o netwotk/weight_S: correction to the weight of competing streams
1603 * Add a long description to the models, that users can see with such
1604 argument on the command line: --cfg=cpu/model:help
1605 * --help-models display the long description of all known models
1607 * config: add the ability to set a default value after registration
1608 Does not override any previously set value (e.g. from cmd line)
1609 * dict: allow to have integer key and data.
1610 When so, you need to use the following functions
1611 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1612 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1613 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1614 In contrary to regular dicts, the key is not malloced before copy.
1615 Mixing scalar and regular elements in the same dict is not tested
1617 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1618 Tracing for Visualization:
1619 * SimGrid is now instrumented in order to generate a trace file for
1620 visualization analysis: to use it, need to compile SimGrid with the
1621 "tracing" option enabled, and instrument the program using SimGrid with
1622 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1623 (among other functions).
1624 * The instrumentation only traces the platform utilization for now
1625 * Documentation to use the tracing functions and how to analyze the
1626 traces with the Triva tool is written.
1627 * More information about: SimGrid FAQ (in the section Tracing Simulations
1630 * We moved to cmake as default build system. Autotools support will
1631 be dropped soon. Check the FAQ for more info about how to use it.
1632 * Greatly improved our cdash/ctest interactions
1633 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1634 * Added memory checking tests with valgrind; lot of memleak fixing.
1635 This may be the first release of simgrid with so few memory issues
1636 * Added code coverage tests.
1637 Our coverage is still improvable, but at least we see it on cdash.
1639 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1641 SimGrid (3.3.4) stable; urgency=low
1643 The "Desktop Grid needs love too" release (also called Xmas release).
1645 Models improvements:
1646 * Major speedup in the maxmin system solving by using lazy evaluation
1647 Instead of solving completely the maxmin system at each iteration,
1648 only invalidate (and recompute) the modified parts.
1649 This new feature is enabled in default models but you can try to
1650 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1651 * Cas01 IMproved as default CPU model
1652 This CPU model is the same Cas01 model, but it uses the
1653 maxmin-selective-update flag and a heap structure to manage
1654 actions on SURF kernel.
1655 It reduces the complexity to find the next action to finish and,
1656 consequently, it's faster than the old Cas01.
1657 This is the new default CPU model (Cas01).
1658 * Rename the old Cas01 model to Cas01_fullupdate
1659 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1660 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1661 * CpuTI (CPU Trace Integration)
1662 A new CPU model whose objective is simulate faster when using
1663 availability trace files.
1664 Instead of using a full featured, over engineered maxmin system for
1665 CPU modeling, this model does the pre-integration of traces files
1666 to calculate the amount of CPU power available, and so, executes
1667 faster than the old CPU models.
1668 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1669 * Use LV08 as default network model since it gives better accuracy
1670 for small messages and shouldn't change things for big ones.
1671 Use --cfg=network_model:CM02 to get the previous behavior.
1674 ******************************************
1675 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1676 ******************************************
1677 * The new CPU model may changes simulations!
1678 The point is that events occurring at the exact same timestamp
1679 are not scheduled in the same order with the old and new
1680 version. This may be enough to completely change the execution
1681 of simulations in some cases.
1682 * The new network model will change simulations!
1683 This new model is more realistic than the previous one, so you
1684 should consider redoing your old experiments with this model.
1685 Sorry for the inconvenience.
1688 * Introduce the supernovae compilation mode
1689 When compiled that way, the whole SimGrid (or almost) is put in a
1690 single compilation unit and compiled in one shoot.
1691 This is to help gcc which has difficulties to inline stuff from one
1693 The speedup seem to be above 15%, althrough more tests are needed on
1694 amd64 to confirm that gain.
1697 * Port of MSG's mailbox on top of SIMIX network
1698 The put/get mechanism was greatly simplified on the way.
1701 * New SIMIX network module. Provides:
1702 - Mailbox: rendez-vous mechanism to find with who you want to speak
1703 - Synchronous send/recv: easier and hopefully faster since the
1704 logic is handled in the maestro process directly now
1705 - Asynchronous send/recv: you dreamt of it? It's here now
1706 Too bad that nobody cared enough to propagate the change to MSG.
1707 * Add semaphores as SIMIX synchronization mechanism.
1710 * new function SD_daxload(char*) to load a DAX file
1711 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1712 * Introduce typed tasks. Specify its kind and cost at creation.
1713 At scheduling, just give where it should be placed, and the cost
1714 for each involved resource is automatically computed.
1715 Existing constructors so far (more to come of course):
1716 - SD_task_create_comm_e2e() for end-to-end communication
1717 - SD_task_create_comp_seq() for sequential computation
1718 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1719 * new function SD_task_dump() for debuging display
1720 * new function SD_task_dotty(task,FILE*) writing to file the info
1721 about the task in dotty format
1722 * SD_task_dependency_exists() can now cope with having one of its
1723 arguments NULL. If so, it tests whether the other argument has any
1725 * Add getters on list of preceding/following tasks:
1726 SD_task_get_parents(task) and SD_task_get_children(task)
1727 * Add getters on amount of workstations and list:
1728 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1729 * Add getter on task kind: SD_task_get_kind(task)
1730 * Update the start_time and finish_time of tasks on completion/failure
1731 * Bugfix: Remove task from state swags when destroyed
1734 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1737 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1738 dynar with one separate thread per value of the dynar.
1739 * Change the prototype of xbt_thread_create(), sorry.
1740 Added a boolean parameter indicating whether we want to join this
1741 thread (used in SG only for now)
1742 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1745 * GTNetS wrappers should now be usable again (and betterly tested too)
1746 * Fix a major regression from 3.2 where the timeout provided to
1747 MSG_task_put_with_timeout() was used as absolute time before which
1748 the comm should be done.
1749 * Start to fix the <cluster> tag.
1750 - Internal links should be good now (beside of the loopback, which
1751 use the private link instead)
1752 - paths to the external world is still rather broken
1753 - the <route:multi> tag is just broken. Actually that's brain-dead.
1754 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1755 to make it less stupid
1756 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1757 * Fix a source-level compatibility glitch from 3.2: after defining
1758 MSG_USE_DEPRECATED, you can use the old name
1759 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1760 * Allow to compile from the SVN with automake 1.11
1761 * Fix some problems when using the "start_time" tag in deployment XMLs.
1762 * Fix #8569: XBT/synchro.h has redundant declarations
1763 * Fix #8563: MSG return values and exceptions
1764 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1765 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1766 Thanks to Robson Peixoto.
1767 * Fix a vicious bug in dictionaries inducing that some elements were
1768 not freed on xbt_dict_free()
1770 Portability report of this version:
1771 * Main portability targets:
1772 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1773 on (amd64/i386/ia64)
1774 - mac leopard on i386
1775 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1776 but nothing critical.
1777 * Other platforms: windows, AIX and others were not tested for this release
1779 Timing report of this version:
1780 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1781 scenario parameters). From 8h to a few seconds in desktop grid settings.
1782 * Supernovae brings about 25% speedup on i386.
1784 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1786 SimGrid (3.3.3) stable; urgency=low
1788 The "Need for Speed" release.
1790 The timings done to validate the 3.3.2 were faulty.
1791 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1793 The problem was a conversion from a manually handled vector to
1794 xbt_dynar_t on the critical path.
1795 xbt_dynar_foreach calls functions, inducing stack management crap.
1797 We inlined these functions and xbt_dynar_foreach is now breath taking.
1798 We also inlined xbt_swag_belong on the way.
1800 Here are some approximate speedup measurements (on master/slaves
1801 simulations lasting between 10s and 20s each):
1802 3.3.1 -> 3.3.2: about same performance
1803 3.3.2 -> 3.3.3: 40% speedup
1804 3.3.1 -> 3.3.3: 40% speedup
1805 3.3.1 with inline patch -> 3.3.3: 30% speedup
1807 Our reading is that the refactoring which occurred in 3.3.2 made us
1808 suffer much more from the xbt_dynar_foreach low performance, but
1809 once we solved this, this refactoring proved to be very performance
1810 effective. From the 40% speedup, somehow, 10% are due to the
1811 inlining and 30% to the refactoring.
1813 That's a pitty that gcc cannot inline functions placed in other files
1814 alone. We have to choose between:
1815 - break the encapsulation (by putting private data structures and
1816 accessors in headers files to help gcc)
1817 - live with low performance
1818 - switch to a decent compiler such as icc (not quite possible).
1820 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1822 SimGrid (3.3.2) stable; urgency=low
1824 The "Simplicity does not preceed complexity, but follows it" release.
1826 The main contributors of this release were (lexical order):
1827 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1830 * Extract the routing logic into its own object.
1831 (was dupplicated in network.c and workstation_LV07.c;
1832 Allows to implement other ways of storing that info)
1833 => kill now useless network_card concept
1834 - Use dynar to represent routes (instead of void** + int*)
1835 - kill link_set (use surf_network_model->resource_set instead)
1836 - Add a command-line option to choose the routing schema to use
1837 - Add three new models:
1838 * Floyd (shortest path computed at initialization)
1839 * Dijikstra (shortest path recomputed all the time)
1840 * Cached Dijikstra (shortest path computed on need)
1841 All these models where contributed by Silas De Munck, and are
1842 described in his ICCS09 paper.
1844 * Simplify model declaration
1845 (less redirections, less function to write when defining a model)
1846 - Factorize stuff between models:
1849 surf_model_resource_set(model)
1850 surf_model_resource_by_name(model, name)
1851 - Unify the types of models in s_surf_model_t (using an union)
1852 - Embeed fields of common_public directly into s_surf_model_t
1853 - Rename model methods:
1854 action_free ~> action_unref
1855 action_change_state ~> action_state_set
1856 action_get_state ~> action_state_get
1857 - Change model methods into functions :
1858 (model)->common_public->action_use ~> surf_action_ref
1860 * Implement a generic resource; use it as ancestor to specific ones
1861 (allows to kill duplicated code in models)
1862 Drawback: timer command don't need no name nor properties;
1863 workstation_CLM03 don't need no properties
1864 (but I guess we can live with those few bytes wasted)
1866 * Improve the action object model
1867 - implement a constructor avoiding dupplicated code about field
1868 initialization in generic_action part.
1870 * Kill the SDP model: it has an external dependency, is deprecated
1871 in flavor of modern lmm models, and didn't compile since a while
1874 * Relocation of the context module from XBT to SIMIX.
1875 (the context were decoupled from the simix processes, duplicating a lot of code)
1876 => a lot of code was factorized
1877 - less overhead is introduced during scheduling
1878 - simpler API for the context factory
1879 - the logic for process creation,destruction and manipulation was simplified
1880 * Simplification of the s_smx_process_t data structure.
1881 => accesing the simix level data associated to a process is faster now,
1882 and the code is a lot more readable.
1885 * Implement some more MPI primitives:
1886 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1887 -implementation: Bcast: flat or 2-ary tree (default),
1888 Barrier: 4-ary tree,
1890 Allreduce: Reduce then Bcast
1891 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1892 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1893 Alltoallv: flat tree, like ompi
1895 * Add support for optimized collectives (Bcast is now binomial by default)
1896 * Port smpirun and smpicc to OS X
1899 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1900 Shout out if you used it.
1903 * Display the list of still queued messages in SG mode when existing
1907 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1908 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1909 * Add xbt_str_from_file(FILE*)
1910 * Add xbt_dict_get_key achieving a linear reverse search
1911 * Remove the context module
1913 Portability report of this version:
1914 * Main portability targets:
1915 - Linux(debian)/x86/context
1916 - Linux(debian)/x86/pthreads
1917 - Linux(debian)/amd64/context
1918 - Linux(debian)/amd64/pthreads
1919 On these, we still have the eratic breakages of gras/pmm and
1920 amok/saturate_sg reported in previous version. We still think
1921 that the tests are the cause of the fault, not the tested code.
1923 - Mac OSX Leopard/x86/context
1924 Still false negative in tesh autotesting.
1925 Smpi still fails, but this time because readlink does not accept -f
1926 Everything seems to work properly beside of that.
1929 - AIX version 5.3 (only tested contexts this time)
1930 Smpi still fails there because mktemp is not installed.
1931 Everything seems to work properly beside of that.
1933 I managed to compile it for the first time, but several breakages.
1934 Won't delay the release for this exotic platform.
1936 * Windows: it's still lagging behind. If you want to help, please
1939 Timing report of this version:
1940 This version seem to be more than 5% faster than 3.3.1 (on linux
1941 64bits with contextes). The gain is less than expected, we are
1942 investigating this for next release.
1944 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1946 SimGrid (3.3.1) stable; urgency=low
1949 * Implement a --cfg-help to show existing configuration variables
1950 * Build chain do not require doxygen in maintainer mode
1953 * fix a bug on struct sizeof computation, which prevented the
1954 exchange of arrays of structs in some conditions
1955 - added a regression test about this in datadesc_usage
1956 * Allow the exchange of 0-long dynamic vectors.
1957 - for that, use -1 as indicator of dynamic size instead of 0
1958 - This implied to change any size from unsigned long to long,
1959 reducing a bit communication abilities, but I guess that with
1960 64bits being quite common, this is more than enough.
1961 - This also induce a protocol change, thus bumping network protocol
1962 version from 0 to 1 (if we have external users, we have to get
1963 clean on that point too ;)
1964 - added two regression tests about this in datadesc_usage
1965 * Be more verbose when propagating local exceptions
1966 This helps debugging.
1967 * Display the status of simulated processes when receiving SIGINT in
1971 * Allow to control the simulation from a trace file.
1972 New functions MSG_action_register() and MSG_action_trace_run()
1973 The first one allows to associate a function execution to each
1974 kind of action while the second one parses a trace file and
1975 triggers the corresponding actions within the system.
1976 For now, only a toy example is provided in examples/msg/actions
1977 * Add an exemple of process migration in examples/msg/migration
1978 * Fix a bug in task exchange which broke MSG_task_get_sender()
1979 Add a teshsuite regression test for that.
1980 [Bug: if MSG_task_get_sender() is called after sender exit,
1982 * Fix a bug which prevented suspend/resume to work properly
1983 * Display the status of simulated processes when receiving SIGINT
1984 This fixes a regression of v3.3. due to the introduction of SIMIX
1985 * Bug fixing in failure management:
1986 - trace could not start by a failure at time 0
1987 - failure during communications were not working
1990 * Add SIMIX_process_set_name() to change the name of the current
1991 process in the log messages.
1992 * Store smx_hosts in a dict since we only retrieve them by name
1993 * Move the configuration infrastructure to surf
1996 * Move the configuration infrastructure to surf
1999 * Massive internal cleanups:
2000 - Store internal structures on processes instead of hosts (allows
2001 to have more than one process per host, in addition of being more
2003 - Cleanup the initialization/finalization process
2004 - Kill a whole bunch of unneeded synchronization:
2005 processes run in exclusive manner within the simulator
2006 - Move queues from global tables to process data fields
2008 - now accept -platform and -hostfile arguments
2009 - Pass the right rank value to processes according to the hostfile
2010 * Compile the examples by default, and use them as regression tests
2011 * Implement MPI_Wtime()
2012 * Change the reference speed to a command line option
2015 * TCP_gamma can now be specified as command line option using
2016 --cfg=TCP_gamma:10000000.0
2017 * Change the --surf-path cmd line option into --cfg=path:
2020 * Also include strbuff from xbt.h public header
2021 * xbt_ex_display(): do not free the exception after displaying
2022 This allows to do more with the given exception afterward.
2023 Users should call xbt_ex_free() themselves.
2027 Portability report of this version:
2028 * Main portability targets:
2029 - Linux(debian)/x86/context
2030 - Linux(debian)/x86/pthreads
2031 - Linux(debian)/amd64/context
2032 - Linux(debian)/amd64/pthreads
2033 These targets fail about 1/10 of times on gras/pmm, but we believe
2034 that this is because of the test, not because of simgrid.
2035 amok/saturate_sg fails even more rarely, and the test may not be
2038 - Mac OSX Leopard/x86/context
2039 The test suite still spits tons of errors because some obscure
2040 force prevents us from removing the temporary directories
2041 arguing that they still contain some metadata I've never heard of.
2042 Smpi fails because seq is not installed.
2043 Everything seems to work properly beside of that.
2046 - AIX version 5.3 (both contexts and pthread)
2047 Smpi still fails there because mktemp is not installed.
2048 XML inclusions seems rosty on AIX.
2050 * Windows: it's still lagging behind. If you want to help, please
2053 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2055 SimGrid (3.3) stable; urgency=high
2059 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2060 [Malek Cherier & Mt]
2062 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2065 Until now, GRAS were using MSG as an interface to SURF. It was
2066 quite difficult because both interface have several differences
2067 (MSG channels vs GRAS sockets were the most notable point).
2069 This also opens the gate to SMPI (which should occur soon) and speed
2070 up simulations by to 40% (even if it were not the main goal).
2072 **************************************
2073 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2074 **************************************
2075 The point is that events occuring at the exact same timestamp are
2076 not scheduled in the same order with the old and new version. This
2077 may be enough to completely change the execution of simulations in
2078 some cases. Sorry for the inconvenience.
2080 * Cleanup and upgrade the XML format to push further scalability
2081 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2083 * Improve the testing infrastructure with tesh. Now a very large part of
2084 the code is tested not only by being run but also by checking that the
2085 output match an expected output [Mt].
2087 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2088 is really less memory-demanding, which should allow you to use
2089 larger files in SimGrid [AL].
2091 * Inform valgrind about our contextes, so that it becomes usable
2092 with the default (and more effecient) version of SimGrid
2093 [contributed by Sékou Diakite, many thanks]
2096 * Introduce a listener thread in charge of receiving incoming
2097 messages from the network. It allows to overlap communication and
2098 computation but most notably, it removes some stupid deadlocks due
2099 to the fact that so far, a process could not send and receive at
2100 the same time. This made most non trivial communication schema
2102 * Convert the PIDs from long int to int to match the MSG ones (and
2103 linux ones too) [Mt]
2104 * New function: gras_agent_spawn() to launch a new process on
2105 current host. Only working in simulation for now. [Mt]
2106 * New function: gras_os_hostport() returning a constant form (ie,
2107 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2110 * Make the backtrace of exceptions more human readable [Mt]
2111 * New module: xbt/str [Mt]
2112 a ton of string utility functions (split, join, printf to a newly
2113 allocated buffer, trim, etc)
2114 * New module: xbt/hash [Mt]
2115 SHA1 hashing algorithm (more to come if needed)
2116 * New module: xbt/synchro [Mt]
2117 synchronization tools (mutex and conditions) working the same way
2118 in simulation and in real life (mainly useful for GRAS, but not
2120 * New module: xbt/queue [Mt]
2121 classical producer/consumer synchronization scheme
2122 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2123 (using the classical functions will get serialized) [Mt]
2124 * Make dictionary internal table dynamic. No need to specify its size
2125 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2127 * Make sure the log channels are organized as a tree under windows
2128 (because of ANSI C compatibility issue, any channel were child of
2129 root directly) [Mt].
2132 * Cleaned many thing in surf and fixed a few bugs [AL].
2133 * Add a nice command line configuration mechanism to compose models [AL].
2134 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2135 the previous one (KCCFLN05). It relies on something that looks like
2136 a max-min sharing mechanism but cannot be written as such. A new solver
2137 was thus designed [AL].
2138 * Add a new solver to lmm. Based on Lagrange optimization and
2139 gradient-based descent, it enables to efficiently maximise systems s.a
2141 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2143 This solver enables to propose two new network models for TCP Reno and
2144 TCP Vegas based on Low's work. These models still need to be fully
2145 tested though [Pedro Velho].
2148 * Bug fix in SD_simulate. Now the time bound given as argument is
2150 * Use the new parallel task model (ptask_L07) as default.
2151 * Use the SURF command line configuration mechanism.
2152 * 0-size tasks (for synchronization) should now work.
2154 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2156 SimGrid (3.2) stable; urgency=high
2160 We still experience issues on this platform, but we believe that at
2161 least MSG is usable.
2163 GRAS API BREAKAGE (for simplification purpose, sorry):
2164 * the gras_msgtype_by_name is not used anymore. Instead of
2165 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2166 you can write (and must)
2167 gras_msg_send(toserver, "request", &request);
2168 - If you still want to pass a gras_msgtype_t to the function (to cache
2169 the type and avoid the lookup time), use the gras_msg_send_() variant.
2170 - Impacted functions:
2171 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2172 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2173 * The callbacks are now expected to return 0 when everything went well
2174 (just like the main() function)
2176 GRAS new features and improvements:
2177 * New module mechanism where user code can use per process globals [Mt]
2178 This is similar to gras_userdata_*() functions, but for libraries. It
2179 factorize some code developped over and over in the examples and AMOK.
2180 It has still to be documented and used (only amok/peermanagement is
2182 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2183 loss when they were small enough to fit into the buffer and sent quickly
2184 enough so that they can all get received in one shoot.
2185 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2186 if not found. Use the *_or_null() variant for the old semantic.
2187 * In gras_msg_handle, do not discard messages without callback.
2188 They are probably messages to be explicitly awaited later (ie, proofs of
2189 mis-synchronization in userland since they are sent before being awaited)
2190 No big deal usually.
2191 * gras_socket_meas_send/recv: semantic changed!
2192 The numerical arguments used to be (1) the total amount of data to send
2193 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2194 messages. This was need for the fool willing to send more than MAXINT
2195 bytes on quite fat pipes.
2198 * Do really rename the hostmanagement module to peermanagement. [Mt]
2199 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2200 but this is rather new and this was documented in the module
2201 documentation (poor excuses, I admit)
2202 * Bandwidth measurement semantic changed! This follows the changes to
2203 gras_socket_meas_send/recv explained above.
2206 * A sequential mode has been added to the workstations. When a workstation
2207 is in sequential mode, it can execute only one task, and the other tasks
2208 are waiting in a FIFO. [Christophe Thiery]
2211 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2212 model for SIMDAG. [Christophe Thiery]
2213 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2214 equal to 0, which caused some bad behaviors in
2215 saturated_constraint_set_update. I now use a threshold mechanism like in
2219 * When running manually src/testall, you select specific units [Mt]
2220 testall is the result of our cunit mechanism, and should replace all
2221 the scripty thingy around since bash don't run easily on billware.
2223 * A mallocator system has been added. [Christophe Thiery]
2224 Mallocators allow you to recycle your unused objects instead of freeing them
2225 and allocating new ones.
2227 Documentation update:
2228 * FAQ reworking + New FAQs:
2229 - "Valgrind spits tons of errors!" [Mt]
2230 - "How to repport bugs" [Mt]
2231 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2232 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2233 - Communication time measurement within MSG [AL]
2234 - I experience weird communication times when I change the latency [AL]
2235 * GRAS tutorial [Mt]
2237 - an introduction to the framework and to the used communication model
2238 - an initiatic tour introducing the most proheminent features:
2240 . Lesson 0: Installing GRAS
2241 . Lesson 1: Setting up your own project
2242 o Part 2: Message passing
2243 . Lesson 2: Exchanging simple messages
2244 . Lesson 3: Passing arguments to the processes (in SG)
2245 . Lesson 4: Attaching callbacks to messages
2246 . Lesson 5: Using globals in processes
2247 . Lesson 6: Logging information properly
2248 . Lesson 7: Using internal timers
2249 . Lesson 8: Handling errors through exceptions
2250 . Lesson 9: Exchanging simple data
2251 . Lesson 10: Remote Procedure Calling (RPC)
2252 . Lesson 11: Explicitely waiting for messages
2253 . Recapping of message passing features in GRAS
2254 - A HOWTO section containing:
2255 o HOWTO design a GRAS application
2256 More are due, of course. They will come latter. In the meanwhile, you can
2257 check the examples which are still here.
2259 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2261 SimGrid (3.1) stable; urgency=high
2265 There was a stack corruption somewhere, visible only when optimizing
2266 with these versions. [Vince]
2269 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2270 module that enabled to play with Directed Acyclic Graphs. It is built
2271 directly on top of SURF and provides an API rather close to the old
2272 SG. Some old codes using SG are currently under rewrite to check that
2273 all needful functions are provided. [Christophe Thiery]
2276 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2277 extension of the classical CLM03 model that gracefully handles
2278 failures. This is now the default model for MSG and GRAS. It doesn't
2279 handle parallel tasks yet though. [AL]
2280 * Bug fix: Weights were not correctly set in the network part.
2281 WARNING: This may have resulted in incorrect results with simulations
2282 where there are more than one flow on a given link. [AL]
2285 * After a (long ?) discussion on simgrid-devel, we have decided that the
2286 convention we had on units was stupid. That is why it has been decided
2287 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2288 WARNING : This means that all previous platform files will not work as
2289 such with this version! A warning is issued to ask users to update
2291 A conversion script can be found in the contrib module of the CVS, under
2292 the name contrib/platform_generation/surfxml_update.pl [MQ]
2295 * Bug fix: Processes were started in reverse order, wrt deployment file.
2296 WARNING: if your code relies on this bug, please fix it. [AL]
2297 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2298 being executed on two different locations. [AL]
2299 * Bug fix: Failures are now better supported thanks to Derrick's tests
2300 (there was many failure situations I hadn't thought of and that weren't
2301 correctly handled). [AL]
2302 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2306 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2307 Exception occurring on server-side are propagated back to client (!).
2309 API CHANGE: the callback changed their prototype. Change:
2310 int my_handler(gras_socket_t expeditor, void *payload_data) {
2312 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2313 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2315 * New! function: gras_msg_handleall to deal with all messages arriving
2316 within a given period.
2317 * New! function: gras_socket_server_range to get a server socket in a
2318 range of port numbers (ease to avoid port number conflicts) [MQ]
2319 * New! gras processes display their backtrace when they get a SIGUSR1
2320 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2321 Sweet to debug RL processes [MQ]
2325 - Do not force experiment sizes to be expressed in kb, or it becomes
2326 impossible to measure the latency this way (needs one byte-long tests)
2327 WARNING: this changes the amok_bw_* function semantic. [MQ]
2328 - Implements the link saturation stuff. [MQ]
2329 * Peer management module:
2330 New! module factorizing code that we wrote over and over [MQ].
2333 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2334 - Test units are placed directly into the library code, they get extracted
2335 automatically and placed into the src/testall binary.
2336 - Convert most of the XBT tests to this system.
2337 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2338 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2339 live in the same C code. This required to change a little bit the API
2340 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2341 * New module: sparse graph structure with basic algorithms (this is work
2342 in progress and the API is not considered to be frozen yet). [AL]
2343 * Display more information on backtraces: source line & function names are
2344 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2345 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2346 * Reworked a little bit some #include statements to load only required
2347 headers. Some user code that relied on SimGrid to include stdlib or
2348 stdio may need to include it by themselves. [AL]
2349 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2351 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2352 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2353 but I guess that almost nobody used those parts. [MQ]
2355 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2357 SimGrid (3.0.1) stable; urgency=low
2360 * Unfortunately, I had missed 5 misnamed functions:
2361 xbt_fifo_item_t xbt_fifo_newitem(void);
2362 void xbt_fifo_freeitem(xbt_fifo_item_t);
2363 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2364 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2365 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2366 They're now deprecated. Please use their new versions:
2367 xbt_fifo_item_t xbt_fifo_new_item(void);
2368 void xbt_fifo_free_item(xbt_fifo_item_t);
2369 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2370 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2371 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2373 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2374 * Documentation: xbt_log module unmercifully reworked [MQ]
2375 * Bugfix: there was a problem with the ending of contexts with
2376 the pthread backend. It caused some weird deadlock or behavior
2377 depending on the pthread implementation. [AL]
2378 * Bugfix: get the exceptions raised in the simulator repport where
2379 and why they come from when they are not catched in time [AL, MQ]
2382 * Bugfix: Do repport the error when two non-connected hosts try to
2383 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2386 * Add additionnal checkings on communications. Assert that two
2387 communicating hosts are connected by a set of links... [AL]
2390 * Add additionnal checkings on channel values in communication [AL]
2391 * New: MSG_task_get_source to see on which host a task was generated [HC]
2392 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2393 the number of tasks waiting to be received on channel and sent
2395 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2396 waits for the first task coming from a given host.. [AL]
2398 GRAS new functionnalities: [MQ]
2399 * Enhance the parsing macro to allow the size of multidimentional objects
2400 to be given thru annotations.
2401 * New example (and documentation): Matrix Multiplication a la RPC
2402 (as when I was young!) and fix a bunch of bugs found on the way.
2404 GRAS performance improvements: [MQ]
2406 * Reduce the amount of cbps creation/destruction by making it static to
2407 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2410 * Change libdata to a set so that we can search for stuff by ID (and thus
2411 reduce the insane amount of dict lookups)
2414 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2415 RL communication on top of SG since it's so uneffective.
2416 It may also be used for inter-thread communication in RL, one day.
2417 * Use gras_datadesc_copy() to exchange messages on top of SG
2419 - improve message exchange performance on top of SG
2420 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2421 * Don't exchange on the network the size of the used part of buffer,
2422 instead, specify the possible buffer size to read().
2424 - reduces the amount of read/write calls (one pair per exchange)
2425 - reduces the amount of exchanged data (the size)
2426 - allows to retrieve all arrived data on receiver side, if we don't need
2427 it right now (subsequent read will peek the buffer)
2428 - allows the receiver to proceed with the begining of the stream before
2429 everything is arrived
2430 - make it possible to build an iov transport (using readv/writev)
2432 - take care of the data with non-stable storage (like stacked data),
2434 * If possible, TCP send uses vector I/O (when writev() is here)
2435 - Don't use it for receive since we send data sizes and data on the
2436 same stream, so we wouldn't be able to chain large amount of chunks
2437 before having to flush the stuff to read the size.
2438 * Rework the transport plugin mechanism to simplify it and reduce the
2439 amount of pointer dereferencement when searching for the right function
2442 * I guess that now, we do almost as few system calls as possible while
2443 doing as few data copy as possible.
2445 To improve it further, we could try to send all the sizes first and then
2446 all the data (to use iov on receiving size), but it's only a partial
2447 solution: when you have 2 dimensional data, the sizes of the second
2448 dimension is data of the first dimension, so you need 3 streams.
2450 I'm not sure the potential performance gains justify the coding burden.
2452 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2454 SimGrid (3.00) stable; urgency=high
2457 * New! Give the possibility to hijack the surf parser and thus bypass
2458 MSG_create_environment and MSG_launch_application. Have a look at
2459 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2461 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2463 SimGrid (2.96) unstable; urgency=low
2468 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2469 This deprecates the xbt_error_t mechanisms.
2470 It modifies (simplifies) all XBT and GRAS API.
2471 MSG API keeps unchanged (exceptions raised by XBT are catched from
2472 within MSG and masked with existing error handling facilities)
2475 * New! Add a FATPIPE model. [AL]
2476 * New! Add a parallel task model. [AL]
2477 * New! Add automatically a loopback interface (in the default
2478 network model) if none was precised.
2481 * Bugfix: MSG_process_resume now works with the current running process.
2483 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2484 * Modification of MSG_task_get_compute_duration. Once a task has been
2485 processed, the value returned by this function is now equal to 0. [AL]
2486 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2487 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2488 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2489 of a m_task. For now, MSG_task_cancel only works with computation
2491 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2492 of the processor (in Mflop/s) regardless of the current load on the
2494 * API Change: use proper naming convention for MSG_getClock and
2495 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2497 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2498 This function changes the priority of a computation task. This priority
2499 doesn't affect the transfer rate. A priority of 2 will make a task
2500 receive two times more cpu power than the other ones. This function
2501 has been added to suit the needs of Nguyen The Loc and hasn't been that
2502 much tested yet. So if it fails, please report it and send me your code.
2504 * API Change: removed all functions and types that were marked "deprecated"
2505 since many months. Renamed MSG_global_init_args to MSG_global_init.
2507 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2509 SimGrid (2.95) unstable; urgency=low
2512 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2513 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2514 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2517 * Bugfix: Allow absolute paths for platform description files [MQ]
2518 * Bugfix: do free the variables after use. Leads to drastic performance
2520 * Implement max_duration (ie, timeouts) on resources [AL]
2523 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2525 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2526 the needed support (timeouts) [AL]
2529 * Implement measurement sockets. You can now get the bandwidth between two
2530 hosts thanks to AMOK (see below). [MQ]
2531 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2532 dynar over the network (yeah) [MQ]
2533 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2534 * simple/static token-ring example. [Alexandre Colucci and MQ]
2535 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2536 in place before [MQ]
2539 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2540 type parsing. "short" and "long" modifiers were also ignored; other
2541 modifier (such as reference level) are still ignored. [MQ]
2542 * Embeed the buffer size within the buffer itself on SG. [MQ]
2543 That way, send() are atomic and cannot get intermixed anymore (at least
2544 the ones which are less than 100k; bigger messages still have the issue)
2545 * Array size pushed by the field, not by the field type (or each
2546 and every long int will push stuff to the cbps) [MQ]
2547 * use select() to sleep since it allows to portably sleep less than one
2550 GRAS (minor cleanups)
2551 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2554 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2555 * type category 'ignored' killed as it was never used and were difficult
2557 * whether a type can cycle or not is now a flag, leaving room for more
2558 flags (as "ignored", if we feel the need one day ;)
2559 * Rename raw sockets to measurement sockets since "raw" has another
2560 meaning in networking community.
2563 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2564 and offers features not belonging to GRAS but that most applications
2565 need. One day, it may be a set of plugins loadable at runtime.
2566 * New module: bandwidth
2567 bandwidth measurement between arbitrary nodes running this module. [MQ]
2569 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2571 SimGrid (2.94) unstable; urgency=low
2573 The first beta release of SimGrid 3 !
2577 * Update the main page and the FAQ. Adding references to gforge.
2580 * Add a gras_os_getpid function.
2583 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2584 * Extend the logs so that they also print PID, hostname, date, ... if
2586 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2587 and kill the old version which were in testsuite/
2588 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2592 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2593 for users. Instead, each programming environment (GRAS, MSG,...) use it
2594 when needed to provide such a feature to users.
2595 Don't shortcut the mechanism or you will also shortcut the virtualization
2596 you need on the simulator.
2600 * Cleanups in configury with regard to compile optimization/warning flags.
2601 Also add -fno-loop-optimize to any powerpc since it's the optimization
2602 killing gcc (< 3.4.0).
2603 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2604 needed by MSG examples complications
2605 * Borrow configury beautifications from PHP
2608 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2609 warning (thanks loris for stumbling into this one).
2610 * Bugfix: stop loading private headers (gras_config.h) from the public
2614 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2615 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2616 RL would imply a DNS resolver, which is *hard* to do in a portable way
2617 (and therefore delayed).
2618 * Implement a real timer mechanism and use it in timing macros. This allows
2619 to avoid rounding errors and get a 0.000005 sec precision in timing
2620 macros. While I was at it, various cleanups:
2621 - allow to declare more than one timed section per file (fix a stupid bug)
2622 - move some private declaration to the right place
2623 - merge conditional execution and timing macros into emulation module
2624 - document the module
2625 - make sure the module cleanups its mess on gras_exit
2626 * Documentation improvements:
2627 - (new) how to compile applications using GRAS
2628 - (new) emulation support (timing macros)
2630 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2632 SimGrid (2.93) unstable; urgency=low
2634 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2637 - Use Paje properly where used. Still to be sanitized properly.
2638 - Portability fix: Add an implementation of the contexts using pthread
2642 - Add xbt_procname(): returns the name of the current process.
2643 Use it to show the current process's name in all logging.
2645 - fix detection of older flex version and the reaction, since we do
2646 depend on modern ones (we use lex_destroy)
2647 - Better separation of SG and RL in the libs: remove all simulation code
2648 from libgras. As a result, this lib is now only 200k when stripped.
2649 Some of the xbt modules may also be duplicated (two sets and such) and
2650 should be cleaned/killed before SG3.
2651 - Insist on using xlC on AIX because of weird problems involving gcc there.
2652 - Cleanup the make remote stuff. This is now done by scripts
2653 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2654 mainly for our private use, but we're working on changing them to user
2657 - Bugfix: flush the socket on close only if there is some *output*.
2658 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2660 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2663 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2665 SimGrid (2.92) unstable; urgency=low
2667 Alpha 3 on the path to SimGrid 3
2671 - New! First try of benchmarking macros.
2672 - New! First try so that gras_stub_generator generate deployment and
2673 remote compilation helpers.
2675 - Bugfix: Initialization fix in msg_test.
2679 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2681 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2683 - API Change: gras_os_sleep to take the amount of seconds as a double.
2684 Accepting an int was error prone since it was the only location where
2685 seconds were coded as such. It leaded to damn rounding errors.
2686 - Bugfix: Hard to belive that timers ever worked before this.
2688 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2690 SimGrid (2.91) unstable; urgency=low
2692 Alpha 2 on the path to SimGrid 3
2696 - Bug fix in the lmm_solver.
2698 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2699 through the function MSG_paje_output.
2700 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2701 - It is possible to bound the rate of a communication in MSG with
2702 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2703 to write it in the changelog).
2704 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2709 - Various cleanups to the autotools stuff
2710 - Begin to move Gras examples to examples/gras/
2711 - Let make distcheck work again (yeah!)
2713 - documentation overhauled using doxygen.
2714 gtk-doc-tools is dead in SimGrid now.
2715 - Automatically extract all existing logging categories, and add the list
2716 to the documentation (long standing one, to say the less)
2718 - Cleanup the known architecture table. Reorder the entries to group what
2719 should be, and use a more consistent naming scheme.
2720 (some of the test dataset are still to be regenerated)
2721 - New! Allow library to register globals on each process just as userdata
2723 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2724 lookup time (for now).
2725 Use it in msg and trp.
2726 This cleans a lot the internals and helps enforcing privacy of the
2727 headers between the gras components.
2728 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2729 - Bugfix: gras_os_time was delirious in RL.
2730 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2732 - Reenable GRAS now that it works.
2734 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2736 SimGrid (2.90) unstable; urgency=low
2738 Alpha 1 on the path to SimGrid 3
2740 * It is a long time since the last release of SimGrid. I'm sorry about
2741 that but as I had told you, I was rewriting a lot of things. I apologize
2742 to those who had been reporting bugs to me and that I had not answered.
2743 If your bug is still in the new version, please tell me. Here is a
2744 summary of the main changes.
2746 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2747 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2748 portability, speed, and a lot more but you'll figure it out later.
2749 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2750 I wanted to release the new MSG as soon as possible and I have
2751 broken GRAS, which is the reason why, for now, only MSG is fully
2752 functional. A laconic description of these projects is available
2753 in the documentation.
2755 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2756 kernel optimized for our needs (called SURF but only the developers
2757 should use it). Hence, MSG is now roughly 30 times faster and I think
2758 that by rewriting a little bit MSG, I could event speed it up a little
2759 bit more. Beside the gain in speed, it is also much easier to encode a
2760 new platform model with SURF than it was with SG. More to come...
2762 * REVOLUTION 3: I have tried to change a little as possible the API of
2763 MSG but a few things really had to disappear. The main differences
2764 with the previous version are :
2765 1) no more m_links_t and the corresponding functions. Platforms are
2766 directly read from a XML description and cannot be hard-coded
2767 anymore. The same format is used for application deployment
2768 description. The new format is described in the documentation.
2769 Have a look in tools/platform_generation. There is a tiny script
2770 that converts from the old platform format to the new one. Concerning
2771 the application deployment format, parsing the old one is tricky.
2772 I think most of you should however be able to convert your files. If
2773 it is really an issue, I can write a C code that does the conversion.
2775 2) the toolbox tbx does not exist anymore. We now have a library
2776 with much more data-structures but without the hash-tables (we have
2777 dictionaries that are much faster).
2779 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2781 *****************************************************************************
2782 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2783 * don't feel like converting the oldies. (Mt) *
2784 *****************************************************************************
2787 Version 2.90: "the long awaited one"
2788 - Finished rewriting and debugging MSG. Rewrote the documentation.
2789 - disable GRAS for now since it needs to be ported to the newest SG
2792 - Finish the port to windows (using mingw32 for cross-compile)
2795 - Main loop and datastructures of SURF. A cpu resource object is
2796 functional. Surf can thus be used to create cpu's with variable
2797 performance on which you can execute some actions.
2799 2004-11-15 Martin Quinson
2800 - Port to ARM. Simply added the alignment and size descriptions. Should
2801 work, but the ARM machines are so slow that I didn't had the opportunity
2802 to 'make check' over there yet.
2804 2004-11-15 Arnaud Legrand
2805 - Trace manager now written. It uses a heap structure and is therefore
2806 expected to be efficient. It may however be speeded up (particularly
2807 when many events occur at the same date) by using red and black
2808 trees. One day maybe...
2809 - Max-min linear system solver written. It uses a sparse matrix
2810 structure taking advantage of its expected use. Most operations are
2811 O(1) and free/calloc are called as few as possible. The computation of
2812 the minimum could however be improved by using a red and black tree
2815 2004-11-03 Arnaud Legrand
2816 - Rename every gras_* function that was in xbt/ to its xbt_
2818 - Add a heap and a doubly-linked list to xbt
2819 - Added a dichotomy to the dictionaries. make check works as well before
2820 so I assume that the patch is correct. I do not know however if things
2821 run effectively faster than before now. :)
2823 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2824 SimGrid, and the version number is bumped to 2.x
2826 2004-10-29 Martin Quinson
2827 - Introduction of the remote errors.
2828 They are the result of a RMI/RPC on the remote machine.
2829 ErrCodes being scalar values, you can't get the host on which those
2830 errors did happen. Extending the error mechanism as in Gnome is possible.
2831 No idea yet whether it is a good idea.
2833 2004-10-28 Martin Quinson
2834 - Interface revolution: the Starred Structure Eradication.
2835 I used to do typedef struct {} toto_t; and then handle *toto_t.
2836 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2837 - ???_t is a valid type (builded with typedef)
2838 - s_toto_t is a structure (access to fields with .)
2839 - s_toto is a structure needing 'struct' keyword to be used
2840 - e_toto_t is an enum
2841 - toto_t is an 'object' (struct*)
2843 typedef struct s_toto {} s_toto_t, *toto_t;
2844 typedef enum {} e_toto_t;
2845 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2846 s_toto_t) is private.
2848 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2849 so that it can set the variable to NULL. It was so for dicts and sets,
2850 it changed for dynars.
2852 - Fix a bunch of memleaks in dict_remove
2853 - Fix a bug in sg/server_socket opening: it failed all the time.
2855 2004-10-07 Martin Quinson
2856 - Speed up dynar lookup operation a bit.
2858 gras_dynar_get is dead.
2860 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2861 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2862 (which gives you the address of the stored data).
2864 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2865 copy of the data using an affectation to do the job and not a memcpy.
2867 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2868 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2870 It does not really speedup the dynar test because they are
2871 setting elements all the time (and look them seldom). But the dict does
2872 far more lookup than setting.
2874 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2876 2004-10-05 Martin Quinson
2877 - Allow to (en/dis)able the cycle detection at run time.
2879 Whether we should check for cycle or not is now a property of each
2880 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2881 datadesc_cycle_unset allow to remove this property when previously set.
2883 Note that the cycle detection is off by default since it impacts the
2884 performance. Watch the data you feed GRAS with ;)
2886 This property is hereditary. Any element embedded in a structure having it
2887 set have it set for the time of this data exchange.
2889 You should set it both on sender and receiver side. If you don't set it on
2890 sender side, it will enter an endless loop. If you forget on receiver
2891 side, the cycles won't be recreated after communication.
2893 - Header reorganization.
2894 Kill gras_private.h, each submodule must load the headers it needs.
2896 2004-10-04 Martin Quinson
2897 - Interface revolution: do not try to survive to malloc failure.
2899 Now, gras_malloc and friends call gras_abort() on failure.
2900 As a conclusion, malloc_error is not a valid error anymore, and all
2901 functions for which it was the only gras_error_t return value are
2902 changed. They now return void, or there result directly.
2903 This simplify the API a lot.
2905 2004-09-29 Martin Quinson
2906 - Re-enable raw sockets.
2907 Created by gras_socket_{client,server}_ext;
2908 Used with gras_raw_{send,recv}
2911 It should allow to kill the last bits of gras first version soon.
2913 This is not completely satisfactory yet (duplicate code with
2914 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2917 - Simplify transport plugin (internal) interface by not passing any
2918 argument to _server and _client, but embedding them in the socket
2921 2004-09-28 Martin Quinson
2922 - Finish the port to AIX.
2923 autoconf was my problem (segfault within the malloc replacement
2924 function. No idea why)
2926 2004-09-16 Martin Quinson
2927 - Fix some size_t madness on 64bit architectures.
2929 2004-09-08 Martin Quinson
2930 - Reduce the number of system headers loaded, overload some more system
2931 calls (such as malloc to cast the result of the system one, and work
2933 - Fix and reintroduce the config support
2935 2004-09-07 Martin Quinson
2936 - Source code reorganization to allow Arnaud to surf all over there.
2937 - Allow to document the logging categories.
2938 - Remove all uppercase from logging categories and useless cleanup in names.
2940 2004-08-18 Martin Quinson
2941 Version 0.6.2 (protocol not changed; API changed)
2942 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2943 gras_error_t), and NULL when not found. Functions expecting a msgtype
2944 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2945 hopefully usefull message.
2946 - Portability to prehistoric sparcs again
2948 2004-08-17 Martin Quinson
2949 Version 0.6.1 (protocol not changed; ABI not changed)
2950 - prealloc some buffers to speed things up
2952 2004-08-11 Martin Quinson
2953 Version 0.6 (protocol not changed; ABI expended)
2954 - The parsing macro can deal with the references, provided that you add
2955 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2957 2004-08-09 Martin Quinson
2958 Version 0.5 (protocol not changed; ABI changed)
2959 - Allow to off turn the cycle detection code in data exchange at
2960 compilation time. It should be at run time, but I'm short of time (and
2961 the config stuff is still broken). That way, we keep dict out of the
2962 critical path, which is good because the performance is poor:
2963 - search not dichotomial yet
2964 - dynar give no way to access their content and memcpy everytime
2965 - In composed data description (struct, ref and so on), stop foolness of
2966 keeping the subtype's ID, but store the type itself. This keeps sets out
2967 of the critical path, which is good since they rely on dynar and
2968 dictionnaries. The only loose of that is that we cannot detect the
2969 redeclaration of a structure/union with another content (but I'm not sure
2970 the code detected well this error before anyway). We still can detect
2971 the redefinition discrepancy for the other types.
2972 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2973 it breaks the code because of type-punning used all over the place).
2974 This breaks on all non-gcc architectures (for now).
2976 All those changes (plus the buffer of last time) allow me to gain 2 order
2977 of magnitude on cruel tests consisting of 800000 array of integers on two
2978 level of a hierarchical structure (200 secondes -> 4 secondes)
2981 - the selector of reference must now return the type it points to, not
2982 the ID of this type.
2984 2004-08-06 Martin Quinson
2985 Version 0.4 (protocol changed; ABI not changed)
2986 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2987 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2988 - Implement a buffer transport, and use it by default (it relies on tcp in
2989 real life and on sg in simulation).
2990 That's a bit hackish since I had a new field to the structure to store
2991 its data without interfering with the subtype ones. Inheritance
2992 is tricky in C. And that's a kind of reverse inheritance with one class
2993 derivating two classes. Or maybe a game with java interfaces. Anyway,
2994 that's damn hard in C (at least).
2995 Moreover, I got tired while trying to ensure plugin separation and
2996 genericity in SG mode. MSG wants me to do weird things, so let's go for
2997 cruel hacks (temporarily of course ;).
2998 See comment in transport_private.h:71
2999 - do not include all the _interface headers in private but in the files
3000 which really need them (to cut the compilation time when they are
3003 2004-07-26 Martin Quinson
3004 Version 0.3 (protocol not changed; ABI changed)
3005 - Major overhault of the datadesc interface to simplify it:
3006 - shorted the function names:
3007 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3008 - add a trivial way to push/pop integers into the cbps without malloc.
3009 This allows to make really generic sub_type description, which simply
3010 pop their size of the stack.
3011 - add a function gras_datadesc_ref_pop_arr() which does what users want
3012 most of the time: Declare a dynamic array (which pops its size of the
3013 stack) and declare a reference to it. Poor name, but anyway.
3014 - kill the post-send callback, add a post-receive one
3016 2004-07-23 Martin Quinson
3017 Version 0.2 (protocol changed; ABI changed)
3018 - add some testing for cpbs in the test cases, and fix some more bugs.
3019 This invalidate again the little64 data file, since I cannot regenerate
3021 - remove an awfull optimization in the logging stuff, allowing me to:
3022 - understand it again
3023 - learn gcc how to check that the argument match the provided format
3024 - fix all errors revealed by gcc after that
3025 - internal keys of dict are not \0 terminated. Deal with it properly in
3026 loggings instead of segfaulting when the user want to see the logs :-/
3028 2004-07-22 Martin Quinson
3029 - Fix some stupid bug preventing cbps (callback postit) from working
3031 2004-07-21 Martin Quinson
3032 - Some documentation cleanups
3033 - remove the useless last argument of msgtype_declare
3034 - rename the Virtu functions to fit into the 'os' namespace
3035 - move headers src/include -> src/include/gras/ and stop fooling with
3036 gras -> . symbolic link
3037 - make distcheck is now successful
3039 2004-07-19 Martin Quinson
3041 - Build shared library also
3042 - Install html doc to the right location
3043 - stop removing maintainer files in make clean
3044 - build tests only on make check
3046 2004-07-13 Martin Quinson
3048 - No major issue in previous version => change versioning schema
3049 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3050 corresponding dataset.
3052 2004-07-11 Martin Quinson
3054 - Get it working with any kind of structure (we can compute the padding
3055 bytes remotely for all the architectures I have access to)
3056 - Implement the structure parsing macro (still not quite robust/complete)
3057 - Improvement to the remote testing toysuite
3059 2004-07-10 Martin Quinson
3060 [autoconf mechanism]
3061 - get ride of a bunch of deprecated macros
3062 - actually run the test for two-compliment, not only compile it :-/
3063 - test whether the structures get packed (and bail out if yes. Damn.
3064 Alignment is a serious matter)
3065 - test whether the structures get compacted (but respecting the alignment
3066 constraints of each types)
3067 - test whether the array fields of structures can straddle alignment boundaries
3069 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3070 (took me 2 hours to find that bug, looking at the wrong place)
3071 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3072 before sending/receiving and are used to compute the offsets of fields
3073 - Given that padding size depend even on compiler options, keep track of
3074 alignment and aligned_size only for the current architecture. Not a big
3075 deal since we send structure fields one after the other (seems
3077 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3078 program, let it work on linux/gcc/little32. portability todo.
3080 2004-07-08 Martin Quinson
3081 - import and improve remote compilation support from FAST
3082 - make sure make check works on half a dozen of machines out there
3084 2004-07-07 Martin Quinson
3085 Let's say it's version 0.0.3 ;)
3086 - Implement conversions (yuhu!)
3087 - Let it work on solaris (beside conversion, of course)
3088 - Stupid me, using rand() to generate the conversion datatests in not wise.
3090 2004-07-06 Martin Quinson
3091 - Let make dist work, since I'm gonna need it to compile on remote hosts
3092 - Let Tests/datadesc_usage write the architecture on which the file was
3093 generated as first byte.
3094 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3095 architecture descriptions.
3096 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3097 on those architectures.
3098 - Optimization: send/recv array of scalar in one shoot
3100 2004-07-05 Martin Quinson
3101 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3103 - Plug a whole bunch of memleaks
3104 - each process now have to call gras_{init,exit}. One day, their log
3105 settings will be separated
3106 - Continue the code factorisation between SG, RL and common in Transport.
3108 2004-07-04 Martin Quinson
3110 - Redistribution between SG and RL.
3111 We wanna have to accept in SG, so move accepted related parts of RL in
3112 the common part. (more precisely, the dynar of all known sockets is no
3113 more a static in transport.c, but part of the process_data)
3115 [gras_stub_generator]
3116 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3118 2004-07-03 Martin Quinson
3119 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3120 them all up in one shot)
3122 - Fix the ugly case of reference to dynamic array.
3123 - New (semi-public) function gras_datadesc_size to allow the messaging
3124 layer to malloc the needed space for the buffer.
3126 - gras_socket_close now expect the socket to close (and not its address to
3127 put NULL in it after it). This is because the socket passed to handlers
3128 is one of their argument (=> not writable).
3130 - propagate the interface cleanup from last week in datadesc, ie remove a
3131 superfluous level of indirection. User pass adress of variable
3132 containing data (both when sending and receiving), and not of a variable
3133 being a pointer to the data. Let's say that I like it better ;)
3134 The price for that is constructs like "int msg=*(int*)payload" in
3135 handlers, but it's a fine price, IMHO.
3137 - Let it work in RL (yuhu)
3139 2004-06-21 Martin Quinson
3141 - porting SG plugin and SG select to new standards (works almost).
3142 - plug memleaks and fix bugs around.
3145 - cleanup the prototype of data recv and force users to specify when they
3146 want to handle references to objects. Test case working even for cycles.
3147 - plug memleaks. Valgrind is perfectly ok with this.
3149 2004-06-12 Martin Quinson
3151 - cleanup the separation between plugin and main code in plugin creation
3153 2004-06-11 Martin Quinson
3155 - Reput hook for raw sockets, needed for BW experiments
3156 - kill a few lines of dead code
3157 [Data description] Interface cleanup
3158 - gras_datadesc_by_name returns the searched type or NULL.
3159 That way, no variable is needed to use a type desc once, which makes
3161 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3162 two parameters were strings (field name, type name), leading to
3164 [Dicos] Interface cleanup
3165 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3166 This is consistant with the dynar API.
3168 2004-04-21 Martin Quinson
3170 - Porting to new standards.
3172 - interface cleanup.
3173 There is no bag anymore, no need to take extra provision to mask the
3174 pointers behind "ID".
3175 Better splitup of functions between files create/exchange/convert.
3176 This is still a bit artificial since convert and receive are so
3177 interleaved, but anyway.
3179 - add a queued message list to procdata (the ones not matching criteria
3181 - factorize some more code between SG and RL wrt procdata
3183 - use gras_exit in example to track memleaks
3184 - get rid of gs_example now that GS is properly integrated into gras
3185 - update run_test to integrate the latest tests (datadesc)
3187 - rename WARNINGn macros to WARNn since it prooved error-prone
3189 2004-04-19 Martin Quinson
3191 - register init/exit functions within gras module mechanism
3192 - send/receive function.
3193 Convertion is not implemented, but short-cutted if not needed.
3194 struct/array elements are sent one by one (instead of block-wise), but
3195 nobody really cares (yet). Get a prototype before optimizing.
3196 - tests (using a file socket) for DD send/receive on:
3197 - base types: int, float
3198 - array: fixed size, string (ie ref to dynamic string)
3199 - structure: homogeneous, heterogeneous
3200 - chained list, graph with cycle
3201 Believe it or not, valgrind is not too unhappy with the results. The
3202 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3206 - Bugfix when using a filename explicitely (instead of '-')
3208 2004-04-09 Martin Quinson
3210 - factorize more code between RL and SG in socket creation
3211 - Complete the implementation and tests of:
3213 o file (only in RL, and mainly for debugging)
3215 I lost 3 days to design a portable address resolver, and then decided
3216 that the prototype mainly have to run on my box.
3217 Addressing portability too early may be like optimizing too early :-/
3219 - use gras_init in the Tests instead of the crappy parse_log_opt
3220 (the latter function is removed)
3221 [Conditional execution]
3222 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3223 [Code reorganisation]
3224 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3225 Build examples against the RL library, since there is no way to disable
3226 its creation for now.
3228 For information, the beginning of coding on GRAS was back in june
3229 2003. I guess that every line has been rewritten at least twice since