1 SimGrid (3.15) UNRELEASED; urgency=low
3 The Ever Green Release.
4 We fixed even the transient bugs on all target architectures:
5 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
6 Windows, FreeBSD, NetBSD.
8 - Rename NetCards to NetPoints.
9 This was intended to help NS3 users, but that's not a netcard.
10 That's a point in the routing algorithm, let's avoid wrong simplifications.
13 - New: MSG_process_yield(). Stop and yield to other processes.
14 - New: MSG_process_daemon(). Daemon processes are automatically killed
15 when the last non-daemon process terminates
16 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
17 to make room for the upcoming network energy plugin.
20 - Ensure that an actor can kill itself with Process::exit()
21 - Kill the obscure NativeException. Nobody want to survive the issues
22 it denotes, so use JniException that is a RuntimeException (not to
23 be catched explicitely).
27 - s4u::onPlatformCreated: right before the simulation starts
28 - s4u::onSimulationEnd: right after the main simulation loop
29 - s4u::onTimeAdvance: right after a clock change (time is discrete)
30 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
31 event from the availability_file changes the avail speed.
32 - Links are now usable from s4u
35 - Backwards Compatibility breaks
36 - The SD_route_get_list and SD_route_get_size functions have been removed.
37 They are replaced by the unique
38 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
39 The route from src to dst is built in the links dynar whose size can
40 be obtained with xbt_dynar_length.
41 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
42 sg_host_route_bandwidth and sg_host_route_latency.
43 Macros ensure the backwards compatibility, but you should fix your code
46 - Bump our claim of support from MPI 1.1 to MPI 2.2. We don't support 100% of
47 it, but it should be enough. Contact us if not.
48 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
49 mapped onto a unique file.
50 - Bugfix : smpirun was sometimes misusing hostfile when wrongly terminated
51 - Fortran : cleanups, fixes, support of user-added operations
52 - RMA : MPI_Accumulate are applied in correct order, and
53 MPI_Win_complete or MPI_Win_post shouldn't miss messages anymore.
54 - Better support for MPI_IN_PLACE
55 - Support MPI_Comm_group_create, MPI_Type_size_x
56 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
57 algorithm by a non failing one from mpich.
60 - Kill the fifo data container: we don't use it anymore.
63 - A <cluster> can now be created with different speed values to represent pstates.
64 see examples/platforms/energy_cluster.xml for an example.
66 -- target_date=March 20 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
68 SimGrid (3.14.159) stable; urgency=low
70 The previous release was only a crude approximation of a Pi release;
71 this one should be a bit better.
73 - Revert a last minute change that broke on Mac OSX.
74 - Fix the changelog and NEWS files.
75 - Improve the documentation, in particular of the routing module.
76 - Integrate some patches flying around in Debian.
78 -- Wed Dec 28 17:14:49 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
80 SimGrid (3.14) stable; urgency=low
83 * The whole documentation was reworked and reorganized. There is
84 still much room for improvement here, but we're on it.
87 * We now need python3 for our testing infrastructure.
88 * Model checking needs libevent
91 * Now works on FreeBSD
94 * Privatization now works on FreeBSD
96 * Privatization is now activated by default in smpirun
97 You can turn it off with -no-privatize if something goes wrong.
99 * Call-location tracing for SMPI.
100 You can add the exact location (filename / linenumber) of an MPI call to
101 your trace files and slow down or speed up the simulation between two
102 consecutive calls by using an adjustment file (see the documentation).
104 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
105 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
106 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
107 values were ignored in some cases. The timings of these functions can now
108 be significantly different.
110 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
111 smpi/running-power is renamed to smpi/host-speed
113 * smpi/grow-injected-times option to enable or disable multiplication of the
114 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
115 make simulation less precise (but also much faster).
117 * smpirun script should be (much) faster for large deployments.
119 * SMPI tracing : fixed issue with poor matching of send/receives.
121 * Replay : Fix broken waitall
123 New functions and features
124 * MSG_parallel_task_execute_with_timeout, to timeout computations.
126 Dropped / renamed functions and features
127 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
128 - MSG_mailbox_is_empty() -> Mailbox::empty()
129 - MSG_mailbox_front() -> Mailbox::front()
130 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
131 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
132 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
133 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
134 * Don't pass the free_f parameter to property related functions:
135 - MSG_host_set_property_value()
136 - MSG_as_router_set_property_value()
137 - MSG_storage_set_property_value()
138 * VM properties. Since msg_vm_t are msg_host_t, just use
139 MSG_host_get_property_value() and friends
140 * VM I/O related things:
141 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
142 - Unimplemented save/restore methods
143 * MSG_as_router_get_property_value() was redundent with
144 MSG_environment_as_get_property_value().
145 - Removed MSG_as_router_*propert*() functions
146 - Added MSG_environment_as_set_property_value() for consistency
147 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
148 Well, they are still there for now, but deprecated with a warning.
149 Please switch to xbt_dict_new_homogeneous() before this is removed
151 * Task affinity. Its intended behavior (that was very badly tested
152 and probably not really working) was deceiving what most users
153 would have hoped here.
154 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
155 * Option network/coordinates is now useless and should be dropped.
158 * Added option storage/max_file_descriptors to allow more than 1024 files opened
161 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
162 * Added Dragonfly topology support
165 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
166 EiBps, EBps, Eibps, Ebps for bandwidth.
167 They may become useful to some lucky ones.
170 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
171 * ex/app_tokenring: new example, very similar to the MSG Token Ring
172 * ex/async_waitAll: new example, on asynchronous communications
175 * Memory usage should be decreased for simulations with a large number
176 of processes. This also helps for SMPI.
178 -- Sat Dec 24 22:34:26 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
180 SimGrid (3.13) stable; urgency=low
182 The Half Release, a.k.a. the Zealous Easter Trim.
184 - We removed half of the lines, that were mostly experimental cruft.
185 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
186 (not counting blanks and comments -- according to openhub.net)
187 - The internals are now compiled in C++ (and will soon be clean C++)
188 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
189 59 klines of C, etc. We added only 29 klines of C++ in replacement.
191 * Backwards Compatibility breaks
192 - Removed Lua simulation bindings (switch to C or Java for that).
193 Lua can still be used to describe platforms
194 - Removed Java kernel plug-ins.
195 Will be reintroduced after the ongoing major internals reorg.
197 - the following functions were removed.
198 They were too specific and should be reimplemented in a generic
199 way, with filter function.
200 - MSG_task_listen_from_host
201 - MSG_mailbox_get_count_host_waiting_tasks
202 - MSG_mailbox_put_with_timeout was removed.
203 Please use MSG_task_send_with_timeout instead.
205 - the SD_application_reinit function was removed. It has been a noop for a while.
206 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
207 replaced by a more flexible mechanism.
208 - The following functions thus do not exist anymore
209 - SD_workstation_get_access_mode
210 - SD_workstation_set_access_mode
211 - SD_workstation_get_current_task
212 - Basic estimation functions have been removed but can easily be replaced
213 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
214 - SD_workstation_get_computation_time => amount / sg_host_speed()
216 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
217 Use VM.getSpeed()*load/100 for the legacy behavior.
219 - option enable_tracing was removed. It was not doing anything for a while.
220 - In the ModelChecker:
221 - the model-checker now ptraces the model-checked process which means
222 you cannot use a debugger on the latter anymore (we might make this
223 optional in the feature);
224 - removed soft-dirty page tracking;
225 - remove model-checked side snapshot management,
226 MC_snapshot() and MC_compare_snapshot();
227 - keep the MC_cut() function as a stub (it was not really working
228 in the previous release).
231 * All options are consistently in kebab-case. Old names are kept as alias.
234 * Switch to platform v4 format.
235 - Rename from 'power' to 'speed' the attributes describing the amount of
236 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
237 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
238 - In <host> and <link>, attributes availability and state are gone.
239 It was redundent with state and availability traces, and with peak values.
240 - In <cluster>, attributes availability_file and state_file are gone.
241 It was too complex and unused.
242 - Kill <gpu>. Was not doing anything.
243 - The DOCTYPE points to the right URL:
244 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
245 (the file at this address now documents the changelog since its v1)
246 - A warning is emitted for unit-less values (they are still accepted).
247 - speed. Default: 'f' or 'flops'. Also defined:
248 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
249 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
250 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
251 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
252 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
253 - latency. Default: 's' second. Also defined:
254 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
256 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
257 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
260 * s4u::Host is now the preferred public interface to the Host features.
261 sg_host_* functions are C bindings to the exact same behavior
262 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
265 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
266 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
267 They were deprecated since a few years, and probably did not even compile anymore.
270 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
271 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
272 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
273 encouraged to update their codes. A list of the modified functions can be found at the end of
274 include/simgrid/simdag.h
277 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
278 the lambda or closure passed as a parameter will run in kernel mode.
279 All the callback functions should be rewritten to that interface at some point.
282 * Reorganizing and cleaning the internals all around the place.
285 * Remove old default barrier/bcast buggy algorithms (see #18407)
286 * Various bug fixes to handle more codes
287 * Remove the need for the --foreground option of smpirun (it is still
288 accepted for backward compatibility).
291 * Kill the setset data container: MC don't use it anymore.
292 * Kill the queue data container: it made more sense with GRAS.
293 * Kill the xbt_peer_t data type: it's useless without GRAS.
294 * Kill rm_cb feature of config sets: it was never useful.
295 * Kill graphxml parsing feature. It was not used.
296 * Kill the deprecated code protected by XBT_USE_DEPRECATED
298 - xbt_dynar_sort_strings(), when the content is char*
299 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
300 They throw exceptions on invalid input;
301 * C++ support for declaring CLI flags (simgrid::config::Flag);
302 * class for abstracting different signal backends (simgrid::xbt::signal).
303 with no external dependencies (we need very simple signals).
306 * refactoring and cleanup of the code;
307 * ongoing process to cleanly separate the model-checking algorithms
308 from the code model-checking support.
310 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
312 SimGrid (3.12) stable; urgency=low
314 The Facelift Release.
317 * Require g++ v4.7 at least to not speak prehistorical C++.
318 * Require Boost 1.48 (for signal2 component).
319 * Java must be version 7 at least when activated.
320 * Builds on Windows again (including Java bindings).
321 * Tracing is now always enabled (no way to turn it out)
322 * Remove GTNetS. It was not working anyway.
323 * Various cleanups in the cmake scripts.
324 * Move headers around to sort them out on installed systems:
325 - instr/instr.h -> simgrid/instr.h
326 - instr/jedule/* -> simgrid/jedule
327 - simdag/datatypes.h was removed
328 - simdag/simdag.h -> simgrid/simdag.h
329 - msg/datatypes.h was removed
330 - msg/msg.h -> simgrid/msg.h
333 * Interface improvement:
334 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
335 - Sanitize the interface in MSG_task_ module:
336 - Merge two functions that were close enough but misleading:
337 set_compute_duration(t) -> set_flops_amount(t)
338 get_remaining_computation(t) -> get_flops_amount(t)
339 - set_data_size(t) -> set_bytes_amount(t)
340 get_data_size(t) -> get_bytes_amount(t)
341 - Massive cleanups in the functions related to the energy
342 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
343 - New: MSG_host_get_pstate()
345 - msg/energy/onoff: switching hosts on and off
347 * Interface improvement:
348 - Rename Host.isAvail() to Host.isOn()
349 - Rename Process.currentProcess() to Process.getCurrentProcess()
350 - Rename Task.setDataSize() to Task.setBytesAmount()
351 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
353 - #18874: Actually allows the GC to reclaim tasks
357 - SIMIX_process_throw: raises an exception in a remote process
358 * Refactoring: Separate sub-modules
359 - libsmx: the public interface, as libc in a real system
360 - popping: the strange dance that converts a user request into a kernel handling
361 - smx_context_*: the virtualization mechanisms that embed the user code
362 - smx_*: the handling of each simcalls
364 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
365 * Rename smx_action_t into smx_synchro_t, making explicit that these
366 things are used to synchronize processes with their environment.
367 For example, a communication is a sort of synchronization involving
368 the communicating processes (that may block until the exchange) and
369 the platform. The same can be said from computations, etc.
371 - #18888: segfault when a process ends before its kill_time
375 - 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.
376 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
377 - MPI_Type_set_name, MPI_Type_get_name
378 - MPI_*_c2f and MPI_*_f2c functions
379 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
380 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
381 - Activate a lot of new tests from the mpich 3 testsuite
383 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
384 - InfiniBand network model added : Based on the works of Jerome Vienne
385 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
386 - When smpi/display_timing is set, also display global simulation time and application times
387 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
388 * Collective communications
389 - 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
390 - 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/).
391 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
393 - Replay now uses algorithms from wanted collective selector
394 - Replay can be used with SMP-aware algorithms
395 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
396 - 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.
398 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
399 - Correctly use loopback on fat-tree clusters
400 - Asynchronous small messages shouldn't trigger deadlocks anymore
401 * Energy/DVFS cleanup and improvement
402 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
403 - new: smpi_get_host_pstate()
406 - "Full" network optimization flag was broken since Surf++
407 - Better handling of precision flags in maxmin
408 - Fix bug causing sometimes "Impossible" errors
409 - Properly pass cluster properties to included hosts
410 * Improvement of the Energy plugin.
411 - Always update the consumption before returning that value
412 - New property: watt_off to denote the dissipation when the host is off
413 - New functions getWattMinAt and getWattMaxAt to retrieve the
414 dissipation of pstates that we are not currently at.
415 * Java: class NetworkLink renamed to Link
416 * New function: simcall_process_get_kill_time()
417 * Massive rename s/workstation/host/
418 - That's intrusive, but that's good for the project consistency. Sorry.
419 - Change config option "workstation/model" into "host/model"
422 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
423 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
424 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
425 - xbt_dirname and xbt_basename for non-POSIX systems
427 * The model checker now runs as a separate process.
428 * The model checker runs must now be launched with the new simgrid-mc program.
429 * Record/Replay: the MC can display a textual representation of a path in the
430 execution graph. It can then be replayed outside of the model checker.
432 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
434 SimGrid (3.11) stable; urgency=low
439 * Normalizing pointers addresses tool for better diff between logs
442 * Add cloud examples using new VMs
443 - examples/msg/cloud/two_tasks_vm.tesh
444 - examples/msg/cloud/simple_vm.tesh
445 - examples/java/cloud/cloud.tesh
446 - examples/java/cloud/migration/migration.tesh
447 * Add java surf examples:
448 - examples/java/surfPlugin/surf_plugin.tesh
449 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
450 - examples/java/surfCpuModel/surf_cpu_model.tesh
451 * Add SMPI+MSG example:
452 - examples/smpi/smpi_msg_masterslave/
457 - msg task destroy cancel test
458 - msg_host on/off test
459 * Move all tests in testsuite to teshsuite (adding tesh files)
460 * Restructure teshsuites
461 - one folder for each kind of test
462 * Restructure AddTests.cmake
464 - structure the order of tests (with sections)
467 * Add virtual machine
468 - creation of a VM on a PM
469 - migration of a VM from a PM to another PM
471 - MSG_process_join(msg_process_t process, double timeout)
472 - msg_bar_t MSG_barrier_init(unsigned int count)
473 - int MSG_barrier_wait(msg_bar_t barrier)
474 - void MSG_barrier_destroy(msg_bar_t barrier)
475 - msg_as_t MSG_environment_as_get_by_name(const char * name)
476 * New option "msg/debug_multiple_use" to help debugging when a task is used
479 - Improvements and finalization of MSG_storage, MSG_file APIs and their
481 - Increase code coverage in test suites
485 * Protect context stacks against stack overflow. The number of protected memory
486 pages allocated on the top of each stack (1 by default) can be configured
487 with the new command line option --cfg=contexts/guard_size:<value>.
488 * Simcalls are now generated by a python script that
489 - generates files included by SimGrid
490 - checks that all the functions exist, and proposes prototypes
492 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
493 host_set_data, host_get_data
495 - simcall_process_join(smx_process_t process, double timeout)
496 * Fix bug where sleeping processing could not be suspended.
499 * Translate surf models from C to C++
500 - Generic classes for all models: Model, Resource, Action
501 - A generic interface for each kind of model (CPU, Network, Storage
502 Workstation, WorkstationVM)
504 * Translate surf routings from C to C++
505 * Add callbacks using sigc++ or boost::signals2
506 - Add callback functions for resource creation/destruction
507 - Add callback functions for action state change
508 - Handle Energy as a plugin
509 * Replace swag by boost::intrusive
510 * Add new routing models for clusters. For documentation, see
511 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
512 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
513 parameters for cluster tag
514 - Fat trees, with topology="FAT_TREE" and
515 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
516 - see examples/platforms/torus_cluster.xml and
517 examples/platforms/fat_tree_cluster.xml
521 * Hostfiles support host:nb_processes construct to deploy several processes on
523 * Collective communication algorithms should not crash if used with
524 improper number of nodes and report the error.
525 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
526 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
527 MPI_Dims_create, MPI_Cart_sub are supported.
528 * New interface to use SMPI programmatically (still depends on MSG for
529 some parts, see examples/smpi/smpi_msg_masterslave) :
530 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
534 * Global variables privatization in MPI executables is now performed at runtime
535 with the option smpi/privatize_global_variables (default:no).
536 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
537 dynamic libraries loaded by the application are not privatized (static
538 linking with these libraries is advised in this case)
541 * Options defined in XML work correctly now.
544 * New cmake option, enable_lib_in_jar, to control whether native libraries are
545 copied into simgrid.jar or not (ON by default). Use this option if you want
546 to reduce the size of the installed simgrid.jar, *and* the native libraries
547 are kept installed elsewhere.
548 * Surf binding with SWIG (code generated in maintainer mode only):
549 - plugin to handle callbacks
550 - CPU model only for the moment
553 * Supernovae build mode is definitively removed. It was used to improve
554 inlining and inter-module optimizations. It is nowadays superseded by
555 link-time optimizations commonly available in compilers.
556 * Update ns-3 find lib. Bindings for ns-3 should work again now.
557 * Add boost dependency for surf++
558 * Add new macro for tests
559 - ADD_TESH(name <tesh_args>)
560 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
564 - xbt_bar_t XBT_barrier_init(unsigned int count)
565 - int XBT_barrier_wait(xbt_bar_t barrier)
566 - void XBT_barrier_destroy(xbt_bar_t barrier)
567 * Make the xbt_os_time module public
569 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
571 SimGrid (3.10) stable; urgency=low
573 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
576 * Reintegrate Java to the main archive as desynchronizing these
577 package is not acceptable anymore (Java is now considered stable)
578 * Add explicit synchronization facilities through semaphores
579 * Bug fix: Task.setDataSize() only changed the C world, not the value
580 cached in the Java world
583 * Dramatically change the way files are handled. API and internals changed, but
584 this part of MSG was not considered as production grade either.
585 * Add explicit synchronization facilities through semaphores
586 * Add a new function MSG_host_get_process_list()
587 * Preliminary DVFS support (see examples/msg/energy/ for details)
590 * SMPI is now included directly in the libsimgrid as the windows
591 linker doesn't force us on splitting it anymore.
592 * Improvements of the SMPI replay tool:
593 - Most of the collective communications are now rooted in the same process as
594 in the original application.
595 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
596 used until now). Multiple data types can now be used in a trace.
597 - The replay tool now supports traces produce either by TAU or a modified
599 - Bug Fix: the compute part of the reduce action is now taken into account.
600 - Gatherv collective is now supported
601 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
602 of smpirun outputs time independent traces for the current run. One file
603 is created per process. If too many processes are simulated, this behavior
604 can be changed to one file for all processes by using the
605 tracing/smpi/format/ti_one_file flag
606 * smpirun generates the host file if needed (with given host count and platform)
607 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
608 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
609 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
610 --cfg=smpi/coll_selector:(mpich/ompi)
611 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
613 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
614 happening inside a collective SMPI call.
615 * Fix the behavior of complex data types handling.
616 * Make MPI_Wtime another synchronization point to take computations into
618 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
619 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
620 * Add all missing Fortran bindings, SMPI should work with Fortran 90
621 (no privatization of global variables yet)
622 * Preliminary DVFS support (see examples/smpi/energy/ for details)
625 * Verification of liveness properties is now available for SMPI applications
626 (in addition to MSG applications)
627 * Bugged examples using SMPI in examples/smpi/mc/
628 * Add --cfg=model-check/visited option. Allows the verification of infinite
629 programs. Detection of loops in the execution thanks to the system state
630 comparison and reduction of the state space to explore. Can be combined with
631 DPOR for safety properties.
634 * Allow to change SimGrid configuration (see --help) within the code
635 thanks to SD_config() as it can be done in MSG.
636 * Add a new function SD_task_set_amount() upon user request.
639 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
640 * Remove rule based routing (no more PCRE dependency)
641 * Add a limiter_link option to cluster tag, to specify a maximum reachable
642 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
643 * Add a loopback_bw and loopback_lat options to cluster tag.
644 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
645 routing. Now peers are encapsulated in an AS and have their own private
646 router but this is transparent.
649 * Our own implementation of getline is renamed xbt_getline, and gets
650 used even if the OS provide a getline(). This should reduce the
651 configuration complexity by using the same code on all platforms.
652 * New type: xbt_cfg_elm_boolean.
653 * Allow to use yes/no for boolean configuration options in the command line.
654 * Allow to disable SimGrid cleanups at exit from command line option.
655 There are situations where one may want a simulation to end with an exit.
656 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
657 annoying when scripting around the simulator. Adding a
658 --cfg=clean_atexit:no allows to circumvent this issue.
661 * Lots of memory leaks were corrected in this release.
662 There is no known memory leaks anymore, in all of our 600+ tests.
663 * New command line option --version, to get SimGrid version information.
664 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
666 * Supernovae builds are deprecated, and expected to be removed in the next
669 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
671 SimGrid (3.9) stable; urgency=low
673 The Grasgory release: GRAS is really dead now.
675 * Complete overhaul of the internal host structures scheme.
678 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
679 was considered as experimental and badly maintained since then.
680 * Keeping it was thus a trap to our potential users, that could take
681 it instead of MSG or SMPI by mistake despite is pity state.
682 * GRAS seems to have very few users (if any), and no one volunteered
683 to maintain it further. It also induces a lot of XBT code (for
684 portability sake), that must be maintained too.
685 * For all these reasons, we killed GRAS. If someone wants to revive it
686 in the future, don't cry, our git history still remembers of GRAS.
689 * Major overhaul. Merge our documentation again as time proved that
690 splitting it was really not helping our users.
691 * Further improve the developer documentation to help newcomers
692 hacking on SimGrid itself. The user documentation (and in
693 particular, the beginner documentation) is still in a sorry state.
696 * Now works on Windows too!
697 * Much more extensive test suite, from MPICH
700 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
701 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
702 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
703 * Bug fix: let task be scheduled when the last dependency to be solved is
704 a control dependency.
705 * Remove SD_load_environment_script function.
706 Use the C sg_platf function if you want to declare a platform
710 * New function: MSG_process_get_number()
711 * Old function documented: MSG_config()
712 * Remove MSG_load_platform_script function
713 Use the C sg_platf function if you want to declare a platform
717 * Change the default value of the TCP_gamma constant (maximal size of TCP
718 congestion window) to a more realistic 4MiB value. If you notice changes in
719 your simulation results, you can fall back to the previous 20k tiny window
720 by adding --cfg=network/TCP_gamma:20000 on command line.
721 * (Hopefully) fix a bug wrt periodic availability and state traces
722 * Bug fix: use default values at start when first event in availability/state
723 trace is not at time 0.
726 * remove the "new_" part of function name sg_platf_new_trace_connect
727 (resulting in sg_platf_trace_connect), since it does not create
731 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
732 use case where it's really mandatory, and maintaining it was a pain in
734 * New: xbt_fifo_search(), search an item with a user-provided
735 comparison function instead of dumb pointer comparison.
738 * Fix the lua deployment:
739 Use `simgrid.init_application()` before deployment instead of
740 `simgrid.msg_register_application()` after.
743 * Transfer the tracing files into the corresponding modules.
745 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
747 SimGrid (3.8.1) stable; urgency=low
749 The "we are told that some people want to also *install* the simgrid
752 * Add missing file "tesh.1" to the archive.
754 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
756 SimGrid (3.8) stable; urgency=low
758 The Psssshiiiit release: SimGrid jumps into the Cloud.
761 * Add an experimental interface to manipulate VMs. They are mainly
762 process groups with very few intrinsic semantic, but they should
763 allow you to build the semantic you want easily.
764 * New function: MSG_host_set_property_value()
765 * New function: MSG_process_on_exit(). To clean memory in all cases.
766 * Bug fixes that made the host (and link) failures unusable.
767 * Add a way to auto-restart process when the host in which they are
768 executing comes back (ON_FAILURE="RESTART" on deployment file,
769 MSG_process_auto_restart_set).
770 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
771 please stop using the old ones, they are DEPRECATED.
773 * Deprecate functions MSG_global_init() / MSG_global_init_args()
774 Please use MSG_init() instead. (reducing the amount of entry
775 points in the library helps us).
776 * Make it impossible to link against the wrong version of the lib
777 * Deprecate MSG_clean(). No need to call it anymore.
778 * Function MSG_get_host_number() is not deprecated anymore.
781 * Split the doc into a user guide and a reference guide.
782 * Start a developper guide to help people hacking on SimGrid.
785 * Enable tracing by default. This modules rocks you should use it.
786 * Remove option custom_flags. Now use environment variables CFLAGS
788 * Use default cmake things to detect lua instead of home grown ones.
789 * New option "enable_mallocators" to disable mallocators, for debugging
790 purpose ("on" by default).
793 * Bug fixes around the resource failures: don't let the processes
794 survive the host they are running onto.
795 * Add an interface to auto-restart processes when the host in which they are
796 executing comes back.
797 * Ensures that SIMIX_clean is called automatically. It's not part of
798 the public interface anymore (bindings should be updated).
801 * Bug fix for when SD_Simulate is called with a positive value: be careful
802 when comparing doubles. Sometimes they are different for non significant
804 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
805 parallel task whose initial work is distributed among host according
806 to the Amdahl's law. Such tasks are created with a parameter alpha
807 that corresponds to the non-parallelizable part of the computation.
808 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
809 between two sets of workstations assuming a 1D block distribution (each
810 workstation owns a similar share of data) on both sides.
812 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
813 Data redistribution will be automatically scheduled once parent and child
814 are both scheduled. The filling of computation_amount and
815 communication_amount structures is now done seamlessly thanks to the chosen
817 * New function SD_workstation_dump to display various information
818 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
819 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
820 bandwidth on the route onto which the task is scheduled and the amount of
822 To divide the nominal bandwidth by 2, the rate then has to be :
823 rate = bandwidth/(2*amount)
824 * Compute tasks that have failed can now be rescheduled and executed again
825 (from their beginning)
826 * Increasing source code coverage (src/simdag is now covered at 95.8%
830 * Re-implement time-independent trace replay using SMPI (at the
831 smpi_smp_* level) instead of MSG. This should replace
832 examples/msg/actions/actions.c
833 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
834 hindexed and structs)
835 * Implement the exchange of non-contiguous data.
836 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
837 * Correct behavior of smpi/sender_gap and set its default value to 0
838 * Add option to asynchronously send small messages to allow better
839 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
840 specifies the size in bytes under which messages will be asynchronously sent.
841 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
842 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
845 * New C interface to define a platform: XML is now optional.
846 For more info, please check include/simgrid/platf.h.
847 * New interface to define random platforms from the C:
848 For more info, please check include/simgrid/platf_generator.h and
849 examples/msg/masterslave/masterslave_platfgen.c
850 * Export a sg_cmdline dynar containing all the arguments we got from
854 * Two new tracing options for adding comments to trace file so you
855 can track your experiments (see --help-tracing for details).
856 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
857 * Adding the SimGrid version that generated the trace file as a comment.
858 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
859 * Fix to avoid key clashes on Paje links
860 * Other minor fixes related to the Paje specification
863 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
864 and renamed to xbt_str_hash() and xbt_str_hash_ext().
865 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
866 * Greatly improve the robustness of mmalloc to user errors (such as
867 using an area after freeing it, or freeing it twice)
869 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
871 SimGrid-java (3.8.1) stable; urgency=low
873 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
874 Warning: all methods are visible, but only some of them are
875 implemented so far. Check the source (src/jtrace.c)
876 for further information.
877 * New module: org.simgrid.msg.File (SimGrid File management functions)
878 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
879 * Change the meaning of Process.restart: now restart the process from
880 the begining, like MSG_process_restart in C.
881 * Add Process.setAutoRestart: handling of process restart when failed
883 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
884 you to retrieve the properties of the processes/hosts
885 * Deprecate Msg.clean(): you can just forget about it now.
886 * New function Process.getCount(), that only works when compiling
887 with the not yet released version 3.9 of the C library.
889 * New context factory based on Coroutines. It mandates a modified JVM
890 but then, the simulations run about five times faster, and there is
891 no limit to the amount of processes (beside of the available memory).
893 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
895 SimGrid (3.7.1) stable; urgency=low
898 * Restore the prototype of MSG_process_create_with_environment() to
899 the pre-3.7 situation by removing the kill_time argument.
900 * Add a MSG_process_set_kill_time() function instead.
903 * Fix weird behaviors when dealing with parallel tasks.
906 * Simgrid is now built as a dll.
907 * Simgrid-java now works on Windows.
908 * Simgrid-Java is now included into Windows package.
911 * First pre-build package for MacOSX.
914 * Fix compilation when using MSG_USE_DEPRECATED.
915 * Fix some compilation issues on Macs and Windows.
916 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
917 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
919 We discovered that the Lua console is broken, but we are missing the
920 manpower to fix it right now. The problem existed in 3.7 too, so we
921 are not blocking the release for that. Sorry if you depended on this
922 feature, any help would be really welcome.
924 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
927 SimGrid-java (3.7.1) stable; urgency=low
929 The "Java aint got to be bloated and slow" release
932 * Various internal cleanups and performance improvement
933 Simulations are expected to run up to twice faster or so
934 * Make Process.kill(process) an instance method, not a static one
935 * User processes are not java.lang.Thread subclasses.
936 This breaks the compatibility (sorry), but previous API was
937 brain-dead, making it impossible to have non-trivial
938 initializations in the process constructor.
939 * Require a full constructor per Process sub-class.
940 Kinda breaks the compatibility (sorry), but this allows a much more
941 efficient way to launch the processes at simulation startup.
942 * Do not embeed our version of semaphores, java 1.5 can be considered
943 as sufficiently prevalent for us to not dupplicate its features.
947 * Add examples for almost every part of the API
948 We spotted and fixed a lot of bugs in the process
950 * New module: asynchronous communication API
951 * New function: Process.sleep()
952 It takes milliseconds as argument, just as java.lang.Thread.sleep()
953 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
954 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
956 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
958 SimGrid (3.7) stable; urgency=low
960 The "spring cleanups (before the next Big Project kicks in)" release.
963 * We can specify the SMPI latency/bandwidth factor with command line
964 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
965 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
966 You can also use the "config tag" from platform file by adding this line
967 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
968 (see "example/platforms/tag_config.xml" to use "config tag").
969 Note that the command line supersedes the platform file configuration.
970 * Change the correction factors used in LMM model, according to
971 the latest experiments described in INRIA RR-7821.
972 Accuracy should be improved this way.
973 * Use the partial invalidation optimization by default for the
974 network too. Should produce the exact same results, only faster.
975 * Major cleanup in surf to merge models and split some optimization
976 mechanisms from the core of the models. As a result you can now
977 specify which model to use (e.g., --cfg=network/model:LV08
978 --cfg=cpu/model:Cas01) and which optimization mode to use
979 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
980 Incompatible combinations should err at initialization. See
981 --help-models for the list of all models and optimization modes.
982 * The CLM03 workstation model was dropped for simplicity because it
983 used the deprecated CM02 network model. Use default instead.
984 * Rename the TCP_gamma configuration option to network/TCP_gamma
985 * Rename the coordinates configuration option to
986 network/coordinates, and document it
987 * Use now crosstraffic keyword instead of the terribly misleading
988 fullduplex keyword. It is activated by default now in the current
989 default model, use --cfg=network/crosstraffic:0 to turn it off.
990 * Ongoing refactoring the model parsing to make XML files optional
991 See include/simgrid/platf.h for details (still to be completed)
994 * Major overhaul of the documentation. Almost instructive now :/
995 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
996 functions and friends. This interface was considered as
997 deprecated since over 2 years, it's time to inform our users that it is.
998 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
999 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1000 use these (crappy) functions in your code.
1001 These functions will be removed soon. Stop using them now.
1002 * Deprecate MSG_get_host_{table,number}
1003 Implement MSG_hosts_as_dynar() instead.
1004 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1005 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1008 * Stabilize the parallel execution mode of user contexts
1009 * Introduce configuration variables to control parallel execution:
1010 - contexts/synchro: Synchronization mode to use when running
1011 contexts in parallel (either futex, posix or busy_wait)
1012 - contexts/parallel_threshold: Minimal number of user contexts
1013 that must be part of a scheduling round to switch to parallel
1014 execution mode (raw contexts only)
1015 * Fix bugs that prevented to use suspend/resume along with
1016 synchronization structures.
1017 * Fix bugs in process termination that lead to invalid memory access
1018 in very specific conditions.
1021 * Introduce a parallel mode for the models (controlled by surf/nthreads
1022 configuration item). In our tests, running the models in parallel
1023 never lead to any speedups because they are so fast that the gain
1024 of computing each model in parallel does not amortizes the
1025 synchronization costs, even when ultra fast futexes are used.
1026 This is released anyway because YMMV.
1029 * Performance boost by using a swag internally to compute the set of
1030 tasks that are finished and should constitute the return value of
1034 * Enable it by default now that it is considered rather stable.
1037 * Documentation of the tracing functions.
1038 * Performance gains when tracing categorized/uncategorized resource
1039 utilization by avoiding calls to get route when updating resource
1040 variables. LMM constraints are being used instead.
1041 * API changed to set task categories. Use MSG_task_set_category instead
1042 of TRACE_msg_set_task_category, and SD_task_set_category instead
1043 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1044 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1045 * Examples for link user variables added (see at examples/msg/tracing/)
1046 * Deprecated function TRACE_msg_set_process_category completely removed
1047 * Trace header updated according to the latest Paje file format
1048 * Tracing network lazy updates, no longer obligate users to use full updates
1049 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1050 * Experimental: let user code declare/set/push/pop application states for hosts
1051 * API changed to allow the manual creation of graph configuration files
1052 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1055 * Improve the API of Lua MSG bindings, using the Lua spirit.
1056 * Each simulated process now lives in its own Lua world (globals are
1057 automatically duplicated). It helps writing simulators. Will allow
1058 to run Splay programs within SimGrid in the future.
1059 * Add a Chord example in Lua, equivalent to the MSG one.
1062 * Start the implementation of a solution to express temporal
1063 properties, not only local assertions. This is still an
1064 experimental work in progress, stay clear from it to be safe.
1068 - Add new runtime parameters --help-logs and --help-log-categories
1069 to display information about supported logging parameters and
1071 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1072 don't exists anymore.
1073 * Mallocators: allow value NULL for the reset function.
1075 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1076 create homogeneous dictionaries, where all the elements share the
1077 same free function. Non homogeneous dictionaries will be
1078 deprecated in the next release.
1079 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1080 - Multi-level dictionaries are deprecated.
1082 - new function xbt_dynar_search_or_negative() that is useful when
1083 you have less than 2 million elements in your dynar and don't
1084 want of the extra complexity of catching exceptions when the
1085 element is not found.
1087 - Make xbt_os_thread module (for thread portability) public.
1088 Documentation is still to come, sorry.
1090 - Cleanups and simplifications to make it maintainable again.
1091 - Exotic features (such as memalign and valloc) were removed.
1092 - The metadata were extended and improved so that the
1093 model-checker becomes able to explore and inspect the heaps.
1094 - This may induce a performance drop when enable_model-checking is
1095 ON in cmake (even if it's not used in the simulation), but it is
1096 necessary at this point to get MC working.
1098 Turn model-checking OFF if simulation performance matters to you.
1099 Not enabling it at runtime is not enough, disable it in cmake.
1101 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1104 SimGrid (3.6.2) stable; urgency=low
1106 The "Not coding new stuff allows to polish old things" release.
1109 * New bindings to the NS3 packet level simulator (experimental)
1110 * Use the raw (efficient) execution contexts instead of the sysv
1111 (portable) ones when possible.
1112 * libpcre is now mandatory in any cases since not using it led to
1113 severe performance loss and possibly other issues
1114 * Update the XML platforms:
1115 - G5K: include the latest machine in Nancy
1116 - GridPP and LCG: new platforms
1117 * Documentation was partially updated, at least (more to come)
1119 Bug fixes, cosmetics and small improvements
1120 * Free terminated processes before the end of the simulation to avoid
1121 exhausting the memory of users having very dynamic amount of
1123 * Bug fix and cosmetics about canceling non-running tasks
1124 * Bug fix about the dot loader's issues when using libcgraph
1127 * Create an installer for windows with nsis (amd64 and win32)
1128 - Add an hello world project to illustrate simgrid project creation.
1129 - Embed libpcre into the Simgrid installer to avoid
1130 its compilation burden
1131 * The raw execution contexts should work on Apple now
1132 * Port to Windows 64 bits
1133 - Sysv contexts now have an implementation for this arch
1134 - GRAS communication features now support this arch
1135 * Drop support for borland compiler on windows
1136 - this code was not maintained, and we kinda depend on gcc nowadays
1137 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1138 * Fix portability issue on unstable ubuntu: linker became picky on
1141 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1143 SimGrid (3.6.1) stable; urgency=low
1145 The "Oops, we broke Macs too" release
1148 * Fixed contexts detection so that raw ones are used when possible
1149 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1150 leads to a strange error, with user code segfaulting sometimes when
1151 the generated code is not perfectly aligned (which is not
1152 controllable from the user side, depends on the amount of code)
1155 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1157 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1159 SimGrid-java (3.6) unstable; urgency=low
1162 * Split of every thing from simgrid v3.5 into a separate package.
1164 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1166 SimGrid (3.6) stable; urgency=medium
1168 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1171 * Bindings now constitute their own package, separated from the main one.
1172 Rationale: reduce our maintenance nightmare by reducing the module coupling
1173 They will soon be released on their own on gforge.
1175 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1176 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1178 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1179 * It's not quite deprecated for now because we have no replacement,
1180 but it may soon become the case.
1183 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1184 * Fortran: New user-level cache variable to store the rank of the running
1185 process. This improves performance by an order of magnitude.
1186 * C: New coccinelle script to automatically locate and modify global and
1187 local static variables.
1188 * Improved SMPI network model with a sender-side gap to account for multiple
1192 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1193 only say if a communication is finished, no matter whether it succeeded or
1194 failed. You can call MSG_comm_get_status() to know the status of a finished
1196 * New function MSG_task_dsend() to send a task and detach it. When a
1197 communication is detached, you are never notified of its success or failure
1198 and the memory is released automatically once it is finished. This function
1199 is useful when you don't care about the end nor the success of a
1201 * Change the prototypes of action replay. Sorry for inconvenience,
1202 but this is really more efficient this way (and to adapt your code,
1203 you just have to fix the initialization, that shouldn't be too long)
1204 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1205 ever managed to do anything useful with it.
1206 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1207 takes a dynar of communications. It returns immediately and gives the
1208 index of a finished communication (if any).
1209 * New example: a basic implementation of the Chord P2P algorithm.
1212 * New model for multi-core CPUs. You can now use the core attribute to
1213 precise the number of cores of a host. This is a basic model. Every
1214 process running on the host receives at most the power provided in
1215 the DTD (throughput<=power). Total throughput of process cannot exceed
1217 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1218 router linked by an up-link and a down-link (possibly asymmetrical).
1219 This kind of pattern allows to easily build last-mile model style platforms.
1220 Aggregating such patterns in a rule-based AS is thus the technique of
1221 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1222 * New model for Vivaldi routing. We transformed the Vivaldi network model
1223 into a Vivaldi routing model (based on the rule-based model). This allows to
1224 combine Vivaldi based latencies with last-mile platforms.
1227 * Added a check for NaN of IEEE754 infinite in the double entries of
1229 * Introduce a new context factory "raw", highly inspirated from the
1230 ucontext factory, but using manually crafted functions in assembly to
1231 do the work in an efficient manner.
1232 * Allow to change the used context factory at run time, not only at
1233 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1234 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1235 contexts at runtime (only with raw contexts or ucontexts).
1236 * Completely rewrote this module to allow parallel execution of user
1237 processes. Use --cfg=contexts/nthreads:N to execute user processes
1238 with N parallel threads (the default is 1, meaning no parallelism).
1239 * Allow to decide dynamically between sequential and parallel modes.
1240 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1241 processes in parallel only when their number is greater than or equal to P
1243 * Added a check for NaN of IEEE754 infinite in the double entries of
1247 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1248 won't output the state of processes when interrupted with Ctrl-C
1249 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1250 NULL-terminated array. This may solve backward compatibility issues
1251 due to the change to return type of SD_simulate. See also:
1252 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1253 * Add new macros with variable number of arguments.
1254 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1255 - in xbt/asserts.h: xbt_assert
1256 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1257 - in xbt/ex.h: THROWF and RETHROWF.
1258 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1260 * Change xbt_die() to accept a format string with arguments, just like printf.
1261 * New data structure: xbt_lib_t, like a dict but more general and with better
1265 * New configuration options
1266 Options triva/categorized and triva/uncategorized can be used to generate
1267 graph configuration files for Triva visualization tool.
1268 * Configuration option tracing/platform is renamed to tracing/categorized
1269 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1270 * New TRACE_user_link_variable interface:
1271 User provides the name of the link and the tracing variable to attach to it
1272 * the declaration of tracing categories must be done after the environment
1274 * simpler tracing interface, just one way to declare categories
1275 TRACE_category or TRACE_category_with_color, it is up to you
1276 * links in the trace file are again identified by their names
1277 * trace contains the full platform hierarchy exactly as declared using the ASes
1278 * Options tracing/msg/[task|process]:1 groups the process by hosts
1279 for both cases, tasks and processes must have names that are unique during the simulation
1280 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1281 * The experimental option tracing/msg/volume is deprecated
1282 its functionality may be reincorporated if needed
1284 The tracing generates a trace file with unordered timestamped events,
1285 because of the way the core simulator (surf) works. A script available
1286 at the tools directory (fix-paje-trace.sh) can be used to put the events
1287 in order. We have changed the tracing so it can generate ordered timestamped
1288 events in the final trace, but depending on the simulator (and how much time
1289 is simulated) that can lead to a huge memory utilization. It is deactivated
1290 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1292 Build Infrastructure
1293 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1294 - We are trying hard to keep the API stable, but it may happen that
1295 some things change (we're a research project after all, not a
1296 nuclear plant operating system). If such things should happen, you
1297 could rely on that macro to adapt.
1298 - current value: 30600 for 3.06.00, aka 3.6
1299 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1300 a number that can be compared with SIMGRID_VERSION.
1301 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1302 assertions and verbose/debug logging events are disabled at compile time.
1304 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1306 SimGrid (3.5) stable; urgency=medium
1309 * New feature to any SimGrid-based simulator: Model-Checking
1310 Check SIN#1 for more details.
1313 * New Model SMPI (three-interval linear regression for correction factors)
1314 See RR-7426, available at http://hal.inria.fr/inria-00527150
1315 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1316 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1317 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1318 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1319 * See also src/smpi/README
1324 - Tracing API changes: TRACE_start and TRACE_end should not be called
1325 by user-code. They are automatically called by simulators created
1326 with SimDAG, MSG and SMPI if the toolkit is compiled with
1327 tracing_enabled=ON. Categories declaration and utilization remain the
1328 same for MSG and SimDag.
1329 - A function was added to the tracing API to declare categories with
1331 - TRACE_category_with_color (char *category, char *color)
1332 where color must be in the following format
1333 "%f %f %f", red, green, blue
1334 and red, green, blue are float values in the interval [0, 1]
1335 - User can specify NULL as color parameter, or continue calling
1336 TRACE_category (cat)
1337 On that case, the tracing system will define random colors
1338 - The following command-line options are supported:
1339 --cfg=tracing/filename:msg.trace
1340 --cfg=tracing:1 (activate tracing, needed to use others)
1341 --cfg=tracing/platform:1 (categorized resource use)
1342 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1343 --cfg=tracing/msg/task:1 (task creation)
1344 --cfg=tracing/msg/process:1 (process creation, migration)
1345 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1346 --cfg=tracing/smpi:1 (SMPI interface tracing)
1347 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1348 - examples of examples/msg/tracing updated
1350 - DAXLoader and DOTLoader functions can generate tasks with categories
1351 - A new function to attribute a category to SD tasks:
1352 TRACE_sd_set_task_category (SD_task_t task, char *category)
1353 * Tracing the MPI interface implemented by SMPI
1354 - Collective operations are traced with states
1355 - Point-to-Point operations are traced with states/links
1356 - Tracing activated by a parameter "-trace filename" passed
1357 to smpirun during execution (considering that simgrid
1358 is compiled with tracing enabled)
1359 - To run the simulation with gdb, the simulator
1360 accepts --cfg=tracing/smpi:1 to trace SMPI
1361 - tesh files to check if smpi tracing is ok
1362 - See examples/smpi/NAS/DT-trace
1363 * GTNetS tracing re-worked
1364 - adaptation to the tracing system of GTNets to cope
1365 with modifications regarding the fullduplex mode
1366 - new tesh files to check if gtnets tracing is ok
1369 * Asynchronous communications through the functions:
1370 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1371 * New function: MSG_load_platform_script()
1372 to make possible using a lua script instead of XML files to set up platforms
1373 * New function: MSG_set_function
1374 to associate functions to processes, used when bypassing the parser
1375 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1377 Platforms: Add some more examples in examples/platforms
1378 * Grid'5000: see www.grid5000.fr
1379 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1382 * Change the XML format. This is a very important modification. SimGrid
1383 3.5 introduces a new hierarchical format based on the notion of
1384 Autonomous Systems. Compatibility with old format is ensured
1385 through the perl script provided in the install bin directory
1386 bin/simgrid_update_xml.
1387 It is now possible to build platforms with specific routing
1388 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1389 connect several platforms together. We will try to provide soon
1390 set of realistic platforms exploiting these properties (have a
1391 look at examples/platforms/ for the moment).
1392 * Take the opportunity of the XML format change to be a good XML citizen:
1393 rename link:ctn to link_ctn and similar changes (also dealt with by
1395 * Add a new routing scheme (rule-based) using regular expressions. It
1396 enables to have an extremely low memory footprint when the
1397 underlying routing is simple and can be compactly described. You
1398 need to have libpcre4-dev (perl regular expressions) installed if
1399 you want to use this routing scheme.
1400 * Revive the cluster TAG and allow to easily and efficiently (both in
1401 term of memory and speed) connect clusters together. Have a look
1402 at teshsuite/simdag/platforms/ to see how this can be done. With
1403 this tag, you can create clusters with thousands of tasks at no
1404 cost (have a look at examples/platforms/).
1405 Note: clusters are implemented as ASes, so there is no need for an
1406 enclosing AS tag if you have only one cluster in your platform.
1407 * Add new generic functions in the public interface that allows the user
1408 to call SURF 'create_resource' methods from your code (same
1409 functionality as the XML bypass mechanism but with a much lighter
1411 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1412 that uses a piecewise linear approximation to produce better
1413 results when exchanging small messages.
1414 * Add a new parameter to handle correctly full duplex link and account
1415 for interferences between uplink and downlink communications
1416 (activate with --cfg=fullduplex:1).
1419 * Rename the SD_READY (all dependencies are satisfied and task is
1420 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1421 dependencies are satisfied) state.
1422 This prevents a confusion between the notion of "ready to schedule"
1423 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1424 simulated" (SD_RUNNABLE) used by the simulation kernel.
1425 * Change the way a task is considered as ready. Instead of removing
1426 dependencies when a task is done, a counter is decreased. This way,
1427 it is always possible to reach ancestors thanks to the
1428 SD_taks_get_parents function (even after the end of the simulation.)
1429 * Change the return type of SD_Simulate from (SD_task_t*) into
1430 xbt_dynar_t. This function was in handling a dynar internally and
1431 converted it into a NULL terminated array for historical reasons.
1432 * New function SD_dotload(char*) to load a DAG described in dot
1433 format. This loader and the corresponding examples require the
1434 installation of the graphviz library.
1435 * Fix a bug in the management of tasks of size 0 in the surf network
1436 models. This problem was only visible with SIMDAG and you should
1437 thus disregard results produced with earlier versions if you
1438 relied on this feature (some tasks were blocked because of this).
1439 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1440 even though your applications were doing only point-to-point
1441 communications and sequential computations. Now you can really use any
1442 model you want (of course, if you create real parallel tasks, which are
1443 not implemented in most models beside ptaskL07, this will abort).
1444 * Add an example that schedules a DAX on an heterogeneous platform
1445 using a Min-Min strategy.
1446 * New function SD_workstation_get_current_task() that returns the kind
1447 of task currently running on a workstation in the sequential access
1449 * Raise some warnings when unexecuted tasks remains at the end of the
1450 simulation. This is usually caused by cycles in the DAG.
1453 * New function: SIMIX_process_set_function() called by MSG_set_function
1454 * Change the underlying waiting queue in semaphores so that a process
1455 can wait on several of them simultaneously (as in waitany).
1456 * Fix the way to handle tokens in semaphores so that all access patterns
1457 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1458 * kill the dirty pimple SIMIX_message_sizes_output()
1459 Please use (proper) visualization instead
1462 * New data container: setset (set of sets of elements)
1463 * New module: mmalloc (mapped malloc, allowing to have several
1464 independent segments of malloc)
1465 * New function: xbt_dict_cursor_set_data()
1466 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1467 * New function: xbt_dynar_is_empty()
1468 * New function: xbt_fifo_get_last_item()
1469 * Fix xbt_dynar_shrink(): use the right element size.
1470 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1471 uninitialized areas during expand.
1472 * Fix semaphores: previous implementation was severely broken.
1473 * Use library init/fini functions for our initialization.
1474 - you can use logs and other feature as soon as you want in your
1475 code (even before the xbt_init / MSG_init)
1476 - xbt_exit is now a no-op and produce a warning when used.
1479 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1480 benefit from the latest improvement to the simulation kernel.
1481 * Kill measurement sockets for now. If you rely on them, sorry. This
1482 release is not for you. This feature will be reintroduced in the
1483 future, but we cannot delay the release any further.
1484 * New function: gras_msgtype_get_name().
1485 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1486 * Fix (at last) the pmm example: it should not randomly fail anymore.
1488 Build chain: bug fixes and overall polishing
1489 * Cmake is now stable enough. Hence, we killed the autotools.
1490 * Port to windows ( TM :)
1491 * Fix the 'make install' target.
1492 No need to use 'make install-simgrid' anymore
1493 * Introduce a 'make dist' target compiling a *source* archive
1494 'make package' compiles a binary archive
1495 * Compile java files only on need
1496 * Add --cd and --setenv command line options to tesh
1497 * Out of source builds are not fully supported yet, but we are close
1498 * Enable supernovae and optimization flags by default for our users
1501 * Add layer to set up environment directly from lua, without XML.
1502 * The effect of gras_stub_generator can be achieved through
1503 lua too (check examples/gras/console/ping_generator.lua)
1505 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1507 SimGrid (3.4.1) stable; urgency=low
1509 The "Polishing easter eggs is probably a good idea" release.
1510 This is a bug fixes release only.
1514 * Fix a bug preventing the tasks from begin garbage collected.
1517 * Fix a bug occuring when a host involved in a communication fails.
1518 This was not detected properly by the other peer involved in the
1519 communication. Now, it's reported as a network error.
1522 * Warn the user about loop dependencies in data flow of DAX files
1523 * Obey the control-flow dependencies of DAX files
1526 * Add option "enable_smpi" allowing to not compile SMPI.
1527 Probably useful for the (Mac) users experiencing a build error here
1528 * Improve the detection of lua5.1 and ruby1.8
1530 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1532 SimGrid (3.4) stable; urgency=low
1534 The "Easter in Cargese" release. Also known as (major changes):
1536 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1537 ~> bindings were greatly improved
1538 ~> new tracing infrastructure for better visualization introduced
1540 * the "Welcome to configury modernity" release.
1541 ~> we switched from autotools to cmake, and improved our cdash
1544 A more detailled list of changes follow (full detail in svn log).
1546 Java Bindings: Various Cleanups
1547 * (install java-gcj-compat-dev on debian-like to use them)
1548 * Remove put/get: no need to export deprecated interface in Java
1549 Use send/receive instead.
1550 * Cleanup the examples and add a README per directory
1551 * Remove example autoDestination (that's the only way to go now)
1552 * Remove example explicitDestination (was a plain copy of basic)
1553 * Make JniException a runtime exception, so that there is no need to
1554 declare the fact that you may encounter such a beast. I guess that
1555 nobody will ever want to survive such error.
1556 * Create specific errors for each MSG case of failure:
1557 host failure, transfer failure, timeout, task cancelled
1558 * Cleanup the exceptions that may get thrown by each function
1559 * Other internal cleanups in Java bindings. Performance still bad :/
1560 Ruby and Lua Bindings: create them
1561 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1562 * That's new and great, you should try them out.
1563 Same functionalities than Java bindings, only even less polished
1565 * Kill the useless "rate" argument of SD_task_get_execution_time()
1566 Everyone used to provide -1 as a value, it was not used, and the
1567 semantic of a possible use wasn't even clear.
1568 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1569 either as comm costs or compute costs to mean that there is no
1570 such thing for that specific task.
1571 * Add a SD_task_set_name() function
1572 * Fix SD_task_unschedule() on typed tasks
1573 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1574 * In DAX loader, accept useless 'level' attributes to <job> since
1575 LIGO DAGs have them (seem to be to ease graphical representation).
1577 * Add an example masterslave_mailbox.c using send/receive and not
1578 the deprecated put/get interface.
1579 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1580 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1582 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1583 (declare MSG_USE_DEPRECATED to still have the old name)
1584 * Add a MSG_task_set_data() function
1585 * About trace replay (see examples/msg/actions):
1587 - Allow to work with splitted trace files for each process
1588 Give the specific trace file as argument of each process,
1589 and call MSG_action_trace_run(NULL)
1590 You can still have one merged file for all processes.
1591 - Fix implementation of collective operations
1592 * Allow task_execute() on 0-sized tasks (closes #10063)
1594 * This is the first release of SimGrid where SMPI is not considered
1595 beta anymore (even if some corners should still be improved)
1596 * Port over the new SIMIX_network submodule (internal refactoring)
1597 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1598 * Implement more missing elements of the standard:
1600 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1602 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1603 MPI_Scan() MPI_Get_processor_name()
1604 - Added implementation of missing case for Alltoall (warning: it's
1605 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1606 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1607 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1608 SMPI_MPI_Allgatherv()
1609 * Bug fixes include:
1610 - MPI_Waitsome() was broken
1611 - Allow relative includes in smpicc
1612 - Command line cfg argument 'reference_speed' was ignored...
1613 - Some functions did not properly lead to auto-benching of user code
1614 - smpicc passes -O2 by default (just like openmpi one)
1616 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1617 * Bug fixes about timeouts during communications
1618 * add SIMIX_message_sizes_output() as a pimple to write to file the
1619 amount of messages per size. Use gnuplot to get histogram.
1620 Pimple because that's the only user-visible function of simix,
1621 defined directly in xbt.h (irk, sorry)
1623 - Add a SIMIX_sem_get_capacity() function
1624 - Fix interactions with processe resume/suspende
1625 - release_forever() was stupidly broken
1626 - Fix SIMIX_display_process_status() for processes in a semaphore
1627 - Make SIMIX_sem_block_onto() user-visible
1628 * Refactoring context stuff:
1629 - Use pseudo-OOP for better modularity
1630 - reimplement SIMIX_process_kill() without process_schedule() so
1631 that the latter can take as invariant that it is called from
1633 - Merge context_start into context_new for sake of simplicity
1635 * Add a Vivaldi network model, coded live during SUD'10 ;)
1636 * Rename configuration variables to start a hierarchy:
1637 o cpu_model -> cpu/model
1638 o network_model -> network/model
1639 o workstation_model -> workstation/model
1640 * New configuration variables:
1641 o network/bandwidth_factor: correction to bandwith
1642 o network/latency_factor: correction to latency
1643 o netwotk/weight_S: correction to the weight of competing streams
1644 * Add a long description to the models, that users can see with such
1645 argument on the command line: --cfg=cpu/model:help
1646 * --help-models display the long description of all known models
1648 * config: add the ability to set a default value after registration
1649 Does not override any previously set value (e.g. from cmd line)
1650 * dict: allow to have integer key and data.
1651 When so, you need to use the following functions
1652 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1653 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1654 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1655 In contrary to regular dicts, the key is not malloced before copy.
1656 Mixing scalar and regular elements in the same dict is not tested
1658 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1659 Tracing for Visualization:
1660 * SimGrid is now instrumented in order to generate a trace file for
1661 visualization analysis: to use it, need to compile SimGrid with the
1662 "tracing" option enabled, and instrument the program using SimGrid with
1663 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1664 (among other functions).
1665 * The instrumentation only traces the platform utilization for now
1666 * Documentation to use the tracing functions and how to analyze the
1667 traces with the Triva tool is written.
1668 * More information about: SimGrid FAQ (in the section Tracing Simulations
1671 * We moved to cmake as default build system. Autotools support will
1672 be dropped soon. Check the FAQ for more info about how to use it.
1673 * Greatly improved our cdash/ctest interactions
1674 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1675 * Added memory checking tests with valgrind; lot of memleak fixing.
1676 This may be the first release of simgrid with so few memory issues
1677 * Added code coverage tests.
1678 Our coverage is still improvable, but at least we see it on cdash.
1680 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1682 SimGrid (3.3.4) stable; urgency=low
1684 The "Desktop Grid needs love too" release (also called Xmas release).
1686 Models improvements:
1687 * Major speedup in the maxmin system solving by using lazy evaluation
1688 Instead of solving completely the maxmin system at each iteration,
1689 only invalidate (and recompute) the modified parts.
1690 This new feature is enabled in default models but you can try to
1691 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1692 * Cas01 IMproved as default CPU model
1693 This CPU model is the same Cas01 model, but it uses the
1694 maxmin-selective-update flag and a heap structure to manage
1695 actions on SURF kernel.
1696 It reduces the complexity to find the next action to finish and,
1697 consequently, it's faster than the old Cas01.
1698 This is the new default CPU model (Cas01).
1699 * Rename the old Cas01 model to Cas01_fullupdate
1700 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1701 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1702 * CpuTI (CPU Trace Integration)
1703 A new CPU model whose objective is simulate faster when using
1704 availability trace files.
1705 Instead of using a full featured, over engineered maxmin system for
1706 CPU modeling, this model does the pre-integration of traces files
1707 to calculate the amount of CPU power available, and so, executes
1708 faster than the old CPU models.
1709 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1710 * Use LV08 as default network model since it gives better accuracy
1711 for small messages and shouldn't change things for big ones.
1712 Use --cfg=network_model:CM02 to get the previous behavior.
1715 ******************************************
1716 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1717 ******************************************
1718 * The new CPU model may changes simulations!
1719 The point is that events occurring at the exact same timestamp
1720 are not scheduled in the same order with the old and new
1721 version. This may be enough to completely change the execution
1722 of simulations in some cases.
1723 * The new network model will change simulations!
1724 This new model is more realistic than the previous one, so you
1725 should consider redoing your old experiments with this model.
1726 Sorry for the inconvenience.
1729 * Introduce the supernovae compilation mode
1730 When compiled that way, the whole SimGrid (or almost) is put in a
1731 single compilation unit and compiled in one shoot.
1732 This is to help gcc which has difficulties to inline stuff from one
1734 The speedup seem to be above 15%, althrough more tests are needed on
1735 amd64 to confirm that gain.
1738 * Port of MSG's mailbox on top of SIMIX network
1739 The put/get mechanism was greatly simplified on the way.
1742 * New SIMIX network module. Provides:
1743 - Mailbox: rendez-vous mechanism to find with who you want to speak
1744 - Synchronous send/recv: easier and hopefully faster since the
1745 logic is handled in the maestro process directly now
1746 - Asynchronous send/recv: you dreamt of it? It's here now
1747 Too bad that nobody cared enough to propagate the change to MSG.
1748 * Add semaphores as SIMIX synchronization mechanism.
1751 * new function SD_daxload(char*) to load a DAX file
1752 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1753 * Introduce typed tasks. Specify its kind and cost at creation.
1754 At scheduling, just give where it should be placed, and the cost
1755 for each involved resource is automatically computed.
1756 Existing constructors so far (more to come of course):
1757 - SD_task_create_comm_e2e() for end-to-end communication
1758 - SD_task_create_comp_seq() for sequential computation
1759 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1760 * new function SD_task_dump() for debuging display
1761 * new function SD_task_dotty(task,FILE*) writing to file the info
1762 about the task in dotty format
1763 * SD_task_dependency_exists() can now cope with having one of its
1764 arguments NULL. If so, it tests whether the other argument has any
1766 * Add getters on list of preceding/following tasks:
1767 SD_task_get_parents(task) and SD_task_get_children(task)
1768 * Add getters on amount of workstations and list:
1769 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1770 * Add getter on task kind: SD_task_get_kind(task)
1771 * Update the start_time and finish_time of tasks on completion/failure
1772 * Bugfix: Remove task from state swags when destroyed
1775 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1778 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1779 dynar with one separate thread per value of the dynar.
1780 * Change the prototype of xbt_thread_create(), sorry.
1781 Added a boolean parameter indicating whether we want to join this
1782 thread (used in SG only for now)
1783 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1786 * GTNetS wrappers should now be usable again (and betterly tested too)
1787 * Fix a major regression from 3.2 where the timeout provided to
1788 MSG_task_put_with_timeout() was used as absolute time before which
1789 the comm should be done.
1790 * Start to fix the <cluster> tag.
1791 - Internal links should be good now (beside of the loopback, which
1792 use the private link instead)
1793 - paths to the external world is still rather broken
1794 - the <route:multi> tag is just broken. Actually that's brain-dead.
1795 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1796 to make it less stupid
1797 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1798 * Fix a source-level compatibility glitch from 3.2: after defining
1799 MSG_USE_DEPRECATED, you can use the old name
1800 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1801 * Allow to compile from the SVN with automake 1.11
1802 * Fix some problems when using the "start_time" tag in deployment XMLs.
1803 * Fix #8569: XBT/synchro.h has redundant declarations
1804 * Fix #8563: MSG return values and exceptions
1805 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1806 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1807 Thanks to Robson Peixoto.
1808 * Fix a vicious bug in dictionaries inducing that some elements were
1809 not freed on xbt_dict_free()
1811 Portability report of this version:
1812 * Main portability targets:
1813 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1814 on (amd64/i386/ia64)
1815 - mac leopard on i386
1816 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1817 but nothing critical.
1818 * Other platforms: windows, AIX and others were not tested for this release
1820 Timing report of this version:
1821 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1822 scenario parameters). From 8h to a few seconds in desktop grid settings.
1823 * Supernovae brings about 25% speedup on i386.
1825 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1827 SimGrid (3.3.3) stable; urgency=low
1829 The "Need for Speed" release.
1831 The timings done to validate the 3.3.2 were faulty.
1832 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1834 The problem was a conversion from a manually handled vector to
1835 xbt_dynar_t on the critical path.
1836 xbt_dynar_foreach calls functions, inducing stack management crap.
1838 We inlined these functions and xbt_dynar_foreach is now breath taking.
1839 We also inlined xbt_swag_belong on the way.
1841 Here are some approximate speedup measurements (on master/slaves
1842 simulations lasting between 10s and 20s each):
1843 3.3.1 -> 3.3.2: about same performance
1844 3.3.2 -> 3.3.3: 40% speedup
1845 3.3.1 -> 3.3.3: 40% speedup
1846 3.3.1 with inline patch -> 3.3.3: 30% speedup
1848 Our reading is that the refactoring which occurred in 3.3.2 made us
1849 suffer much more from the xbt_dynar_foreach low performance, but
1850 once we solved this, this refactoring proved to be very performance
1851 effective. From the 40% speedup, somehow, 10% are due to the
1852 inlining and 30% to the refactoring.
1854 That's a pitty that gcc cannot inline functions placed in other files
1855 alone. We have to choose between:
1856 - break the encapsulation (by putting private data structures and
1857 accessors in headers files to help gcc)
1858 - live with low performance
1859 - switch to a decent compiler such as icc (not quite possible).
1861 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1863 SimGrid (3.3.2) stable; urgency=low
1865 The "Simplicity does not preceed complexity, but follows it" release.
1867 The main contributors of this release were (lexical order):
1868 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1871 * Extract the routing logic into its own object.
1872 (was dupplicated in network.c and workstation_LV07.c;
1873 Allows to implement other ways of storing that info)
1874 => kill now useless network_card concept
1875 - Use dynar to represent routes (instead of void** + int*)
1876 - kill link_set (use surf_network_model->resource_set instead)
1877 - Add a command-line option to choose the routing schema to use
1878 - Add three new models:
1879 * Floyd (shortest path computed at initialization)
1880 * Dijikstra (shortest path recomputed all the time)
1881 * Cached Dijikstra (shortest path computed on need)
1882 All these models where contributed by Silas De Munck, and are
1883 described in his ICCS09 paper.
1885 * Simplify model declaration
1886 (less redirections, less function to write when defining a model)
1887 - Factorize stuff between models:
1890 surf_model_resource_set(model)
1891 surf_model_resource_by_name(model, name)
1892 - Unify the types of models in s_surf_model_t (using an union)
1893 - Embeed fields of common_public directly into s_surf_model_t
1894 - Rename model methods:
1895 action_free ~> action_unref
1896 action_change_state ~> action_state_set
1897 action_get_state ~> action_state_get
1898 - Change model methods into functions :
1899 (model)->common_public->action_use ~> surf_action_ref
1901 * Implement a generic resource; use it as ancestor to specific ones
1902 (allows to kill duplicated code in models)
1903 Drawback: timer command don't need no name nor properties;
1904 workstation_CLM03 don't need no properties
1905 (but I guess we can live with those few bytes wasted)
1907 * Improve the action object model
1908 - implement a constructor avoiding dupplicated code about field
1909 initialization in generic_action part.
1911 * Kill the SDP model: it has an external dependency, is deprecated
1912 in flavor of modern lmm models, and didn't compile since a while
1915 * Relocation of the context module from XBT to SIMIX.
1916 (the context were decoupled from the simix processes, duplicating a lot of code)
1917 => a lot of code was factorized
1918 - less overhead is introduced during scheduling
1919 - simpler API for the context factory
1920 - the logic for process creation,destruction and manipulation was simplified
1921 * Simplification of the s_smx_process_t data structure.
1922 => accesing the simix level data associated to a process is faster now,
1923 and the code is a lot more readable.
1926 * Implement some more MPI primitives:
1927 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1928 -implementation: Bcast: flat or 2-ary tree (default),
1929 Barrier: 4-ary tree,
1931 Allreduce: Reduce then Bcast
1932 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1933 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1934 Alltoallv: flat tree, like ompi
1936 * Add support for optimized collectives (Bcast is now binomial by default)
1937 * Port smpirun and smpicc to OS X
1940 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1941 Shout out if you used it.
1944 * Display the list of still queued messages in SG mode when existing
1948 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1949 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1950 * Add xbt_str_from_file(FILE*)
1951 * Add xbt_dict_get_key achieving a linear reverse search
1952 * Remove the context module
1954 Portability report of this version:
1955 * Main portability targets:
1956 - Linux(debian)/x86/context
1957 - Linux(debian)/x86/pthreads
1958 - Linux(debian)/amd64/context
1959 - Linux(debian)/amd64/pthreads
1960 On these, we still have the eratic breakages of gras/pmm and
1961 amok/saturate_sg reported in previous version. We still think
1962 that the tests are the cause of the fault, not the tested code.
1964 - Mac OSX Leopard/x86/context
1965 Still false negative in tesh autotesting.
1966 Smpi still fails, but this time because readlink does not accept -f
1967 Everything seems to work properly beside of that.
1970 - AIX version 5.3 (only tested contexts this time)
1971 Smpi still fails there because mktemp is not installed.
1972 Everything seems to work properly beside of that.
1974 I managed to compile it for the first time, but several breakages.
1975 Won't delay the release for this exotic platform.
1977 * Windows: it's still lagging behind. If you want to help, please
1980 Timing report of this version:
1981 This version seem to be more than 5% faster than 3.3.1 (on linux
1982 64bits with contextes). The gain is less than expected, we are
1983 investigating this for next release.
1985 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1987 SimGrid (3.3.1) stable; urgency=low
1990 * Implement a --cfg-help to show existing configuration variables
1991 * Build chain do not require doxygen in maintainer mode
1994 * fix a bug on struct sizeof computation, which prevented the
1995 exchange of arrays of structs in some conditions
1996 - added a regression test about this in datadesc_usage
1997 * Allow the exchange of 0-long dynamic vectors.
1998 - for that, use -1 as indicator of dynamic size instead of 0
1999 - This implied to change any size from unsigned long to long,
2000 reducing a bit communication abilities, but I guess that with
2001 64bits being quite common, this is more than enough.
2002 - This also induce a protocol change, thus bumping network protocol
2003 version from 0 to 1 (if we have external users, we have to get
2004 clean on that point too ;)
2005 - added two regression tests about this in datadesc_usage
2006 * Be more verbose when propagating local exceptions
2007 This helps debugging.
2008 * Display the status of simulated processes when receiving SIGINT in
2012 * Allow to control the simulation from a trace file.
2013 New functions MSG_action_register() and MSG_action_trace_run()
2014 The first one allows to associate a function execution to each
2015 kind of action while the second one parses a trace file and
2016 triggers the corresponding actions within the system.
2017 For now, only a toy example is provided in examples/msg/actions
2018 * Add an exemple of process migration in examples/msg/migration
2019 * Fix a bug in task exchange which broke MSG_task_get_sender()
2020 Add a teshsuite regression test for that.
2021 [Bug: if MSG_task_get_sender() is called after sender exit,
2023 * Fix a bug which prevented suspend/resume to work properly
2024 * Display the status of simulated processes when receiving SIGINT
2025 This fixes a regression of v3.3. due to the introduction of SIMIX
2026 * Bug fixing in failure management:
2027 - trace could not start by a failure at time 0
2028 - failure during communications were not working
2031 * Add SIMIX_process_set_name() to change the name of the current
2032 process in the log messages.
2033 * Store smx_hosts in a dict since we only retrieve them by name
2034 * Move the configuration infrastructure to surf
2037 * Move the configuration infrastructure to surf
2040 * Massive internal cleanups:
2041 - Store internal structures on processes instead of hosts (allows
2042 to have more than one process per host, in addition of being more
2044 - Cleanup the initialization/finalization process
2045 - Kill a whole bunch of unneeded synchronization:
2046 processes run in exclusive manner within the simulator
2047 - Move queues from global tables to process data fields
2049 - now accept -platform and -hostfile arguments
2050 - Pass the right rank value to processes according to the hostfile
2051 * Compile the examples by default, and use them as regression tests
2052 * Implement MPI_Wtime()
2053 * Change the reference speed to a command line option
2056 * TCP_gamma can now be specified as command line option using
2057 --cfg=TCP_gamma:10000000.0
2058 * Change the --surf-path cmd line option into --cfg=path:
2061 * Also include strbuff from xbt.h public header
2062 * xbt_ex_display(): do not free the exception after displaying
2063 This allows to do more with the given exception afterward.
2064 Users should call xbt_ex_free() themselves.
2068 Portability report of this version:
2069 * Main portability targets:
2070 - Linux(debian)/x86/context
2071 - Linux(debian)/x86/pthreads
2072 - Linux(debian)/amd64/context
2073 - Linux(debian)/amd64/pthreads
2074 These targets fail about 1/10 of times on gras/pmm, but we believe
2075 that this is because of the test, not because of simgrid.
2076 amok/saturate_sg fails even more rarely, and the test may not be
2079 - Mac OSX Leopard/x86/context
2080 The test suite still spits tons of errors because some obscure
2081 force prevents us from removing the temporary directories
2082 arguing that they still contain some metadata I've never heard of.
2083 Smpi fails because seq is not installed.
2084 Everything seems to work properly beside of that.
2087 - AIX version 5.3 (both contexts and pthread)
2088 Smpi still fails there because mktemp is not installed.
2089 XML inclusions seems rosty on AIX.
2091 * Windows: it's still lagging behind. If you want to help, please
2094 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2096 SimGrid (3.3) stable; urgency=high
2100 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2101 [Malek Cherier & Mt]
2103 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2106 Until now, GRAS were using MSG as an interface to SURF. It was
2107 quite difficult because both interface have several differences
2108 (MSG channels vs GRAS sockets were the most notable point).
2110 This also opens the gate to SMPI (which should occur soon) and speed
2111 up simulations by to 40% (even if it were not the main goal).
2113 **************************************
2114 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2115 **************************************
2116 The point is that events occuring at the exact same timestamp are
2117 not scheduled in the same order with the old and new version. This
2118 may be enough to completely change the execution of simulations in
2119 some cases. Sorry for the inconvenience.
2121 * Cleanup and upgrade the XML format to push further scalability
2122 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2124 * Improve the testing infrastructure with tesh. Now a very large part of
2125 the code is tested not only by being run but also by checking that the
2126 output match an expected output [Mt].
2128 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2129 is really less memory-demanding, which should allow you to use
2130 larger files in SimGrid [AL].
2132 * Inform valgrind about our contextes, so that it becomes usable
2133 with the default (and more effecient) version of SimGrid
2134 [contributed by Sékou Diakite, many thanks]
2137 * Introduce a listener thread in charge of receiving incoming
2138 messages from the network. It allows to overlap communication and
2139 computation but most notably, it removes some stupid deadlocks due
2140 to the fact that so far, a process could not send and receive at
2141 the same time. This made most non trivial communication schema
2143 * Convert the PIDs from long int to int to match the MSG ones (and
2144 linux ones too) [Mt]
2145 * New function: gras_agent_spawn() to launch a new process on
2146 current host. Only working in simulation for now. [Mt]
2147 * New function: gras_os_hostport() returning a constant form (ie,
2148 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2151 * Make the backtrace of exceptions more human readable [Mt]
2152 * New module: xbt/str [Mt]
2153 a ton of string utility functions (split, join, printf to a newly
2154 allocated buffer, trim, etc)
2155 * New module: xbt/hash [Mt]
2156 SHA1 hashing algorithm (more to come if needed)
2157 * New module: xbt/synchro [Mt]
2158 synchronization tools (mutex and conditions) working the same way
2159 in simulation and in real life (mainly useful for GRAS, but not
2161 * New module: xbt/queue [Mt]
2162 classical producer/consumer synchronization scheme
2163 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2164 (using the classical functions will get serialized) [Mt]
2165 * Make dictionary internal table dynamic. No need to specify its size
2166 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2168 * Make sure the log channels are organized as a tree under windows
2169 (because of ANSI C compatibility issue, any channel were child of
2170 root directly) [Mt].
2173 * Cleaned many thing in surf and fixed a few bugs [AL].
2174 * Add a nice command line configuration mechanism to compose models [AL].
2175 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2176 the previous one (KCCFLN05). It relies on something that looks like
2177 a max-min sharing mechanism but cannot be written as such. A new solver
2178 was thus designed [AL].
2179 * Add a new solver to lmm. Based on Lagrange optimization and
2180 gradient-based descent, it enables to efficiently maximise systems s.a
2182 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2184 This solver enables to propose two new network models for TCP Reno and
2185 TCP Vegas based on Low's work. These models still need to be fully
2186 tested though [Pedro Velho].
2189 * Bug fix in SD_simulate. Now the time bound given as argument is
2191 * Use the new parallel task model (ptask_L07) as default.
2192 * Use the SURF command line configuration mechanism.
2193 * 0-size tasks (for synchronization) should now work.
2195 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2197 SimGrid (3.2) stable; urgency=high
2201 We still experience issues on this platform, but we believe that at
2202 least MSG is usable.
2204 GRAS API BREAKAGE (for simplification purpose, sorry):
2205 * the gras_msgtype_by_name is not used anymore. Instead of
2206 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2207 you can write (and must)
2208 gras_msg_send(toserver, "request", &request);
2209 - If you still want to pass a gras_msgtype_t to the function (to cache
2210 the type and avoid the lookup time), use the gras_msg_send_() variant.
2211 - Impacted functions:
2212 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2213 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2214 * The callbacks are now expected to return 0 when everything went well
2215 (just like the main() function)
2217 GRAS new features and improvements:
2218 * New module mechanism where user code can use per process globals [Mt]
2219 This is similar to gras_userdata_*() functions, but for libraries. It
2220 factorize some code developped over and over in the examples and AMOK.
2221 It has still to be documented and used (only amok/peermanagement is
2223 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2224 loss when they were small enough to fit into the buffer and sent quickly
2225 enough so that they can all get received in one shoot.
2226 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2227 if not found. Use the *_or_null() variant for the old semantic.
2228 * In gras_msg_handle, do not discard messages without callback.
2229 They are probably messages to be explicitly awaited later (ie, proofs of
2230 mis-synchronization in userland since they are sent before being awaited)
2231 No big deal usually.
2232 * gras_socket_meas_send/recv: semantic changed!
2233 The numerical arguments used to be (1) the total amount of data to send
2234 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2235 messages. This was need for the fool willing to send more than MAXINT
2236 bytes on quite fat pipes.
2239 * Do really rename the hostmanagement module to peermanagement. [Mt]
2240 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2241 but this is rather new and this was documented in the module
2242 documentation (poor excuses, I admit)
2243 * Bandwidth measurement semantic changed! This follows the changes to
2244 gras_socket_meas_send/recv explained above.
2247 * A sequential mode has been added to the workstations. When a workstation
2248 is in sequential mode, it can execute only one task, and the other tasks
2249 are waiting in a FIFO. [Christophe Thiery]
2252 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2253 model for SIMDAG. [Christophe Thiery]
2254 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2255 equal to 0, which caused some bad behaviors in
2256 saturated_constraint_set_update. I now use a threshold mechanism like in
2260 * When running manually src/testall, you select specific units [Mt]
2261 testall is the result of our cunit mechanism, and should replace all
2262 the scripty thingy around since bash don't run easily on billware.
2264 * A mallocator system has been added. [Christophe Thiery]
2265 Mallocators allow you to recycle your unused objects instead of freeing them
2266 and allocating new ones.
2268 Documentation update:
2269 * FAQ reworking + New FAQs:
2270 - "Valgrind spits tons of errors!" [Mt]
2271 - "How to repport bugs" [Mt]
2272 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2273 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2274 - Communication time measurement within MSG [AL]
2275 - I experience weird communication times when I change the latency [AL]
2276 * GRAS tutorial [Mt]
2278 - an introduction to the framework and to the used communication model
2279 - an initiatic tour introducing the most proheminent features:
2281 . Lesson 0: Installing GRAS
2282 . Lesson 1: Setting up your own project
2283 o Part 2: Message passing
2284 . Lesson 2: Exchanging simple messages
2285 . Lesson 3: Passing arguments to the processes (in SG)
2286 . Lesson 4: Attaching callbacks to messages
2287 . Lesson 5: Using globals in processes
2288 . Lesson 6: Logging information properly
2289 . Lesson 7: Using internal timers
2290 . Lesson 8: Handling errors through exceptions
2291 . Lesson 9: Exchanging simple data
2292 . Lesson 10: Remote Procedure Calling (RPC)
2293 . Lesson 11: Explicitely waiting for messages
2294 . Recapping of message passing features in GRAS
2295 - A HOWTO section containing:
2296 o HOWTO design a GRAS application
2297 More are due, of course. They will come latter. In the meanwhile, you can
2298 check the examples which are still here.
2300 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2302 SimGrid (3.1) stable; urgency=high
2306 There was a stack corruption somewhere, visible only when optimizing
2307 with these versions. [Vince]
2310 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2311 module that enabled to play with Directed Acyclic Graphs. It is built
2312 directly on top of SURF and provides an API rather close to the old
2313 SG. Some old codes using SG are currently under rewrite to check that
2314 all needful functions are provided. [Christophe Thiery]
2317 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2318 extension of the classical CLM03 model that gracefully handles
2319 failures. This is now the default model for MSG and GRAS. It doesn't
2320 handle parallel tasks yet though. [AL]
2321 * Bug fix: Weights were not correctly set in the network part.
2322 WARNING: This may have resulted in incorrect results with simulations
2323 where there are more than one flow on a given link. [AL]
2326 * After a (long ?) discussion on simgrid-devel, we have decided that the
2327 convention we had on units was stupid. That is why it has been decided
2328 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2329 WARNING : This means that all previous platform files will not work as
2330 such with this version! A warning is issued to ask users to update
2332 A conversion script can be found in the contrib module of the CVS, under
2333 the name contrib/platform_generation/surfxml_update.pl [MQ]
2336 * Bug fix: Processes were started in reverse order, wrt deployment file.
2337 WARNING: if your code relies on this bug, please fix it. [AL]
2338 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2339 being executed on two different locations. [AL]
2340 * Bug fix: Failures are now better supported thanks to Derrick's tests
2341 (there was many failure situations I hadn't thought of and that weren't
2342 correctly handled). [AL]
2343 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2347 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2348 Exception occurring on server-side are propagated back to client (!).
2350 API CHANGE: the callback changed their prototype. Change:
2351 int my_handler(gras_socket_t expeditor, void *payload_data) {
2353 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2354 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2356 * New! function: gras_msg_handleall to deal with all messages arriving
2357 within a given period.
2358 * New! function: gras_socket_server_range to get a server socket in a
2359 range of port numbers (ease to avoid port number conflicts) [MQ]
2360 * New! gras processes display their backtrace when they get a SIGUSR1
2361 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2362 Sweet to debug RL processes [MQ]
2366 - Do not force experiment sizes to be expressed in kb, or it becomes
2367 impossible to measure the latency this way (needs one byte-long tests)
2368 WARNING: this changes the amok_bw_* function semantic. [MQ]
2369 - Implements the link saturation stuff. [MQ]
2370 * Peer management module:
2371 New! module factorizing code that we wrote over and over [MQ].
2374 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2375 - Test units are placed directly into the library code, they get extracted
2376 automatically and placed into the src/testall binary.
2377 - Convert most of the XBT tests to this system.
2378 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2379 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2380 live in the same C code. This required to change a little bit the API
2381 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2382 * New module: sparse graph structure with basic algorithms (this is work
2383 in progress and the API is not considered to be frozen yet). [AL]
2384 * Display more information on backtraces: source line & function names are
2385 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2386 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2387 * Reworked a little bit some #include statements to load only required
2388 headers. Some user code that relied on SimGrid to include stdlib or
2389 stdio may need to include it by themselves. [AL]
2390 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2392 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2393 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2394 but I guess that almost nobody used those parts. [MQ]
2396 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2398 SimGrid (3.0.1) stable; urgency=low
2401 * Unfortunately, I had missed 5 misnamed functions:
2402 xbt_fifo_item_t xbt_fifo_newitem(void);
2403 void xbt_fifo_freeitem(xbt_fifo_item_t);
2404 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2405 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2406 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2407 They're now deprecated. Please use their new versions:
2408 xbt_fifo_item_t xbt_fifo_new_item(void);
2409 void xbt_fifo_free_item(xbt_fifo_item_t);
2410 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2411 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2412 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2414 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2415 * Documentation: xbt_log module unmercifully reworked [MQ]
2416 * Bugfix: there was a problem with the ending of contexts with
2417 the pthread backend. It caused some weird deadlock or behavior
2418 depending on the pthread implementation. [AL]
2419 * Bugfix: get the exceptions raised in the simulator repport where
2420 and why they come from when they are not catched in time [AL, MQ]
2423 * Bugfix: Do repport the error when two non-connected hosts try to
2424 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2427 * Add additionnal checkings on communications. Assert that two
2428 communicating hosts are connected by a set of links... [AL]
2431 * Add additionnal checkings on channel values in communication [AL]
2432 * New: MSG_task_get_source to see on which host a task was generated [HC]
2433 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2434 the number of tasks waiting to be received on channel and sent
2436 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2437 waits for the first task coming from a given host.. [AL]
2439 GRAS new functionnalities: [MQ]
2440 * Enhance the parsing macro to allow the size of multidimentional objects
2441 to be given thru annotations.
2442 * New example (and documentation): Matrix Multiplication a la RPC
2443 (as when I was young!) and fix a bunch of bugs found on the way.
2445 GRAS performance improvements: [MQ]
2447 * Reduce the amount of cbps creation/destruction by making it static to
2448 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2451 * Change libdata to a set so that we can search for stuff by ID (and thus
2452 reduce the insane amount of dict lookups)
2455 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2456 RL communication on top of SG since it's so uneffective.
2457 It may also be used for inter-thread communication in RL, one day.
2458 * Use gras_datadesc_copy() to exchange messages on top of SG
2460 - improve message exchange performance on top of SG
2461 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2462 * Don't exchange on the network the size of the used part of buffer,
2463 instead, specify the possible buffer size to read().
2465 - reduces the amount of read/write calls (one pair per exchange)
2466 - reduces the amount of exchanged data (the size)
2467 - allows to retrieve all arrived data on receiver side, if we don't need
2468 it right now (subsequent read will peek the buffer)
2469 - allows the receiver to proceed with the begining of the stream before
2470 everything is arrived
2471 - make it possible to build an iov transport (using readv/writev)
2473 - take care of the data with non-stable storage (like stacked data),
2475 * If possible, TCP send uses vector I/O (when writev() is here)
2476 - Don't use it for receive since we send data sizes and data on the
2477 same stream, so we wouldn't be able to chain large amount of chunks
2478 before having to flush the stuff to read the size.
2479 * Rework the transport plugin mechanism to simplify it and reduce the
2480 amount of pointer dereferencement when searching for the right function
2483 * I guess that now, we do almost as few system calls as possible while
2484 doing as few data copy as possible.
2486 To improve it further, we could try to send all the sizes first and then
2487 all the data (to use iov on receiving size), but it's only a partial
2488 solution: when you have 2 dimensional data, the sizes of the second
2489 dimension is data of the first dimension, so you need 3 streams.
2491 I'm not sure the potential performance gains justify the coding burden.
2493 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2495 SimGrid (3.00) stable; urgency=high
2498 * New! Give the possibility to hijack the surf parser and thus bypass
2499 MSG_create_environment and MSG_launch_application. Have a look at
2500 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2502 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2504 SimGrid (2.96) unstable; urgency=low
2509 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2510 This deprecates the xbt_error_t mechanisms.
2511 It modifies (simplifies) all XBT and GRAS API.
2512 MSG API keeps unchanged (exceptions raised by XBT are catched from
2513 within MSG and masked with existing error handling facilities)
2516 * New! Add a FATPIPE model. [AL]
2517 * New! Add a parallel task model. [AL]
2518 * New! Add automatically a loopback interface (in the default
2519 network model) if none was precised.
2522 * Bugfix: MSG_process_resume now works with the current running process.
2524 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2525 * Modification of MSG_task_get_compute_duration. Once a task has been
2526 processed, the value returned by this function is now equal to 0. [AL]
2527 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2528 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2529 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2530 of a m_task. For now, MSG_task_cancel only works with computation
2532 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2533 of the processor (in Mflop/s) regardless of the current load on the
2535 * API Change: use proper naming convention for MSG_getClock and
2536 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2538 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2539 This function changes the priority of a computation task. This priority
2540 doesn't affect the transfer rate. A priority of 2 will make a task
2541 receive two times more cpu power than the other ones. This function
2542 has been added to suit the needs of Nguyen The Loc and hasn't been that
2543 much tested yet. So if it fails, please report it and send me your code.
2545 * API Change: removed all functions and types that were marked "deprecated"
2546 since many months. Renamed MSG_global_init_args to MSG_global_init.
2548 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2550 SimGrid (2.95) unstable; urgency=low
2553 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2554 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2555 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2558 * Bugfix: Allow absolute paths for platform description files [MQ]
2559 * Bugfix: do free the variables after use. Leads to drastic performance
2561 * Implement max_duration (ie, timeouts) on resources [AL]
2564 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2566 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2567 the needed support (timeouts) [AL]
2570 * Implement measurement sockets. You can now get the bandwidth between two
2571 hosts thanks to AMOK (see below). [MQ]
2572 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2573 dynar over the network (yeah) [MQ]
2574 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2575 * simple/static token-ring example. [Alexandre Colucci and MQ]
2576 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2577 in place before [MQ]
2580 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2581 type parsing. "short" and "long" modifiers were also ignored; other
2582 modifier (such as reference level) are still ignored. [MQ]
2583 * Embeed the buffer size within the buffer itself on SG. [MQ]
2584 That way, send() are atomic and cannot get intermixed anymore (at least
2585 the ones which are less than 100k; bigger messages still have the issue)
2586 * Array size pushed by the field, not by the field type (or each
2587 and every long int will push stuff to the cbps) [MQ]
2588 * use select() to sleep since it allows to portably sleep less than one
2591 GRAS (minor cleanups)
2592 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2595 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2596 * type category 'ignored' killed as it was never used and were difficult
2598 * whether a type can cycle or not is now a flag, leaving room for more
2599 flags (as "ignored", if we feel the need one day ;)
2600 * Rename raw sockets to measurement sockets since "raw" has another
2601 meaning in networking community.
2604 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2605 and offers features not belonging to GRAS but that most applications
2606 need. One day, it may be a set of plugins loadable at runtime.
2607 * New module: bandwidth
2608 bandwidth measurement between arbitrary nodes running this module. [MQ]
2610 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2612 SimGrid (2.94) unstable; urgency=low
2614 The first beta release of SimGrid 3 !
2618 * Update the main page and the FAQ. Adding references to gforge.
2621 * Add a gras_os_getpid function.
2624 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2625 * Extend the logs so that they also print PID, hostname, date, ... if
2627 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2628 and kill the old version which were in testsuite/
2629 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2633 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2634 for users. Instead, each programming environment (GRAS, MSG,...) use it
2635 when needed to provide such a feature to users.
2636 Don't shortcut the mechanism or you will also shortcut the virtualization
2637 you need on the simulator.
2641 * Cleanups in configury with regard to compile optimization/warning flags.
2642 Also add -fno-loop-optimize to any powerpc since it's the optimization
2643 killing gcc (< 3.4.0).
2644 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2645 needed by MSG examples complications
2646 * Borrow configury beautifications from PHP
2649 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2650 warning (thanks loris for stumbling into this one).
2651 * Bugfix: stop loading private headers (gras_config.h) from the public
2655 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2656 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2657 RL would imply a DNS resolver, which is *hard* to do in a portable way
2658 (and therefore delayed).
2659 * Implement a real timer mechanism and use it in timing macros. This allows
2660 to avoid rounding errors and get a 0.000005 sec precision in timing
2661 macros. While I was at it, various cleanups:
2662 - allow to declare more than one timed section per file (fix a stupid bug)
2663 - move some private declaration to the right place
2664 - merge conditional execution and timing macros into emulation module
2665 - document the module
2666 - make sure the module cleanups its mess on gras_exit
2667 * Documentation improvements:
2668 - (new) how to compile applications using GRAS
2669 - (new) emulation support (timing macros)
2671 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2673 SimGrid (2.93) unstable; urgency=low
2675 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2678 - Use Paje properly where used. Still to be sanitized properly.
2679 - Portability fix: Add an implementation of the contexts using pthread
2683 - Add xbt_procname(): returns the name of the current process.
2684 Use it to show the current process's name in all logging.
2686 - fix detection of older flex version and the reaction, since we do
2687 depend on modern ones (we use lex_destroy)
2688 - Better separation of SG and RL in the libs: remove all simulation code
2689 from libgras. As a result, this lib is now only 200k when stripped.
2690 Some of the xbt modules may also be duplicated (two sets and such) and
2691 should be cleaned/killed before SG3.
2692 - Insist on using xlC on AIX because of weird problems involving gcc there.
2693 - Cleanup the make remote stuff. This is now done by scripts
2694 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2695 mainly for our private use, but we're working on changing them to user
2698 - Bugfix: flush the socket on close only if there is some *output*.
2699 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2701 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2704 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2706 SimGrid (2.92) unstable; urgency=low
2708 Alpha 3 on the path to SimGrid 3
2712 - New! First try of benchmarking macros.
2713 - New! First try so that gras_stub_generator generate deployment and
2714 remote compilation helpers.
2716 - Bugfix: Initialization fix in msg_test.
2720 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2722 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2724 - API Change: gras_os_sleep to take the amount of seconds as a double.
2725 Accepting an int was error prone since it was the only location where
2726 seconds were coded as such. It leaded to damn rounding errors.
2727 - Bugfix: Hard to belive that timers ever worked before this.
2729 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2731 SimGrid (2.91) unstable; urgency=low
2733 Alpha 2 on the path to SimGrid 3
2737 - Bug fix in the lmm_solver.
2739 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2740 through the function MSG_paje_output.
2741 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2742 - It is possible to bound the rate of a communication in MSG with
2743 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2744 to write it in the changelog).
2745 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2750 - Various cleanups to the autotools stuff
2751 - Begin to move Gras examples to examples/gras/
2752 - Let make distcheck work again (yeah!)
2754 - documentation overhauled using doxygen.
2755 gtk-doc-tools is dead in SimGrid now.
2756 - Automatically extract all existing logging categories, and add the list
2757 to the documentation (long standing one, to say the less)
2759 - Cleanup the known architecture table. Reorder the entries to group what
2760 should be, and use a more consistent naming scheme.
2761 (some of the test dataset are still to be regenerated)
2762 - New! Allow library to register globals on each process just as userdata
2764 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2765 lookup time (for now).
2766 Use it in msg and trp.
2767 This cleans a lot the internals and helps enforcing privacy of the
2768 headers between the gras components.
2769 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2770 - Bugfix: gras_os_time was delirious in RL.
2771 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2773 - Reenable GRAS now that it works.
2775 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2777 SimGrid (2.90) unstable; urgency=low
2779 Alpha 1 on the path to SimGrid 3
2781 * It is a long time since the last release of SimGrid. I'm sorry about
2782 that but as I had told you, I was rewriting a lot of things. I apologize
2783 to those who had been reporting bugs to me and that I had not answered.
2784 If your bug is still in the new version, please tell me. Here is a
2785 summary of the main changes.
2787 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2788 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2789 portability, speed, and a lot more but you'll figure it out later.
2790 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2791 I wanted to release the new MSG as soon as possible and I have
2792 broken GRAS, which is the reason why, for now, only MSG is fully
2793 functional. A laconic description of these projects is available
2794 in the documentation.
2796 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2797 kernel optimized for our needs (called SURF but only the developers
2798 should use it). Hence, MSG is now roughly 30 times faster and I think
2799 that by rewriting a little bit MSG, I could event speed it up a little
2800 bit more. Beside the gain in speed, it is also much easier to encode a
2801 new platform model with SURF than it was with SG. More to come...
2803 * REVOLUTION 3: I have tried to change a little as possible the API of
2804 MSG but a few things really had to disappear. The main differences
2805 with the previous version are :
2806 1) no more m_links_t and the corresponding functions. Platforms are
2807 directly read from a XML description and cannot be hard-coded
2808 anymore. The same format is used for application deployment
2809 description. The new format is described in the documentation.
2810 Have a look in tools/platform_generation. There is a tiny script
2811 that converts from the old platform format to the new one. Concerning
2812 the application deployment format, parsing the old one is tricky.
2813 I think most of you should however be able to convert your files. If
2814 it is really an issue, I can write a C code that does the conversion.
2816 2) the toolbox tbx does not exist anymore. We now have a library
2817 with much more data-structures but without the hash-tables (we have
2818 dictionaries that are much faster).
2820 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2822 *****************************************************************************
2823 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2824 * don't feel like converting the oldies. (Mt) *
2825 *****************************************************************************
2828 Version 2.90: "the long awaited one"
2829 - Finished rewriting and debugging MSG. Rewrote the documentation.
2830 - disable GRAS for now since it needs to be ported to the newest SG
2833 - Finish the port to windows (using mingw32 for cross-compile)
2836 - Main loop and datastructures of SURF. A cpu resource object is
2837 functional. Surf can thus be used to create cpu's with variable
2838 performance on which you can execute some actions.
2840 2004-11-15 Martin Quinson
2841 - Port to ARM. Simply added the alignment and size descriptions. Should
2842 work, but the ARM machines are so slow that I didn't had the opportunity
2843 to 'make check' over there yet.
2845 2004-11-15 Arnaud Legrand
2846 - Trace manager now written. It uses a heap structure and is therefore
2847 expected to be efficient. It may however be speeded up (particularly
2848 when many events occur at the same date) by using red and black
2849 trees. One day maybe...
2850 - Max-min linear system solver written. It uses a sparse matrix
2851 structure taking advantage of its expected use. Most operations are
2852 O(1) and free/calloc are called as few as possible. The computation of
2853 the minimum could however be improved by using a red and black tree
2856 2004-11-03 Arnaud Legrand
2857 - Rename every gras_* function that was in xbt/ to its xbt_
2859 - Add a heap and a doubly-linked list to xbt
2860 - Added a dichotomy to the dictionaries. make check works as well before
2861 so I assume that the patch is correct. I do not know however if things
2862 run effectively faster than before now. :)
2864 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2865 SimGrid, and the version number is bumped to 2.x
2867 2004-10-29 Martin Quinson
2868 - Introduction of the remote errors.
2869 They are the result of a RMI/RPC on the remote machine.
2870 ErrCodes being scalar values, you can't get the host on which those
2871 errors did happen. Extending the error mechanism as in Gnome is possible.
2872 No idea yet whether it is a good idea.
2874 2004-10-28 Martin Quinson
2875 - Interface revolution: the Starred Structure Eradication.
2876 I used to do typedef struct {} toto_t; and then handle *toto_t.
2877 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2878 - ???_t is a valid type (builded with typedef)
2879 - s_toto_t is a structure (access to fields with .)
2880 - s_toto is a structure needing 'struct' keyword to be used
2881 - e_toto_t is an enum
2882 - toto_t is an 'object' (struct*)
2884 typedef struct s_toto {} s_toto_t, *toto_t;
2885 typedef enum {} e_toto_t;
2886 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2887 s_toto_t) is private.
2889 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2890 so that it can set the variable to NULL. It was so for dicts and sets,
2891 it changed for dynars.
2893 - Fix a bunch of memleaks in dict_remove
2894 - Fix a bug in sg/server_socket opening: it failed all the time.
2896 2004-10-07 Martin Quinson
2897 - Speed up dynar lookup operation a bit.
2899 gras_dynar_get is dead.
2901 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2902 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2903 (which gives you the address of the stored data).
2905 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2906 copy of the data using an affectation to do the job and not a memcpy.
2908 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2909 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2911 It does not really speedup the dynar test because they are
2912 setting elements all the time (and look them seldom). But the dict does
2913 far more lookup than setting.
2915 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2917 2004-10-05 Martin Quinson
2918 - Allow to (en/dis)able the cycle detection at run time.
2920 Whether we should check for cycle or not is now a property of each
2921 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2922 datadesc_cycle_unset allow to remove this property when previously set.
2924 Note that the cycle detection is off by default since it impacts the
2925 performance. Watch the data you feed GRAS with ;)
2927 This property is hereditary. Any element embedded in a structure having it
2928 set have it set for the time of this data exchange.
2930 You should set it both on sender and receiver side. If you don't set it on
2931 sender side, it will enter an endless loop. If you forget on receiver
2932 side, the cycles won't be recreated after communication.
2934 - Header reorganization.
2935 Kill gras_private.h, each submodule must load the headers it needs.
2937 2004-10-04 Martin Quinson
2938 - Interface revolution: do not try to survive to malloc failure.
2940 Now, gras_malloc and friends call gras_abort() on failure.
2941 As a conclusion, malloc_error is not a valid error anymore, and all
2942 functions for which it was the only gras_error_t return value are
2943 changed. They now return void, or there result directly.
2944 This simplify the API a lot.
2946 2004-09-29 Martin Quinson
2947 - Re-enable raw sockets.
2948 Created by gras_socket_{client,server}_ext;
2949 Used with gras_raw_{send,recv}
2952 It should allow to kill the last bits of gras first version soon.
2954 This is not completely satisfactory yet (duplicate code with
2955 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2958 - Simplify transport plugin (internal) interface by not passing any
2959 argument to _server and _client, but embedding them in the socket
2962 2004-09-28 Martin Quinson
2963 - Finish the port to AIX.
2964 autoconf was my problem (segfault within the malloc replacement
2965 function. No idea why)
2967 2004-09-16 Martin Quinson
2968 - Fix some size_t madness on 64bit architectures.
2970 2004-09-08 Martin Quinson
2971 - Reduce the number of system headers loaded, overload some more system
2972 calls (such as malloc to cast the result of the system one, and work
2974 - Fix and reintroduce the config support
2976 2004-09-07 Martin Quinson
2977 - Source code reorganization to allow Arnaud to surf all over there.
2978 - Allow to document the logging categories.
2979 - Remove all uppercase from logging categories and useless cleanup in names.
2981 2004-08-18 Martin Quinson
2982 Version 0.6.2 (protocol not changed; API changed)
2983 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2984 gras_error_t), and NULL when not found. Functions expecting a msgtype
2985 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2986 hopefully usefull message.
2987 - Portability to prehistoric sparcs again
2989 2004-08-17 Martin Quinson
2990 Version 0.6.1 (protocol not changed; ABI not changed)
2991 - prealloc some buffers to speed things up
2993 2004-08-11 Martin Quinson
2994 Version 0.6 (protocol not changed; ABI expended)
2995 - The parsing macro can deal with the references, provided that you add
2996 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2998 2004-08-09 Martin Quinson
2999 Version 0.5 (protocol not changed; ABI changed)
3000 - Allow to off turn the cycle detection code in data exchange at
3001 compilation time. It should be at run time, but I'm short of time (and
3002 the config stuff is still broken). That way, we keep dict out of the
3003 critical path, which is good because the performance is poor:
3004 - search not dichotomial yet
3005 - dynar give no way to access their content and memcpy everytime
3006 - In composed data description (struct, ref and so on), stop foolness of
3007 keeping the subtype's ID, but store the type itself. This keeps sets out
3008 of the critical path, which is good since they rely on dynar and
3009 dictionnaries. The only loose of that is that we cannot detect the
3010 redeclaration of a structure/union with another content (but I'm not sure
3011 the code detected well this error before anyway). We still can detect
3012 the redefinition discrepancy for the other types.
3013 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3014 it breaks the code because of type-punning used all over the place).
3015 This breaks on all non-gcc architectures (for now).
3017 All those changes (plus the buffer of last time) allow me to gain 2 order
3018 of magnitude on cruel tests consisting of 800000 array of integers on two
3019 level of a hierarchical structure (200 secondes -> 4 secondes)
3022 - the selector of reference must now return the type it points to, not
3023 the ID of this type.
3025 2004-08-06 Martin Quinson
3026 Version 0.4 (protocol changed; ABI not changed)
3027 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3028 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3029 - Implement a buffer transport, and use it by default (it relies on tcp in
3030 real life and on sg in simulation).
3031 That's a bit hackish since I had a new field to the structure to store
3032 its data without interfering with the subtype ones. Inheritance
3033 is tricky in C. And that's a kind of reverse inheritance with one class
3034 derivating two classes. Or maybe a game with java interfaces. Anyway,
3035 that's damn hard in C (at least).
3036 Moreover, I got tired while trying to ensure plugin separation and
3037 genericity in SG mode. MSG wants me to do weird things, so let's go for
3038 cruel hacks (temporarily of course ;).
3039 See comment in transport_private.h:71
3040 - do not include all the _interface headers in private but in the files
3041 which really need them (to cut the compilation time when they are
3044 2004-07-26 Martin Quinson
3045 Version 0.3 (protocol not changed; ABI changed)
3046 - Major overhault of the datadesc interface to simplify it:
3047 - shorted the function names:
3048 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3049 - add a trivial way to push/pop integers into the cbps without malloc.
3050 This allows to make really generic sub_type description, which simply
3051 pop their size of the stack.
3052 - add a function gras_datadesc_ref_pop_arr() which does what users want
3053 most of the time: Declare a dynamic array (which pops its size of the
3054 stack) and declare a reference to it. Poor name, but anyway.
3055 - kill the post-send callback, add a post-receive one
3057 2004-07-23 Martin Quinson
3058 Version 0.2 (protocol changed; ABI changed)
3059 - add some testing for cpbs in the test cases, and fix some more bugs.
3060 This invalidate again the little64 data file, since I cannot regenerate
3062 - remove an awfull optimization in the logging stuff, allowing me to:
3063 - understand it again
3064 - learn gcc how to check that the argument match the provided format
3065 - fix all errors revealed by gcc after that
3066 - internal keys of dict are not \0 terminated. Deal with it properly in
3067 loggings instead of segfaulting when the user want to see the logs :-/
3069 2004-07-22 Martin Quinson
3070 - Fix some stupid bug preventing cbps (callback postit) from working
3072 2004-07-21 Martin Quinson
3073 - Some documentation cleanups
3074 - remove the useless last argument of msgtype_declare
3075 - rename the Virtu functions to fit into the 'os' namespace
3076 - move headers src/include -> src/include/gras/ and stop fooling with
3077 gras -> . symbolic link
3078 - make distcheck is now successful
3080 2004-07-19 Martin Quinson
3082 - Build shared library also
3083 - Install html doc to the right location
3084 - stop removing maintainer files in make clean
3085 - build tests only on make check
3087 2004-07-13 Martin Quinson
3089 - No major issue in previous version => change versioning schema
3090 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3091 corresponding dataset.
3093 2004-07-11 Martin Quinson
3095 - Get it working with any kind of structure (we can compute the padding
3096 bytes remotely for all the architectures I have access to)
3097 - Implement the structure parsing macro (still not quite robust/complete)
3098 - Improvement to the remote testing toysuite
3100 2004-07-10 Martin Quinson
3101 [autoconf mechanism]
3102 - get ride of a bunch of deprecated macros
3103 - actually run the test for two-compliment, not only compile it :-/
3104 - test whether the structures get packed (and bail out if yes. Damn.
3105 Alignment is a serious matter)
3106 - test whether the structures get compacted (but respecting the alignment
3107 constraints of each types)
3108 - test whether the array fields of structures can straddle alignment boundaries
3110 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3111 (took me 2 hours to find that bug, looking at the wrong place)
3112 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3113 before sending/receiving and are used to compute the offsets of fields
3114 - Given that padding size depend even on compiler options, keep track of
3115 alignment and aligned_size only for the current architecture. Not a big
3116 deal since we send structure fields one after the other (seems
3118 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3119 program, let it work on linux/gcc/little32. portability todo.
3121 2004-07-08 Martin Quinson
3122 - import and improve remote compilation support from FAST
3123 - make sure make check works on half a dozen of machines out there
3125 2004-07-07 Martin Quinson
3126 Let's say it's version 0.0.3 ;)
3127 - Implement conversions (yuhu!)
3128 - Let it work on solaris (beside conversion, of course)
3129 - Stupid me, using rand() to generate the conversion datatests in not wise.
3131 2004-07-06 Martin Quinson
3132 - Let make dist work, since I'm gonna need it to compile on remote hosts
3133 - Let Tests/datadesc_usage write the architecture on which the file was
3134 generated as first byte.
3135 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3136 architecture descriptions.
3137 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3138 on those architectures.
3139 - Optimization: send/recv array of scalar in one shoot
3141 2004-07-05 Martin Quinson
3142 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3144 - Plug a whole bunch of memleaks
3145 - each process now have to call gras_{init,exit}. One day, their log
3146 settings will be separated
3147 - Continue the code factorisation between SG, RL and common in Transport.
3149 2004-07-04 Martin Quinson
3151 - Redistribution between SG and RL.
3152 We wanna have to accept in SG, so move accepted related parts of RL in
3153 the common part. (more precisely, the dynar of all known sockets is no
3154 more a static in transport.c, but part of the process_data)
3156 [gras_stub_generator]
3157 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3159 2004-07-03 Martin Quinson
3160 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3161 them all up in one shot)
3163 - Fix the ugly case of reference to dynamic array.
3164 - New (semi-public) function gras_datadesc_size to allow the messaging
3165 layer to malloc the needed space for the buffer.
3167 - gras_socket_close now expect the socket to close (and not its address to
3168 put NULL in it after it). This is because the socket passed to handlers
3169 is one of their argument (=> not writable).
3171 - propagate the interface cleanup from last week in datadesc, ie remove a
3172 superfluous level of indirection. User pass adress of variable
3173 containing data (both when sending and receiving), and not of a variable
3174 being a pointer to the data. Let's say that I like it better ;)
3175 The price for that is constructs like "int msg=*(int*)payload" in
3176 handlers, but it's a fine price, IMHO.
3178 - Let it work in RL (yuhu)
3180 2004-06-21 Martin Quinson
3182 - porting SG plugin and SG select to new standards (works almost).
3183 - plug memleaks and fix bugs around.
3186 - cleanup the prototype of data recv and force users to specify when they
3187 want to handle references to objects. Test case working even for cycles.
3188 - plug memleaks. Valgrind is perfectly ok with this.
3190 2004-06-12 Martin Quinson
3192 - cleanup the separation between plugin and main code in plugin creation
3194 2004-06-11 Martin Quinson
3196 - Reput hook for raw sockets, needed for BW experiments
3197 - kill a few lines of dead code
3198 [Data description] Interface cleanup
3199 - gras_datadesc_by_name returns the searched type or NULL.
3200 That way, no variable is needed to use a type desc once, which makes
3202 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3203 two parameters were strings (field name, type name), leading to
3205 [Dicos] Interface cleanup
3206 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3207 This is consistant with the dynar API.
3209 2004-04-21 Martin Quinson
3211 - Porting to new standards.
3213 - interface cleanup.
3214 There is no bag anymore, no need to take extra provision to mask the
3215 pointers behind "ID".
3216 Better splitup of functions between files create/exchange/convert.
3217 This is still a bit artificial since convert and receive are so
3218 interleaved, but anyway.
3220 - add a queued message list to procdata (the ones not matching criteria
3222 - factorize some more code between SG and RL wrt procdata
3224 - use gras_exit in example to track memleaks
3225 - get rid of gs_example now that GS is properly integrated into gras
3226 - update run_test to integrate the latest tests (datadesc)
3228 - rename WARNINGn macros to WARNn since it prooved error-prone
3230 2004-04-19 Martin Quinson
3232 - register init/exit functions within gras module mechanism
3233 - send/receive function.
3234 Convertion is not implemented, but short-cutted if not needed.
3235 struct/array elements are sent one by one (instead of block-wise), but
3236 nobody really cares (yet). Get a prototype before optimizing.
3237 - tests (using a file socket) for DD send/receive on:
3238 - base types: int, float
3239 - array: fixed size, string (ie ref to dynamic string)
3240 - structure: homogeneous, heterogeneous
3241 - chained list, graph with cycle
3242 Believe it or not, valgrind is not too unhappy with the results. The
3243 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3247 - Bugfix when using a filename explicitely (instead of '-')
3249 2004-04-09 Martin Quinson
3251 - factorize more code between RL and SG in socket creation
3252 - Complete the implementation and tests of:
3254 o file (only in RL, and mainly for debugging)
3256 I lost 3 days to design a portable address resolver, and then decided
3257 that the prototype mainly have to run on my box.
3258 Addressing portability too early may be like optimizing too early :-/
3260 - use gras_init in the Tests instead of the crappy parse_log_opt
3261 (the latter function is removed)
3262 [Conditional execution]
3263 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3264 [Code reorganisation]
3265 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3266 Build examples against the RL library, since there is no way to disable
3267 its creation for now.
3269 For information, the beginning of coding on GRAS was back in june
3270 2003. I guess that every line has been rewritten at least twice since