1 SimGrid (3.15) UNRELEASED; urgency=low
3 - Rename NetCards to NetPoints.
4 This was intended to help NS3 users, but that's not a netcard.
5 That's a point in the routing algorithm, let's avoid do wrong simplifications.
8 - New function: MSG_process_yield()
11 - Ensure that an actor can kill itself with Process::exit()
12 - Kill the obscure NativeException. Nobody want to survive the issues
13 it denotes, so use JniException that is a RuntimeException (not to
14 be catched explicitely).
16 -- target_date=March 20 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
18 SimGrid (3.14.159) stable; urgency=low
20 The previous release was only a crude approximation of a Pi release;
21 this one should be a bit better.
23 - Revert a last minute change that broke on Mac OSX.
24 - Fix the changelog and NEWS files.
25 - Improve the documentation, in particular of the routing module.
26 - Integrate some patches flying around in Debian.
28 -- Wed Dec 28 17:14:49 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
30 SimGrid (3.14) stable; urgency=low
33 * The whole documentation was reworked and reorganized. There is
34 still much room for improvement here, but we're on it.
37 * We now need python3 for our testing infrastructure.
38 * Model checking needs libevent
41 * Privatization is now activated by default in smpirun
42 You can turn it off with -no-privatize if something goes wrong.
44 * Call-location tracing for SMPI.
45 You can add the exact location (filename / linenumber) of an MPI call to
46 your trace files and slow down or speed up the simulation between two
47 consecutive calls by using an adjustment file (see the documentation).
49 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
50 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
51 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
52 values were ignored in some cases. The timings of these functions can now
53 be significantly different.
55 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
56 smpi/running-power is renamed to smpi/host-speed
58 * smpi/grow-injected-times option to enable or disable multiplication of the
59 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
60 make simulation less precise (but also much faster).
62 * smpirun script should be (much) faster for large deployments.
64 * SMPI tracing : fixed issue with poor matching of send/receives.
66 * Replay : Fix broken waitall
68 New functions and features
69 * MSG_parallel_task_execute_with_timeout, to timeout computations.
71 Dropped / renamed functions and features
72 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
73 - MSG_mailbox_is_empty() -> Mailbox::empty()
74 - MSG_mailbox_front() -> Mailbox::front()
75 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
76 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
77 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
78 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
79 * Don't pass the free_f parameter to property related functions:
80 - MSG_host_set_property_value()
81 - MSG_as_router_set_property_value()
82 - MSG_storage_set_property_value()
83 * VM properties. Since msg_vm_t are msg_host_t, just use
84 MSG_host_get_property_value() and friends
85 * VM I/O related things:
86 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
87 - Unimplemented save/restore methods
88 * MSG_as_router_get_property_value() was redundent with
89 MSG_environment_as_get_property_value().
90 - Removed MSG_as_router_*propert*() functions
91 - Added MSG_environment_as_set_property_value() for consistency
92 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
93 Well, they are still there for now, but deprecated with a warning.
94 Please switch to xbt_dict_new_homogeneous() before this is removed
96 * Task affinity. Its intended behavior (that was very badly tested
97 and probably not really working) was deceiving what most users
98 would have hoped here.
99 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
100 * Option network/coordinates is now useless and should be dropped.
103 * Added option storage/max_file_descriptors to allow more than 1024 files opened
106 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
107 * Added Dragonfly topology support
110 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
111 EiBps, EBps, Eibps, Ebps for bandwidth.
112 They may become useful to some lucky ones.
115 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
116 * ex/app_tokenring: new example, very similar to the MSG Token Ring
117 * ex/async_waitAll: new example, on asynchronous communications
120 * Memory usage should be decreased for simulations with a large number
121 of processes. This also helps for SMPI.
123 -- Sat Dec 24 22:34:26 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
125 SimGrid (3.13) stable; urgency=low
127 The Half Release, a.k.a. the Zealous Easter Trim.
129 - We removed half of the lines, that were mostly experimental cruft.
130 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
131 (not counting blanks and comments -- according to openhub.net)
132 - The internals are now compiled in C++ (and will soon be clean C++)
133 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
134 59 klines of C, etc. We added only 29 klines of C++ in replacement.
136 * Backwards Compatibility breaks
137 - Removed Lua simulation bindings (switch to C or Java for that).
138 Lua can still be used to describe platforms
139 - Removed Java kernel plug-ins.
140 Will be reintroduced after the ongoing major internals reorg.
142 - the following functions were removed.
143 They were too specific and should be reimplemented in a generic
144 way, with filter function.
145 - MSG_task_listen_from_host
146 - MSG_mailbox_get_count_host_waiting_tasks
147 - MSG_mailbox_put_with_timeout was removed.
148 Please use MSG_task_send_with_timeout instead.
150 - the SD_application_reinit function was removed. It has been a noop for a while.
151 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
152 replaced by a more flexible mechanism.
153 - The following functions thus do not exist anymore
154 - SD_workstation_get_access_mode
155 - SD_workstation_set_access_mode
156 - SD_workstation_get_current_task
157 - Basic estimation functions have been removed but can easily be replaced
158 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
159 - SD_workstation_get_computation_time => amount / sg_host_speed()
161 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
162 Use VM.getSpeed()*load/100 for the legacy behavior.
164 - option enable_tracing was removed. It was not doing anything for a while.
165 - In the ModelChecker:
166 - the model-checker now ptraces the model-checked process which means
167 you cannot use a debugger on the latter anymore (we might make this
168 optional in the feature);
169 - removed soft-dirty page tracking;
170 - remove model-checked side snapshot management,
171 MC_snapshot() and MC_compare_snapshot();
172 - keep the MC_cut() function as a stub (it was not really working
173 in the previous release).
176 * All options are consistently in kebab-case. Old names are kept as alias.
179 * Switch to platform v4 format.
180 - Rename from 'power' to 'speed' the attributes describing the amount of
181 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
182 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
183 - In <host> and <link>, attributes availability and state are gone.
184 It was redundent with state and availability traces, and with peak values.
185 - In <cluster>, attributes availability_file and state_file are gone.
186 It was too complex and unused.
187 - Kill <gpu>. Was not doing anything.
188 - The DOCTYPE points to the right URL:
189 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
190 (the file at this address now documents the changelog since its v1)
191 - A warning is emitted for unit-less values (they are still accepted).
192 - speed. Default: 'f' or 'flops'. Also defined:
193 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
194 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
195 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
196 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
197 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
198 - latency. Default: 's' second. Also defined:
199 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
201 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
202 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
205 * s4u::Host is now the preferred public interface to the Host features.
206 sg_host_* functions are C bindings to the exact same behavior
207 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
210 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
211 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
212 They were deprecated since a few years, and probably did not even compile anymore.
215 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
216 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
217 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
218 encouraged to update their codes. A list of the modified functions can be found at the end of
219 include/simgrid/simdag.h
222 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
223 the lambda or closure passed as a parameter will run in kernel mode.
224 All the callback functions should be rewritten to that interface at some point.
227 * Reorganizing and cleaning the internals all around the place.
230 * Remove old default barrier/bcast buggy algorithms (see #18407)
231 * Various bug fixes to handle more codes
232 * Remove the need for the --foreground option of smpirun (it is still
233 accepted for backward compatibility).
236 * Kill the setset data container: MC don't use it anymore.
237 * Kill the queue data container: it made more sense with GRAS.
238 * Kill the xbt_peer_t data type: it's useless without GRAS.
239 * Kill rm_cb feature of config sets: it was never useful.
240 * Kill graphxml parsing feature. It was not used.
241 * Kill the deprecated code protected by XBT_USE_DEPRECATED
243 - xbt_dynar_sort_strings(), when the content is char*
244 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
245 They throw exceptions on invalid input;
246 * C++ support for declaring CLI flags (simgrid::config::Flag);
247 * class for abstracting different signal backends (simgrid::xbt::signal).
248 with no external dependencies (we need very simple signals).
251 * refactoring and cleanup of the code;
252 * ongoing process to cleanly separate the model-checking algorithms
253 from the code model-checking support.
255 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
257 SimGrid (3.12) stable; urgency=low
259 The Facelift Release.
262 * Require g++ v4.7 at least to not speak prehistorical C++.
263 * Require Boost 1.48 (for signal2 component).
264 * Java must be version 7 at least when activated.
265 * Builds on Windows again (including Java bindings).
266 * Tracing is now always enabled (no way to turn it out)
267 * Remove GTNetS. It was not working anyway.
268 * Various cleanups in the cmake scripts.
269 * Move headers around to sort them out on installed systems:
270 - instr/instr.h -> simgrid/instr.h
271 - instr/jedule/* -> simgrid/jedule
272 - simdag/datatypes.h was removed
273 - simdag/simdag.h -> simgrid/simdag.h
274 - msg/datatypes.h was removed
275 - msg/msg.h -> simgrid/msg.h
278 * Interface improvement:
279 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
280 - Sanitize the interface in MSG_task_ module:
281 - Merge two functions that were close enough but misleading:
282 set_compute_duration(t) -> set_flops_amount(t)
283 get_remaining_computation(t) -> get_flops_amount(t)
284 - set_data_size(t) -> set_bytes_amount(t)
285 get_data_size(t) -> get_bytes_amount(t)
286 - Massive cleanups in the functions related to the energy
287 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
288 - New: MSG_host_get_pstate()
290 - msg/energy/onoff: switching hosts on and off
292 * Interface improvement:
293 - Rename Host.isAvail() to Host.isOn()
294 - Rename Process.currentProcess() to Process.getCurrentProcess()
295 - Rename Task.setDataSize() to Task.setBytesAmount()
296 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
298 - #18874: Actually allows the GC to reclaim tasks
302 - SIMIX_process_throw: raises an exception in a remote process
303 * Refactoring: Separate sub-modules
304 - libsmx: the public interface, as libc in a real system
305 - popping: the strange dance that converts a user request into a kernel handling
306 - smx_context_*: the virtualization mechanisms that embed the user code
307 - smx_*: the handling of each simcalls
309 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
310 * Rename smx_action_t into smx_synchro_t, making explicit that these
311 things are used to synchronize processes with their environment.
312 For example, a communication is a sort of synchronization involving
313 the communicating processes (that may block until the exchange) and
314 the platform. The same can be said from computations, etc.
316 - #18888: segfault when a process ends before its kill_time
320 - 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.
321 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
322 - MPI_Type_set_name, MPI_Type_get_name
323 - MPI_*_c2f and MPI_*_f2c functions
324 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
325 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
326 - Activate a lot of new tests from the mpich 3 testsuite
328 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
329 - InfiniBand network model added : Based on the works of Jerome Vienne
330 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
331 - When smpi/display_timing is set, also display global simulation time and application times
332 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
333 * Collective communications
334 - 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
335 - 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/).
336 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
338 - Replay now uses algorithms from wanted collective selector
339 - Replay can be used with SMP-aware algorithms
340 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
341 - 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.
343 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
344 - Correctly use loopback on fat-tree clusters
345 - Asynchronous small messages shouldn't trigger deadlocks anymore
346 * Energy/DVFS cleanup and improvement
347 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
348 - new: smpi_get_host_pstate()
351 - "Full" network optimization flag was broken since Surf++
352 - Better handling of precision flags in maxmin
353 - Fix bug causing sometimes "Impossible" errors
354 - Properly pass cluster properties to included hosts
355 * Improvement of the Energy plugin.
356 - Always update the consumption before returning that value
357 - New property: watt_off to denote the dissipation when the host is off
358 - New functions getWattMinAt and getWattMaxAt to retrieve the
359 dissipation of pstates that we are not currently at.
360 * Java: class NetworkLink renamed to Link
361 * New function: simcall_process_get_kill_time()
362 * Massive rename s/workstation/host/
363 - That's intrusive, but that's good for the project consistency. Sorry.
364 - Change config option "workstation/model" into "host/model"
367 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
368 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
369 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
370 - xbt_dirname and xbt_basename for non-POSIX systems
372 * The model checker now runs as a separate process.
373 * The model checker runs must now be launched with the new simgrid-mc program.
374 * Record/Replay: the MC can display a textual representation of a path in the
375 execution graph. It can then be replayed outside of the model checker.
377 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
379 SimGrid (3.11) stable; urgency=low
384 * Normalizing pointers addresses tool for better diff between logs
387 * Add cloud examples using new VMs
388 - examples/msg/cloud/two_tasks_vm.tesh
389 - examples/msg/cloud/simple_vm.tesh
390 - examples/java/cloud/cloud.tesh
391 - examples/java/cloud/migration/migration.tesh
392 * Add java surf examples:
393 - examples/java/surfPlugin/surf_plugin.tesh
394 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
395 - examples/java/surfCpuModel/surf_cpu_model.tesh
396 * Add SMPI+MSG example:
397 - examples/smpi/smpi_msg_masterslave/
402 - msg task destroy cancel test
403 - msg_host on/off test
404 * Move all tests in testsuite to teshsuite (adding tesh files)
405 * Restructure teshsuites
406 - one folder for each kind of test
407 * Restructure AddTests.cmake
409 - structure the order of tests (with sections)
412 * Add virtual machine
413 - creation of a VM on a PM
414 - migration of a VM from a PM to another PM
416 - MSG_process_join(msg_process_t process, double timeout)
417 - msg_bar_t MSG_barrier_init(unsigned int count)
418 - int MSG_barrier_wait(msg_bar_t barrier)
419 - void MSG_barrier_destroy(msg_bar_t barrier)
420 - msg_as_t MSG_environment_as_get_by_name(const char * name)
421 * New option "msg/debug_multiple_use" to help debugging when a task is used
424 - Improvements and finalization of MSG_storage, MSG_file APIs and their
426 - Increase code coverage in test suites
430 * Protect context stacks against stack overflow. The number of protected memory
431 pages allocated on the top of each stack (1 by default) can be configured
432 with the new command line option --cfg=contexts/guard_size:<value>.
433 * Simcalls are now generated by a python script that
434 - generates files included by SimGrid
435 - checks that all the functions exist, and proposes prototypes
437 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
438 host_set_data, host_get_data
440 - simcall_process_join(smx_process_t process, double timeout)
441 * Fix bug where sleeping processing could not be suspended.
444 * Translate surf models from C to C++
445 - Generic classes for all models: Model, Resource, Action
446 - A generic interface for each kind of model (CPU, Network, Storage
447 Workstation, WorkstationVM)
449 * Translate surf routings from C to C++
450 * Add callbacks using sigc++ or boost::signals2
451 - Add callback functions for resource creation/destruction
452 - Add callback functions for action state change
453 - Handle Energy as a plugin
454 * Replace swag by boost::intrusive
455 * Add new routing models for clusters. For documentation, see
456 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
457 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
458 parameters for cluster tag
459 - Fat trees, with topology="FAT_TREE" and
460 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
461 - see examples/platforms/torus_cluster.xml and
462 examples/platforms/fat_tree_cluster.xml
466 * Hostfiles support host:nb_processes construct to deploy several processes on
468 * Collective communication algorithms should not crash if used with
469 improper number of nodes and report the error.
470 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
471 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
472 MPI_Dims_create, MPI_Cart_sub are supported.
473 * New interface to use SMPI programmatically (still depends on MSG for
474 some parts, see examples/smpi/smpi_msg_masterslave) :
475 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
479 * Global variables privatization in MPI executables is now performed at runtime
480 with the option smpi/privatize_global_variables (default:no).
481 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
482 dynamic libraries loaded by the application are not privatized (static
483 linking with these libraries is advised in this case)
486 * Options defined in XML work correctly now.
489 * New cmake option, enable_lib_in_jar, to control whether native libraries are
490 copied into simgrid.jar or not (ON by default). Use this option if you want
491 to reduce the size of the installed simgrid.jar, *and* the native libraries
492 are kept installed elsewhere.
493 * Surf binding with SWIG (code generated in maintainer mode only):
494 - plugin to handle callbacks
495 - CPU model only for the moment
498 * Supernovae build mode is definitively removed. It was used to improve
499 inlining and inter-module optimizations. It is nowadays superseded by
500 link-time optimizations commonly available in compilers.
501 * Update ns-3 find lib. Bindings for ns-3 should work again now.
502 * Add boost dependency for surf++
503 * Add new macro for tests
504 - ADD_TESH(name <tesh_args>)
505 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
509 - xbt_bar_t XBT_barrier_init(unsigned int count)
510 - int XBT_barrier_wait(xbt_bar_t barrier)
511 - void XBT_barrier_destroy(xbt_bar_t barrier)
512 * Make the xbt_os_time module public
514 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
516 SimGrid (3.10) stable; urgency=low
518 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
521 * Reintegrate Java to the main archive as desynchronizing these
522 package is not acceptable anymore (Java is now considered stable)
523 * Add explicit synchronization facilities through semaphores
524 * Bug fix: Task.setDataSize() only changed the C world, not the value
525 cached in the Java world
528 * Dramatically change the way files are handled. API and internals changed, but
529 this part of MSG was not considered as production grade either.
530 * Add explicit synchronization facilities through semaphores
531 * Add a new function MSG_host_get_process_list()
532 * Preliminary DVFS support (see examples/msg/energy/ for details)
535 * SMPI is now included directly in the libsimgrid as the windows
536 linker doesn't force us on splitting it anymore.
537 * Improvements of the SMPI replay tool:
538 - Most of the collective communications are now rooted in the same process as
539 in the original application.
540 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
541 used until now). Multiple data types can now be used in a trace.
542 - The replay tool now supports traces produce either by TAU or a modified
544 - Bug Fix: the compute part of the reduce action is now taken into account.
545 - Gatherv collective is now supported
546 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
547 of smpirun outputs time independent traces for the current run. One file
548 is created per process. If too many processes are simulated, this behavior
549 can be changed to one file for all processes by using the
550 tracing/smpi/format/ti_one_file flag
551 * smpirun generates the host file if needed (with given host count and platform)
552 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
553 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
554 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
555 --cfg=smpi/coll_selector:(mpich/ompi)
556 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
558 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
559 happening inside a collective SMPI call.
560 * Fix the behavior of complex data types handling.
561 * Make MPI_Wtime another synchronization point to take computations into
563 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
564 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
565 * Add all missing Fortran bindings, SMPI should work with Fortran 90
566 (no privatization of global variables yet)
567 * Preliminary DVFS support (see examples/smpi/energy/ for details)
570 * Verification of liveness properties is now available for SMPI applications
571 (in addition to MSG applications)
572 * Bugged examples using SMPI in examples/smpi/mc/
573 * Add --cfg=model-check/visited option. Allows the verification of infinite
574 programs. Detection of loops in the execution thanks to the system state
575 comparison and reduction of the state space to explore. Can be combined with
576 DPOR for safety properties.
579 * Allow to change SimGrid configuration (see --help) within the code
580 thanks to SD_config() as it can be done in MSG.
581 * Add a new function SD_task_set_amount() upon user request.
584 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
585 * Remove rule based routing (no more PCRE dependency)
586 * Add a limiter_link option to cluster tag, to specify a maximum reachable
587 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
588 * Add a loopback_bw and loopback_lat options to cluster tag.
589 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
590 routing. Now peers are encapsulated in an AS and have their own private
591 router but this is transparent.
594 * Our own implementation of getline is renamed xbt_getline, and gets
595 used even if the OS provide a getline(). This should reduce the
596 configuration complexity by using the same code on all platforms.
597 * New type: xbt_cfg_elm_boolean.
598 * Allow to use yes/no for boolean configuration options in the command line.
599 * Allow to disable SimGrid cleanups at exit from command line option.
600 There are situations where one may want a simulation to end with an exit.
601 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
602 annoying when scripting around the simulator. Adding a
603 --cfg=clean_atexit:no allows to circumvent this issue.
606 * Lots of memory leaks were corrected in this release.
607 There is no known memory leaks anymore, in all of our 600+ tests.
608 * New command line option --version, to get SimGrid version information.
609 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
611 * Supernovae builds are deprecated, and expected to be removed in the next
614 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
616 SimGrid (3.9) stable; urgency=low
618 The Grasgory release: GRAS is really dead now.
620 * Complete overhaul of the internal host structures scheme.
623 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
624 was considered as experimental and badly maintained since then.
625 * Keeping it was thus a trap to our potential users, that could take
626 it instead of MSG or SMPI by mistake despite is pity state.
627 * GRAS seems to have very few users (if any), and no one volunteered
628 to maintain it further. It also induces a lot of XBT code (for
629 portability sake), that must be maintained too.
630 * For all these reasons, we killed GRAS. If someone wants to revive it
631 in the future, don't cry, our git history still remembers of GRAS.
634 * Major overhaul. Merge our documentation again as time proved that
635 splitting it was really not helping our users.
636 * Further improve the developer documentation to help newcomers
637 hacking on SimGrid itself. The user documentation (and in
638 particular, the beginner documentation) is still in a sorry state.
641 * Now works on Windows too!
642 * Much more extensive test suite, from MPICH
645 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
646 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
647 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
648 * Bug fix: let task be scheduled when the last dependency to be solved is
649 a control dependency.
650 * Remove SD_load_environment_script function.
651 Use the C sg_platf function if you want to declare a platform
655 * New function: MSG_process_get_number()
656 * Old function documented: MSG_config()
657 * Remove MSG_load_platform_script function
658 Use the C sg_platf function if you want to declare a platform
662 * Change the default value of the TCP_gamma constant (maximal size of TCP
663 congestion window) to a more realistic 4MiB value. If you notice changes in
664 your simulation results, you can fall back to the previous 20k tiny window
665 by adding --cfg=network/TCP_gamma:20000 on command line.
666 * (Hopefully) fix a bug wrt periodic availability and state traces
667 * Bug fix: use default values at start when first event in availability/state
668 trace is not at time 0.
671 * remove the "new_" part of function name sg_platf_new_trace_connect
672 (resulting in sg_platf_trace_connect), since it does not create
676 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
677 use case where it's really mandatory, and maintaining it was a pain in
679 * New: xbt_fifo_search(), search an item with a user-provided
680 comparison function instead of dumb pointer comparison.
683 * Fix the lua deployment:
684 Use `simgrid.init_application()` before deployment instead of
685 `simgrid.msg_register_application()` after.
688 * Transfer the tracing files into the corresponding modules.
690 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
692 SimGrid (3.8.1) stable; urgency=low
694 The "we are told that some people want to also *install* the simgrid
697 * Add missing file "tesh.1" to the archive.
699 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
701 SimGrid (3.8) stable; urgency=low
703 The Psssshiiiit release: SimGrid jumps into the Cloud.
706 * Add an experimental interface to manipulate VMs. They are mainly
707 process groups with very few intrinsic semantic, but they should
708 allow you to build the semantic you want easily.
709 * New function: MSG_host_set_property_value()
710 * New function: MSG_process_on_exit(). To clean memory in all cases.
711 * Bug fixes that made the host (and link) failures unusable.
712 * Add a way to auto-restart process when the host in which they are
713 executing comes back (ON_FAILURE="RESTART" on deployment file,
714 MSG_process_auto_restart_set).
715 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
716 please stop using the old ones, they are DEPRECATED.
718 * Deprecate functions MSG_global_init() / MSG_global_init_args()
719 Please use MSG_init() instead. (reducing the amount of entry
720 points in the library helps us).
721 * Make it impossible to link against the wrong version of the lib
722 * Deprecate MSG_clean(). No need to call it anymore.
723 * Function MSG_get_host_number() is not deprecated anymore.
726 * Split the doc into a user guide and a reference guide.
727 * Start a developper guide to help people hacking on SimGrid.
730 * Enable tracing by default. This modules rocks you should use it.
731 * Remove option custom_flags. Now use environment variables CFLAGS
733 * Use default cmake things to detect lua instead of home grown ones.
734 * New option "enable_mallocators" to disable mallocators, for debugging
735 purpose ("on" by default).
738 * Bug fixes around the resource failures: don't let the processes
739 survive the host they are running onto.
740 * Add an interface to auto-restart processes when the host in which they are
741 executing comes back.
742 * Ensures that SIMIX_clean is called automatically. It's not part of
743 the public interface anymore (bindings should be updated).
746 * Bug fix for when SD_Simulate is called with a positive value: be careful
747 when comparing doubles. Sometimes they are different for non significant
749 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
750 parallel task whose initial work is distributed among host according
751 to the Amdahl's law. Such tasks are created with a parameter alpha
752 that corresponds to the non-parallelizable part of the computation.
753 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
754 between two sets of workstations assuming a 1D block distribution (each
755 workstation owns a similar share of data) on both sides.
757 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
758 Data redistribution will be automatically scheduled once parent and child
759 are both scheduled. The filling of computation_amount and
760 communication_amount structures is now done seamlessly thanks to the chosen
762 * New function SD_workstation_dump to display various information
763 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
764 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
765 bandwidth on the route onto which the task is scheduled and the amount of
767 To divide the nominal bandwidth by 2, the rate then has to be :
768 rate = bandwidth/(2*amount)
769 * Compute tasks that have failed can now be rescheduled and executed again
770 (from their beginning)
771 * Increasing source code coverage (src/simdag is now covered at 95.8%
775 * Re-implement time-independent trace replay using SMPI (at the
776 smpi_smp_* level) instead of MSG. This should replace
777 examples/msg/actions/actions.c
778 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
779 hindexed and structs)
780 * Implement the exchange of non-contiguous data.
781 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
782 * Correct behavior of smpi/sender_gap and set its default value to 0
783 * Add option to asynchronously send small messages to allow better
784 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
785 specifies the size in bytes under which messages will be asynchronously sent.
786 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
787 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
790 * New C interface to define a platform: XML is now optional.
791 For more info, please check include/simgrid/platf.h.
792 * New interface to define random platforms from the C:
793 For more info, please check include/simgrid/platf_generator.h and
794 examples/msg/masterslave/masterslave_platfgen.c
795 * Export a sg_cmdline dynar containing all the arguments we got from
799 * Two new tracing options for adding comments to trace file so you
800 can track your experiments (see --help-tracing for details).
801 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
802 * Adding the SimGrid version that generated the trace file as a comment.
803 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
804 * Fix to avoid key clashes on Paje links
805 * Other minor fixes related to the Paje specification
808 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
809 and renamed to xbt_str_hash() and xbt_str_hash_ext().
810 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
811 * Greatly improve the robustness of mmalloc to user errors (such as
812 using an area after freeing it, or freeing it twice)
814 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
816 SimGrid-java (3.8.1) stable; urgency=low
818 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
819 Warning: all methods are visible, but only some of them are
820 implemented so far. Check the source (src/jtrace.c)
821 for further information.
822 * New module: org.simgrid.msg.File (SimGrid File management functions)
823 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
824 * Change the meaning of Process.restart: now restart the process from
825 the begining, like MSG_process_restart in C.
826 * Add Process.setAutoRestart: handling of process restart when failed
828 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
829 you to retrieve the properties of the processes/hosts
830 * Deprecate Msg.clean(): you can just forget about it now.
831 * New function Process.getCount(), that only works when compiling
832 with the not yet released version 3.9 of the C library.
834 * New context factory based on Coroutines. It mandates a modified JVM
835 but then, the simulations run about five times faster, and there is
836 no limit to the amount of processes (beside of the available memory).
838 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
840 SimGrid (3.7.1) stable; urgency=low
843 * Restore the prototype of MSG_process_create_with_environment() to
844 the pre-3.7 situation by removing the kill_time argument.
845 * Add a MSG_process_set_kill_time() function instead.
848 * Fix weird behaviors when dealing with parallel tasks.
851 * Simgrid is now built as a dll.
852 * Simgrid-java now works on Windows.
853 * Simgrid-Java is now included into Windows package.
856 * First pre-build package for MacOSX.
859 * Fix compilation when using MSG_USE_DEPRECATED.
860 * Fix some compilation issues on Macs and Windows.
861 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
862 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
864 We discovered that the Lua console is broken, but we are missing the
865 manpower to fix it right now. The problem existed in 3.7 too, so we
866 are not blocking the release for that. Sorry if you depended on this
867 feature, any help would be really welcome.
869 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
872 SimGrid-java (3.7.1) stable; urgency=low
874 The "Java aint got to be bloated and slow" release
877 * Various internal cleanups and performance improvement
878 Simulations are expected to run up to twice faster or so
879 * Make Process.kill(process) an instance method, not a static one
880 * User processes are not java.lang.Thread subclasses.
881 This breaks the compatibility (sorry), but previous API was
882 brain-dead, making it impossible to have non-trivial
883 initializations in the process constructor.
884 * Require a full constructor per Process sub-class.
885 Kinda breaks the compatibility (sorry), but this allows a much more
886 efficient way to launch the processes at simulation startup.
887 * Do not embeed our version of semaphores, java 1.5 can be considered
888 as sufficiently prevalent for us to not dupplicate its features.
892 * Add examples for almost every part of the API
893 We spotted and fixed a lot of bugs in the process
895 * New module: asynchronous communication API
896 * New function: Process.sleep()
897 It takes milliseconds as argument, just as java.lang.Thread.sleep()
898 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
899 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
901 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
903 SimGrid (3.7) stable; urgency=low
905 The "spring cleanups (before the next Big Project kicks in)" release.
908 * We can specify the SMPI latency/bandwidth factor with command line
909 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
910 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
911 You can also use the "config tag" from platform file by adding this line
912 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
913 (see "example/platforms/tag_config.xml" to use "config tag").
914 Note that the command line supersedes the platform file configuration.
915 * Change the correction factors used in LMM model, according to
916 the latest experiments described in INRIA RR-7821.
917 Accuracy should be improved this way.
918 * Use the partial invalidation optimization by default for the
919 network too. Should produce the exact same results, only faster.
920 * Major cleanup in surf to merge models and split some optimization
921 mechanisms from the core of the models. As a result you can now
922 specify which model to use (e.g., --cfg=network/model:LV08
923 --cfg=cpu/model:Cas01) and which optimization mode to use
924 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
925 Incompatible combinations should err at initialization. See
926 --help-models for the list of all models and optimization modes.
927 * The CLM03 workstation model was dropped for simplicity because it
928 used the deprecated CM02 network model. Use default instead.
929 * Rename the TCP_gamma configuration option to network/TCP_gamma
930 * Rename the coordinates configuration option to
931 network/coordinates, and document it
932 * Use now crosstraffic keyword instead of the terribly misleading
933 fullduplex keyword. It is activated by default now in the current
934 default model, use --cfg=network/crosstraffic:0 to turn it off.
935 * Ongoing refactoring the model parsing to make XML files optional
936 See include/simgrid/platf.h for details (still to be completed)
939 * Major overhaul of the documentation. Almost instructive now :/
940 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
941 functions and friends. This interface was considered as
942 deprecated since over 2 years, it's time to inform our users that it is.
943 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
944 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
945 use these (crappy) functions in your code.
946 These functions will be removed soon. Stop using them now.
947 * Deprecate MSG_get_host_{table,number}
948 Implement MSG_hosts_as_dynar() instead.
949 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
950 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
953 * Stabilize the parallel execution mode of user contexts
954 * Introduce configuration variables to control parallel execution:
955 - contexts/synchro: Synchronization mode to use when running
956 contexts in parallel (either futex, posix or busy_wait)
957 - contexts/parallel_threshold: Minimal number of user contexts
958 that must be part of a scheduling round to switch to parallel
959 execution mode (raw contexts only)
960 * Fix bugs that prevented to use suspend/resume along with
961 synchronization structures.
962 * Fix bugs in process termination that lead to invalid memory access
963 in very specific conditions.
966 * Introduce a parallel mode for the models (controlled by surf/nthreads
967 configuration item). In our tests, running the models in parallel
968 never lead to any speedups because they are so fast that the gain
969 of computing each model in parallel does not amortizes the
970 synchronization costs, even when ultra fast futexes are used.
971 This is released anyway because YMMV.
974 * Performance boost by using a swag internally to compute the set of
975 tasks that are finished and should constitute the return value of
979 * Enable it by default now that it is considered rather stable.
982 * Documentation of the tracing functions.
983 * Performance gains when tracing categorized/uncategorized resource
984 utilization by avoiding calls to get route when updating resource
985 variables. LMM constraints are being used instead.
986 * API changed to set task categories. Use MSG_task_set_category instead
987 of TRACE_msg_set_task_category, and SD_task_set_category instead
988 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
989 * Bugfix for graphicator, routes not correctly obtained, memory leaks
990 * Examples for link user variables added (see at examples/msg/tracing/)
991 * Deprecated function TRACE_msg_set_process_category completely removed
992 * Trace header updated according to the latest Paje file format
993 * Tracing network lazy updates, no longer obligate users to use full updates
994 * --cfg=tracing/platform:1 also registers power/bandwidth variables
995 * Experimental: let user code declare/set/push/pop application states for hosts
996 * API changed to allow the manual creation of graph configuration files
997 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1000 * Improve the API of Lua MSG bindings, using the Lua spirit.
1001 * Each simulated process now lives in its own Lua world (globals are
1002 automatically duplicated). It helps writing simulators. Will allow
1003 to run Splay programs within SimGrid in the future.
1004 * Add a Chord example in Lua, equivalent to the MSG one.
1007 * Start the implementation of a solution to express temporal
1008 properties, not only local assertions. This is still an
1009 experimental work in progress, stay clear from it to be safe.
1013 - Add new runtime parameters --help-logs and --help-log-categories
1014 to display information about supported logging parameters and
1016 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1017 don't exists anymore.
1018 * Mallocators: allow value NULL for the reset function.
1020 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1021 create homogeneous dictionaries, where all the elements share the
1022 same free function. Non homogeneous dictionaries will be
1023 deprecated in the next release.
1024 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1025 - Multi-level dictionaries are deprecated.
1027 - new function xbt_dynar_search_or_negative() that is useful when
1028 you have less than 2 million elements in your dynar and don't
1029 want of the extra complexity of catching exceptions when the
1030 element is not found.
1032 - Make xbt_os_thread module (for thread portability) public.
1033 Documentation is still to come, sorry.
1035 - Cleanups and simplifications to make it maintainable again.
1036 - Exotic features (such as memalign and valloc) were removed.
1037 - The metadata were extended and improved so that the
1038 model-checker becomes able to explore and inspect the heaps.
1039 - This may induce a performance drop when enable_model-checking is
1040 ON in cmake (even if it's not used in the simulation), but it is
1041 necessary at this point to get MC working.
1043 Turn model-checking OFF if simulation performance matters to you.
1044 Not enabling it at runtime is not enough, disable it in cmake.
1046 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1049 SimGrid (3.6.2) stable; urgency=low
1051 The "Not coding new stuff allows to polish old things" release.
1054 * New bindings to the NS3 packet level simulator (experimental)
1055 * Use the raw (efficient) execution contexts instead of the sysv
1056 (portable) ones when possible.
1057 * libpcre is now mandatory in any cases since not using it led to
1058 severe performance loss and possibly other issues
1059 * Update the XML platforms:
1060 - G5K: include the latest machine in Nancy
1061 - GridPP and LCG: new platforms
1062 * Documentation was partially updated, at least (more to come)
1064 Bug fixes, cosmetics and small improvements
1065 * Free terminated processes before the end of the simulation to avoid
1066 exhausting the memory of users having very dynamic amount of
1068 * Bug fix and cosmetics about canceling non-running tasks
1069 * Bug fix about the dot loader's issues when using libcgraph
1072 * Create an installer for windows with nsis (amd64 and win32)
1073 - Add an hello world project to illustrate simgrid project creation.
1074 - Embed libpcre into the Simgrid installer to avoid
1075 its compilation burden
1076 * The raw execution contexts should work on Apple now
1077 * Port to Windows 64 bits
1078 - Sysv contexts now have an implementation for this arch
1079 - GRAS communication features now support this arch
1080 * Drop support for borland compiler on windows
1081 - this code was not maintained, and we kinda depend on gcc nowadays
1082 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1083 * Fix portability issue on unstable ubuntu: linker became picky on
1086 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1088 SimGrid (3.6.1) stable; urgency=low
1090 The "Oops, we broke Macs too" release
1093 * Fixed contexts detection so that raw ones are used when possible
1094 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1095 leads to a strange error, with user code segfaulting sometimes when
1096 the generated code is not perfectly aligned (which is not
1097 controllable from the user side, depends on the amount of code)
1100 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1102 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1104 SimGrid-java (3.6) unstable; urgency=low
1107 * Split of every thing from simgrid v3.5 into a separate package.
1109 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1111 SimGrid (3.6) stable; urgency=medium
1113 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1116 * Bindings now constitute their own package, separated from the main one.
1117 Rationale: reduce our maintenance nightmare by reducing the module coupling
1118 They will soon be released on their own on gforge.
1120 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1121 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1123 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1124 * It's not quite deprecated for now because we have no replacement,
1125 but it may soon become the case.
1128 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1129 * Fortran: New user-level cache variable to store the rank of the running
1130 process. This improves performance by an order of magnitude.
1131 * C: New coccinelle script to automatically locate and modify global and
1132 local static variables.
1133 * Improved SMPI network model with a sender-side gap to account for multiple
1137 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1138 only say if a communication is finished, no matter whether it succeeded or
1139 failed. You can call MSG_comm_get_status() to know the status of a finished
1141 * New function MSG_task_dsend() to send a task and detach it. When a
1142 communication is detached, you are never notified of its success or failure
1143 and the memory is released automatically once it is finished. This function
1144 is useful when you don't care about the end nor the success of a
1146 * Change the prototypes of action replay. Sorry for inconvenience,
1147 but this is really more efficient this way (and to adapt your code,
1148 you just have to fix the initialization, that shouldn't be too long)
1149 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1150 ever managed to do anything useful with it.
1151 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1152 takes a dynar of communications. It returns immediately and gives the
1153 index of a finished communication (if any).
1154 * New example: a basic implementation of the Chord P2P algorithm.
1157 * New model for multi-core CPUs. You can now use the core attribute to
1158 precise the number of cores of a host. This is a basic model. Every
1159 process running on the host receives at most the power provided in
1160 the DTD (throughput<=power). Total throughput of process cannot exceed
1162 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1163 router linked by an up-link and a down-link (possibly asymmetrical).
1164 This kind of pattern allows to easily build last-mile model style platforms.
1165 Aggregating such patterns in a rule-based AS is thus the technique of
1166 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1167 * New model for Vivaldi routing. We transformed the Vivaldi network model
1168 into a Vivaldi routing model (based on the rule-based model). This allows to
1169 combine Vivaldi based latencies with last-mile platforms.
1172 * Added a check for NaN of IEEE754 infinite in the double entries of
1174 * Introduce a new context factory "raw", highly inspirated from the
1175 ucontext factory, but using manually crafted functions in assembly to
1176 do the work in an efficient manner.
1177 * Allow to change the used context factory at run time, not only at
1178 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1179 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1180 contexts at runtime (only with raw contexts or ucontexts).
1181 * Completely rewrote this module to allow parallel execution of user
1182 processes. Use --cfg=contexts/nthreads:N to execute user processes
1183 with N parallel threads (the default is 1, meaning no parallelism).
1184 * Allow to decide dynamically between sequential and parallel modes.
1185 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1186 processes in parallel only when their number is greater than or equal to P
1188 * Added a check for NaN of IEEE754 infinite in the double entries of
1192 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1193 won't output the state of processes when interrupted with Ctrl-C
1194 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1195 NULL-terminated array. This may solve backward compatibility issues
1196 due to the change to return type of SD_simulate. See also:
1197 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1198 * Add new macros with variable number of arguments.
1199 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1200 - in xbt/asserts.h: xbt_assert
1201 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1202 - in xbt/ex.h: THROWF and RETHROWF.
1203 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1205 * Change xbt_die() to accept a format string with arguments, just like printf.
1206 * New data structure: xbt_lib_t, like a dict but more general and with better
1210 * New configuration options
1211 Options triva/categorized and triva/uncategorized can be used to generate
1212 graph configuration files for Triva visualization tool.
1213 * Configuration option tracing/platform is renamed to tracing/categorized
1214 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1215 * New TRACE_user_link_variable interface:
1216 User provides the name of the link and the tracing variable to attach to it
1217 * the declaration of tracing categories must be done after the environment
1219 * simpler tracing interface, just one way to declare categories
1220 TRACE_category or TRACE_category_with_color, it is up to you
1221 * links in the trace file are again identified by their names
1222 * trace contains the full platform hierarchy exactly as declared using the ASes
1223 * Options tracing/msg/[task|process]:1 groups the process by hosts
1224 for both cases, tasks and processes must have names that are unique during the simulation
1225 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1226 * The experimental option tracing/msg/volume is deprecated
1227 its functionality may be reincorporated if needed
1229 The tracing generates a trace file with unordered timestamped events,
1230 because of the way the core simulator (surf) works. A script available
1231 at the tools directory (fix-paje-trace.sh) can be used to put the events
1232 in order. We have changed the tracing so it can generate ordered timestamped
1233 events in the final trace, but depending on the simulator (and how much time
1234 is simulated) that can lead to a huge memory utilization. It is deactivated
1235 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1237 Build Infrastructure
1238 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1239 - We are trying hard to keep the API stable, but it may happen that
1240 some things change (we're a research project after all, not a
1241 nuclear plant operating system). If such things should happen, you
1242 could rely on that macro to adapt.
1243 - current value: 30600 for 3.06.00, aka 3.6
1244 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1245 a number that can be compared with SIMGRID_VERSION.
1246 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1247 assertions and verbose/debug logging events are disabled at compile time.
1249 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1251 SimGrid (3.5) stable; urgency=medium
1254 * New feature to any SimGrid-based simulator: Model-Checking
1255 Check SIN#1 for more details.
1258 * New Model SMPI (three-interval linear regression for correction factors)
1259 See RR-7426, available at http://hal.inria.fr/inria-00527150
1260 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1261 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1262 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1263 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1264 * See also src/smpi/README
1269 - Tracing API changes: TRACE_start and TRACE_end should not be called
1270 by user-code. They are automatically called by simulators created
1271 with SimDAG, MSG and SMPI if the toolkit is compiled with
1272 tracing_enabled=ON. Categories declaration and utilization remain the
1273 same for MSG and SimDag.
1274 - A function was added to the tracing API to declare categories with
1276 - TRACE_category_with_color (char *category, char *color)
1277 where color must be in the following format
1278 "%f %f %f", red, green, blue
1279 and red, green, blue are float values in the interval [0, 1]
1280 - User can specify NULL as color parameter, or continue calling
1281 TRACE_category (cat)
1282 On that case, the tracing system will define random colors
1283 - The following command-line options are supported:
1284 --cfg=tracing/filename:msg.trace
1285 --cfg=tracing:1 (activate tracing, needed to use others)
1286 --cfg=tracing/platform:1 (categorized resource use)
1287 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1288 --cfg=tracing/msg/task:1 (task creation)
1289 --cfg=tracing/msg/process:1 (process creation, migration)
1290 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1291 --cfg=tracing/smpi:1 (SMPI interface tracing)
1292 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1293 - examples of examples/msg/tracing updated
1295 - DAXLoader and DOTLoader functions can generate tasks with categories
1296 - A new function to attribute a category to SD tasks:
1297 TRACE_sd_set_task_category (SD_task_t task, char *category)
1298 * Tracing the MPI interface implemented by SMPI
1299 - Collective operations are traced with states
1300 - Point-to-Point operations are traced with states/links
1301 - Tracing activated by a parameter "-trace filename" passed
1302 to smpirun during execution (considering that simgrid
1303 is compiled with tracing enabled)
1304 - To run the simulation with gdb, the simulator
1305 accepts --cfg=tracing/smpi:1 to trace SMPI
1306 - tesh files to check if smpi tracing is ok
1307 - See examples/smpi/NAS/DT-trace
1308 * GTNetS tracing re-worked
1309 - adaptation to the tracing system of GTNets to cope
1310 with modifications regarding the fullduplex mode
1311 - new tesh files to check if gtnets tracing is ok
1314 * Asynchronous communications through the functions:
1315 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1316 * New function: MSG_load_platform_script()
1317 to make possible using a lua script instead of XML files to set up platforms
1318 * New function: MSG_set_function
1319 to associate functions to processes, used when bypassing the parser
1320 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1322 Platforms: Add some more examples in examples/platforms
1323 * Grid'5000: see www.grid5000.fr
1324 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1327 * Change the XML format. This is a very important modification. SimGrid
1328 3.5 introduces a new hierarchical format based on the notion of
1329 Autonomous Systems. Compatibility with old format is ensured
1330 through the perl script provided in the install bin directory
1331 bin/simgrid_update_xml.
1332 It is now possible to build platforms with specific routing
1333 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1334 connect several platforms together. We will try to provide soon
1335 set of realistic platforms exploiting these properties (have a
1336 look at examples/platforms/ for the moment).
1337 * Take the opportunity of the XML format change to be a good XML citizen:
1338 rename link:ctn to link_ctn and similar changes (also dealt with by
1340 * Add a new routing scheme (rule-based) using regular expressions. It
1341 enables to have an extremely low memory footprint when the
1342 underlying routing is simple and can be compactly described. You
1343 need to have libpcre4-dev (perl regular expressions) installed if
1344 you want to use this routing scheme.
1345 * Revive the cluster TAG and allow to easily and efficiently (both in
1346 term of memory and speed) connect clusters together. Have a look
1347 at teshsuite/simdag/platforms/ to see how this can be done. With
1348 this tag, you can create clusters with thousands of tasks at no
1349 cost (have a look at examples/platforms/).
1350 Note: clusters are implemented as ASes, so there is no need for an
1351 enclosing AS tag if you have only one cluster in your platform.
1352 * Add new generic functions in the public interface that allows the user
1353 to call SURF 'create_resource' methods from your code (same
1354 functionality as the XML bypass mechanism but with a much lighter
1356 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1357 that uses a piecewise linear approximation to produce better
1358 results when exchanging small messages.
1359 * Add a new parameter to handle correctly full duplex link and account
1360 for interferences between uplink and downlink communications
1361 (activate with --cfg=fullduplex:1).
1364 * Rename the SD_READY (all dependencies are satisfied and task is
1365 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1366 dependencies are satisfied) state.
1367 This prevents a confusion between the notion of "ready to schedule"
1368 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1369 simulated" (SD_RUNNABLE) used by the simulation kernel.
1370 * Change the way a task is considered as ready. Instead of removing
1371 dependencies when a task is done, a counter is decreased. This way,
1372 it is always possible to reach ancestors thanks to the
1373 SD_taks_get_parents function (even after the end of the simulation.)
1374 * Change the return type of SD_Simulate from (SD_task_t*) into
1375 xbt_dynar_t. This function was in handling a dynar internally and
1376 converted it into a NULL terminated array for historical reasons.
1377 * New function SD_dotload(char*) to load a DAG described in dot
1378 format. This loader and the corresponding examples require the
1379 installation of the graphviz library.
1380 * Fix a bug in the management of tasks of size 0 in the surf network
1381 models. This problem was only visible with SIMDAG and you should
1382 thus disregard results produced with earlier versions if you
1383 relied on this feature (some tasks were blocked because of this).
1384 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1385 even though your applications were doing only point-to-point
1386 communications and sequential computations. Now you can really use any
1387 model you want (of course, if you create real parallel tasks, which are
1388 not implemented in most models beside ptaskL07, this will abort).
1389 * Add an example that schedules a DAX on an heterogeneous platform
1390 using a Min-Min strategy.
1391 * New function SD_workstation_get_current_task() that returns the kind
1392 of task currently running on a workstation in the sequential access
1394 * Raise some warnings when unexecuted tasks remains at the end of the
1395 simulation. This is usually caused by cycles in the DAG.
1398 * New function: SIMIX_process_set_function() called by MSG_set_function
1399 * Change the underlying waiting queue in semaphores so that a process
1400 can wait on several of them simultaneously (as in waitany).
1401 * Fix the way to handle tokens in semaphores so that all access patterns
1402 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1403 * kill the dirty pimple SIMIX_message_sizes_output()
1404 Please use (proper) visualization instead
1407 * New data container: setset (set of sets of elements)
1408 * New module: mmalloc (mapped malloc, allowing to have several
1409 independent segments of malloc)
1410 * New function: xbt_dict_cursor_set_data()
1411 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1412 * New function: xbt_dynar_is_empty()
1413 * New function: xbt_fifo_get_last_item()
1414 * Fix xbt_dynar_shrink(): use the right element size.
1415 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1416 uninitialized areas during expand.
1417 * Fix semaphores: previous implementation was severely broken.
1418 * Use library init/fini functions for our initialization.
1419 - you can use logs and other feature as soon as you want in your
1420 code (even before the xbt_init / MSG_init)
1421 - xbt_exit is now a no-op and produce a warning when used.
1424 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1425 benefit from the latest improvement to the simulation kernel.
1426 * Kill measurement sockets for now. If you rely on them, sorry. This
1427 release is not for you. This feature will be reintroduced in the
1428 future, but we cannot delay the release any further.
1429 * New function: gras_msgtype_get_name().
1430 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1431 * Fix (at last) the pmm example: it should not randomly fail anymore.
1433 Build chain: bug fixes and overall polishing
1434 * Cmake is now stable enough. Hence, we killed the autotools.
1435 * Port to windows ( TM :)
1436 * Fix the 'make install' target.
1437 No need to use 'make install-simgrid' anymore
1438 * Introduce a 'make dist' target compiling a *source* archive
1439 'make package' compiles a binary archive
1440 * Compile java files only on need
1441 * Add --cd and --setenv command line options to tesh
1442 * Out of source builds are not fully supported yet, but we are close
1443 * Enable supernovae and optimization flags by default for our users
1446 * Add layer to set up environment directly from lua, without XML.
1447 * The effect of gras_stub_generator can be achieved through
1448 lua too (check examples/gras/console/ping_generator.lua)
1450 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1452 SimGrid (3.4.1) stable; urgency=low
1454 The "Polishing easter eggs is probably a good idea" release.
1455 This is a bug fixes release only.
1459 * Fix a bug preventing the tasks from begin garbage collected.
1462 * Fix a bug occuring when a host involved in a communication fails.
1463 This was not detected properly by the other peer involved in the
1464 communication. Now, it's reported as a network error.
1467 * Warn the user about loop dependencies in data flow of DAX files
1468 * Obey the control-flow dependencies of DAX files
1471 * Add option "enable_smpi" allowing to not compile SMPI.
1472 Probably useful for the (Mac) users experiencing a build error here
1473 * Improve the detection of lua5.1 and ruby1.8
1475 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1477 SimGrid (3.4) stable; urgency=low
1479 The "Easter in Cargese" release. Also known as (major changes):
1481 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1482 ~> bindings were greatly improved
1483 ~> new tracing infrastructure for better visualization introduced
1485 * the "Welcome to configury modernity" release.
1486 ~> we switched from autotools to cmake, and improved our cdash
1489 A more detailled list of changes follow (full detail in svn log).
1491 Java Bindings: Various Cleanups
1492 * (install java-gcj-compat-dev on debian-like to use them)
1493 * Remove put/get: no need to export deprecated interface in Java
1494 Use send/receive instead.
1495 * Cleanup the examples and add a README per directory
1496 * Remove example autoDestination (that's the only way to go now)
1497 * Remove example explicitDestination (was a plain copy of basic)
1498 * Make JniException a runtime exception, so that there is no need to
1499 declare the fact that you may encounter such a beast. I guess that
1500 nobody will ever want to survive such error.
1501 * Create specific errors for each MSG case of failure:
1502 host failure, transfer failure, timeout, task cancelled
1503 * Cleanup the exceptions that may get thrown by each function
1504 * Other internal cleanups in Java bindings. Performance still bad :/
1505 Ruby and Lua Bindings: create them
1506 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1507 * That's new and great, you should try them out.
1508 Same functionalities than Java bindings, only even less polished
1510 * Kill the useless "rate" argument of SD_task_get_execution_time()
1511 Everyone used to provide -1 as a value, it was not used, and the
1512 semantic of a possible use wasn't even clear.
1513 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1514 either as comm costs or compute costs to mean that there is no
1515 such thing for that specific task.
1516 * Add a SD_task_set_name() function
1517 * Fix SD_task_unschedule() on typed tasks
1518 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1519 * In DAX loader, accept useless 'level' attributes to <job> since
1520 LIGO DAGs have them (seem to be to ease graphical representation).
1522 * Add an example masterslave_mailbox.c using send/receive and not
1523 the deprecated put/get interface.
1524 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1525 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1527 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1528 (declare MSG_USE_DEPRECATED to still have the old name)
1529 * Add a MSG_task_set_data() function
1530 * About trace replay (see examples/msg/actions):
1532 - Allow to work with splitted trace files for each process
1533 Give the specific trace file as argument of each process,
1534 and call MSG_action_trace_run(NULL)
1535 You can still have one merged file for all processes.
1536 - Fix implementation of collective operations
1537 * Allow task_execute() on 0-sized tasks (closes #10063)
1539 * This is the first release of SimGrid where SMPI is not considered
1540 beta anymore (even if some corners should still be improved)
1541 * Port over the new SIMIX_network submodule (internal refactoring)
1542 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1543 * Implement more missing elements of the standard:
1545 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1547 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1548 MPI_Scan() MPI_Get_processor_name()
1549 - Added implementation of missing case for Alltoall (warning: it's
1550 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1551 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1552 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1553 SMPI_MPI_Allgatherv()
1554 * Bug fixes include:
1555 - MPI_Waitsome() was broken
1556 - Allow relative includes in smpicc
1557 - Command line cfg argument 'reference_speed' was ignored...
1558 - Some functions did not properly lead to auto-benching of user code
1559 - smpicc passes -O2 by default (just like openmpi one)
1561 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1562 * Bug fixes about timeouts during communications
1563 * add SIMIX_message_sizes_output() as a pimple to write to file the
1564 amount of messages per size. Use gnuplot to get histogram.
1565 Pimple because that's the only user-visible function of simix,
1566 defined directly in xbt.h (irk, sorry)
1568 - Add a SIMIX_sem_get_capacity() function
1569 - Fix interactions with processe resume/suspende
1570 - release_forever() was stupidly broken
1571 - Fix SIMIX_display_process_status() for processes in a semaphore
1572 - Make SIMIX_sem_block_onto() user-visible
1573 * Refactoring context stuff:
1574 - Use pseudo-OOP for better modularity
1575 - reimplement SIMIX_process_kill() without process_schedule() so
1576 that the latter can take as invariant that it is called from
1578 - Merge context_start into context_new for sake of simplicity
1580 * Add a Vivaldi network model, coded live during SUD'10 ;)
1581 * Rename configuration variables to start a hierarchy:
1582 o cpu_model -> cpu/model
1583 o network_model -> network/model
1584 o workstation_model -> workstation/model
1585 * New configuration variables:
1586 o network/bandwidth_factor: correction to bandwith
1587 o network/latency_factor: correction to latency
1588 o netwotk/weight_S: correction to the weight of competing streams
1589 * Add a long description to the models, that users can see with such
1590 argument on the command line: --cfg=cpu/model:help
1591 * --help-models display the long description of all known models
1593 * config: add the ability to set a default value after registration
1594 Does not override any previously set value (e.g. from cmd line)
1595 * dict: allow to have integer key and data.
1596 When so, you need to use the following functions
1597 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1598 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1599 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1600 In contrary to regular dicts, the key is not malloced before copy.
1601 Mixing scalar and regular elements in the same dict is not tested
1603 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1604 Tracing for Visualization:
1605 * SimGrid is now instrumented in order to generate a trace file for
1606 visualization analysis: to use it, need to compile SimGrid with the
1607 "tracing" option enabled, and instrument the program using SimGrid with
1608 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1609 (among other functions).
1610 * The instrumentation only traces the platform utilization for now
1611 * Documentation to use the tracing functions and how to analyze the
1612 traces with the Triva tool is written.
1613 * More information about: SimGrid FAQ (in the section Tracing Simulations
1616 * We moved to cmake as default build system. Autotools support will
1617 be dropped soon. Check the FAQ for more info about how to use it.
1618 * Greatly improved our cdash/ctest interactions
1619 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1620 * Added memory checking tests with valgrind; lot of memleak fixing.
1621 This may be the first release of simgrid with so few memory issues
1622 * Added code coverage tests.
1623 Our coverage is still improvable, but at least we see it on cdash.
1625 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1627 SimGrid (3.3.4) stable; urgency=low
1629 The "Desktop Grid needs love too" release (also called Xmas release).
1631 Models improvements:
1632 * Major speedup in the maxmin system solving by using lazy evaluation
1633 Instead of solving completely the maxmin system at each iteration,
1634 only invalidate (and recompute) the modified parts.
1635 This new feature is enabled in default models but you can try to
1636 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1637 * Cas01 IMproved as default CPU model
1638 This CPU model is the same Cas01 model, but it uses the
1639 maxmin-selective-update flag and a heap structure to manage
1640 actions on SURF kernel.
1641 It reduces the complexity to find the next action to finish and,
1642 consequently, it's faster than the old Cas01.
1643 This is the new default CPU model (Cas01).
1644 * Rename the old Cas01 model to Cas01_fullupdate
1645 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1646 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1647 * CpuTI (CPU Trace Integration)
1648 A new CPU model whose objective is simulate faster when using
1649 availability trace files.
1650 Instead of using a full featured, over engineered maxmin system for
1651 CPU modeling, this model does the pre-integration of traces files
1652 to calculate the amount of CPU power available, and so, executes
1653 faster than the old CPU models.
1654 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1655 * Use LV08 as default network model since it gives better accuracy
1656 for small messages and shouldn't change things for big ones.
1657 Use --cfg=network_model:CM02 to get the previous behavior.
1660 ******************************************
1661 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1662 ******************************************
1663 * The new CPU model may changes simulations!
1664 The point is that events occurring at the exact same timestamp
1665 are not scheduled in the same order with the old and new
1666 version. This may be enough to completely change the execution
1667 of simulations in some cases.
1668 * The new network model will change simulations!
1669 This new model is more realistic than the previous one, so you
1670 should consider redoing your old experiments with this model.
1671 Sorry for the inconvenience.
1674 * Introduce the supernovae compilation mode
1675 When compiled that way, the whole SimGrid (or almost) is put in a
1676 single compilation unit and compiled in one shoot.
1677 This is to help gcc which has difficulties to inline stuff from one
1679 The speedup seem to be above 15%, althrough more tests are needed on
1680 amd64 to confirm that gain.
1683 * Port of MSG's mailbox on top of SIMIX network
1684 The put/get mechanism was greatly simplified on the way.
1687 * New SIMIX network module. Provides:
1688 - Mailbox: rendez-vous mechanism to find with who you want to speak
1689 - Synchronous send/recv: easier and hopefully faster since the
1690 logic is handled in the maestro process directly now
1691 - Asynchronous send/recv: you dreamt of it? It's here now
1692 Too bad that nobody cared enough to propagate the change to MSG.
1693 * Add semaphores as SIMIX synchronization mechanism.
1696 * new function SD_daxload(char*) to load a DAX file
1697 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1698 * Introduce typed tasks. Specify its kind and cost at creation.
1699 At scheduling, just give where it should be placed, and the cost
1700 for each involved resource is automatically computed.
1701 Existing constructors so far (more to come of course):
1702 - SD_task_create_comm_e2e() for end-to-end communication
1703 - SD_task_create_comp_seq() for sequential computation
1704 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1705 * new function SD_task_dump() for debuging display
1706 * new function SD_task_dotty(task,FILE*) writing to file the info
1707 about the task in dotty format
1708 * SD_task_dependency_exists() can now cope with having one of its
1709 arguments NULL. If so, it tests whether the other argument has any
1711 * Add getters on list of preceding/following tasks:
1712 SD_task_get_parents(task) and SD_task_get_children(task)
1713 * Add getters on amount of workstations and list:
1714 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1715 * Add getter on task kind: SD_task_get_kind(task)
1716 * Update the start_time and finish_time of tasks on completion/failure
1717 * Bugfix: Remove task from state swags when destroyed
1720 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1723 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1724 dynar with one separate thread per value of the dynar.
1725 * Change the prototype of xbt_thread_create(), sorry.
1726 Added a boolean parameter indicating whether we want to join this
1727 thread (used in SG only for now)
1728 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1731 * GTNetS wrappers should now be usable again (and betterly tested too)
1732 * Fix a major regression from 3.2 where the timeout provided to
1733 MSG_task_put_with_timeout() was used as absolute time before which
1734 the comm should be done.
1735 * Start to fix the <cluster> tag.
1736 - Internal links should be good now (beside of the loopback, which
1737 use the private link instead)
1738 - paths to the external world is still rather broken
1739 - the <route:multi> tag is just broken. Actually that's brain-dead.
1740 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1741 to make it less stupid
1742 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1743 * Fix a source-level compatibility glitch from 3.2: after defining
1744 MSG_USE_DEPRECATED, you can use the old name
1745 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1746 * Allow to compile from the SVN with automake 1.11
1747 * Fix some problems when using the "start_time" tag in deployment XMLs.
1748 * Fix #8569: XBT/synchro.h has redundant declarations
1749 * Fix #8563: MSG return values and exceptions
1750 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1751 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1752 Thanks to Robson Peixoto.
1753 * Fix a vicious bug in dictionaries inducing that some elements were
1754 not freed on xbt_dict_free()
1756 Portability report of this version:
1757 * Main portability targets:
1758 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1759 on (amd64/i386/ia64)
1760 - mac leopard on i386
1761 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1762 but nothing critical.
1763 * Other platforms: windows, AIX and others were not tested for this release
1765 Timing report of this version:
1766 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1767 scenario parameters). From 8h to a few seconds in desktop grid settings.
1768 * Supernovae brings about 25% speedup on i386.
1770 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1772 SimGrid (3.3.3) stable; urgency=low
1774 The "Need for Speed" release.
1776 The timings done to validate the 3.3.2 were faulty.
1777 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1779 The problem was a conversion from a manually handled vector to
1780 xbt_dynar_t on the critical path.
1781 xbt_dynar_foreach calls functions, inducing stack management crap.
1783 We inlined these functions and xbt_dynar_foreach is now breath taking.
1784 We also inlined xbt_swag_belong on the way.
1786 Here are some approximate speedup measurements (on master/slaves
1787 simulations lasting between 10s and 20s each):
1788 3.3.1 -> 3.3.2: about same performance
1789 3.3.2 -> 3.3.3: 40% speedup
1790 3.3.1 -> 3.3.3: 40% speedup
1791 3.3.1 with inline patch -> 3.3.3: 30% speedup
1793 Our reading is that the refactoring which occurred in 3.3.2 made us
1794 suffer much more from the xbt_dynar_foreach low performance, but
1795 once we solved this, this refactoring proved to be very performance
1796 effective. From the 40% speedup, somehow, 10% are due to the
1797 inlining and 30% to the refactoring.
1799 That's a pitty that gcc cannot inline functions placed in other files
1800 alone. We have to choose between:
1801 - break the encapsulation (by putting private data structures and
1802 accessors in headers files to help gcc)
1803 - live with low performance
1804 - switch to a decent compiler such as icc (not quite possible).
1806 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1808 SimGrid (3.3.2) stable; urgency=low
1810 The "Simplicity does not preceed complexity, but follows it" release.
1812 The main contributors of this release were (lexical order):
1813 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1816 * Extract the routing logic into its own object.
1817 (was dupplicated in network.c and workstation_LV07.c;
1818 Allows to implement other ways of storing that info)
1819 => kill now useless network_card concept
1820 - Use dynar to represent routes (instead of void** + int*)
1821 - kill link_set (use surf_network_model->resource_set instead)
1822 - Add a command-line option to choose the routing schema to use
1823 - Add three new models:
1824 * Floyd (shortest path computed at initialization)
1825 * Dijikstra (shortest path recomputed all the time)
1826 * Cached Dijikstra (shortest path computed on need)
1827 All these models where contributed by Silas De Munck, and are
1828 described in his ICCS09 paper.
1830 * Simplify model declaration
1831 (less redirections, less function to write when defining a model)
1832 - Factorize stuff between models:
1835 surf_model_resource_set(model)
1836 surf_model_resource_by_name(model, name)
1837 - Unify the types of models in s_surf_model_t (using an union)
1838 - Embeed fields of common_public directly into s_surf_model_t
1839 - Rename model methods:
1840 action_free ~> action_unref
1841 action_change_state ~> action_state_set
1842 action_get_state ~> action_state_get
1843 - Change model methods into functions :
1844 (model)->common_public->action_use ~> surf_action_ref
1846 * Implement a generic resource; use it as ancestor to specific ones
1847 (allows to kill duplicated code in models)
1848 Drawback: timer command don't need no name nor properties;
1849 workstation_CLM03 don't need no properties
1850 (but I guess we can live with those few bytes wasted)
1852 * Improve the action object model
1853 - implement a constructor avoiding dupplicated code about field
1854 initialization in generic_action part.
1856 * Kill the SDP model: it has an external dependency, is deprecated
1857 in flavor of modern lmm models, and didn't compile since a while
1860 * Relocation of the context module from XBT to SIMIX.
1861 (the context were decoupled from the simix processes, duplicating a lot of code)
1862 => a lot of code was factorized
1863 - less overhead is introduced during scheduling
1864 - simpler API for the context factory
1865 - the logic for process creation,destruction and manipulation was simplified
1866 * Simplification of the s_smx_process_t data structure.
1867 => accesing the simix level data associated to a process is faster now,
1868 and the code is a lot more readable.
1871 * Implement some more MPI primitives:
1872 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1873 -implementation: Bcast: flat or 2-ary tree (default),
1874 Barrier: 4-ary tree,
1876 Allreduce: Reduce then Bcast
1877 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1878 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1879 Alltoallv: flat tree, like ompi
1881 * Add support for optimized collectives (Bcast is now binomial by default)
1882 * Port smpirun and smpicc to OS X
1885 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1886 Shout out if you used it.
1889 * Display the list of still queued messages in SG mode when existing
1893 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1894 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1895 * Add xbt_str_from_file(FILE*)
1896 * Add xbt_dict_get_key achieving a linear reverse search
1897 * Remove the context module
1899 Portability report of this version:
1900 * Main portability targets:
1901 - Linux(debian)/x86/context
1902 - Linux(debian)/x86/pthreads
1903 - Linux(debian)/amd64/context
1904 - Linux(debian)/amd64/pthreads
1905 On these, we still have the eratic breakages of gras/pmm and
1906 amok/saturate_sg reported in previous version. We still think
1907 that the tests are the cause of the fault, not the tested code.
1909 - Mac OSX Leopard/x86/context
1910 Still false negative in tesh autotesting.
1911 Smpi still fails, but this time because readlink does not accept -f
1912 Everything seems to work properly beside of that.
1915 - AIX version 5.3 (only tested contexts this time)
1916 Smpi still fails there because mktemp is not installed.
1917 Everything seems to work properly beside of that.
1919 I managed to compile it for the first time, but several breakages.
1920 Won't delay the release for this exotic platform.
1922 * Windows: it's still lagging behind. If you want to help, please
1925 Timing report of this version:
1926 This version seem to be more than 5% faster than 3.3.1 (on linux
1927 64bits with contextes). The gain is less than expected, we are
1928 investigating this for next release.
1930 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1932 SimGrid (3.3.1) stable; urgency=low
1935 * Implement a --cfg-help to show existing configuration variables
1936 * Build chain do not require doxygen in maintainer mode
1939 * fix a bug on struct sizeof computation, which prevented the
1940 exchange of arrays of structs in some conditions
1941 - added a regression test about this in datadesc_usage
1942 * Allow the exchange of 0-long dynamic vectors.
1943 - for that, use -1 as indicator of dynamic size instead of 0
1944 - This implied to change any size from unsigned long to long,
1945 reducing a bit communication abilities, but I guess that with
1946 64bits being quite common, this is more than enough.
1947 - This also induce a protocol change, thus bumping network protocol
1948 version from 0 to 1 (if we have external users, we have to get
1949 clean on that point too ;)
1950 - added two regression tests about this in datadesc_usage
1951 * Be more verbose when propagating local exceptions
1952 This helps debugging.
1953 * Display the status of simulated processes when receiving SIGINT in
1957 * Allow to control the simulation from a trace file.
1958 New functions MSG_action_register() and MSG_action_trace_run()
1959 The first one allows to associate a function execution to each
1960 kind of action while the second one parses a trace file and
1961 triggers the corresponding actions within the system.
1962 For now, only a toy example is provided in examples/msg/actions
1963 * Add an exemple of process migration in examples/msg/migration
1964 * Fix a bug in task exchange which broke MSG_task_get_sender()
1965 Add a teshsuite regression test for that.
1966 [Bug: if MSG_task_get_sender() is called after sender exit,
1968 * Fix a bug which prevented suspend/resume to work properly
1969 * Display the status of simulated processes when receiving SIGINT
1970 This fixes a regression of v3.3. due to the introduction of SIMIX
1971 * Bug fixing in failure management:
1972 - trace could not start by a failure at time 0
1973 - failure during communications were not working
1976 * Add SIMIX_process_set_name() to change the name of the current
1977 process in the log messages.
1978 * Store smx_hosts in a dict since we only retrieve them by name
1979 * Move the configuration infrastructure to surf
1982 * Move the configuration infrastructure to surf
1985 * Massive internal cleanups:
1986 - Store internal structures on processes instead of hosts (allows
1987 to have more than one process per host, in addition of being more
1989 - Cleanup the initialization/finalization process
1990 - Kill a whole bunch of unneeded synchronization:
1991 processes run in exclusive manner within the simulator
1992 - Move queues from global tables to process data fields
1994 - now accept -platform and -hostfile arguments
1995 - Pass the right rank value to processes according to the hostfile
1996 * Compile the examples by default, and use them as regression tests
1997 * Implement MPI_Wtime()
1998 * Change the reference speed to a command line option
2001 * TCP_gamma can now be specified as command line option using
2002 --cfg=TCP_gamma:10000000.0
2003 * Change the --surf-path cmd line option into --cfg=path:
2006 * Also include strbuff from xbt.h public header
2007 * xbt_ex_display(): do not free the exception after displaying
2008 This allows to do more with the given exception afterward.
2009 Users should call xbt_ex_free() themselves.
2013 Portability report of this version:
2014 * Main portability targets:
2015 - Linux(debian)/x86/context
2016 - Linux(debian)/x86/pthreads
2017 - Linux(debian)/amd64/context
2018 - Linux(debian)/amd64/pthreads
2019 These targets fail about 1/10 of times on gras/pmm, but we believe
2020 that this is because of the test, not because of simgrid.
2021 amok/saturate_sg fails even more rarely, and the test may not be
2024 - Mac OSX Leopard/x86/context
2025 The test suite still spits tons of errors because some obscure
2026 force prevents us from removing the temporary directories
2027 arguing that they still contain some metadata I've never heard of.
2028 Smpi fails because seq is not installed.
2029 Everything seems to work properly beside of that.
2032 - AIX version 5.3 (both contexts and pthread)
2033 Smpi still fails there because mktemp is not installed.
2034 XML inclusions seems rosty on AIX.
2036 * Windows: it's still lagging behind. If you want to help, please
2039 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2041 SimGrid (3.3) stable; urgency=high
2045 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2046 [Malek Cherier & Mt]
2048 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2051 Until now, GRAS were using MSG as an interface to SURF. It was
2052 quite difficult because both interface have several differences
2053 (MSG channels vs GRAS sockets were the most notable point).
2055 This also opens the gate to SMPI (which should occur soon) and speed
2056 up simulations by to 40% (even if it were not the main goal).
2058 **************************************
2059 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2060 **************************************
2061 The point is that events occuring at the exact same timestamp are
2062 not scheduled in the same order with the old and new version. This
2063 may be enough to completely change the execution of simulations in
2064 some cases. Sorry for the inconvenience.
2066 * Cleanup and upgrade the XML format to push further scalability
2067 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2069 * Improve the testing infrastructure with tesh. Now a very large part of
2070 the code is tested not only by being run but also by checking that the
2071 output match an expected output [Mt].
2073 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2074 is really less memory-demanding, which should allow you to use
2075 larger files in SimGrid [AL].
2077 * Inform valgrind about our contextes, so that it becomes usable
2078 with the default (and more effecient) version of SimGrid
2079 [contributed by Sékou Diakite, many thanks]
2082 * Introduce a listener thread in charge of receiving incoming
2083 messages from the network. It allows to overlap communication and
2084 computation but most notably, it removes some stupid deadlocks due
2085 to the fact that so far, a process could not send and receive at
2086 the same time. This made most non trivial communication schema
2088 * Convert the PIDs from long int to int to match the MSG ones (and
2089 linux ones too) [Mt]
2090 * New function: gras_agent_spawn() to launch a new process on
2091 current host. Only working in simulation for now. [Mt]
2092 * New function: gras_os_hostport() returning a constant form (ie,
2093 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2096 * Make the backtrace of exceptions more human readable [Mt]
2097 * New module: xbt/str [Mt]
2098 a ton of string utility functions (split, join, printf to a newly
2099 allocated buffer, trim, etc)
2100 * New module: xbt/hash [Mt]
2101 SHA1 hashing algorithm (more to come if needed)
2102 * New module: xbt/synchro [Mt]
2103 synchronization tools (mutex and conditions) working the same way
2104 in simulation and in real life (mainly useful for GRAS, but not
2106 * New module: xbt/queue [Mt]
2107 classical producer/consumer synchronization scheme
2108 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2109 (using the classical functions will get serialized) [Mt]
2110 * Make dictionary internal table dynamic. No need to specify its size
2111 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2113 * Make sure the log channels are organized as a tree under windows
2114 (because of ANSI C compatibility issue, any channel were child of
2115 root directly) [Mt].
2118 * Cleaned many thing in surf and fixed a few bugs [AL].
2119 * Add a nice command line configuration mechanism to compose models [AL].
2120 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2121 the previous one (KCCFLN05). It relies on something that looks like
2122 a max-min sharing mechanism but cannot be written as such. A new solver
2123 was thus designed [AL].
2124 * Add a new solver to lmm. Based on Lagrange optimization and
2125 gradient-based descent, it enables to efficiently maximise systems s.a
2127 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2129 This solver enables to propose two new network models for TCP Reno and
2130 TCP Vegas based on Low's work. These models still need to be fully
2131 tested though [Pedro Velho].
2134 * Bug fix in SD_simulate. Now the time bound given as argument is
2136 * Use the new parallel task model (ptask_L07) as default.
2137 * Use the SURF command line configuration mechanism.
2138 * 0-size tasks (for synchronization) should now work.
2140 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2142 SimGrid (3.2) stable; urgency=high
2146 We still experience issues on this platform, but we believe that at
2147 least MSG is usable.
2149 GRAS API BREAKAGE (for simplification purpose, sorry):
2150 * the gras_msgtype_by_name is not used anymore. Instead of
2151 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2152 you can write (and must)
2153 gras_msg_send(toserver, "request", &request);
2154 - If you still want to pass a gras_msgtype_t to the function (to cache
2155 the type and avoid the lookup time), use the gras_msg_send_() variant.
2156 - Impacted functions:
2157 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2158 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2159 * The callbacks are now expected to return 0 when everything went well
2160 (just like the main() function)
2162 GRAS new features and improvements:
2163 * New module mechanism where user code can use per process globals [Mt]
2164 This is similar to gras_userdata_*() functions, but for libraries. It
2165 factorize some code developped over and over in the examples and AMOK.
2166 It has still to be documented and used (only amok/peermanagement is
2168 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2169 loss when they were small enough to fit into the buffer and sent quickly
2170 enough so that they can all get received in one shoot.
2171 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2172 if not found. Use the *_or_null() variant for the old semantic.
2173 * In gras_msg_handle, do not discard messages without callback.
2174 They are probably messages to be explicitly awaited later (ie, proofs of
2175 mis-synchronization in userland since they are sent before being awaited)
2176 No big deal usually.
2177 * gras_socket_meas_send/recv: semantic changed!
2178 The numerical arguments used to be (1) the total amount of data to send
2179 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2180 messages. This was need for the fool willing to send more than MAXINT
2181 bytes on quite fat pipes.
2184 * Do really rename the hostmanagement module to peermanagement. [Mt]
2185 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2186 but this is rather new and this was documented in the module
2187 documentation (poor excuses, I admit)
2188 * Bandwidth measurement semantic changed! This follows the changes to
2189 gras_socket_meas_send/recv explained above.
2192 * A sequential mode has been added to the workstations. When a workstation
2193 is in sequential mode, it can execute only one task, and the other tasks
2194 are waiting in a FIFO. [Christophe Thiery]
2197 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2198 model for SIMDAG. [Christophe Thiery]
2199 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2200 equal to 0, which caused some bad behaviors in
2201 saturated_constraint_set_update. I now use a threshold mechanism like in
2205 * When running manually src/testall, you select specific units [Mt]
2206 testall is the result of our cunit mechanism, and should replace all
2207 the scripty thingy around since bash don't run easily on billware.
2209 * A mallocator system has been added. [Christophe Thiery]
2210 Mallocators allow you to recycle your unused objects instead of freeing them
2211 and allocating new ones.
2213 Documentation update:
2214 * FAQ reworking + New FAQs:
2215 - "Valgrind spits tons of errors!" [Mt]
2216 - "How to repport bugs" [Mt]
2217 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2218 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2219 - Communication time measurement within MSG [AL]
2220 - I experience weird communication times when I change the latency [AL]
2221 * GRAS tutorial [Mt]
2223 - an introduction to the framework and to the used communication model
2224 - an initiatic tour introducing the most proheminent features:
2226 . Lesson 0: Installing GRAS
2227 . Lesson 1: Setting up your own project
2228 o Part 2: Message passing
2229 . Lesson 2: Exchanging simple messages
2230 . Lesson 3: Passing arguments to the processes (in SG)
2231 . Lesson 4: Attaching callbacks to messages
2232 . Lesson 5: Using globals in processes
2233 . Lesson 6: Logging information properly
2234 . Lesson 7: Using internal timers
2235 . Lesson 8: Handling errors through exceptions
2236 . Lesson 9: Exchanging simple data
2237 . Lesson 10: Remote Procedure Calling (RPC)
2238 . Lesson 11: Explicitely waiting for messages
2239 . Recapping of message passing features in GRAS
2240 - A HOWTO section containing:
2241 o HOWTO design a GRAS application
2242 More are due, of course. They will come latter. In the meanwhile, you can
2243 check the examples which are still here.
2245 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2247 SimGrid (3.1) stable; urgency=high
2251 There was a stack corruption somewhere, visible only when optimizing
2252 with these versions. [Vince]
2255 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2256 module that enabled to play with Directed Acyclic Graphs. It is built
2257 directly on top of SURF and provides an API rather close to the old
2258 SG. Some old codes using SG are currently under rewrite to check that
2259 all needful functions are provided. [Christophe Thiery]
2262 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2263 extension of the classical CLM03 model that gracefully handles
2264 failures. This is now the default model for MSG and GRAS. It doesn't
2265 handle parallel tasks yet though. [AL]
2266 * Bug fix: Weights were not correctly set in the network part.
2267 WARNING: This may have resulted in incorrect results with simulations
2268 where there are more than one flow on a given link. [AL]
2271 * After a (long ?) discussion on simgrid-devel, we have decided that the
2272 convention we had on units was stupid. That is why it has been decided
2273 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2274 WARNING : This means that all previous platform files will not work as
2275 such with this version! A warning is issued to ask users to update
2277 A conversion script can be found in the contrib module of the CVS, under
2278 the name contrib/platform_generation/surfxml_update.pl [MQ]
2281 * Bug fix: Processes were started in reverse order, wrt deployment file.
2282 WARNING: if your code relies on this bug, please fix it. [AL]
2283 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2284 being executed on two different locations. [AL]
2285 * Bug fix: Failures are now better supported thanks to Derrick's tests
2286 (there was many failure situations I hadn't thought of and that weren't
2287 correctly handled). [AL]
2288 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2292 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2293 Exception occurring on server-side are propagated back to client (!).
2295 API CHANGE: the callback changed their prototype. Change:
2296 int my_handler(gras_socket_t expeditor, void *payload_data) {
2298 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2299 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2301 * New! function: gras_msg_handleall to deal with all messages arriving
2302 within a given period.
2303 * New! function: gras_socket_server_range to get a server socket in a
2304 range of port numbers (ease to avoid port number conflicts) [MQ]
2305 * New! gras processes display their backtrace when they get a SIGUSR1
2306 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2307 Sweet to debug RL processes [MQ]
2311 - Do not force experiment sizes to be expressed in kb, or it becomes
2312 impossible to measure the latency this way (needs one byte-long tests)
2313 WARNING: this changes the amok_bw_* function semantic. [MQ]
2314 - Implements the link saturation stuff. [MQ]
2315 * Peer management module:
2316 New! module factorizing code that we wrote over and over [MQ].
2319 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2320 - Test units are placed directly into the library code, they get extracted
2321 automatically and placed into the src/testall binary.
2322 - Convert most of the XBT tests to this system.
2323 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2324 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2325 live in the same C code. This required to change a little bit the API
2326 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2327 * New module: sparse graph structure with basic algorithms (this is work
2328 in progress and the API is not considered to be frozen yet). [AL]
2329 * Display more information on backtraces: source line & function names are
2330 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2331 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2332 * Reworked a little bit some #include statements to load only required
2333 headers. Some user code that relied on SimGrid to include stdlib or
2334 stdio may need to include it by themselves. [AL]
2335 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2337 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2338 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2339 but I guess that almost nobody used those parts. [MQ]
2341 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2343 SimGrid (3.0.1) stable; urgency=low
2346 * Unfortunately, I had missed 5 misnamed functions:
2347 xbt_fifo_item_t xbt_fifo_newitem(void);
2348 void xbt_fifo_freeitem(xbt_fifo_item_t);
2349 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2350 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2351 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2352 They're now deprecated. Please use their new versions:
2353 xbt_fifo_item_t xbt_fifo_new_item(void);
2354 void xbt_fifo_free_item(xbt_fifo_item_t);
2355 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2356 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2357 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2359 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2360 * Documentation: xbt_log module unmercifully reworked [MQ]
2361 * Bugfix: there was a problem with the ending of contexts with
2362 the pthread backend. It caused some weird deadlock or behavior
2363 depending on the pthread implementation. [AL]
2364 * Bugfix: get the exceptions raised in the simulator repport where
2365 and why they come from when they are not catched in time [AL, MQ]
2368 * Bugfix: Do repport the error when two non-connected hosts try to
2369 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2372 * Add additionnal checkings on communications. Assert that two
2373 communicating hosts are connected by a set of links... [AL]
2376 * Add additionnal checkings on channel values in communication [AL]
2377 * New: MSG_task_get_source to see on which host a task was generated [HC]
2378 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2379 the number of tasks waiting to be received on channel and sent
2381 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2382 waits for the first task coming from a given host.. [AL]
2384 GRAS new functionnalities: [MQ]
2385 * Enhance the parsing macro to allow the size of multidimentional objects
2386 to be given thru annotations.
2387 * New example (and documentation): Matrix Multiplication a la RPC
2388 (as when I was young!) and fix a bunch of bugs found on the way.
2390 GRAS performance improvements: [MQ]
2392 * Reduce the amount of cbps creation/destruction by making it static to
2393 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2396 * Change libdata to a set so that we can search for stuff by ID (and thus
2397 reduce the insane amount of dict lookups)
2400 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2401 RL communication on top of SG since it's so uneffective.
2402 It may also be used for inter-thread communication in RL, one day.
2403 * Use gras_datadesc_copy() to exchange messages on top of SG
2405 - improve message exchange performance on top of SG
2406 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2407 * Don't exchange on the network the size of the used part of buffer,
2408 instead, specify the possible buffer size to read().
2410 - reduces the amount of read/write calls (one pair per exchange)
2411 - reduces the amount of exchanged data (the size)
2412 - allows to retrieve all arrived data on receiver side, if we don't need
2413 it right now (subsequent read will peek the buffer)
2414 - allows the receiver to proceed with the begining of the stream before
2415 everything is arrived
2416 - make it possible to build an iov transport (using readv/writev)
2418 - take care of the data with non-stable storage (like stacked data),
2420 * If possible, TCP send uses vector I/O (when writev() is here)
2421 - Don't use it for receive since we send data sizes and data on the
2422 same stream, so we wouldn't be able to chain large amount of chunks
2423 before having to flush the stuff to read the size.
2424 * Rework the transport plugin mechanism to simplify it and reduce the
2425 amount of pointer dereferencement when searching for the right function
2428 * I guess that now, we do almost as few system calls as possible while
2429 doing as few data copy as possible.
2431 To improve it further, we could try to send all the sizes first and then
2432 all the data (to use iov on receiving size), but it's only a partial
2433 solution: when you have 2 dimensional data, the sizes of the second
2434 dimension is data of the first dimension, so you need 3 streams.
2436 I'm not sure the potential performance gains justify the coding burden.
2438 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2440 SimGrid (3.00) stable; urgency=high
2443 * New! Give the possibility to hijack the surf parser and thus bypass
2444 MSG_create_environment and MSG_launch_application. Have a look at
2445 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2447 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2449 SimGrid (2.96) unstable; urgency=low
2454 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2455 This deprecates the xbt_error_t mechanisms.
2456 It modifies (simplifies) all XBT and GRAS API.
2457 MSG API keeps unchanged (exceptions raised by XBT are catched from
2458 within MSG and masked with existing error handling facilities)
2461 * New! Add a FATPIPE model. [AL]
2462 * New! Add a parallel task model. [AL]
2463 * New! Add automatically a loopback interface (in the default
2464 network model) if none was precised.
2467 * Bugfix: MSG_process_resume now works with the current running process.
2469 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2470 * Modification of MSG_task_get_compute_duration. Once a task has been
2471 processed, the value returned by this function is now equal to 0. [AL]
2472 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2473 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2474 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2475 of a m_task. For now, MSG_task_cancel only works with computation
2477 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2478 of the processor (in Mflop/s) regardless of the current load on the
2480 * API Change: use proper naming convention for MSG_getClock and
2481 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2483 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2484 This function changes the priority of a computation task. This priority
2485 doesn't affect the transfer rate. A priority of 2 will make a task
2486 receive two times more cpu power than the other ones. This function
2487 has been added to suit the needs of Nguyen The Loc and hasn't been that
2488 much tested yet. So if it fails, please report it and send me your code.
2490 * API Change: removed all functions and types that were marked "deprecated"
2491 since many months. Renamed MSG_global_init_args to MSG_global_init.
2493 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2495 SimGrid (2.95) unstable; urgency=low
2498 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2499 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2500 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2503 * Bugfix: Allow absolute paths for platform description files [MQ]
2504 * Bugfix: do free the variables after use. Leads to drastic performance
2506 * Implement max_duration (ie, timeouts) on resources [AL]
2509 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2511 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2512 the needed support (timeouts) [AL]
2515 * Implement measurement sockets. You can now get the bandwidth between two
2516 hosts thanks to AMOK (see below). [MQ]
2517 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2518 dynar over the network (yeah) [MQ]
2519 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2520 * simple/static token-ring example. [Alexandre Colucci and MQ]
2521 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2522 in place before [MQ]
2525 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2526 type parsing. "short" and "long" modifiers were also ignored; other
2527 modifier (such as reference level) are still ignored. [MQ]
2528 * Embeed the buffer size within the buffer itself on SG. [MQ]
2529 That way, send() are atomic and cannot get intermixed anymore (at least
2530 the ones which are less than 100k; bigger messages still have the issue)
2531 * Array size pushed by the field, not by the field type (or each
2532 and every long int will push stuff to the cbps) [MQ]
2533 * use select() to sleep since it allows to portably sleep less than one
2536 GRAS (minor cleanups)
2537 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2540 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2541 * type category 'ignored' killed as it was never used and were difficult
2543 * whether a type can cycle or not is now a flag, leaving room for more
2544 flags (as "ignored", if we feel the need one day ;)
2545 * Rename raw sockets to measurement sockets since "raw" has another
2546 meaning in networking community.
2549 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2550 and offers features not belonging to GRAS but that most applications
2551 need. One day, it may be a set of plugins loadable at runtime.
2552 * New module: bandwidth
2553 bandwidth measurement between arbitrary nodes running this module. [MQ]
2555 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2557 SimGrid (2.94) unstable; urgency=low
2559 The first beta release of SimGrid 3 !
2563 * Update the main page and the FAQ. Adding references to gforge.
2566 * Add a gras_os_getpid function.
2569 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2570 * Extend the logs so that they also print PID, hostname, date, ... if
2572 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2573 and kill the old version which were in testsuite/
2574 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2578 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2579 for users. Instead, each programming environment (GRAS, MSG,...) use it
2580 when needed to provide such a feature to users.
2581 Don't shortcut the mechanism or you will also shortcut the virtualization
2582 you need on the simulator.
2586 * Cleanups in configury with regard to compile optimization/warning flags.
2587 Also add -fno-loop-optimize to any powerpc since it's the optimization
2588 killing gcc (< 3.4.0).
2589 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2590 needed by MSG examples complications
2591 * Borrow configury beautifications from PHP
2594 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2595 warning (thanks loris for stumbling into this one).
2596 * Bugfix: stop loading private headers (gras_config.h) from the public
2600 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2601 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2602 RL would imply a DNS resolver, which is *hard* to do in a portable way
2603 (and therefore delayed).
2604 * Implement a real timer mechanism and use it in timing macros. This allows
2605 to avoid rounding errors and get a 0.000005 sec precision in timing
2606 macros. While I was at it, various cleanups:
2607 - allow to declare more than one timed section per file (fix a stupid bug)
2608 - move some private declaration to the right place
2609 - merge conditional execution and timing macros into emulation module
2610 - document the module
2611 - make sure the module cleanups its mess on gras_exit
2612 * Documentation improvements:
2613 - (new) how to compile applications using GRAS
2614 - (new) emulation support (timing macros)
2616 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2618 SimGrid (2.93) unstable; urgency=low
2620 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2623 - Use Paje properly where used. Still to be sanitized properly.
2624 - Portability fix: Add an implementation of the contexts using pthread
2628 - Add xbt_procname(): returns the name of the current process.
2629 Use it to show the current process's name in all logging.
2631 - fix detection of older flex version and the reaction, since we do
2632 depend on modern ones (we use lex_destroy)
2633 - Better separation of SG and RL in the libs: remove all simulation code
2634 from libgras. As a result, this lib is now only 200k when stripped.
2635 Some of the xbt modules may also be duplicated (two sets and such) and
2636 should be cleaned/killed before SG3.
2637 - Insist on using xlC on AIX because of weird problems involving gcc there.
2638 - Cleanup the make remote stuff. This is now done by scripts
2639 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2640 mainly for our private use, but we're working on changing them to user
2643 - Bugfix: flush the socket on close only if there is some *output*.
2644 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2646 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2649 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2651 SimGrid (2.92) unstable; urgency=low
2653 Alpha 3 on the path to SimGrid 3
2657 - New! First try of benchmarking macros.
2658 - New! First try so that gras_stub_generator generate deployment and
2659 remote compilation helpers.
2661 - Bugfix: Initialization fix in msg_test.
2665 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2667 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2669 - API Change: gras_os_sleep to take the amount of seconds as a double.
2670 Accepting an int was error prone since it was the only location where
2671 seconds were coded as such. It leaded to damn rounding errors.
2672 - Bugfix: Hard to belive that timers ever worked before this.
2674 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2676 SimGrid (2.91) unstable; urgency=low
2678 Alpha 2 on the path to SimGrid 3
2682 - Bug fix in the lmm_solver.
2684 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2685 through the function MSG_paje_output.
2686 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2687 - It is possible to bound the rate of a communication in MSG with
2688 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2689 to write it in the changelog).
2690 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2695 - Various cleanups to the autotools stuff
2696 - Begin to move Gras examples to examples/gras/
2697 - Let make distcheck work again (yeah!)
2699 - documentation overhauled using doxygen.
2700 gtk-doc-tools is dead in SimGrid now.
2701 - Automatically extract all existing logging categories, and add the list
2702 to the documentation (long standing one, to say the less)
2704 - Cleanup the known architecture table. Reorder the entries to group what
2705 should be, and use a more consistent naming scheme.
2706 (some of the test dataset are still to be regenerated)
2707 - New! Allow library to register globals on each process just as userdata
2709 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2710 lookup time (for now).
2711 Use it in msg and trp.
2712 This cleans a lot the internals and helps enforcing privacy of the
2713 headers between the gras components.
2714 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2715 - Bugfix: gras_os_time was delirious in RL.
2716 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2718 - Reenable GRAS now that it works.
2720 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2722 SimGrid (2.90) unstable; urgency=low
2724 Alpha 1 on the path to SimGrid 3
2726 * It is a long time since the last release of SimGrid. I'm sorry about
2727 that but as I had told you, I was rewriting a lot of things. I apologize
2728 to those who had been reporting bugs to me and that I had not answered.
2729 If your bug is still in the new version, please tell me. Here is a
2730 summary of the main changes.
2732 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2733 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2734 portability, speed, and a lot more but you'll figure it out later.
2735 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2736 I wanted to release the new MSG as soon as possible and I have
2737 broken GRAS, which is the reason why, for now, only MSG is fully
2738 functional. A laconic description of these projects is available
2739 in the documentation.
2741 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2742 kernel optimized for our needs (called SURF but only the developers
2743 should use it). Hence, MSG is now roughly 30 times faster and I think
2744 that by rewriting a little bit MSG, I could event speed it up a little
2745 bit more. Beside the gain in speed, it is also much easier to encode a
2746 new platform model with SURF than it was with SG. More to come...
2748 * REVOLUTION 3: I have tried to change a little as possible the API of
2749 MSG but a few things really had to disappear. The main differences
2750 with the previous version are :
2751 1) no more m_links_t and the corresponding functions. Platforms are
2752 directly read from a XML description and cannot be hard-coded
2753 anymore. The same format is used for application deployment
2754 description. The new format is described in the documentation.
2755 Have a look in tools/platform_generation. There is a tiny script
2756 that converts from the old platform format to the new one. Concerning
2757 the application deployment format, parsing the old one is tricky.
2758 I think most of you should however be able to convert your files. If
2759 it is really an issue, I can write a C code that does the conversion.
2761 2) the toolbox tbx does not exist anymore. We now have a library
2762 with much more data-structures but without the hash-tables (we have
2763 dictionaries that are much faster).
2765 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2767 *****************************************************************************
2768 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2769 * don't feel like converting the oldies. (Mt) *
2770 *****************************************************************************
2773 Version 2.90: "the long awaited one"
2774 - Finished rewriting and debugging MSG. Rewrote the documentation.
2775 - disable GRAS for now since it needs to be ported to the newest SG
2778 - Finish the port to windows (using mingw32 for cross-compile)
2781 - Main loop and datastructures of SURF. A cpu resource object is
2782 functional. Surf can thus be used to create cpu's with variable
2783 performance on which you can execute some actions.
2785 2004-11-15 Martin Quinson
2786 - Port to ARM. Simply added the alignment and size descriptions. Should
2787 work, but the ARM machines are so slow that I didn't had the opportunity
2788 to 'make check' over there yet.
2790 2004-11-15 Arnaud Legrand
2791 - Trace manager now written. It uses a heap structure and is therefore
2792 expected to be efficient. It may however be speeded up (particularly
2793 when many events occur at the same date) by using red and black
2794 trees. One day maybe...
2795 - Max-min linear system solver written. It uses a sparse matrix
2796 structure taking advantage of its expected use. Most operations are
2797 O(1) and free/calloc are called as few as possible. The computation of
2798 the minimum could however be improved by using a red and black tree
2801 2004-11-03 Arnaud Legrand
2802 - Rename every gras_* function that was in xbt/ to its xbt_
2804 - Add a heap and a doubly-linked list to xbt
2805 - Added a dichotomy to the dictionaries. make check works as well before
2806 so I assume that the patch is correct. I do not know however if things
2807 run effectively faster than before now. :)
2809 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2810 SimGrid, and the version number is bumped to 2.x
2812 2004-10-29 Martin Quinson
2813 - Introduction of the remote errors.
2814 They are the result of a RMI/RPC on the remote machine.
2815 ErrCodes being scalar values, you can't get the host on which those
2816 errors did happen. Extending the error mechanism as in Gnome is possible.
2817 No idea yet whether it is a good idea.
2819 2004-10-28 Martin Quinson
2820 - Interface revolution: the Starred Structure Eradication.
2821 I used to do typedef struct {} toto_t; and then handle *toto_t.
2822 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2823 - ???_t is a valid type (builded with typedef)
2824 - s_toto_t is a structure (access to fields with .)
2825 - s_toto is a structure needing 'struct' keyword to be used
2826 - e_toto_t is an enum
2827 - toto_t is an 'object' (struct*)
2829 typedef struct s_toto {} s_toto_t, *toto_t;
2830 typedef enum {} e_toto_t;
2831 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2832 s_toto_t) is private.
2834 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2835 so that it can set the variable to NULL. It was so for dicts and sets,
2836 it changed for dynars.
2838 - Fix a bunch of memleaks in dict_remove
2839 - Fix a bug in sg/server_socket opening: it failed all the time.
2841 2004-10-07 Martin Quinson
2842 - Speed up dynar lookup operation a bit.
2844 gras_dynar_get is dead.
2846 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2847 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2848 (which gives you the address of the stored data).
2850 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2851 copy of the data using an affectation to do the job and not a memcpy.
2853 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2854 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2856 It does not really speedup the dynar test because they are
2857 setting elements all the time (and look them seldom). But the dict does
2858 far more lookup than setting.
2860 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2862 2004-10-05 Martin Quinson
2863 - Allow to (en/dis)able the cycle detection at run time.
2865 Whether we should check for cycle or not is now a property of each
2866 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2867 datadesc_cycle_unset allow to remove this property when previously set.
2869 Note that the cycle detection is off by default since it impacts the
2870 performance. Watch the data you feed GRAS with ;)
2872 This property is hereditary. Any element embedded in a structure having it
2873 set have it set for the time of this data exchange.
2875 You should set it both on sender and receiver side. If you don't set it on
2876 sender side, it will enter an endless loop. If you forget on receiver
2877 side, the cycles won't be recreated after communication.
2879 - Header reorganization.
2880 Kill gras_private.h, each submodule must load the headers it needs.
2882 2004-10-04 Martin Quinson
2883 - Interface revolution: do not try to survive to malloc failure.
2885 Now, gras_malloc and friends call gras_abort() on failure.
2886 As a conclusion, malloc_error is not a valid error anymore, and all
2887 functions for which it was the only gras_error_t return value are
2888 changed. They now return void, or there result directly.
2889 This simplify the API a lot.
2891 2004-09-29 Martin Quinson
2892 - Re-enable raw sockets.
2893 Created by gras_socket_{client,server}_ext;
2894 Used with gras_raw_{send,recv}
2897 It should allow to kill the last bits of gras first version soon.
2899 This is not completely satisfactory yet (duplicate code with
2900 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2903 - Simplify transport plugin (internal) interface by not passing any
2904 argument to _server and _client, but embedding them in the socket
2907 2004-09-28 Martin Quinson
2908 - Finish the port to AIX.
2909 autoconf was my problem (segfault within the malloc replacement
2910 function. No idea why)
2912 2004-09-16 Martin Quinson
2913 - Fix some size_t madness on 64bit architectures.
2915 2004-09-08 Martin Quinson
2916 - Reduce the number of system headers loaded, overload some more system
2917 calls (such as malloc to cast the result of the system one, and work
2919 - Fix and reintroduce the config support
2921 2004-09-07 Martin Quinson
2922 - Source code reorganization to allow Arnaud to surf all over there.
2923 - Allow to document the logging categories.
2924 - Remove all uppercase from logging categories and useless cleanup in names.
2926 2004-08-18 Martin Quinson
2927 Version 0.6.2 (protocol not changed; API changed)
2928 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2929 gras_error_t), and NULL when not found. Functions expecting a msgtype
2930 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2931 hopefully usefull message.
2932 - Portability to prehistoric sparcs again
2934 2004-08-17 Martin Quinson
2935 Version 0.6.1 (protocol not changed; ABI not changed)
2936 - prealloc some buffers to speed things up
2938 2004-08-11 Martin Quinson
2939 Version 0.6 (protocol not changed; ABI expended)
2940 - The parsing macro can deal with the references, provided that you add
2941 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2943 2004-08-09 Martin Quinson
2944 Version 0.5 (protocol not changed; ABI changed)
2945 - Allow to off turn the cycle detection code in data exchange at
2946 compilation time. It should be at run time, but I'm short of time (and
2947 the config stuff is still broken). That way, we keep dict out of the
2948 critical path, which is good because the performance is poor:
2949 - search not dichotomial yet
2950 - dynar give no way to access their content and memcpy everytime
2951 - In composed data description (struct, ref and so on), stop foolness of
2952 keeping the subtype's ID, but store the type itself. This keeps sets out
2953 of the critical path, which is good since they rely on dynar and
2954 dictionnaries. The only loose of that is that we cannot detect the
2955 redeclaration of a structure/union with another content (but I'm not sure
2956 the code detected well this error before anyway). We still can detect
2957 the redefinition discrepancy for the other types.
2958 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2959 it breaks the code because of type-punning used all over the place).
2960 This breaks on all non-gcc architectures (for now).
2962 All those changes (plus the buffer of last time) allow me to gain 2 order
2963 of magnitude on cruel tests consisting of 800000 array of integers on two
2964 level of a hierarchical structure (200 secondes -> 4 secondes)
2967 - the selector of reference must now return the type it points to, not
2968 the ID of this type.
2970 2004-08-06 Martin Quinson
2971 Version 0.4 (protocol changed; ABI not changed)
2972 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2973 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2974 - Implement a buffer transport, and use it by default (it relies on tcp in
2975 real life and on sg in simulation).
2976 That's a bit hackish since I had a new field to the structure to store
2977 its data without interfering with the subtype ones. Inheritance
2978 is tricky in C. And that's a kind of reverse inheritance with one class
2979 derivating two classes. Or maybe a game with java interfaces. Anyway,
2980 that's damn hard in C (at least).
2981 Moreover, I got tired while trying to ensure plugin separation and
2982 genericity in SG mode. MSG wants me to do weird things, so let's go for
2983 cruel hacks (temporarily of course ;).
2984 See comment in transport_private.h:71
2985 - do not include all the _interface headers in private but in the files
2986 which really need them (to cut the compilation time when they are
2989 2004-07-26 Martin Quinson
2990 Version 0.3 (protocol not changed; ABI changed)
2991 - Major overhault of the datadesc interface to simplify it:
2992 - shorted the function names:
2993 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2994 - add a trivial way to push/pop integers into the cbps without malloc.
2995 This allows to make really generic sub_type description, which simply
2996 pop their size of the stack.
2997 - add a function gras_datadesc_ref_pop_arr() which does what users want
2998 most of the time: Declare a dynamic array (which pops its size of the
2999 stack) and declare a reference to it. Poor name, but anyway.
3000 - kill the post-send callback, add a post-receive one
3002 2004-07-23 Martin Quinson
3003 Version 0.2 (protocol changed; ABI changed)
3004 - add some testing for cpbs in the test cases, and fix some more bugs.
3005 This invalidate again the little64 data file, since I cannot regenerate
3007 - remove an awfull optimization in the logging stuff, allowing me to:
3008 - understand it again
3009 - learn gcc how to check that the argument match the provided format
3010 - fix all errors revealed by gcc after that
3011 - internal keys of dict are not \0 terminated. Deal with it properly in
3012 loggings instead of segfaulting when the user want to see the logs :-/
3014 2004-07-22 Martin Quinson
3015 - Fix some stupid bug preventing cbps (callback postit) from working
3017 2004-07-21 Martin Quinson
3018 - Some documentation cleanups
3019 - remove the useless last argument of msgtype_declare
3020 - rename the Virtu functions to fit into the 'os' namespace
3021 - move headers src/include -> src/include/gras/ and stop fooling with
3022 gras -> . symbolic link
3023 - make distcheck is now successful
3025 2004-07-19 Martin Quinson
3027 - Build shared library also
3028 - Install html doc to the right location
3029 - stop removing maintainer files in make clean
3030 - build tests only on make check
3032 2004-07-13 Martin Quinson
3034 - No major issue in previous version => change versioning schema
3035 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3036 corresponding dataset.
3038 2004-07-11 Martin Quinson
3040 - Get it working with any kind of structure (we can compute the padding
3041 bytes remotely for all the architectures I have access to)
3042 - Implement the structure parsing macro (still not quite robust/complete)
3043 - Improvement to the remote testing toysuite
3045 2004-07-10 Martin Quinson
3046 [autoconf mechanism]
3047 - get ride of a bunch of deprecated macros
3048 - actually run the test for two-compliment, not only compile it :-/
3049 - test whether the structures get packed (and bail out if yes. Damn.
3050 Alignment is a serious matter)
3051 - test whether the structures get compacted (but respecting the alignment
3052 constraints of each types)
3053 - test whether the array fields of structures can straddle alignment boundaries
3055 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3056 (took me 2 hours to find that bug, looking at the wrong place)
3057 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3058 before sending/receiving and are used to compute the offsets of fields
3059 - Given that padding size depend even on compiler options, keep track of
3060 alignment and aligned_size only for the current architecture. Not a big
3061 deal since we send structure fields one after the other (seems
3063 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3064 program, let it work on linux/gcc/little32. portability todo.
3066 2004-07-08 Martin Quinson
3067 - import and improve remote compilation support from FAST
3068 - make sure make check works on half a dozen of machines out there
3070 2004-07-07 Martin Quinson
3071 Let's say it's version 0.0.3 ;)
3072 - Implement conversions (yuhu!)
3073 - Let it work on solaris (beside conversion, of course)
3074 - Stupid me, using rand() to generate the conversion datatests in not wise.
3076 2004-07-06 Martin Quinson
3077 - Let make dist work, since I'm gonna need it to compile on remote hosts
3078 - Let Tests/datadesc_usage write the architecture on which the file was
3079 generated as first byte.
3080 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3081 architecture descriptions.
3082 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3083 on those architectures.
3084 - Optimization: send/recv array of scalar in one shoot
3086 2004-07-05 Martin Quinson
3087 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3089 - Plug a whole bunch of memleaks
3090 - each process now have to call gras_{init,exit}. One day, their log
3091 settings will be separated
3092 - Continue the code factorisation between SG, RL and common in Transport.
3094 2004-07-04 Martin Quinson
3096 - Redistribution between SG and RL.
3097 We wanna have to accept in SG, so move accepted related parts of RL in
3098 the common part. (more precisely, the dynar of all known sockets is no
3099 more a static in transport.c, but part of the process_data)
3101 [gras_stub_generator]
3102 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3104 2004-07-03 Martin Quinson
3105 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3106 them all up in one shot)
3108 - Fix the ugly case of reference to dynamic array.
3109 - New (semi-public) function gras_datadesc_size to allow the messaging
3110 layer to malloc the needed space for the buffer.
3112 - gras_socket_close now expect the socket to close (and not its address to
3113 put NULL in it after it). This is because the socket passed to handlers
3114 is one of their argument (=> not writable).
3116 - propagate the interface cleanup from last week in datadesc, ie remove a
3117 superfluous level of indirection. User pass adress of variable
3118 containing data (both when sending and receiving), and not of a variable
3119 being a pointer to the data. Let's say that I like it better ;)
3120 The price for that is constructs like "int msg=*(int*)payload" in
3121 handlers, but it's a fine price, IMHO.
3123 - Let it work in RL (yuhu)
3125 2004-06-21 Martin Quinson
3127 - porting SG plugin and SG select to new standards (works almost).
3128 - plug memleaks and fix bugs around.
3131 - cleanup the prototype of data recv and force users to specify when they
3132 want to handle references to objects. Test case working even for cycles.
3133 - plug memleaks. Valgrind is perfectly ok with this.
3135 2004-06-12 Martin Quinson
3137 - cleanup the separation between plugin and main code in plugin creation
3139 2004-06-11 Martin Quinson
3141 - Reput hook for raw sockets, needed for BW experiments
3142 - kill a few lines of dead code
3143 [Data description] Interface cleanup
3144 - gras_datadesc_by_name returns the searched type or NULL.
3145 That way, no variable is needed to use a type desc once, which makes
3147 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3148 two parameters were strings (field name, type name), leading to
3150 [Dicos] Interface cleanup
3151 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3152 This is consistant with the dynar API.
3154 2004-04-21 Martin Quinson
3156 - Porting to new standards.
3158 - interface cleanup.
3159 There is no bag anymore, no need to take extra provision to mask the
3160 pointers behind "ID".
3161 Better splitup of functions between files create/exchange/convert.
3162 This is still a bit artificial since convert and receive are so
3163 interleaved, but anyway.
3165 - add a queued message list to procdata (the ones not matching criteria
3167 - factorize some more code between SG and RL wrt procdata
3169 - use gras_exit in example to track memleaks
3170 - get rid of gs_example now that GS is properly integrated into gras
3171 - update run_test to integrate the latest tests (datadesc)
3173 - rename WARNINGn macros to WARNn since it prooved error-prone
3175 2004-04-19 Martin Quinson
3177 - register init/exit functions within gras module mechanism
3178 - send/receive function.
3179 Convertion is not implemented, but short-cutted if not needed.
3180 struct/array elements are sent one by one (instead of block-wise), but
3181 nobody really cares (yet). Get a prototype before optimizing.
3182 - tests (using a file socket) for DD send/receive on:
3183 - base types: int, float
3184 - array: fixed size, string (ie ref to dynamic string)
3185 - structure: homogeneous, heterogeneous
3186 - chained list, graph with cycle
3187 Believe it or not, valgrind is not too unhappy with the results. The
3188 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3192 - Bugfix when using a filename explicitely (instead of '-')
3194 2004-04-09 Martin Quinson
3196 - factorize more code between RL and SG in socket creation
3197 - Complete the implementation and tests of:
3199 o file (only in RL, and mainly for debugging)
3201 I lost 3 days to design a portable address resolver, and then decided
3202 that the prototype mainly have to run on my box.
3203 Addressing portability too early may be like optimizing too early :-/
3205 - use gras_init in the Tests instead of the crappy parse_log_opt
3206 (the latter function is removed)
3207 [Conditional execution]
3208 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3209 [Code reorganisation]
3210 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3211 Build examples against the RL library, since there is no way to disable
3212 its creation for now.
3214 For information, the beginning of coding on GRAS was back in june
3215 2003. I guess that every line has been rewritten at least twice since