1 SimGrid (3.14) UNRELEASED; urgency=low
4 * The whole documentation was reworked and reorganized. There is
5 still many room for improvement here, but we're on it.
8 * Privatization is now activated by default in smpirun
9 You can turn it off with -no-privatize if something goes wrong.
11 * Call-location tracing for SMPI.
12 You can add the exact location (filename / linenumber) of an MPI call to
13 your trace files and slow down or speed up the simulation between two
14 consecutive calls by using an adjustment file (see the documentation).
16 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
17 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
18 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
19 values were ignored in some cases. The timings of these functions can now
20 be significantly different.
22 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
24 Dropped functions and features
25 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
26 - MSG_mailbox_is_empty() -> Mailbox::empty()
27 - MSG_mailbox_front() -> Mailbox::front()
28 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
29 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
30 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
31 * Task affinity. Its intended behavior (that was very badly tested
32 and probably not really working) was deceiving what most users
33 would have hoped here.
36 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
37 EiBps, EBps, Eibps, Ebps for bandwidth.
38 They may become useful to some lucky ones.
41 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
42 * ex/app_tokenring: new example, very similar to the MSG Token Ring
43 * ex/async_waitAll: new example, on asynchronous communications
45 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
47 SimGrid (3.13) stable; urgency=low
49 The Half Release, a.k.a. the Zealous Easter Trim.
51 - We removed half of the lines, that were mostly experimental cruft.
52 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
53 (not counting blanks and comments -- according to openhub.net)
54 - The internals are now compiled in C++ (and will soon be clean C++)
55 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
56 59 klines of C, etc. We added only 29 klines of C++ in replacement.
58 * Backwards Compatibility breaks
59 - Removed Lua simulation bindings (switch to C or Java for that).
60 Lua can still be used to describe platforms
61 - Removed Java kernel plug-ins.
62 Will be reintroduced after the ongoing major internals reorg.
64 - the following functions were removed.
65 They were too specific and should be reimplemented in a generic
66 way, with filter function.
67 - MSG_task_listen_from_host
68 - MSG_mailbox_get_count_host_waiting_tasks
69 - MSG_mailbox_put_with_timeout was removed.
70 Please use MSG_task_send_with_timeout instead.
72 - the SD_application_reinit function was removed. It has been a noop for a while.
73 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
74 replaced by a more flexible mechanism.
75 - The following functions thus do not exist anymore
76 - SD_workstation_get_access_mode
77 - SD_workstation_set_access_mode
78 - SD_workstation_get_current_task
79 - Basic estimation functions have been removed but can easily be replaced
80 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
81 - SD_workstation_get_computation_time => amount / sg_host_speed()
83 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
84 Use VM.getSpeed()*load/100 for the legacy behavior.
86 - option enable_tracing was removed. It was not doing anything for a while.
87 - In the ModelChecker:
88 - the model-checker now ptraces the model-checked process which means
89 you cannot use a debugger on the latter anymore (we might make this
90 optional in the feature);
91 - removed soft-dirty page tracking;
92 - remove model-checked side snapshot management,
93 MC_snapshot() and MC_compare_snapshot();
94 - keep the MC_cut() function as a stub (it was not really working
95 in the previous release).
98 * All options are consistently in kebab-case. Old names are kept as alias.
101 * Switch to platform v4 format.
102 - Rename from 'power' to 'speed' the attributes describing the amount of
103 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
104 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
105 - In <host> and <link>, attributes availability and state are gone.
106 It was redundent with state and availability traces, and with peak values.
107 - In <cluster>, attributes availability_file and state_file are gone.
108 It was too complex and unused.
109 - Kill <gpu>. Was not doing anything.
110 - The DOCTYPE points to the right URL:
111 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
112 (the file at this address now documents the changelog since its v1)
113 - A warning is emitted for unit-less values (they are still accepted).
114 - speed. Default: 'f' or 'flops'. Also defined:
115 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
116 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
117 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
118 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
119 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
120 - latency. Default: 's' second. Also defined:
121 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
123 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
124 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
127 * s4u::Host is now the preferred public interface to the Host features.
128 sg_host_* functions are C bindings to the exact same behavior
129 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
132 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
133 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
134 They were deprecated since a few years, and probably did not even compile anymore.
137 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
138 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
139 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
140 encouraged to update their codes. A list of the modified functions can be found at the end of
141 include/simgrid/simdag.h
144 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
145 the lambda or closure passed as a parameter will run in kernel mode.
146 All the callback functions should be rewritten to that interface at some point.
149 * Reorganizing and cleaning the internals all around the place.
152 * Remove old default barrier/bcast buggy algorithms (see #18407)
153 * Various bug fixes to handle more codes
154 * Remove the need for the --foreground option of smpirun (it is still
155 accepted for backward compatibility).
158 * Kill the setset data container: MC don't use it anymore.
159 * Kill the queue data container: it made more sense with GRAS.
160 * Kill the xbt_peer_t data type: it's useless without GRAS.
161 * Kill rm_cb feature of config sets: it was never useful.
162 * Kill graphxml parsing feature. It was not used.
163 * Kill the deprecated code protected by XBT_USE_DEPRECATED
165 - xbt_dynar_sort_strings(), when the content is char*
166 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
167 They throw exceptions on invalid input;
168 * C++ support for declaring CLI flags (simgrid::config::Flag);
169 * class for abstracting different signal backends (simgrid::xbt::signal).
170 with no external dependencies (we need very simple signals).
173 * refactoring and cleanup of the code;
174 * ongoing process to cleanly separate the model-checking algorithms
175 from the code model-checking support.
177 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
179 SimGrid (3.12) stable; urgency=low
181 The Facelift Release.
184 * Require g++ v4.7 at least to not speak prehistorical C++.
185 * Require Boost 1.48 (for signal2 component).
186 * Java must be version 7 at least when activated.
187 * Builds on Windows again (including Java bindings).
188 * Tracing is now always enabled (no way to turn it out)
189 * Remove GTNetS. It was not working anyway.
190 * Various cleanups in the cmake scripts.
191 * Move headers around to sort them out on installed systems:
192 - instr/instr.h -> simgrid/instr.h
193 - instr/jedule/* -> simgrid/jedule
194 - simdag/datatypes.h was removed
195 - simdag/simdag.h -> simgrid/simdag.h
196 - msg/datatypes.h was removed
197 - msg/msg.h -> simgrid/msg.h
200 * Interface improvement:
201 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
202 - Sanitize the interface in MSG_task_ module:
203 - Merge two functions that were close enough but misleading:
204 set_compute_duration(t) -> set_flops_amount(t)
205 get_remaining_computation(t) -> get_flops_amount(t)
206 - set_data_size(t) -> set_bytes_amount(t)
207 get_data_size(t) -> get_bytes_amount(t)
208 - Massive cleanups in the functions related to the energy
209 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
210 - New: MSG_host_get_pstate()
212 - msg/energy/onoff: switching hosts on and off
214 * Interface improvement:
215 - Rename Host.isAvail() to Host.isOn()
216 - Rename Process.currentProcess() to Process.getCurrentProcess()
217 - Rename Task.setDataSize() to Task.setBytesAmount()
218 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
220 - #18874: Actually allows the GC to reclaim tasks
224 - SIMIX_process_throw: raises an exception in a remote process
225 * Refactoring: Separate sub-modules
226 - libsmx: the public interface, as libc in a real system
227 - popping: the strange dance that converts a user request into a kernel handling
228 - smx_context_*: the virtualization mechanisms that embed the user code
229 - smx_*: the handling of each simcalls
231 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
232 * Rename smx_action_t into smx_synchro_t, making explicit that these
233 things are used to synchronize processes with their environment.
234 For example, a communication is a sort of synchronization involving
235 the communicating processes (that may block until the exchange) and
236 the platform. The same can be said from computations, etc.
238 - #18888: segfault when a process ends before its kill_time
242 - 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.
243 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
244 - MPI_Type_set_name, MPI_Type_get_name
245 - MPI_*_c2f and MPI_*_f2c functions
246 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
247 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
248 - Activate a lot of new tests from the mpich 3 testsuite
250 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
251 - InfiniBand network model added : Based on the works of Jerome Vienne
252 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
253 - When smpi/display_timing is set, also display global simulation time and application times
254 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
255 * Collective communications
256 - 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
257 - 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/).
258 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
260 - Replay now uses algorithms from wanted collective selector
261 - Replay can be used with SMP-aware algorithms
262 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
263 - 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.
265 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
266 - Correctly use loopback on fat-tree clusters
267 - Asynchronous small messages shouldn't trigger deadlocks anymore
268 * Energy/DVFS cleanup and improvement
269 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
270 - new: smpi_get_host_pstate()
273 - "Full" network optimization flag was broken since Surf++
274 - Better handling of precision flags in maxmin
275 - Fix bug causing sometimes "Impossible" errors
276 - Properly pass cluster properties to included hosts
277 * Improvement of the Energy plugin.
278 - Always update the consumption before returning that value
279 - New property: watt_off to denote the dissipation when the host is off
280 - New functions getWattMinAt and getWattMaxAt to retrieve the
281 dissipation of pstates that we are not currently at.
282 * Java: class NetworkLink renamed to Link
283 * New function: simcall_process_get_kill_time()
284 * Massive rename s/workstation/host/
285 - That's intrusive, but that's good for the project consistency. Sorry.
286 - Change config option "workstation/model" into "host/model"
289 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
290 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
291 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
292 - xbt_dirname and xbt_basename for non-POSIX systems
294 * The model checker now runs as a separate process.
295 * The model checker runs must now be launched with the new simgrid-mc program.
296 * Record/Replay: the MC can display a textual representation of a path in the
297 execution graph. It can then be replayed outside of the model checker.
299 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
301 SimGrid (3.11) stable; urgency=low
306 * Normalizing pointers addresses tool for better diff between logs
309 * Add cloud examples using new VMs
310 - examples/msg/cloud/two_tasks_vm.tesh
311 - examples/msg/cloud/simple_vm.tesh
312 - examples/java/cloud/cloud.tesh
313 - examples/java/cloud/migration/migration.tesh
314 * Add java surf examples:
315 - examples/java/surfPlugin/surf_plugin.tesh
316 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
317 - examples/java/surfCpuModel/surf_cpu_model.tesh
318 * Add SMPI+MSG example:
319 - examples/smpi/smpi_msg_masterslave/
324 - msg task destroy cancel test
325 - msg_host on/off test
326 * Move all tests in testsuite to teshsuite (adding tesh files)
327 * Restructure teshsuites
328 - one folder for each kind of test
329 * Restructure AddTests.cmake
331 - structure the order of tests (with sections)
334 * Add virtual machine
335 - creation of a VM on a PM
336 - migration of a VM from a PM to another PM
338 - MSG_process_join(msg_process_t process, double timeout)
339 - msg_bar_t MSG_barrier_init(unsigned int count)
340 - int MSG_barrier_wait(msg_bar_t barrier)
341 - void MSG_barrier_destroy(msg_bar_t barrier)
342 - msg_as_t MSG_environment_as_get_by_name(const char * name)
343 * New option "msg/debug_multiple_use" to help debugging when a task is used
346 - Improvements and finalization of MSG_storage, MSG_file APIs and their
348 - Increase code coverage in test suites
352 * Protect context stacks against stack overflow. The number of protected memory
353 pages allocated on the top of each stack (1 by default) can be configured
354 with the new command line option --cfg=contexts/guard_size:<value>.
355 * Simcalls are now generated by a python script that
356 - generates files included by SimGrid
357 - checks that all the functions exist, and proposes prototypes
359 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
360 host_set_data, host_get_data
362 - simcall_process_join(smx_process_t process, double timeout)
363 * Fix bug where sleeping processing could not be suspended.
366 * Translate surf models from C to C++
367 - Generic classes for all models: Model, Resource, Action
368 - A generic interface for each kind of model (CPU, Network, Storage
369 Workstation, WorkstationVM)
371 * Translate surf routings from C to C++
372 * Add callbacks using sigc++ or boost::signals2
373 - Add callback functions for resource creation/destruction
374 - Add callback functions for action state change
375 - Handle Energy as a plugin
376 * Replace swag by boost::intrusive
377 * Add new routing models for clusters. For documentation, see
378 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
379 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
380 parameters for cluster tag
381 - Fat trees, with topology="FAT_TREE" and
382 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
383 - see examples/platforms/torus_cluster.xml and
384 examples/platforms/fat_tree_cluster.xml
388 * Hostfiles support host:nb_processes construct to deploy several processes on
390 * Collective communication algorithms should not crash if used with
391 improper number of nodes and report the error.
392 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
393 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
394 MPI_Dims_create, MPI_Cart_sub are supported.
395 * New interface to use SMPI programmatically (still depends on MSG for
396 some parts, see examples/smpi/smpi_msg_masterslave) :
397 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
401 * Global variables privatization in MPI executables is now performed at runtime
402 with the option smpi/privatize_global_variables (default:no).
403 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
404 dynamic libraries loaded by the application are not privatized (static
405 linking with these libraries is advised in this case)
408 * Options defined in XML work correctly now.
411 * New cmake option, enable_lib_in_jar, to control whether native libraries are
412 copied into simgrid.jar or not (ON by default). Use this option if you want
413 to reduce the size of the installed simgrid.jar, *and* the native libraries
414 are kept installed elsewhere.
415 * Surf binding with SWIG (code generated in maintainer mode only):
416 - plugin to handle callbacks
417 - CPU model only for the moment
420 * Supernovae build mode is definitively removed. It was used to improve
421 inlining and inter-module optimizations. It is nowadays superseded by
422 link-time optimizations commonly available in compilers.
423 * Update ns-3 find lib. Bindings for ns-3 should work again now.
424 * Add boost dependency for surf++
425 * Add new macro for tests
426 - ADD_TESH(name <tesh_args>)
427 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
431 - xbt_bar_t XBT_barrier_init(unsigned int count)
432 - int XBT_barrier_wait(xbt_bar_t barrier)
433 - void XBT_barrier_destroy(xbt_bar_t barrier)
434 * Make the xbt_os_time module public
436 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
438 SimGrid (3.10) stable; urgency=low
440 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
443 * Reintegrate Java to the main archive as desynchronizing these
444 package is not acceptable anymore (Java is now considered stable)
445 * Add explicit synchronization facilities through semaphores
446 * Bug fix: Task.setDataSize() only changed the C world, not the value
447 cached in the Java world
450 * Dramatically change the way files are handled. API and internals changed, but
451 this part of MSG was not considered as production grade either.
452 * Add explicit synchronization facilities through semaphores
453 * Add a new function MSG_host_get_process_list()
454 * Preliminary DVFS support (see examples/msg/energy/ for details)
457 * SMPI is now included directly in the libsimgrid as the windows
458 linker doesn't force us on splitting it anymore.
459 * Improvements of the SMPI replay tool:
460 - Most of the collective communications are now rooted in the same process as
461 in the original application.
462 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
463 used until now). Multiple data types can now be used in a trace.
464 - The replay tool now supports traces produce either by TAU or a modified
466 - Bug Fix: the compute part of the reduce action is now taken into account.
467 - Gatherv collective is now supported
468 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
469 of smpirun outputs time independent traces for the current run. One file
470 is created per process. If too many processes are simulated, this behavior
471 can be changed to one file for all processes by using the
472 tracing/smpi/format/ti_one_file flag
473 * smpirun generates the host file if needed (with given host count and platform)
474 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
475 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
476 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
477 --cfg=smpi/coll_selector:(mpich/ompi)
478 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
480 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
481 happening inside a collective SMPI call.
482 * Fix the behavior of complex data types handling.
483 * Make MPI_Wtime another synchronization point to take computations into
485 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
486 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
487 * Add all missing Fortran bindings, SMPI should work with Fortran 90
488 (no privatization of global variables yet)
489 * Preliminary DVFS support (see examples/smpi/energy/ for details)
492 * Verification of liveness properties is now available for SMPI applications
493 (in addition to MSG applications)
494 * Bugged examples using SMPI in examples/smpi/mc/
495 * Add --cfg=model-check/visited option. Allows the verification of infinite
496 programs. Detection of loops in the execution thanks to the system state
497 comparison and reduction of the state space to explore. Can be combined with
498 DPOR for safety properties.
501 * Allow to change SimGrid configuration (see --help) within the code
502 thanks to SD_config() as it can be done in MSG.
503 * Add a new function SD_task_set_amount() upon user request.
506 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
507 * Remove rule based routing (no more PCRE dependency)
508 * Add a limiter_link option to cluster tag, to specify a maximum reachable
509 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
510 * Add a loopback_bw and loopback_lat options to cluster tag.
511 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
512 routing. Now peers are encapsulated in an AS and have their own private
513 router but this is transparent.
516 * Our own implementation of getline is renamed xbt_getline, and gets
517 used even if the OS provide a getline(). This should reduce the
518 configuration complexity by using the same code on all platforms.
519 * New type: xbt_cfg_elm_boolean.
520 * Allow to use yes/no for boolean configuration options in the command line.
521 * Allow to disable SimGrid cleanups at exit from command line option.
522 There are situations where one may want a simulation to end with an exit.
523 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
524 annoying when scripting around the simulator. Adding a
525 --cfg=clean_atexit:no allows to circumvent this issue.
528 * Lots of memory leaks were corrected in this release.
529 There is no known memory leaks anymore, in all of our 600+ tests.
530 * New command line option --version, to get SimGrid version information.
531 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
533 * Supernovae builds are deprecated, and expected to be removed in the next
536 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
538 SimGrid (3.9) stable; urgency=low
540 The Grasgory release: GRAS is really dead now.
542 * Complete overhaul of the internal host structures scheme.
545 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
546 was considered as experimental and badly maintained since then.
547 * Keeping it was thus a trap to our potential users, that could take
548 it instead of MSG or SMPI by mistake despite is pity state.
549 * GRAS seems to have very few users (if any), and no one volunteered
550 to maintain it further. It also induces a lot of XBT code (for
551 portability sake), that must be maintained too.
552 * For all these reasons, we killed GRAS. If someone wants to revive it
553 in the future, don't cry, our git history still remembers of GRAS.
556 * Major overhaul. Merge our documentation again as time proved that
557 splitting it was really not helping our users.
558 * Further improve the developer documentation to help newcomers
559 hacking on SimGrid itself. The user documentation (and in
560 particular, the beginner documentation) is still in a sorry state.
563 * Now works on Windows too!
564 * Much more extensive test suite, from MPICH
567 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
568 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
569 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
570 * Bug fix: let task be scheduled when the last dependency to be solved is
571 a control dependency.
572 * Remove SD_load_environment_script function.
573 Use the C sg_platf function if you want to declare a platform
577 * New function: MSG_process_get_number()
578 * Old function documented: MSG_config()
579 * Remove MSG_load_platform_script function
580 Use the C sg_platf function if you want to declare a platform
584 * Change the default value of the TCP_gamma constant (maximal size of TCP
585 congestion window) to a more realistic 4MiB value. If you notice changes in
586 your simulation results, you can fall back to the previous 20k tiny window
587 by adding --cfg=network/TCP_gamma:20000 on command line.
588 * (Hopefully) fix a bug wrt periodic availability and state traces
589 * Bug fix: use default values at start when first event in availability/state
590 trace is not at time 0.
593 * remove the "new_" part of function name sg_platf_new_trace_connect
594 (resulting in sg_platf_trace_connect), since it does not create
598 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
599 use case where it's really mandatory, and maintaining it was a pain in
601 * New: xbt_fifo_search(), search an item with a user-provided
602 comparison function instead of dumb pointer comparison.
605 * Fix the lua deployment:
606 Use `simgrid.init_application()` before deployment instead of
607 `simgrid.msg_register_application()` after.
610 * Transfer the tracing files into the corresponding modules.
612 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
614 SimGrid (3.8.1) stable; urgency=low
616 The "we are told that some people want to also *install* the simgrid
619 * Add missing file "tesh.1" to the archive.
621 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
623 SimGrid (3.8) stable; urgency=low
625 The Psssshiiiit release: SimGrid jumps into the Cloud.
628 * Add an experimental interface to manipulate VMs. They are mainly
629 process groups with very few intrinsic semantic, but they should
630 allow you to build the semantic you want easily.
631 * New function: MSG_host_set_property_value()
632 * New function: MSG_process_on_exit(). To clean memory in all cases.
633 * Bug fixes that made the host (and link) failures unusable.
634 * Add a way to auto-restart process when the host in which they are
635 executing comes back (ON_FAILURE="RESTART" on deployment file,
636 MSG_process_auto_restart_set).
637 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
638 please stop using the old ones, they are DEPRECATED.
640 * Deprecate functions MSG_global_init() / MSG_global_init_args()
641 Please use MSG_init() instead. (reducing the amount of entry
642 points in the library helps us).
643 * Make it impossible to link against the wrong version of the lib
644 * Deprecate MSG_clean(). No need to call it anymore.
645 * Function MSG_get_host_number() is not deprecated anymore.
648 * Split the doc into a user guide and a reference guide.
649 * Start a developper guide to help people hacking on SimGrid.
652 * Enable tracing by default. This modules rocks you should use it.
653 * Remove option custom_flags. Now use environment variables CFLAGS
655 * Use default cmake things to detect lua instead of home grown ones.
656 * New option "enable_mallocators" to disable mallocators, for debugging
657 purpose ("on" by default).
660 * Bug fixes around the resource failures: don't let the processes
661 survive the host they are running onto.
662 * Add an interface to auto-restart processes when the host in which they are
663 executing comes back.
664 * Ensures that SIMIX_clean is called automatically. It's not part of
665 the public interface anymore (bindings should be updated).
668 * Bug fix for when SD_Simulate is called with a positive value: be careful
669 when comparing doubles. Sometimes they are different for non significant
671 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
672 parallel task whose initial work is distributed among host according
673 to the Amdahl's law. Such tasks are created with a parameter alpha
674 that corresponds to the non-parallelizable part of the computation.
675 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
676 between two sets of workstations assuming a 1D block distribution (each
677 workstation owns a similar share of data) on both sides.
679 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
680 Data redistribution will be automatically scheduled once parent and child
681 are both scheduled. The filling of computation_amount and
682 communication_amount structures is now done seamlessly thanks to the chosen
684 * New function SD_workstation_dump to display various information
685 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
686 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
687 bandwidth on the route onto which the task is scheduled and the amount of
689 To divide the nominal bandwidth by 2, the rate then has to be :
690 rate = bandwidth/(2*amount)
691 * Compute tasks that have failed can now be rescheduled and executed again
692 (from their beginning)
693 * Increasing source code coverage (src/simdag is now covered at 95.8%
697 * Re-implement time-independent trace replay using SMPI (at the
698 smpi_smp_* level) instead of MSG. This should replace
699 examples/msg/actions/actions.c
700 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
701 hindexed and structs)
702 * Implement the exchange of non-contiguous data.
703 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
704 * Correct behavior of smpi/sender_gap and set its default value to 0
705 * Add option to asynchronously send small messages to allow better
706 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
707 specifies the size in bytes under which messages will be asynchronously sent.
708 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
709 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
712 * New C interface to define a platform: XML is now optional.
713 For more info, please check include/simgrid/platf.h.
714 * New interface to define random platforms from the C:
715 For more info, please check include/simgrid/platf_generator.h and
716 examples/msg/masterslave/masterslave_platfgen.c
717 * Export a sg_cmdline dynar containing all the arguments we got from
721 * Two new tracing options for adding comments to trace file so you
722 can track your experiments (see --help-tracing for details).
723 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
724 * Adding the SimGrid version that generated the trace file as a comment.
725 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
726 * Fix to avoid key clashes on Paje links
727 * Other minor fixes related to the Paje specification
730 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
731 and renamed to xbt_str_hash() and xbt_str_hash_ext().
732 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
733 * Greatly improve the robustness of mmalloc to user errors (such as
734 using an area after freeing it, or freeing it twice)
736 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
738 SimGrid-java (3.8.1) stable; urgency=low
740 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
741 Warning: all methods are visible, but only some of them are
742 implemented so far. Check the source (src/jtrace.c)
743 for further information.
744 * New module: org.simgrid.msg.File (SimGrid File management functions)
745 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
746 * Change the meaning of Process.restart: now restart the process from
747 the begining, like MSG_process_restart in C.
748 * Add Process.setAutoRestart: handling of process restart when failed
750 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
751 you to retrieve the properties of the processes/hosts
752 * Deprecate Msg.clean(): you can just forget about it now.
753 * New function Process.getCount(), that only works when compiling
754 with the not yet released version 3.9 of the C library.
756 * New context factory based on Coroutines. It mandates a modified JVM
757 but then, the simulations run about five times faster, and there is
758 no limit to the amount of processes (beside of the available memory).
760 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
762 SimGrid (3.7.1) stable; urgency=low
765 * Restore the prototype of MSG_process_create_with_environment() to
766 the pre-3.7 situation by removing the kill_time argument.
767 * Add a MSG_process_set_kill_time() function instead.
770 * Fix weird behaviors when dealing with parallel tasks.
773 * Simgrid is now built as a dll.
774 * Simgrid-java now works on Windows.
775 * Simgrid-Java is now included into Windows package.
778 * First pre-build package for MacOSX.
781 * Fix compilation when using MSG_USE_DEPRECATED.
782 * Fix some compilation issues on Macs and Windows.
783 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
784 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
786 We discovered that the Lua console is broken, but we are missing the
787 manpower to fix it right now. The problem existed in 3.7 too, so we
788 are not blocking the release for that. Sorry if you depended on this
789 feature, any help would be really welcome.
791 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
794 SimGrid-java (3.7.1) stable; urgency=low
796 The "Java aint got to be bloated and slow" release
799 * Various internal cleanups and performance improvement
800 Simulations are expected to run up to twice faster or so
801 * Make Process.kill(process) an instance method, not a static one
802 * User processes are not java.lang.Thread subclasses.
803 This breaks the compatibility (sorry), but previous API was
804 brain-dead, making it impossible to have non-trivial
805 initializations in the process constructor.
806 * Require a full constructor per Process sub-class.
807 Kinda breaks the compatibility (sorry), but this allows a much more
808 efficient way to launch the processes at simulation startup.
809 * Do not embeed our version of semaphores, java 1.5 can be considered
810 as sufficiently prevalent for us to not dupplicate its features.
814 * Add examples for almost every part of the API
815 We spotted and fixed a lot of bugs in the process
817 * New module: asynchronous communication API
818 * New function: Process.sleep()
819 It takes milliseconds as argument, just as java.lang.Thread.sleep()
820 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
821 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
823 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
825 SimGrid (3.7) stable; urgency=low
827 The "spring cleanups (before the next Big Project kicks in)" release.
830 * We can specify the SMPI latency/bandwidth factor with command line
831 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
832 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
833 You can also use the "config tag" from platform file by adding this line
834 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
835 (see "example/platforms/tag_config.xml" to use "config tag").
836 Note that the command line supersedes the platform file configuration.
837 * Change the correction factors used in LMM model, according to
838 the latest experiments described in INRIA RR-7821.
839 Accuracy should be improved this way.
840 * Use the partial invalidation optimization by default for the
841 network too. Should produce the exact same results, only faster.
842 * Major cleanup in surf to merge models and split some optimization
843 mechanisms from the core of the models. As a result you can now
844 specify which model to use (e.g., --cfg=network/model:LV08
845 --cfg=cpu/model:Cas01) and which optimization mode to use
846 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
847 Incompatible combinations should err at initialization. See
848 --help-models for the list of all models and optimization modes.
849 * The CLM03 workstation model was dropped for simplicity because it
850 used the deprecated CM02 network model. Use default instead.
851 * Rename the TCP_gamma configuration option to network/TCP_gamma
852 * Rename the coordinates configuration option to
853 network/coordinates, and document it
854 * Use now crosstraffic keyword instead of the terribly misleading
855 fullduplex keyword. It is activated by default now in the current
856 default model, use --cfg=network/crosstraffic:0 to turn it off.
857 * Ongoing refactoring the model parsing to make XML files optional
858 See include/simgrid/platf.h for details (still to be completed)
861 * Major overhaul of the documentation. Almost instructive now :/
862 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
863 functions and friends. This interface was considered as
864 deprecated since over 2 years, it's time to inform our users that it is.
865 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
866 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
867 use these (crappy) functions in your code.
868 These functions will be removed soon. Stop using them now.
869 * Deprecate MSG_get_host_{table,number}
870 Implement MSG_hosts_as_dynar() instead.
871 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
872 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
875 * Stabilize the parallel execution mode of user contexts
876 * Introduce configuration variables to control parallel execution:
877 - contexts/synchro: Synchronization mode to use when running
878 contexts in parallel (either futex, posix or busy_wait)
879 - contexts/parallel_threshold: Minimal number of user contexts
880 that must be part of a scheduling round to switch to parallel
881 execution mode (raw contexts only)
882 * Fix bugs that prevented to use suspend/resume along with
883 synchronization structures.
884 * Fix bugs in process termination that lead to invalid memory access
885 in very specific conditions.
888 * Introduce a parallel mode for the models (controlled by surf/nthreads
889 configuration item). In our tests, running the models in parallel
890 never lead to any speedups because they are so fast that the gain
891 of computing each model in parallel does not amortizes the
892 synchronization costs, even when ultra fast futexes are used.
893 This is released anyway because YMMV.
896 * Performance boost by using a swag internally to compute the set of
897 tasks that are finished and should constitute the return value of
901 * Enable it by default now that it is considered rather stable.
904 * Documentation of the tracing functions.
905 * Performance gains when tracing categorized/uncategorized resource
906 utilization by avoiding calls to get route when updating resource
907 variables. LMM constraints are being used instead.
908 * API changed to set task categories. Use MSG_task_set_category instead
909 of TRACE_msg_set_task_category, and SD_task_set_category instead
910 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
911 * Bugfix for graphicator, routes not correctly obtained, memory leaks
912 * Examples for link user variables added (see at examples/msg/tracing/)
913 * Deprecated function TRACE_msg_set_process_category completely removed
914 * Trace header updated according to the latest Paje file format
915 * Tracing network lazy updates, no longer obligate users to use full updates
916 * --cfg=tracing/platform:1 also registers power/bandwidth variables
917 * Experimental: let user code declare/set/push/pop application states for hosts
918 * API changed to allow the manual creation of graph configuration files
919 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
922 * Improve the API of Lua MSG bindings, using the Lua spirit.
923 * Each simulated process now lives in its own Lua world (globals are
924 automatically duplicated). It helps writing simulators. Will allow
925 to run Splay programs within SimGrid in the future.
926 * Add a Chord example in Lua, equivalent to the MSG one.
929 * Start the implementation of a solution to express temporal
930 properties, not only local assertions. This is still an
931 experimental work in progress, stay clear from it to be safe.
935 - Add new runtime parameters --help-logs and --help-log-categories
936 to display information about supported logging parameters and
938 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
939 don't exists anymore.
940 * Mallocators: allow value NULL for the reset function.
942 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
943 create homogeneous dictionaries, where all the elements share the
944 same free function. Non homogeneous dictionaries will be
945 deprecated in the next release.
946 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
947 - Multi-level dictionaries are deprecated.
949 - new function xbt_dynar_search_or_negative() that is useful when
950 you have less than 2 million elements in your dynar and don't
951 want of the extra complexity of catching exceptions when the
952 element is not found.
954 - Make xbt_os_thread module (for thread portability) public.
955 Documentation is still to come, sorry.
957 - Cleanups and simplifications to make it maintainable again.
958 - Exotic features (such as memalign and valloc) were removed.
959 - The metadata were extended and improved so that the
960 model-checker becomes able to explore and inspect the heaps.
961 - This may induce a performance drop when enable_model-checking is
962 ON in cmake (even if it's not used in the simulation), but it is
963 necessary at this point to get MC working.
965 Turn model-checking OFF if simulation performance matters to you.
966 Not enabling it at runtime is not enough, disable it in cmake.
968 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
971 SimGrid (3.6.2) stable; urgency=low
973 The "Not coding new stuff allows to polish old things" release.
976 * New bindings to the NS3 packet level simulator (experimental)
977 * Use the raw (efficient) execution contexts instead of the sysv
978 (portable) ones when possible.
979 * libpcre is now mandatory in any cases since not using it led to
980 severe performance loss and possibly other issues
981 * Update the XML platforms:
982 - G5K: include the latest machine in Nancy
983 - GridPP and LCG: new platforms
984 * Documentation was partially updated, at least (more to come)
986 Bug fixes, cosmetics and small improvements
987 * Free terminated processes before the end of the simulation to avoid
988 exhausting the memory of users having very dynamic amount of
990 * Bug fix and cosmetics about canceling non-running tasks
991 * Bug fix about the dot loader's issues when using libcgraph
994 * Create an installer for windows with nsis (amd64 and win32)
995 - Add an hello world project to illustrate simgrid project creation.
996 - Embed libpcre into the Simgrid installer to avoid
997 its compilation burden
998 * The raw execution contexts should work on Apple now
999 * Port to Windows 64 bits
1000 - Sysv contexts now have an implementation for this arch
1001 - GRAS communication features now support this arch
1002 * Drop support for borland compiler on windows
1003 - this code was not maintained, and we kinda depend on gcc nowadays
1004 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1005 * Fix portability issue on unstable ubuntu: linker became picky on
1008 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1010 SimGrid (3.6.1) stable; urgency=low
1012 The "Oops, we broke Macs too" release
1015 * Fixed contexts detection so that raw ones are used when possible
1016 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1017 leads to a strange error, with user code segfaulting sometimes when
1018 the generated code is not perfectly aligned (which is not
1019 controllable from the user side, depends on the amount of code)
1022 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1024 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1026 SimGrid-java (3.6) unstable; urgency=low
1029 * Split of every thing from simgrid v3.5 into a separate package.
1031 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1033 SimGrid (3.6) stable; urgency=medium
1035 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1038 * Bindings now constitute their own package, separated from the main one.
1039 Rationale: reduce our maintenance nightmare by reducing the module coupling
1040 They will soon be released on their own on gforge.
1042 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1043 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1045 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1046 * It's not quite deprecated for now because we have no replacement,
1047 but it may soon become the case.
1050 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1051 * Fortran: New user-level cache variable to store the rank of the running
1052 process. This improves performance by an order of magnitude.
1053 * C: New coccinelle script to automatically locate and modify global and
1054 local static variables.
1055 * Improved SMPI network model with a sender-side gap to account for multiple
1059 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1060 only say if a communication is finished, no matter whether it succeeded or
1061 failed. You can call MSG_comm_get_status() to know the status of a finished
1063 * New function MSG_task_dsend() to send a task and detach it. When a
1064 communication is detached, you are never notified of its success or failure
1065 and the memory is released automatically once it is finished. This function
1066 is useful when you don't care about the end nor the success of a
1068 * Change the prototypes of action replay. Sorry for inconvenience,
1069 but this is really more efficient this way (and to adapt your code,
1070 you just have to fix the initialization, that shouldn't be too long)
1071 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1072 ever managed to do anything useful with it.
1073 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1074 takes a dynar of communications. It returns immediately and gives the
1075 index of a finished communication (if any).
1076 * New example: a basic implementation of the Chord P2P algorithm.
1079 * New model for multi-core CPUs. You can now use the core attribute to
1080 precise the number of cores of a host. This is a basic model. Every
1081 process running on the host receives at most the power provided in
1082 the DTD (throughput<=power). Total throughput of process cannot exceed
1084 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1085 router linked by an up-link and a down-link (possibly asymmetrical).
1086 This kind of pattern allows to easily build last-mile model style platforms.
1087 Aggregating such patterns in a rule-based AS is thus the technique of
1088 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1089 * New model for Vivaldi routing. We transformed the Vivaldi network model
1090 into a Vivaldi routing model (based on the rule-based model). This allows to
1091 combine Vivaldi based latencies with last-mile platforms.
1094 * Added a check for NaN of IEEE754 infinite in the double entries of
1096 * Introduce a new context factory "raw", highly inspirated from the
1097 ucontext factory, but using manually crafted functions in assembly to
1098 do the work in an efficient manner.
1099 * Allow to change the used context factory at run time, not only at
1100 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1101 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1102 contexts at runtime (only with raw contexts or ucontexts).
1103 * Completely rewrote this module to allow parallel execution of user
1104 processes. Use --cfg=contexts/nthreads:N to execute user processes
1105 with N parallel threads (the default is 1, meaning no parallelism).
1106 * Allow to decide dynamically between sequential and parallel modes.
1107 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1108 processes in parallel only when their number is greater than or equal to P
1110 * Added a check for NaN of IEEE754 infinite in the double entries of
1114 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1115 won't output the state of processes when interrupted with Ctrl-C
1116 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1117 NULL-terminated array. This may solve backward compatibility issues
1118 due to the change to return type of SD_simulate. See also:
1119 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1120 * Add new macros with variable number of arguments.
1121 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1122 - in xbt/asserts.h: xbt_assert
1123 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1124 - in xbt/ex.h: THROWF and RETHROWF.
1125 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1127 * Change xbt_die() to accept a format string with arguments, just like printf.
1128 * New data structure: xbt_lib_t, like a dict but more general and with better
1132 * New configuration options
1133 Options triva/categorized and triva/uncategorized can be used to generate
1134 graph configuration files for Triva visualization tool.
1135 * Configuration option tracing/platform is renamed to tracing/categorized
1136 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1137 * New TRACE_user_link_variable interface:
1138 User provides the name of the link and the tracing variable to attach to it
1139 * the declaration of tracing categories must be done after the environment
1141 * simpler tracing interface, just one way to declare categories
1142 TRACE_category or TRACE_category_with_color, it is up to you
1143 * links in the trace file are again identified by their names
1144 * trace contains the full platform hierarchy exactly as declared using the ASes
1145 * Options tracing/msg/[task|process]:1 groups the process by hosts
1146 for both cases, tasks and processes must have names that are unique during the simulation
1147 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1148 * The experimental option tracing/msg/volume is deprecated
1149 its functionality may be reincorporated if needed
1151 The tracing generates a trace file with unordered timestamped events,
1152 because of the way the core simulator (surf) works. A script available
1153 at the tools directory (fix-paje-trace.sh) can be used to put the events
1154 in order. We have changed the tracing so it can generate ordered timestamped
1155 events in the final trace, but depending on the simulator (and how much time
1156 is simulated) that can lead to a huge memory utilization. It is deactivated
1157 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1159 Build Infrastructure
1160 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1161 - We are trying hard to keep the API stable, but it may happen that
1162 some things change (we're a research project after all, not a
1163 nuclear plant operating system). If such things should happen, you
1164 could rely on that macro to adapt.
1165 - current value: 30600 for 3.06.00, aka 3.6
1166 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1167 a number that can be compared with SIMGRID_VERSION.
1168 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1169 assertions and verbose/debug logging events are disabled at compile time.
1171 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1173 SimGrid (3.5) stable; urgency=medium
1176 * New feature to any SimGrid-based simulator: Model-Checking
1177 Check SIN#1 for more details.
1180 * New Model SMPI (three-interval linear regression for correction factors)
1181 See RR-7426, available at http://hal.inria.fr/inria-00527150
1182 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1183 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1184 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1185 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1186 * See also src/smpi/README
1191 - Tracing API changes: TRACE_start and TRACE_end should not be called
1192 by user-code. They are automatically called by simulators created
1193 with SimDAG, MSG and SMPI if the toolkit is compiled with
1194 tracing_enabled=ON. Categories declaration and utilization remain the
1195 same for MSG and SimDag.
1196 - A function was added to the tracing API to declare categories with
1198 - TRACE_category_with_color (char *category, char *color)
1199 where color must be in the following format
1200 "%f %f %f", red, green, blue
1201 and red, green, blue are float values in the interval [0, 1]
1202 - User can specify NULL as color parameter, or continue calling
1203 TRACE_category (cat)
1204 On that case, the tracing system will define random colors
1205 - The following command-line options are supported:
1206 --cfg=tracing/filename:msg.trace
1207 --cfg=tracing:1 (activate tracing, needed to use others)
1208 --cfg=tracing/platform:1 (categorized resource use)
1209 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1210 --cfg=tracing/msg/task:1 (task creation)
1211 --cfg=tracing/msg/process:1 (process creation, migration)
1212 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1213 --cfg=tracing/smpi:1 (SMPI interface tracing)
1214 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1215 - examples of examples/msg/tracing updated
1217 - DAXLoader and DOTLoader functions can generate tasks with categories
1218 - A new function to attribute a category to SD tasks:
1219 TRACE_sd_set_task_category (SD_task_t task, char *category)
1220 * Tracing the MPI interface implemented by SMPI
1221 - Collective operations are traced with states
1222 - Point-to-Point operations are traced with states/links
1223 - Tracing activated by a parameter "-trace filename" passed
1224 to smpirun during execution (considering that simgrid
1225 is compiled with tracing enabled)
1226 - To run the simulation with gdb, the simulator
1227 accepts --cfg=tracing/smpi:1 to trace SMPI
1228 - tesh files to check if smpi tracing is ok
1229 - See examples/smpi/NAS/DT-trace
1230 * GTNetS tracing re-worked
1231 - adaptation to the tracing system of GTNets to cope
1232 with modifications regarding the fullduplex mode
1233 - new tesh files to check if gtnets tracing is ok
1236 * Asynchronous communications through the functions:
1237 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1238 * New function: MSG_load_platform_script()
1239 to make possible using a lua script instead of XML files to set up platforms
1240 * New function: MSG_set_function
1241 to associate functions to processes, used when bypassing the parser
1242 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1244 Platforms: Add some more examples in examples/platforms
1245 * Grid'5000: see www.grid5000.fr
1246 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1249 * Change the XML format. This is a very important modification. SimGrid
1250 3.5 introduces a new hierarchical format based on the notion of
1251 Autonomous Systems. Compatibility with old format is ensured
1252 through the perl script provided in the install bin directory
1253 bin/simgrid_update_xml.
1254 It is now possible to build platforms with specific routing
1255 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1256 connect several platforms together. We will try to provide soon
1257 set of realistic platforms exploiting these properties (have a
1258 look at examples/platforms/ for the moment).
1259 * Take the opportunity of the XML format change to be a good XML citizen:
1260 rename link:ctn to link_ctn and similar changes (also dealt with by
1262 * Add a new routing scheme (rule-based) using regular expressions. It
1263 enables to have an extremely low memory footprint when the
1264 underlying routing is simple and can be compactly described. You
1265 need to have libpcre4-dev (perl regular expressions) installed if
1266 you want to use this routing scheme.
1267 * Revive the cluster TAG and allow to easily and efficiently (both in
1268 term of memory and speed) connect clusters together. Have a look
1269 at teshsuite/simdag/platforms/ to see how this can be done. With
1270 this tag, you can create clusters with thousands of tasks at no
1271 cost (have a look at examples/platforms/).
1272 Note: clusters are implemented as ASes, so there is no need for an
1273 enclosing AS tag if you have only one cluster in your platform.
1274 * Add new generic functions in the public interface that allows the user
1275 to call SURF 'create_resource' methods from your code (same
1276 functionality as the XML bypass mechanism but with a much lighter
1278 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1279 that uses a piecewise linear approximation to produce better
1280 results when exchanging small messages.
1281 * Add a new parameter to handle correctly full duplex link and account
1282 for interferences between uplink and downlink communications
1283 (activate with --cfg=fullduplex:1).
1286 * Rename the SD_READY (all dependencies are satisfied and task is
1287 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1288 dependencies are satisfied) state.
1289 This prevents a confusion between the notion of "ready to schedule"
1290 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1291 simulated" (SD_RUNNABLE) used by the simulation kernel.
1292 * Change the way a task is considered as ready. Instead of removing
1293 dependencies when a task is done, a counter is decreased. This way,
1294 it is always possible to reach ancestors thanks to the
1295 SD_taks_get_parents function (even after the end of the simulation.)
1296 * Change the return type of SD_Simulate from (SD_task_t*) into
1297 xbt_dynar_t. This function was in handling a dynar internally and
1298 converted it into a NULL terminated array for historical reasons.
1299 * New function SD_dotload(char*) to load a DAG described in dot
1300 format. This loader and the corresponding examples require the
1301 installation of the graphviz library.
1302 * Fix a bug in the management of tasks of size 0 in the surf network
1303 models. This problem was only visible with SIMDAG and you should
1304 thus disregard results produced with earlier versions if you
1305 relied on this feature (some tasks were blocked because of this).
1306 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1307 even though your applications were doing only point-to-point
1308 communications and sequential computations. Now you can really use any
1309 model you want (of course, if you create real parallel tasks, which are
1310 not implemented in most models beside ptaskL07, this will abort).
1311 * Add an example that schedules a DAX on an heterogeneous platform
1312 using a Min-Min strategy.
1313 * New function SD_workstation_get_current_task() that returns the kind
1314 of task currently running on a workstation in the sequential access
1316 * Raise some warnings when unexecuted tasks remains at the end of the
1317 simulation. This is usually caused by cycles in the DAG.
1320 * New function: SIMIX_process_set_function() called by MSG_set_function
1321 * Change the underlying waiting queue in semaphores so that a process
1322 can wait on several of them simultaneously (as in waitany).
1323 * Fix the way to handle tokens in semaphores so that all access patterns
1324 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1325 * kill the dirty pimple SIMIX_message_sizes_output()
1326 Please use (proper) visualization instead
1329 * New data container: setset (set of sets of elements)
1330 * New module: mmalloc (mapped malloc, allowing to have several
1331 independent segments of malloc)
1332 * New function: xbt_dict_cursor_set_data()
1333 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1334 * New function: xbt_dynar_is_empty()
1335 * New function: xbt_fifo_get_last_item()
1336 * Fix xbt_dynar_shrink(): use the right element size.
1337 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1338 uninitialized areas during expand.
1339 * Fix semaphores: previous implementation was severely broken.
1340 * Use library init/fini functions for our initialization.
1341 - you can use logs and other feature as soon as you want in your
1342 code (even before the xbt_init / MSG_init)
1343 - xbt_exit is now a no-op and produce a warning when used.
1346 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1347 benefit from the latest improvement to the simulation kernel.
1348 * Kill measurement sockets for now. If you rely on them, sorry. This
1349 release is not for you. This feature will be reintroduced in the
1350 future, but we cannot delay the release any further.
1351 * New function: gras_msgtype_get_name().
1352 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1353 * Fix (at last) the pmm example: it should not randomly fail anymore.
1355 Build chain: bug fixes and overall polishing
1356 * Cmake is now stable enough. Hence, we killed the autotools.
1357 * Port to windows ( TM :)
1358 * Fix the 'make install' target.
1359 No need to use 'make install-simgrid' anymore
1360 * Introduce a 'make dist' target compiling a *source* archive
1361 'make package' compiles a binary archive
1362 * Compile java files only on need
1363 * Add --cd and --setenv command line options to tesh
1364 * Out of source builds are not fully supported yet, but we are close
1365 * Enable supernovae and optimization flags by default for our users
1368 * Add layer to set up environment directly from lua, without XML.
1369 * The effect of gras_stub_generator can be achieved through
1370 lua too (check examples/gras/console/ping_generator.lua)
1372 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1374 SimGrid (3.4.1) stable; urgency=low
1376 The "Polishing easter eggs is probably a good idea" release.
1377 This is a bug fixes release only.
1381 * Fix a bug preventing the tasks from begin garbage collected.
1384 * Fix a bug occuring when a host involved in a communication fails.
1385 This was not detected properly by the other peer involved in the
1386 communication. Now, it's reported as a network error.
1389 * Warn the user about loop dependencies in data flow of DAX files
1390 * Obey the control-flow dependencies of DAX files
1393 * Add option "enable_smpi" allowing to not compile SMPI.
1394 Probably useful for the (Mac) users experiencing a build error here
1395 * Improve the detection of lua5.1 and ruby1.8
1397 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1399 SimGrid (3.4) stable; urgency=low
1401 The "Easter in Cargese" release. Also known as (major changes):
1403 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1404 ~> bindings were greatly improved
1405 ~> new tracing infrastructure for better visualization introduced
1407 * the "Welcome to configury modernity" release.
1408 ~> we switched from autotools to cmake, and improved our cdash
1411 A more detailled list of changes follow (full detail in svn log).
1413 Java Bindings: Various Cleanups
1414 * (install java-gcj-compat-dev on debian-like to use them)
1415 * Remove put/get: no need to export deprecated interface in Java
1416 Use send/receive instead.
1417 * Cleanup the examples and add a README per directory
1418 * Remove example autoDestination (that's the only way to go now)
1419 * Remove example explicitDestination (was a plain copy of basic)
1420 * Make JniException a runtime exception, so that there is no need to
1421 declare the fact that you may encounter such a beast. I guess that
1422 nobody will ever want to survive such error.
1423 * Create specific errors for each MSG case of failure:
1424 host failure, transfer failure, timeout, task cancelled
1425 * Cleanup the exceptions that may get thrown by each function
1426 * Other internal cleanups in Java bindings. Performance still bad :/
1427 Ruby and Lua Bindings: create them
1428 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1429 * That's new and great, you should try them out.
1430 Same functionalities than Java bindings, only even less polished
1432 * Kill the useless "rate" argument of SD_task_get_execution_time()
1433 Everyone used to provide -1 as a value, it was not used, and the
1434 semantic of a possible use wasn't even clear.
1435 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1436 either as comm costs or compute costs to mean that there is no
1437 such thing for that specific task.
1438 * Add a SD_task_set_name() function
1439 * Fix SD_task_unschedule() on typed tasks
1440 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1441 * In DAX loader, accept useless 'level' attributes to <job> since
1442 LIGO DAGs have them (seem to be to ease graphical representation).
1444 * Add an example masterslave_mailbox.c using send/receive and not
1445 the deprecated put/get interface.
1446 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1447 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1449 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1450 (declare MSG_USE_DEPRECATED to still have the old name)
1451 * Add a MSG_task_set_data() function
1452 * About trace replay (see examples/msg/actions):
1454 - Allow to work with splitted trace files for each process
1455 Give the specific trace file as argument of each process,
1456 and call MSG_action_trace_run(NULL)
1457 You can still have one merged file for all processes.
1458 - Fix implementation of collective operations
1459 * Allow task_execute() on 0-sized tasks (closes #10063)
1461 * This is the first release of SimGrid where SMPI is not considered
1462 beta anymore (even if some corners should still be improved)
1463 * Port over the new SIMIX_network submodule (internal refactoring)
1464 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1465 * Implement more missing elements of the standard:
1467 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1469 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1470 MPI_Scan() MPI_Get_processor_name()
1471 - Added implementation of missing case for Alltoall (warning: it's
1472 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1473 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1474 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1475 SMPI_MPI_Allgatherv()
1476 * Bug fixes include:
1477 - MPI_Waitsome() was broken
1478 - Allow relative includes in smpicc
1479 - Command line cfg argument 'reference_speed' was ignored...
1480 - Some functions did not properly lead to auto-benching of user code
1481 - smpicc passes -O2 by default (just like openmpi one)
1483 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1484 * Bug fixes about timeouts during communications
1485 * add SIMIX_message_sizes_output() as a pimple to write to file the
1486 amount of messages per size. Use gnuplot to get histogram.
1487 Pimple because that's the only user-visible function of simix,
1488 defined directly in xbt.h (irk, sorry)
1490 - Add a SIMIX_sem_get_capacity() function
1491 - Fix interactions with processe resume/suspende
1492 - release_forever() was stupidly broken
1493 - Fix SIMIX_display_process_status() for processes in a semaphore
1494 - Make SIMIX_sem_block_onto() user-visible
1495 * Refactoring context stuff:
1496 - Use pseudo-OOP for better modularity
1497 - reimplement SIMIX_process_kill() without process_schedule() so
1498 that the latter can take as invariant that it is called from
1500 - Merge context_start into context_new for sake of simplicity
1502 * Add a Vivaldi network model, coded live during SUD'10 ;)
1503 * Rename configuration variables to start a hierarchy:
1504 o cpu_model -> cpu/model
1505 o network_model -> network/model
1506 o workstation_model -> workstation/model
1507 * New configuration variables:
1508 o network/bandwidth_factor: correction to bandwith
1509 o network/latency_factor: correction to latency
1510 o netwotk/weight_S: correction to the weight of competing streams
1511 * Add a long description to the models, that users can see with such
1512 argument on the command line: --cfg=cpu/model:help
1513 * --help-models display the long description of all known models
1515 * config: add the ability to set a default value after registration
1516 Does not override any previously set value (e.g. from cmd line)
1517 * dict: allow to have integer key and data.
1518 When so, you need to use the following functions
1519 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1520 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1521 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1522 In contrary to regular dicts, the key is not malloced before copy.
1523 Mixing scalar and regular elements in the same dict is not tested
1525 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1526 Tracing for Visualization:
1527 * SimGrid is now instrumented in order to generate a trace file for
1528 visualization analysis: to use it, need to compile SimGrid with the
1529 "tracing" option enabled, and instrument the program using SimGrid with
1530 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1531 (among other functions).
1532 * The instrumentation only traces the platform utilization for now
1533 * Documentation to use the tracing functions and how to analyze the
1534 traces with the Triva tool is written.
1535 * More information about: SimGrid FAQ (in the section Tracing Simulations
1538 * We moved to cmake as default build system. Autotools support will
1539 be dropped soon. Check the FAQ for more info about how to use it.
1540 * Greatly improved our cdash/ctest interactions
1541 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1542 * Added memory checking tests with valgrind; lot of memleak fixing.
1543 This may be the first release of simgrid with so few memory issues
1544 * Added code coverage tests.
1545 Our coverage is still improvable, but at least we see it on cdash.
1547 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1549 SimGrid (3.3.4) stable; urgency=low
1551 The "Desktop Grid needs love too" release (also called Xmas release).
1553 Models improvements:
1554 * Major speedup in the maxmin system solving by using lazy evaluation
1555 Instead of solving completely the maxmin system at each iteration,
1556 only invalidate (and recompute) the modified parts.
1557 This new feature is enabled in default models but you can try to
1558 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1559 * Cas01 IMproved as default CPU model
1560 This CPU model is the same Cas01 model, but it uses the
1561 maxmin-selective-update flag and a heap structure to manage
1562 actions on SURF kernel.
1563 It reduces the complexity to find the next action to finish and,
1564 consequently, it's faster than the old Cas01.
1565 This is the new default CPU model (Cas01).
1566 * Rename the old Cas01 model to Cas01_fullupdate
1567 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1568 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1569 * CpuTI (CPU Trace Integration)
1570 A new CPU model whose objective is simulate faster when using
1571 availability trace files.
1572 Instead of using a full featured, over engineered maxmin system for
1573 CPU modeling, this model does the pre-integration of traces files
1574 to calculate the amount of CPU power available, and so, executes
1575 faster than the old CPU models.
1576 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1577 * Use LV08 as default network model since it gives better accuracy
1578 for small messages and shouldn't change things for big ones.
1579 Use --cfg=network_model:CM02 to get the previous behavior.
1582 ******************************************
1583 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1584 ******************************************
1585 * The new CPU model may changes simulations!
1586 The point is that events occurring at the exact same timestamp
1587 are not scheduled in the same order with the old and new
1588 version. This may be enough to completely change the execution
1589 of simulations in some cases.
1590 * The new network model will change simulations!
1591 This new model is more realistic than the previous one, so you
1592 should consider redoing your old experiments with this model.
1593 Sorry for the inconvenience.
1596 * Introduce the supernovae compilation mode
1597 When compiled that way, the whole SimGrid (or almost) is put in a
1598 single compilation unit and compiled in one shoot.
1599 This is to help gcc which has difficulties to inline stuff from one
1601 The speedup seem to be above 15%, althrough more tests are needed on
1602 amd64 to confirm that gain.
1605 * Port of MSG's mailbox on top of SIMIX network
1606 The put/get mechanism was greatly simplified on the way.
1609 * New SIMIX network module. Provides:
1610 - Mailbox: rendez-vous mechanism to find with who you want to speak
1611 - Synchronous send/recv: easier and hopefully faster since the
1612 logic is handled in the maestro process directly now
1613 - Asynchronous send/recv: you dreamt of it? It's here now
1614 Too bad that nobody cared enough to propagate the change to MSG.
1615 * Add semaphores as SIMIX synchronization mechanism.
1618 * new function SD_daxload(char*) to load a DAX file
1619 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1620 * Introduce typed tasks. Specify its kind and cost at creation.
1621 At scheduling, just give where it should be placed, and the cost
1622 for each involved resource is automatically computed.
1623 Existing constructors so far (more to come of course):
1624 - SD_task_create_comm_e2e() for end-to-end communication
1625 - SD_task_create_comp_seq() for sequential computation
1626 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1627 * new function SD_task_dump() for debuging display
1628 * new function SD_task_dotty(task,FILE*) writing to file the info
1629 about the task in dotty format
1630 * SD_task_dependency_exists() can now cope with having one of its
1631 arguments NULL. If so, it tests whether the other argument has any
1633 * Add getters on list of preceding/following tasks:
1634 SD_task_get_parents(task) and SD_task_get_children(task)
1635 * Add getters on amount of workstations and list:
1636 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1637 * Add getter on task kind: SD_task_get_kind(task)
1638 * Update the start_time and finish_time of tasks on completion/failure
1639 * Bugfix: Remove task from state swags when destroyed
1642 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1645 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1646 dynar with one separate thread per value of the dynar.
1647 * Change the prototype of xbt_thread_create(), sorry.
1648 Added a boolean parameter indicating whether we want to join this
1649 thread (used in SG only for now)
1650 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1653 * GTNetS wrappers should now be usable again (and betterly tested too)
1654 * Fix a major regression from 3.2 where the timeout provided to
1655 MSG_task_put_with_timeout() was used as absolute time before which
1656 the comm should be done.
1657 * Start to fix the <cluster> tag.
1658 - Internal links should be good now (beside of the loopback, which
1659 use the private link instead)
1660 - paths to the external world is still rather broken
1661 - the <route:multi> tag is just broken. Actually that's brain-dead.
1662 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1663 to make it less stupid
1664 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1665 * Fix a source-level compatibility glitch from 3.2: after defining
1666 MSG_USE_DEPRECATED, you can use the old name
1667 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1668 * Allow to compile from the SVN with automake 1.11
1669 * Fix some problems when using the "start_time" tag in deployment XMLs.
1670 * Fix #8569: XBT/synchro.h has redundant declarations
1671 * Fix #8563: MSG return values and exceptions
1672 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1673 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1674 Thanks to Robson Peixoto.
1675 * Fix a vicious bug in dictionaries inducing that some elements were
1676 not freed on xbt_dict_free()
1678 Portability report of this version:
1679 * Main portability targets:
1680 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1681 on (amd64/i386/ia64)
1682 - mac leopard on i386
1683 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1684 but nothing critical.
1685 * Other platforms: windows, AIX and others were not tested for this release
1687 Timing report of this version:
1688 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1689 scenario parameters). From 8h to a few seconds in desktop grid settings.
1690 * Supernovae brings about 25% speedup on i386.
1692 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1694 SimGrid (3.3.3) stable; urgency=low
1696 The "Need for Speed" release.
1698 The timings done to validate the 3.3.2 were faulty.
1699 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1701 The problem was a conversion from a manually handled vector to
1702 xbt_dynar_t on the critical path.
1703 xbt_dynar_foreach calls functions, inducing stack management crap.
1705 We inlined these functions and xbt_dynar_foreach is now breath taking.
1706 We also inlined xbt_swag_belong on the way.
1708 Here are some approximate speedup measurements (on master/slaves
1709 simulations lasting between 10s and 20s each):
1710 3.3.1 -> 3.3.2: about same performance
1711 3.3.2 -> 3.3.3: 40% speedup
1712 3.3.1 -> 3.3.3: 40% speedup
1713 3.3.1 with inline patch -> 3.3.3: 30% speedup
1715 Our reading is that the refactoring which occurred in 3.3.2 made us
1716 suffer much more from the xbt_dynar_foreach low performance, but
1717 once we solved this, this refactoring proved to be very performance
1718 effective. From the 40% speedup, somehow, 10% are due to the
1719 inlining and 30% to the refactoring.
1721 That's a pitty that gcc cannot inline functions placed in other files
1722 alone. We have to choose between:
1723 - break the encapsulation (by putting private data structures and
1724 accessors in headers files to help gcc)
1725 - live with low performance
1726 - switch to a decent compiler such as icc (not quite possible).
1728 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1730 SimGrid (3.3.2) stable; urgency=low
1732 The "Simplicity does not preceed complexity, but follows it" release.
1734 The main contributors of this release were (lexical order):
1735 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1738 * Extract the routing logic into its own object.
1739 (was dupplicated in network.c and workstation_LV07.c;
1740 Allows to implement other ways of storing that info)
1741 => kill now useless network_card concept
1742 - Use dynar to represent routes (instead of void** + int*)
1743 - kill link_set (use surf_network_model->resource_set instead)
1744 - Add a command-line option to choose the routing schema to use
1745 - Add three new models:
1746 * Floyd (shortest path computed at initialization)
1747 * Dijikstra (shortest path recomputed all the time)
1748 * Cached Dijikstra (shortest path computed on need)
1749 All these models where contributed by Silas De Munck, and are
1750 described in his ICCS09 paper.
1752 * Simplify model declaration
1753 (less redirections, less function to write when defining a model)
1754 - Factorize stuff between models:
1757 surf_model_resource_set(model)
1758 surf_model_resource_by_name(model, name)
1759 - Unify the types of models in s_surf_model_t (using an union)
1760 - Embeed fields of common_public directly into s_surf_model_t
1761 - Rename model methods:
1762 action_free ~> action_unref
1763 action_change_state ~> action_state_set
1764 action_get_state ~> action_state_get
1765 - Change model methods into functions :
1766 (model)->common_public->action_use ~> surf_action_ref
1768 * Implement a generic resource; use it as ancestor to specific ones
1769 (allows to kill duplicated code in models)
1770 Drawback: timer command don't need no name nor properties;
1771 workstation_CLM03 don't need no properties
1772 (but I guess we can live with those few bytes wasted)
1774 * Improve the action object model
1775 - implement a constructor avoiding dupplicated code about field
1776 initialization in generic_action part.
1778 * Kill the SDP model: it has an external dependency, is deprecated
1779 in flavor of modern lmm models, and didn't compile since a while
1782 * Relocation of the context module from XBT to SIMIX.
1783 (the context were decoupled from the simix processes, duplicating a lot of code)
1784 => a lot of code was factorized
1785 - less overhead is introduced during scheduling
1786 - simpler API for the context factory
1787 - the logic for process creation,destruction and manipulation was simplified
1788 * Simplification of the s_smx_process_t data structure.
1789 => accesing the simix level data associated to a process is faster now,
1790 and the code is a lot more readable.
1793 * Implement some more MPI primitives:
1794 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1795 -implementation: Bcast: flat or 2-ary tree (default),
1796 Barrier: 4-ary tree,
1798 Allreduce: Reduce then Bcast
1799 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1800 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1801 Alltoallv: flat tree, like ompi
1803 * Add support for optimized collectives (Bcast is now binomial by default)
1804 * Port smpirun and smpicc to OS X
1807 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1808 Shout out if you used it.
1811 * Display the list of still queued messages in SG mode when existing
1815 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1816 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1817 * Add xbt_str_from_file(FILE*)
1818 * Add xbt_dict_get_key achieving a linear reverse search
1819 * Remove the context module
1821 Portability report of this version:
1822 * Main portability targets:
1823 - Linux(debian)/x86/context
1824 - Linux(debian)/x86/pthreads
1825 - Linux(debian)/amd64/context
1826 - Linux(debian)/amd64/pthreads
1827 On these, we still have the eratic breakages of gras/pmm and
1828 amok/saturate_sg reported in previous version. We still think
1829 that the tests are the cause of the fault, not the tested code.
1831 - Mac OSX Leopard/x86/context
1832 Still false negative in tesh autotesting.
1833 Smpi still fails, but this time because readlink does not accept -f
1834 Everything seems to work properly beside of that.
1837 - AIX version 5.3 (only tested contexts this time)
1838 Smpi still fails there because mktemp is not installed.
1839 Everything seems to work properly beside of that.
1841 I managed to compile it for the first time, but several breakages.
1842 Won't delay the release for this exotic platform.
1844 * Windows: it's still lagging behind. If you want to help, please
1847 Timing report of this version:
1848 This version seem to be more than 5% faster than 3.3.1 (on linux
1849 64bits with contextes). The gain is less than expected, we are
1850 investigating this for next release.
1852 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1854 SimGrid (3.3.1) stable; urgency=low
1857 * Implement a --cfg-help to show existing configuration variables
1858 * Build chain do not require doxygen in maintainer mode
1861 * fix a bug on struct sizeof computation, which prevented the
1862 exchange of arrays of structs in some conditions
1863 - added a regression test about this in datadesc_usage
1864 * Allow the exchange of 0-long dynamic vectors.
1865 - for that, use -1 as indicator of dynamic size instead of 0
1866 - This implied to change any size from unsigned long to long,
1867 reducing a bit communication abilities, but I guess that with
1868 64bits being quite common, this is more than enough.
1869 - This also induce a protocol change, thus bumping network protocol
1870 version from 0 to 1 (if we have external users, we have to get
1871 clean on that point too ;)
1872 - added two regression tests about this in datadesc_usage
1873 * Be more verbose when propagating local exceptions
1874 This helps debugging.
1875 * Display the status of simulated processes when receiving SIGINT in
1879 * Allow to control the simulation from a trace file.
1880 New functions MSG_action_register() and MSG_action_trace_run()
1881 The first one allows to associate a function execution to each
1882 kind of action while the second one parses a trace file and
1883 triggers the corresponding actions within the system.
1884 For now, only a toy example is provided in examples/msg/actions
1885 * Add an exemple of process migration in examples/msg/migration
1886 * Fix a bug in task exchange which broke MSG_task_get_sender()
1887 Add a teshsuite regression test for that.
1888 [Bug: if MSG_task_get_sender() is called after sender exit,
1890 * Fix a bug which prevented suspend/resume to work properly
1891 * Display the status of simulated processes when receiving SIGINT
1892 This fixes a regression of v3.3. due to the introduction of SIMIX
1893 * Bug fixing in failure management:
1894 - trace could not start by a failure at time 0
1895 - failure during communications were not working
1898 * Add SIMIX_process_set_name() to change the name of the current
1899 process in the log messages.
1900 * Store smx_hosts in a dict since we only retrieve them by name
1901 * Move the configuration infrastructure to surf
1904 * Move the configuration infrastructure to surf
1907 * Massive internal cleanups:
1908 - Store internal structures on processes instead of hosts (allows
1909 to have more than one process per host, in addition of being more
1911 - Cleanup the initialization/finalization process
1912 - Kill a whole bunch of unneeded synchronization:
1913 processes run in exclusive manner within the simulator
1914 - Move queues from global tables to process data fields
1916 - now accept -platform and -hostfile arguments
1917 - Pass the right rank value to processes according to the hostfile
1918 * Compile the examples by default, and use them as regression tests
1919 * Implement MPI_Wtime()
1920 * Change the reference speed to a command line option
1923 * TCP_gamma can now be specified as command line option using
1924 --cfg=TCP_gamma:10000000.0
1925 * Change the --surf-path cmd line option into --cfg=path:
1928 * Also include strbuff from xbt.h public header
1929 * xbt_ex_display(): do not free the exception after displaying
1930 This allows to do more with the given exception afterward.
1931 Users should call xbt_ex_free() themselves.
1935 Portability report of this version:
1936 * Main portability targets:
1937 - Linux(debian)/x86/context
1938 - Linux(debian)/x86/pthreads
1939 - Linux(debian)/amd64/context
1940 - Linux(debian)/amd64/pthreads
1941 These targets fail about 1/10 of times on gras/pmm, but we believe
1942 that this is because of the test, not because of simgrid.
1943 amok/saturate_sg fails even more rarely, and the test may not be
1946 - Mac OSX Leopard/x86/context
1947 The test suite still spits tons of errors because some obscure
1948 force prevents us from removing the temporary directories
1949 arguing that they still contain some metadata I've never heard of.
1950 Smpi fails because seq is not installed.
1951 Everything seems to work properly beside of that.
1954 - AIX version 5.3 (both contexts and pthread)
1955 Smpi still fails there because mktemp is not installed.
1956 XML inclusions seems rosty on AIX.
1958 * Windows: it's still lagging behind. If you want to help, please
1961 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1963 SimGrid (3.3) stable; urgency=high
1967 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1968 [Malek Cherier & Mt]
1970 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1973 Until now, GRAS were using MSG as an interface to SURF. It was
1974 quite difficult because both interface have several differences
1975 (MSG channels vs GRAS sockets were the most notable point).
1977 This also opens the gate to SMPI (which should occur soon) and speed
1978 up simulations by to 40% (even if it were not the main goal).
1980 **************************************
1981 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1982 **************************************
1983 The point is that events occuring at the exact same timestamp are
1984 not scheduled in the same order with the old and new version. This
1985 may be enough to completely change the execution of simulations in
1986 some cases. Sorry for the inconvenience.
1988 * Cleanup and upgrade the XML format to push further scalability
1989 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1991 * Improve the testing infrastructure with tesh. Now a very large part of
1992 the code is tested not only by being run but also by checking that the
1993 output match an expected output [Mt].
1995 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1996 is really less memory-demanding, which should allow you to use
1997 larger files in SimGrid [AL].
1999 * Inform valgrind about our contextes, so that it becomes usable
2000 with the default (and more effecient) version of SimGrid
2001 [contributed by Sékou Diakite, many thanks]
2004 * Introduce a listener thread in charge of receiving incoming
2005 messages from the network. It allows to overlap communication and
2006 computation but most notably, it removes some stupid deadlocks due
2007 to the fact that so far, a process could not send and receive at
2008 the same time. This made most non trivial communication schema
2010 * Convert the PIDs from long int to int to match the MSG ones (and
2011 linux ones too) [Mt]
2012 * New function: gras_agent_spawn() to launch a new process on
2013 current host. Only working in simulation for now. [Mt]
2014 * New function: gras_os_hostport() returning a constant form (ie,
2015 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2018 * Make the backtrace of exceptions more human readable [Mt]
2019 * New module: xbt/str [Mt]
2020 a ton of string utility functions (split, join, printf to a newly
2021 allocated buffer, trim, etc)
2022 * New module: xbt/hash [Mt]
2023 SHA1 hashing algorithm (more to come if needed)
2024 * New module: xbt/synchro [Mt]
2025 synchronization tools (mutex and conditions) working the same way
2026 in simulation and in real life (mainly useful for GRAS, but not
2028 * New module: xbt/queue [Mt]
2029 classical producer/consumer synchronization scheme
2030 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2031 (using the classical functions will get serialized) [Mt]
2032 * Make dictionary internal table dynamic. No need to specify its size
2033 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2035 * Make sure the log channels are organized as a tree under windows
2036 (because of ANSI C compatibility issue, any channel were child of
2037 root directly) [Mt].
2040 * Cleaned many thing in surf and fixed a few bugs [AL].
2041 * Add a nice command line configuration mechanism to compose models [AL].
2042 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2043 the previous one (KCCFLN05). It relies on something that looks like
2044 a max-min sharing mechanism but cannot be written as such. A new solver
2045 was thus designed [AL].
2046 * Add a new solver to lmm. Based on Lagrange optimization and
2047 gradient-based descent, it enables to efficiently maximise systems s.a
2049 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2051 This solver enables to propose two new network models for TCP Reno and
2052 TCP Vegas based on Low's work. These models still need to be fully
2053 tested though [Pedro Velho].
2056 * Bug fix in SD_simulate. Now the time bound given as argument is
2058 * Use the new parallel task model (ptask_L07) as default.
2059 * Use the SURF command line configuration mechanism.
2060 * 0-size tasks (for synchronization) should now work.
2062 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2064 SimGrid (3.2) stable; urgency=high
2068 We still experience issues on this platform, but we believe that at
2069 least MSG is usable.
2071 GRAS API BREAKAGE (for simplification purpose, sorry):
2072 * the gras_msgtype_by_name is not used anymore. Instead of
2073 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2074 you can write (and must)
2075 gras_msg_send(toserver, "request", &request);
2076 - If you still want to pass a gras_msgtype_t to the function (to cache
2077 the type and avoid the lookup time), use the gras_msg_send_() variant.
2078 - Impacted functions:
2079 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2080 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2081 * The callbacks are now expected to return 0 when everything went well
2082 (just like the main() function)
2084 GRAS new features and improvements:
2085 * New module mechanism where user code can use per process globals [Mt]
2086 This is similar to gras_userdata_*() functions, but for libraries. It
2087 factorize some code developped over and over in the examples and AMOK.
2088 It has still to be documented and used (only amok/peermanagement is
2090 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2091 loss when they were small enough to fit into the buffer and sent quickly
2092 enough so that they can all get received in one shoot.
2093 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2094 if not found. Use the *_or_null() variant for the old semantic.
2095 * In gras_msg_handle, do not discard messages without callback.
2096 They are probably messages to be explicitly awaited later (ie, proofs of
2097 mis-synchronization in userland since they are sent before being awaited)
2098 No big deal usually.
2099 * gras_socket_meas_send/recv: semantic changed!
2100 The numerical arguments used to be (1) the total amount of data to send
2101 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2102 messages. This was need for the fool willing to send more than MAXINT
2103 bytes on quite fat pipes.
2106 * Do really rename the hostmanagement module to peermanagement. [Mt]
2107 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2108 but this is rather new and this was documented in the module
2109 documentation (poor excuses, I admit)
2110 * Bandwidth measurement semantic changed! This follows the changes to
2111 gras_socket_meas_send/recv explained above.
2114 * A sequential mode has been added to the workstations. When a workstation
2115 is in sequential mode, it can execute only one task, and the other tasks
2116 are waiting in a FIFO. [Christophe Thiery]
2119 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2120 model for SIMDAG. [Christophe Thiery]
2121 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2122 equal to 0, which caused some bad behaviors in
2123 saturated_constraint_set_update. I now use a threshold mechanism like in
2127 * When running manually src/testall, you select specific units [Mt]
2128 testall is the result of our cunit mechanism, and should replace all
2129 the scripty thingy around since bash don't run easily on billware.
2131 * A mallocator system has been added. [Christophe Thiery]
2132 Mallocators allow you to recycle your unused objects instead of freeing them
2133 and allocating new ones.
2135 Documentation update:
2136 * FAQ reworking + New FAQs:
2137 - "Valgrind spits tons of errors!" [Mt]
2138 - "How to repport bugs" [Mt]
2139 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2140 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2141 - Communication time measurement within MSG [AL]
2142 - I experience weird communication times when I change the latency [AL]
2143 * GRAS tutorial [Mt]
2145 - an introduction to the framework and to the used communication model
2146 - an initiatic tour introducing the most proheminent features:
2148 . Lesson 0: Installing GRAS
2149 . Lesson 1: Setting up your own project
2150 o Part 2: Message passing
2151 . Lesson 2: Exchanging simple messages
2152 . Lesson 3: Passing arguments to the processes (in SG)
2153 . Lesson 4: Attaching callbacks to messages
2154 . Lesson 5: Using globals in processes
2155 . Lesson 6: Logging information properly
2156 . Lesson 7: Using internal timers
2157 . Lesson 8: Handling errors through exceptions
2158 . Lesson 9: Exchanging simple data
2159 . Lesson 10: Remote Procedure Calling (RPC)
2160 . Lesson 11: Explicitely waiting for messages
2161 . Recapping of message passing features in GRAS
2162 - A HOWTO section containing:
2163 o HOWTO design a GRAS application
2164 More are due, of course. They will come latter. In the meanwhile, you can
2165 check the examples which are still here.
2167 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2169 SimGrid (3.1) stable; urgency=high
2173 There was a stack corruption somewhere, visible only when optimizing
2174 with these versions. [Vince]
2177 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2178 module that enabled to play with Directed Acyclic Graphs. It is built
2179 directly on top of SURF and provides an API rather close to the old
2180 SG. Some old codes using SG are currently under rewrite to check that
2181 all needful functions are provided. [Christophe Thiery]
2184 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2185 extension of the classical CLM03 model that gracefully handles
2186 failures. This is now the default model for MSG and GRAS. It doesn't
2187 handle parallel tasks yet though. [AL]
2188 * Bug fix: Weights were not correctly set in the network part.
2189 WARNING: This may have resulted in incorrect results with simulations
2190 where there are more than one flow on a given link. [AL]
2193 * After a (long ?) discussion on simgrid-devel, we have decided that the
2194 convention we had on units was stupid. That is why it has been decided
2195 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2196 WARNING : This means that all previous platform files will not work as
2197 such with this version! A warning is issued to ask users to update
2199 A conversion script can be found in the contrib module of the CVS, under
2200 the name contrib/platform_generation/surfxml_update.pl [MQ]
2203 * Bug fix: Processes were started in reverse order, wrt deployment file.
2204 WARNING: if your code relies on this bug, please fix it. [AL]
2205 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2206 being executed on two different locations. [AL]
2207 * Bug fix: Failures are now better supported thanks to Derrick's tests
2208 (there was many failure situations I hadn't thought of and that weren't
2209 correctly handled). [AL]
2210 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2214 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2215 Exception occurring on server-side are propagated back to client (!).
2217 API CHANGE: the callback changed their prototype. Change:
2218 int my_handler(gras_socket_t expeditor, void *payload_data) {
2220 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2221 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2223 * New! function: gras_msg_handleall to deal with all messages arriving
2224 within a given period.
2225 * New! function: gras_socket_server_range to get a server socket in a
2226 range of port numbers (ease to avoid port number conflicts) [MQ]
2227 * New! gras processes display their backtrace when they get a SIGUSR1
2228 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2229 Sweet to debug RL processes [MQ]
2233 - Do not force experiment sizes to be expressed in kb, or it becomes
2234 impossible to measure the latency this way (needs one byte-long tests)
2235 WARNING: this changes the amok_bw_* function semantic. [MQ]
2236 - Implements the link saturation stuff. [MQ]
2237 * Peer management module:
2238 New! module factorizing code that we wrote over and over [MQ].
2241 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2242 - Test units are placed directly into the library code, they get extracted
2243 automatically and placed into the src/testall binary.
2244 - Convert most of the XBT tests to this system.
2245 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2246 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2247 live in the same C code. This required to change a little bit the API
2248 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2249 * New module: sparse graph structure with basic algorithms (this is work
2250 in progress and the API is not considered to be frozen yet). [AL]
2251 * Display more information on backtraces: source line & function names are
2252 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2253 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2254 * Reworked a little bit some #include statements to load only required
2255 headers. Some user code that relied on SimGrid to include stdlib or
2256 stdio may need to include it by themselves. [AL]
2257 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2259 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2260 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2261 but I guess that almost nobody used those parts. [MQ]
2263 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2265 SimGrid (3.0.1) stable; urgency=low
2268 * Unfortunately, I had missed 5 misnamed functions:
2269 xbt_fifo_item_t xbt_fifo_newitem(void);
2270 void xbt_fifo_freeitem(xbt_fifo_item_t);
2271 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2272 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2273 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2274 They're now deprecated. Please use their new versions:
2275 xbt_fifo_item_t xbt_fifo_new_item(void);
2276 void xbt_fifo_free_item(xbt_fifo_item_t);
2277 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2278 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2279 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2281 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2282 * Documentation: xbt_log module unmercifully reworked [MQ]
2283 * Bugfix: there was a problem with the ending of contexts with
2284 the pthread backend. It caused some weird deadlock or behavior
2285 depending on the pthread implementation. [AL]
2286 * Bugfix: get the exceptions raised in the simulator repport where
2287 and why they come from when they are not catched in time [AL, MQ]
2290 * Bugfix: Do repport the error when two non-connected hosts try to
2291 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2294 * Add additionnal checkings on communications. Assert that two
2295 communicating hosts are connected by a set of links... [AL]
2298 * Add additionnal checkings on channel values in communication [AL]
2299 * New: MSG_task_get_source to see on which host a task was generated [HC]
2300 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2301 the number of tasks waiting to be received on channel and sent
2303 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2304 waits for the first task coming from a given host.. [AL]
2306 GRAS new functionnalities: [MQ]
2307 * Enhance the parsing macro to allow the size of multidimentional objects
2308 to be given thru annotations.
2309 * New example (and documentation): Matrix Multiplication a la RPC
2310 (as when I was young!) and fix a bunch of bugs found on the way.
2312 GRAS performance improvements: [MQ]
2314 * Reduce the amount of cbps creation/destruction by making it static to
2315 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2318 * Change libdata to a set so that we can search for stuff by ID (and thus
2319 reduce the insane amount of dict lookups)
2322 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2323 RL communication on top of SG since it's so uneffective.
2324 It may also be used for inter-thread communication in RL, one day.
2325 * Use gras_datadesc_copy() to exchange messages on top of SG
2327 - improve message exchange performance on top of SG
2328 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2329 * Don't exchange on the network the size of the used part of buffer,
2330 instead, specify the possible buffer size to read().
2332 - reduces the amount of read/write calls (one pair per exchange)
2333 - reduces the amount of exchanged data (the size)
2334 - allows to retrieve all arrived data on receiver side, if we don't need
2335 it right now (subsequent read will peek the buffer)
2336 - allows the receiver to proceed with the begining of the stream before
2337 everything is arrived
2338 - make it possible to build an iov transport (using readv/writev)
2340 - take care of the data with non-stable storage (like stacked data),
2342 * If possible, TCP send uses vector I/O (when writev() is here)
2343 - Don't use it for receive since we send data sizes and data on the
2344 same stream, so we wouldn't be able to chain large amount of chunks
2345 before having to flush the stuff to read the size.
2346 * Rework the transport plugin mechanism to simplify it and reduce the
2347 amount of pointer dereferencement when searching for the right function
2350 * I guess that now, we do almost as few system calls as possible while
2351 doing as few data copy as possible.
2353 To improve it further, we could try to send all the sizes first and then
2354 all the data (to use iov on receiving size), but it's only a partial
2355 solution: when you have 2 dimensional data, the sizes of the second
2356 dimension is data of the first dimension, so you need 3 streams.
2358 I'm not sure the potential performance gains justify the coding burden.
2360 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2362 SimGrid (3.00) stable; urgency=high
2365 * New! Give the possibility to hijack the surf parser and thus bypass
2366 MSG_create_environment and MSG_launch_application. Have a look at
2367 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2369 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2371 SimGrid (2.96) unstable; urgency=low
2376 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2377 This deprecates the xbt_error_t mechanisms.
2378 It modifies (simplifies) all XBT and GRAS API.
2379 MSG API keeps unchanged (exceptions raised by XBT are catched from
2380 within MSG and masked with existing error handling facilities)
2383 * New! Add a FATPIPE model. [AL]
2384 * New! Add a parallel task model. [AL]
2385 * New! Add automatically a loopback interface (in the default
2386 network model) if none was precised.
2389 * Bugfix: MSG_process_resume now works with the current running process.
2391 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2392 * Modification of MSG_task_get_compute_duration. Once a task has been
2393 processed, the value returned by this function is now equal to 0. [AL]
2394 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2395 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2396 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2397 of a m_task. For now, MSG_task_cancel only works with computation
2399 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2400 of the processor (in Mflop/s) regardless of the current load on the
2402 * API Change: use proper naming convention for MSG_getClock and
2403 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2405 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2406 This function changes the priority of a computation task. This priority
2407 doesn't affect the transfer rate. A priority of 2 will make a task
2408 receive two times more cpu power than the other ones. This function
2409 has been added to suit the needs of Nguyen The Loc and hasn't been that
2410 much tested yet. So if it fails, please report it and send me your code.
2412 * API Change: removed all functions and types that were marked "deprecated"
2413 since many months. Renamed MSG_global_init_args to MSG_global_init.
2415 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2417 SimGrid (2.95) unstable; urgency=low
2420 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2421 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2422 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2425 * Bugfix: Allow absolute paths for platform description files [MQ]
2426 * Bugfix: do free the variables after use. Leads to drastic performance
2428 * Implement max_duration (ie, timeouts) on resources [AL]
2431 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2433 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2434 the needed support (timeouts) [AL]
2437 * Implement measurement sockets. You can now get the bandwidth between two
2438 hosts thanks to AMOK (see below). [MQ]
2439 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2440 dynar over the network (yeah) [MQ]
2441 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2442 * simple/static token-ring example. [Alexandre Colucci and MQ]
2443 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2444 in place before [MQ]
2447 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2448 type parsing. "short" and "long" modifiers were also ignored; other
2449 modifier (such as reference level) are still ignored. [MQ]
2450 * Embeed the buffer size within the buffer itself on SG. [MQ]
2451 That way, send() are atomic and cannot get intermixed anymore (at least
2452 the ones which are less than 100k; bigger messages still have the issue)
2453 * Array size pushed by the field, not by the field type (or each
2454 and every long int will push stuff to the cbps) [MQ]
2455 * use select() to sleep since it allows to portably sleep less than one
2458 GRAS (minor cleanups)
2459 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2462 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2463 * type category 'ignored' killed as it was never used and were difficult
2465 * whether a type can cycle or not is now a flag, leaving room for more
2466 flags (as "ignored", if we feel the need one day ;)
2467 * Rename raw sockets to measurement sockets since "raw" has another
2468 meaning in networking community.
2471 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2472 and offers features not belonging to GRAS but that most applications
2473 need. One day, it may be a set of plugins loadable at runtime.
2474 * New module: bandwidth
2475 bandwidth measurement between arbitrary nodes running this module. [MQ]
2477 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2479 SimGrid (2.94) unstable; urgency=low
2481 The first beta release of SimGrid 3 !
2485 * Update the main page and the FAQ. Adding references to gforge.
2488 * Add a gras_os_getpid function.
2491 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2492 * Extend the logs so that they also print PID, hostname, date, ... if
2494 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2495 and kill the old version which were in testsuite/
2496 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2500 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2501 for users. Instead, each programming environment (GRAS, MSG,...) use it
2502 when needed to provide such a feature to users.
2503 Don't shortcut the mechanism or you will also shortcut the virtualization
2504 you need on the simulator.
2508 * Cleanups in configury with regard to compile optimization/warning flags.
2509 Also add -fno-loop-optimize to any powerpc since it's the optimization
2510 killing gcc (< 3.4.0).
2511 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2512 needed by MSG examples complications
2513 * Borrow configury beautifications from PHP
2516 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2517 warning (thanks loris for stumbling into this one).
2518 * Bugfix: stop loading private headers (gras_config.h) from the public
2522 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2523 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2524 RL would imply a DNS resolver, which is *hard* to do in a portable way
2525 (and therefore delayed).
2526 * Implement a real timer mechanism and use it in timing macros. This allows
2527 to avoid rounding errors and get a 0.000005 sec precision in timing
2528 macros. While I was at it, various cleanups:
2529 - allow to declare more than one timed section per file (fix a stupid bug)
2530 - move some private declaration to the right place
2531 - merge conditional execution and timing macros into emulation module
2532 - document the module
2533 - make sure the module cleanups its mess on gras_exit
2534 * Documentation improvements:
2535 - (new) how to compile applications using GRAS
2536 - (new) emulation support (timing macros)
2538 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2540 SimGrid (2.93) unstable; urgency=low
2542 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2545 - Use Paje properly where used. Still to be sanitized properly.
2546 - Portability fix: Add an implementation of the contexts using pthread
2550 - Add xbt_procname(): returns the name of the current process.
2551 Use it to show the current process's name in all logging.
2553 - fix detection of older flex version and the reaction, since we do
2554 depend on modern ones (we use lex_destroy)
2555 - Better separation of SG and RL in the libs: remove all simulation code
2556 from libgras. As a result, this lib is now only 200k when stripped.
2557 Some of the xbt modules may also be duplicated (two sets and such) and
2558 should be cleaned/killed before SG3.
2559 - Insist on using xlC on AIX because of weird problems involving gcc there.
2560 - Cleanup the make remote stuff. This is now done by scripts
2561 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2562 mainly for our private use, but we're working on changing them to user
2565 - Bugfix: flush the socket on close only if there is some *output*.
2566 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2568 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2571 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2573 SimGrid (2.92) unstable; urgency=low
2575 Alpha 3 on the path to SimGrid 3
2579 - New! First try of benchmarking macros.
2580 - New! First try so that gras_stub_generator generate deployment and
2581 remote compilation helpers.
2583 - Bugfix: Initialization fix in msg_test.
2587 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2589 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2591 - API Change: gras_os_sleep to take the amount of seconds as a double.
2592 Accepting an int was error prone since it was the only location where
2593 seconds were coded as such. It leaded to damn rounding errors.
2594 - Bugfix: Hard to belive that timers ever worked before this.
2596 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2598 SimGrid (2.91) unstable; urgency=low
2600 Alpha 2 on the path to SimGrid 3
2604 - Bug fix in the lmm_solver.
2606 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2607 through the function MSG_paje_output.
2608 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2609 - It is possible to bound the rate of a communication in MSG with
2610 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2611 to write it in the changelog).
2612 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2617 - Various cleanups to the autotools stuff
2618 - Begin to move Gras examples to examples/gras/
2619 - Let make distcheck work again (yeah!)
2621 - documentation overhauled using doxygen.
2622 gtk-doc-tools is dead in SimGrid now.
2623 - Automatically extract all existing logging categories, and add the list
2624 to the documentation (long standing one, to say the less)
2626 - Cleanup the known architecture table. Reorder the entries to group what
2627 should be, and use a more consistent naming scheme.
2628 (some of the test dataset are still to be regenerated)
2629 - New! Allow library to register globals on each process just as userdata
2631 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2632 lookup time (for now).
2633 Use it in msg and trp.
2634 This cleans a lot the internals and helps enforcing privacy of the
2635 headers between the gras components.
2636 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2637 - Bugfix: gras_os_time was delirious in RL.
2638 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2640 - Reenable GRAS now that it works.
2642 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2644 SimGrid (2.90) unstable; urgency=low
2646 Alpha 1 on the path to SimGrid 3
2648 * It is a long time since the last release of SimGrid. I'm sorry about
2649 that but as I had told you, I was rewriting a lot of things. I apologize
2650 to those who had been reporting bugs to me and that I had not answered.
2651 If your bug is still in the new version, please tell me. Here is a
2652 summary of the main changes.
2654 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2655 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2656 portability, speed, and a lot more but you'll figure it out later.
2657 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2658 I wanted to release the new MSG as soon as possible and I have
2659 broken GRAS, which is the reason why, for now, only MSG is fully
2660 functional. A laconic description of these projects is available
2661 in the documentation.
2663 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2664 kernel optimized for our needs (called SURF but only the developers
2665 should use it). Hence, MSG is now roughly 30 times faster and I think
2666 that by rewriting a little bit MSG, I could event speed it up a little
2667 bit more. Beside the gain in speed, it is also much easier to encode a
2668 new platform model with SURF than it was with SG. More to come...
2670 * REVOLUTION 3: I have tried to change a little as possible the API of
2671 MSG but a few things really had to disappear. The main differences
2672 with the previous version are :
2673 1) no more m_links_t and the corresponding functions. Platforms are
2674 directly read from a XML description and cannot be hard-coded
2675 anymore. The same format is used for application deployment
2676 description. The new format is described in the documentation.
2677 Have a look in tools/platform_generation. There is a tiny script
2678 that converts from the old platform format to the new one. Concerning
2679 the application deployment format, parsing the old one is tricky.
2680 I think most of you should however be able to convert your files. If
2681 it is really an issue, I can write a C code that does the conversion.
2683 2) the toolbox tbx does not exist anymore. We now have a library
2684 with much more data-structures but without the hash-tables (we have
2685 dictionaries that are much faster).
2687 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2689 *****************************************************************************
2690 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2691 * don't feel like converting the oldies. (Mt) *
2692 *****************************************************************************
2695 Version 2.90: "the long awaited one"
2696 - Finished rewriting and debugging MSG. Rewrote the documentation.
2697 - disable GRAS for now since it needs to be ported to the newest SG
2700 - Finish the port to windows (using mingw32 for cross-compile)
2703 - Main loop and datastructures of SURF. A cpu resource object is
2704 functional. Surf can thus be used to create cpu's with variable
2705 performance on which you can execute some actions.
2707 2004-11-15 Martin Quinson
2708 - Port to ARM. Simply added the alignment and size descriptions. Should
2709 work, but the ARM machines are so slow that I didn't had the opportunity
2710 to 'make check' over there yet.
2712 2004-11-15 Arnaud Legrand
2713 - Trace manager now written. It uses a heap structure and is therefore
2714 expected to be efficient. It may however be speeded up (particularly
2715 when many events occur at the same date) by using red and black
2716 trees. One day maybe...
2717 - Max-min linear system solver written. It uses a sparse matrix
2718 structure taking advantage of its expected use. Most operations are
2719 O(1) and free/calloc are called as few as possible. The computation of
2720 the minimum could however be improved by using a red and black tree
2723 2004-11-03 Arnaud Legrand
2724 - Rename every gras_* function that was in xbt/ to its xbt_
2726 - Add a heap and a doubly-linked list to xbt
2727 - Added a dichotomy to the dictionaries. make check works as well before
2728 so I assume that the patch is correct. I do not know however if things
2729 run effectively faster than before now. :)
2731 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2732 SimGrid, and the version number is bumped to 2.x
2734 2004-10-29 Martin Quinson
2735 - Introduction of the remote errors.
2736 They are the result of a RMI/RPC on the remote machine.
2737 ErrCodes being scalar values, you can't get the host on which those
2738 errors did happen. Extending the error mechanism as in Gnome is possible.
2739 No idea yet whether it is a good idea.
2741 2004-10-28 Martin Quinson
2742 - Interface revolution: the Starred Structure Eradication.
2743 I used to do typedef struct {} toto_t; and then handle *toto_t.
2744 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2745 - ???_t is a valid type (builded with typedef)
2746 - s_toto_t is a structure (access to fields with .)
2747 - s_toto is a structure needing 'struct' keyword to be used
2748 - e_toto_t is an enum
2749 - toto_t is an 'object' (struct*)
2751 typedef struct s_toto {} s_toto_t, *toto_t;
2752 typedef enum {} e_toto_t;
2753 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2754 s_toto_t) is private.
2756 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2757 so that it can set the variable to NULL. It was so for dicts and sets,
2758 it changed for dynars.
2760 - Fix a bunch of memleaks in dict_remove
2761 - Fix a bug in sg/server_socket opening: it failed all the time.
2763 2004-10-07 Martin Quinson
2764 - Speed up dynar lookup operation a bit.
2766 gras_dynar_get is dead.
2768 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2769 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2770 (which gives you the address of the stored data).
2772 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2773 copy of the data using an affectation to do the job and not a memcpy.
2775 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2776 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2778 It does not really speedup the dynar test because they are
2779 setting elements all the time (and look them seldom). But the dict does
2780 far more lookup than setting.
2782 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2784 2004-10-05 Martin Quinson
2785 - Allow to (en/dis)able the cycle detection at run time.
2787 Whether we should check for cycle or not is now a property of each
2788 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2789 datadesc_cycle_unset allow to remove this property when previously set.
2791 Note that the cycle detection is off by default since it impacts the
2792 performance. Watch the data you feed GRAS with ;)
2794 This property is hereditary. Any element embedded in a structure having it
2795 set have it set for the time of this data exchange.
2797 You should set it both on sender and receiver side. If you don't set it on
2798 sender side, it will enter an endless loop. If you forget on receiver
2799 side, the cycles won't be recreated after communication.
2801 - Header reorganization.
2802 Kill gras_private.h, each submodule must load the headers it needs.
2804 2004-10-04 Martin Quinson
2805 - Interface revolution: do not try to survive to malloc failure.
2807 Now, gras_malloc and friends call gras_abort() on failure.
2808 As a conclusion, malloc_error is not a valid error anymore, and all
2809 functions for which it was the only gras_error_t return value are
2810 changed. They now return void, or there result directly.
2811 This simplify the API a lot.
2813 2004-09-29 Martin Quinson
2814 - Re-enable raw sockets.
2815 Created by gras_socket_{client,server}_ext;
2816 Used with gras_raw_{send,recv}
2819 It should allow to kill the last bits of gras first version soon.
2821 This is not completely satisfactory yet (duplicate code with
2822 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2825 - Simplify transport plugin (internal) interface by not passing any
2826 argument to _server and _client, but embedding them in the socket
2829 2004-09-28 Martin Quinson
2830 - Finish the port to AIX.
2831 autoconf was my problem (segfault within the malloc replacement
2832 function. No idea why)
2834 2004-09-16 Martin Quinson
2835 - Fix some size_t madness on 64bit architectures.
2837 2004-09-08 Martin Quinson
2838 - Reduce the number of system headers loaded, overload some more system
2839 calls (such as malloc to cast the result of the system one, and work
2841 - Fix and reintroduce the config support
2843 2004-09-07 Martin Quinson
2844 - Source code reorganization to allow Arnaud to surf all over there.
2845 - Allow to document the logging categories.
2846 - Remove all uppercase from logging categories and useless cleanup in names.
2848 2004-08-18 Martin Quinson
2849 Version 0.6.2 (protocol not changed; API changed)
2850 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2851 gras_error_t), and NULL when not found. Functions expecting a msgtype
2852 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2853 hopefully usefull message.
2854 - Portability to prehistoric sparcs again
2856 2004-08-17 Martin Quinson
2857 Version 0.6.1 (protocol not changed; ABI not changed)
2858 - prealloc some buffers to speed things up
2860 2004-08-11 Martin Quinson
2861 Version 0.6 (protocol not changed; ABI expended)
2862 - The parsing macro can deal with the references, provided that you add
2863 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2865 2004-08-09 Martin Quinson
2866 Version 0.5 (protocol not changed; ABI changed)
2867 - Allow to off turn the cycle detection code in data exchange at
2868 compilation time. It should be at run time, but I'm short of time (and
2869 the config stuff is still broken). That way, we keep dict out of the
2870 critical path, which is good because the performance is poor:
2871 - search not dichotomial yet
2872 - dynar give no way to access their content and memcpy everytime
2873 - In composed data description (struct, ref and so on), stop foolness of
2874 keeping the subtype's ID, but store the type itself. This keeps sets out
2875 of the critical path, which is good since they rely on dynar and
2876 dictionnaries. The only loose of that is that we cannot detect the
2877 redeclaration of a structure/union with another content (but I'm not sure
2878 the code detected well this error before anyway). We still can detect
2879 the redefinition discrepancy for the other types.
2880 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2881 it breaks the code because of type-punning used all over the place).
2882 This breaks on all non-gcc architectures (for now).
2884 All those changes (plus the buffer of last time) allow me to gain 2 order
2885 of magnitude on cruel tests consisting of 800000 array of integers on two
2886 level of a hierarchical structure (200 secondes -> 4 secondes)
2889 - the selector of reference must now return the type it points to, not
2890 the ID of this type.
2892 2004-08-06 Martin Quinson
2893 Version 0.4 (protocol changed; ABI not changed)
2894 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2895 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2896 - Implement a buffer transport, and use it by default (it relies on tcp in
2897 real life and on sg in simulation).
2898 That's a bit hackish since I had a new field to the structure to store
2899 its data without interfering with the subtype ones. Inheritance
2900 is tricky in C. And that's a kind of reverse inheritance with one class
2901 derivating two classes. Or maybe a game with java interfaces. Anyway,
2902 that's damn hard in C (at least).
2903 Moreover, I got tired while trying to ensure plugin separation and
2904 genericity in SG mode. MSG wants me to do weird things, so let's go for
2905 cruel hacks (temporarily of course ;).
2906 See comment in transport_private.h:71
2907 - do not include all the _interface headers in private but in the files
2908 which really need them (to cut the compilation time when they are
2911 2004-07-26 Martin Quinson
2912 Version 0.3 (protocol not changed; ABI changed)
2913 - Major overhault of the datadesc interface to simplify it:
2914 - shorted the function names:
2915 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2916 - add a trivial way to push/pop integers into the cbps without malloc.
2917 This allows to make really generic sub_type description, which simply
2918 pop their size of the stack.
2919 - add a function gras_datadesc_ref_pop_arr() which does what users want
2920 most of the time: Declare a dynamic array (which pops its size of the
2921 stack) and declare a reference to it. Poor name, but anyway.
2922 - kill the post-send callback, add a post-receive one
2924 2004-07-23 Martin Quinson
2925 Version 0.2 (protocol changed; ABI changed)
2926 - add some testing for cpbs in the test cases, and fix some more bugs.
2927 This invalidate again the little64 data file, since I cannot regenerate
2929 - remove an awfull optimization in the logging stuff, allowing me to:
2930 - understand it again
2931 - learn gcc how to check that the argument match the provided format
2932 - fix all errors revealed by gcc after that
2933 - internal keys of dict are not \0 terminated. Deal with it properly in
2934 loggings instead of segfaulting when the user want to see the logs :-/
2936 2004-07-22 Martin Quinson
2937 - Fix some stupid bug preventing cbps (callback postit) from working
2939 2004-07-21 Martin Quinson
2940 - Some documentation cleanups
2941 - remove the useless last argument of msgtype_declare
2942 - rename the Virtu functions to fit into the 'os' namespace
2943 - move headers src/include -> src/include/gras/ and stop fooling with
2944 gras -> . symbolic link
2945 - make distcheck is now successful
2947 2004-07-19 Martin Quinson
2949 - Build shared library also
2950 - Install html doc to the right location
2951 - stop removing maintainer files in make clean
2952 - build tests only on make check
2954 2004-07-13 Martin Quinson
2956 - No major issue in previous version => change versioning schema
2957 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2958 corresponding dataset.
2960 2004-07-11 Martin Quinson
2962 - Get it working with any kind of structure (we can compute the padding
2963 bytes remotely for all the architectures I have access to)
2964 - Implement the structure parsing macro (still not quite robust/complete)
2965 - Improvement to the remote testing toysuite
2967 2004-07-10 Martin Quinson
2968 [autoconf mechanism]
2969 - get ride of a bunch of deprecated macros
2970 - actually run the test for two-compliment, not only compile it :-/
2971 - test whether the structures get packed (and bail out if yes. Damn.
2972 Alignment is a serious matter)
2973 - test whether the structures get compacted (but respecting the alignment
2974 constraints of each types)
2975 - test whether the array fields of structures can straddle alignment boundaries
2977 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2978 (took me 2 hours to find that bug, looking at the wrong place)
2979 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2980 before sending/receiving and are used to compute the offsets of fields
2981 - Given that padding size depend even on compiler options, keep track of
2982 alignment and aligned_size only for the current architecture. Not a big
2983 deal since we send structure fields one after the other (seems
2985 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2986 program, let it work on linux/gcc/little32. portability todo.
2988 2004-07-08 Martin Quinson
2989 - import and improve remote compilation support from FAST
2990 - make sure make check works on half a dozen of machines out there
2992 2004-07-07 Martin Quinson
2993 Let's say it's version 0.0.3 ;)
2994 - Implement conversions (yuhu!)
2995 - Let it work on solaris (beside conversion, of course)
2996 - Stupid me, using rand() to generate the conversion datatests in not wise.
2998 2004-07-06 Martin Quinson
2999 - Let make dist work, since I'm gonna need it to compile on remote hosts
3000 - Let Tests/datadesc_usage write the architecture on which the file was
3001 generated as first byte.
3002 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3003 architecture descriptions.
3004 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3005 on those architectures.
3006 - Optimization: send/recv array of scalar in one shoot
3008 2004-07-05 Martin Quinson
3009 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3011 - Plug a whole bunch of memleaks
3012 - each process now have to call gras_{init,exit}. One day, their log
3013 settings will be separated
3014 - Continue the code factorisation between SG, RL and common in Transport.
3016 2004-07-04 Martin Quinson
3018 - Redistribution between SG and RL.
3019 We wanna have to accept in SG, so move accepted related parts of RL in
3020 the common part. (more precisely, the dynar of all known sockets is no
3021 more a static in transport.c, but part of the process_data)
3023 [gras_stub_generator]
3024 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3026 2004-07-03 Martin Quinson
3027 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3028 them all up in one shot)
3030 - Fix the ugly case of reference to dynamic array.
3031 - New (semi-public) function gras_datadesc_size to allow the messaging
3032 layer to malloc the needed space for the buffer.
3034 - gras_socket_close now expect the socket to close (and not its address to
3035 put NULL in it after it). This is because the socket passed to handlers
3036 is one of their argument (=> not writable).
3038 - propagate the interface cleanup from last week in datadesc, ie remove a
3039 superfluous level of indirection. User pass adress of variable
3040 containing data (both when sending and receiving), and not of a variable
3041 being a pointer to the data. Let's say that I like it better ;)
3042 The price for that is constructs like "int msg=*(int*)payload" in
3043 handlers, but it's a fine price, IMHO.
3045 - Let it work in RL (yuhu)
3047 2004-06-21 Martin Quinson
3049 - porting SG plugin and SG select to new standards (works almost).
3050 - plug memleaks and fix bugs around.
3053 - cleanup the prototype of data recv and force users to specify when they
3054 want to handle references to objects. Test case working even for cycles.
3055 - plug memleaks. Valgrind is perfectly ok with this.
3057 2004-06-12 Martin Quinson
3059 - cleanup the separation between plugin and main code in plugin creation
3061 2004-06-11 Martin Quinson
3063 - Reput hook for raw sockets, needed for BW experiments
3064 - kill a few lines of dead code
3065 [Data description] Interface cleanup
3066 - gras_datadesc_by_name returns the searched type or NULL.
3067 That way, no variable is needed to use a type desc once, which makes
3069 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3070 two parameters were strings (field name, type name), leading to
3072 [Dicos] Interface cleanup
3073 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3074 This is consistant with the dynar API.
3076 2004-04-21 Martin Quinson
3078 - Porting to new standards.
3080 - interface cleanup.
3081 There is no bag anymore, no need to take extra provision to mask the
3082 pointers behind "ID".
3083 Better splitup of functions between files create/exchange/convert.
3084 This is still a bit artificial since convert and receive are so
3085 interleaved, but anyway.
3087 - add a queued message list to procdata (the ones not matching criteria
3089 - factorize some more code between SG and RL wrt procdata
3091 - use gras_exit in example to track memleaks
3092 - get rid of gs_example now that GS is properly integrated into gras
3093 - update run_test to integrate the latest tests (datadesc)
3095 - rename WARNINGn macros to WARNn since it prooved error-prone
3097 2004-04-19 Martin Quinson
3099 - register init/exit functions within gras module mechanism
3100 - send/receive function.
3101 Convertion is not implemented, but short-cutted if not needed.
3102 struct/array elements are sent one by one (instead of block-wise), but
3103 nobody really cares (yet). Get a prototype before optimizing.
3104 - tests (using a file socket) for DD send/receive on:
3105 - base types: int, float
3106 - array: fixed size, string (ie ref to dynamic string)
3107 - structure: homogeneous, heterogeneous
3108 - chained list, graph with cycle
3109 Believe it or not, valgrind is not too unhappy with the results. The
3110 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3114 - Bugfix when using a filename explicitely (instead of '-')
3116 2004-04-09 Martin Quinson
3118 - factorize more code between RL and SG in socket creation
3119 - Complete the implementation and tests of:
3121 o file (only in RL, and mainly for debugging)
3123 I lost 3 days to design a portable address resolver, and then decided
3124 that the prototype mainly have to run on my box.
3125 Addressing portability too early may be like optimizing too early :-/
3127 - use gras_init in the Tests instead of the crappy parse_log_opt
3128 (the latter function is removed)
3129 [Conditional execution]
3130 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3131 [Code reorganisation]
3132 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3133 Build examples against the RL library, since there is no way to disable
3134 its creation for now.
3136 For information, the beginning of coding on GRAS was back in june
3137 2003. I guess that every line has been rewritten at least twice since