1 SimGrid (3.14) UNRELEASED; urgency=low
4 * The whole documentation was reworked and reorganized. There is
5 still much room for improvement here, but we're on it.
8 * We now need python3 for our testing infrastructure.
11 * Privatization is now activated by default in smpirun
12 You can turn it off with -no-privatize if something goes wrong.
14 * Call-location tracing for SMPI.
15 You can add the exact location (filename / linenumber) of an MPI call to
16 your trace files and slow down or speed up the simulation between two
17 consecutive calls by using an adjustment file (see the documentation).
19 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
20 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
21 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
22 values were ignored in some cases. The timings of these functions can now
23 be significantly different.
25 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
26 smpi/running-power is renamed to smpi/host-speed
28 New functions and features
29 * MSG_parallel_task_execute_with_timeout, to timeout computations.
31 Dropped / renamed functions and features
32 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
33 - MSG_mailbox_is_empty() -> Mailbox::empty()
34 - MSG_mailbox_front() -> Mailbox::front()
35 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
36 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
37 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
38 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
39 * Don't pass the free_f parameter to property related functions:
40 - MSG_host_set_property_value()
41 - MSG_as_router_set_property_value()
42 - MSG_storage_set_property_value()
43 * VM properties. Since msg_vm_t are msg_host_t, just use
44 MSG_host_get_property_value() and friends
45 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
46 Well, they are still there for now, but deprecated with a warning.
47 Please switch to xbt_dict_new_homogeneous() before this is removed
49 * Task affinity. Its intended behavior (that was very badly tested
50 and probably not really working) was deceiving what most users
51 would have hoped here.
52 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
53 * Option network/coordinates is now useless and should be dropped.
56 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
57 EiBps, EBps, Eibps, Ebps for bandwidth.
58 They may become useful to some lucky ones.
61 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
62 * ex/app_tokenring: new example, very similar to the MSG Token Ring
63 * ex/async_waitAll: new example, on asynchronous communications
65 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
67 SimGrid (3.13) stable; urgency=low
69 The Half Release, a.k.a. the Zealous Easter Trim.
71 - We removed half of the lines, that were mostly experimental cruft.
72 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
73 (not counting blanks and comments -- according to openhub.net)
74 - The internals are now compiled in C++ (and will soon be clean C++)
75 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
76 59 klines of C, etc. We added only 29 klines of C++ in replacement.
78 * Backwards Compatibility breaks
79 - Removed Lua simulation bindings (switch to C or Java for that).
80 Lua can still be used to describe platforms
81 - Removed Java kernel plug-ins.
82 Will be reintroduced after the ongoing major internals reorg.
84 - the following functions were removed.
85 They were too specific and should be reimplemented in a generic
86 way, with filter function.
87 - MSG_task_listen_from_host
88 - MSG_mailbox_get_count_host_waiting_tasks
89 - MSG_mailbox_put_with_timeout was removed.
90 Please use MSG_task_send_with_timeout instead.
92 - the SD_application_reinit function was removed. It has been a noop for a while.
93 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
94 replaced by a more flexible mechanism.
95 - The following functions thus do not exist anymore
96 - SD_workstation_get_access_mode
97 - SD_workstation_set_access_mode
98 - SD_workstation_get_current_task
99 - Basic estimation functions have been removed but can easily be replaced
100 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
101 - SD_workstation_get_computation_time => amount / sg_host_speed()
103 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
104 Use VM.getSpeed()*load/100 for the legacy behavior.
106 - option enable_tracing was removed. It was not doing anything for a while.
107 - In the ModelChecker:
108 - the model-checker now ptraces the model-checked process which means
109 you cannot use a debugger on the latter anymore (we might make this
110 optional in the feature);
111 - removed soft-dirty page tracking;
112 - remove model-checked side snapshot management,
113 MC_snapshot() and MC_compare_snapshot();
114 - keep the MC_cut() function as a stub (it was not really working
115 in the previous release).
118 * All options are consistently in kebab-case. Old names are kept as alias.
121 * Switch to platform v4 format.
122 - Rename from 'power' to 'speed' the attributes describing the amount of
123 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
124 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
125 - In <host> and <link>, attributes availability and state are gone.
126 It was redundent with state and availability traces, and with peak values.
127 - In <cluster>, attributes availability_file and state_file are gone.
128 It was too complex and unused.
129 - Kill <gpu>. Was not doing anything.
130 - The DOCTYPE points to the right URL:
131 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
132 (the file at this address now documents the changelog since its v1)
133 - A warning is emitted for unit-less values (they are still accepted).
134 - speed. Default: 'f' or 'flops'. Also defined:
135 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
136 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
137 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
138 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
139 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
140 - latency. Default: 's' second. Also defined:
141 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
143 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
144 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
147 * s4u::Host is now the preferred public interface to the Host features.
148 sg_host_* functions are C bindings to the exact same behavior
149 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
152 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
153 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
154 They were deprecated since a few years, and probably did not even compile anymore.
157 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
158 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
159 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
160 encouraged to update their codes. A list of the modified functions can be found at the end of
161 include/simgrid/simdag.h
164 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
165 the lambda or closure passed as a parameter will run in kernel mode.
166 All the callback functions should be rewritten to that interface at some point.
169 * Reorganizing and cleaning the internals all around the place.
172 * Remove old default barrier/bcast buggy algorithms (see #18407)
173 * Various bug fixes to handle more codes
174 * Remove the need for the --foreground option of smpirun (it is still
175 accepted for backward compatibility).
178 * Kill the setset data container: MC don't use it anymore.
179 * Kill the queue data container: it made more sense with GRAS.
180 * Kill the xbt_peer_t data type: it's useless without GRAS.
181 * Kill rm_cb feature of config sets: it was never useful.
182 * Kill graphxml parsing feature. It was not used.
183 * Kill the deprecated code protected by XBT_USE_DEPRECATED
185 - xbt_dynar_sort_strings(), when the content is char*
186 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
187 They throw exceptions on invalid input;
188 * C++ support for declaring CLI flags (simgrid::config::Flag);
189 * class for abstracting different signal backends (simgrid::xbt::signal).
190 with no external dependencies (we need very simple signals).
193 * refactoring and cleanup of the code;
194 * ongoing process to cleanly separate the model-checking algorithms
195 from the code model-checking support.
197 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
199 SimGrid (3.12) stable; urgency=low
201 The Facelift Release.
204 * Require g++ v4.7 at least to not speak prehistorical C++.
205 * Require Boost 1.48 (for signal2 component).
206 * Java must be version 7 at least when activated.
207 * Builds on Windows again (including Java bindings).
208 * Tracing is now always enabled (no way to turn it out)
209 * Remove GTNetS. It was not working anyway.
210 * Various cleanups in the cmake scripts.
211 * Move headers around to sort them out on installed systems:
212 - instr/instr.h -> simgrid/instr.h
213 - instr/jedule/* -> simgrid/jedule
214 - simdag/datatypes.h was removed
215 - simdag/simdag.h -> simgrid/simdag.h
216 - msg/datatypes.h was removed
217 - msg/msg.h -> simgrid/msg.h
220 * Interface improvement:
221 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
222 - Sanitize the interface in MSG_task_ module:
223 - Merge two functions that were close enough but misleading:
224 set_compute_duration(t) -> set_flops_amount(t)
225 get_remaining_computation(t) -> get_flops_amount(t)
226 - set_data_size(t) -> set_bytes_amount(t)
227 get_data_size(t) -> get_bytes_amount(t)
228 - Massive cleanups in the functions related to the energy
229 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
230 - New: MSG_host_get_pstate()
232 - msg/energy/onoff: switching hosts on and off
234 * Interface improvement:
235 - Rename Host.isAvail() to Host.isOn()
236 - Rename Process.currentProcess() to Process.getCurrentProcess()
237 - Rename Task.setDataSize() to Task.setBytesAmount()
238 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
240 - #18874: Actually allows the GC to reclaim tasks
244 - SIMIX_process_throw: raises an exception in a remote process
245 * Refactoring: Separate sub-modules
246 - libsmx: the public interface, as libc in a real system
247 - popping: the strange dance that converts a user request into a kernel handling
248 - smx_context_*: the virtualization mechanisms that embed the user code
249 - smx_*: the handling of each simcalls
251 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
252 * Rename smx_action_t into smx_synchro_t, making explicit that these
253 things are used to synchronize processes with their environment.
254 For example, a communication is a sort of synchronization involving
255 the communicating processes (that may block until the exchange) and
256 the platform. The same can be said from computations, etc.
258 - #18888: segfault when a process ends before its kill_time
262 - 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.
263 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
264 - MPI_Type_set_name, MPI_Type_get_name
265 - MPI_*_c2f and MPI_*_f2c functions
266 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
267 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
268 - Activate a lot of new tests from the mpich 3 testsuite
270 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
271 - InfiniBand network model added : Based on the works of Jerome Vienne
272 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
273 - When smpi/display_timing is set, also display global simulation time and application times
274 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
275 * Collective communications
276 - 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
277 - 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/).
278 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
280 - Replay now uses algorithms from wanted collective selector
281 - Replay can be used with SMP-aware algorithms
282 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
283 - 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.
285 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
286 - Correctly use loopback on fat-tree clusters
287 - Asynchronous small messages shouldn't trigger deadlocks anymore
288 * Energy/DVFS cleanup and improvement
289 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
290 - new: smpi_get_host_pstate()
293 - "Full" network optimization flag was broken since Surf++
294 - Better handling of precision flags in maxmin
295 - Fix bug causing sometimes "Impossible" errors
296 - Properly pass cluster properties to included hosts
297 * Improvement of the Energy plugin.
298 - Always update the consumption before returning that value
299 - New property: watt_off to denote the dissipation when the host is off
300 - New functions getWattMinAt and getWattMaxAt to retrieve the
301 dissipation of pstates that we are not currently at.
302 * Java: class NetworkLink renamed to Link
303 * New function: simcall_process_get_kill_time()
304 * Massive rename s/workstation/host/
305 - That's intrusive, but that's good for the project consistency. Sorry.
306 - Change config option "workstation/model" into "host/model"
309 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
310 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
311 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
312 - xbt_dirname and xbt_basename for non-POSIX systems
314 * The model checker now runs as a separate process.
315 * The model checker runs must now be launched with the new simgrid-mc program.
316 * Record/Replay: the MC can display a textual representation of a path in the
317 execution graph. It can then be replayed outside of the model checker.
319 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
321 SimGrid (3.11) stable; urgency=low
326 * Normalizing pointers addresses tool for better diff between logs
329 * Add cloud examples using new VMs
330 - examples/msg/cloud/two_tasks_vm.tesh
331 - examples/msg/cloud/simple_vm.tesh
332 - examples/java/cloud/cloud.tesh
333 - examples/java/cloud/migration/migration.tesh
334 * Add java surf examples:
335 - examples/java/surfPlugin/surf_plugin.tesh
336 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
337 - examples/java/surfCpuModel/surf_cpu_model.tesh
338 * Add SMPI+MSG example:
339 - examples/smpi/smpi_msg_masterslave/
344 - msg task destroy cancel test
345 - msg_host on/off test
346 * Move all tests in testsuite to teshsuite (adding tesh files)
347 * Restructure teshsuites
348 - one folder for each kind of test
349 * Restructure AddTests.cmake
351 - structure the order of tests (with sections)
354 * Add virtual machine
355 - creation of a VM on a PM
356 - migration of a VM from a PM to another PM
358 - MSG_process_join(msg_process_t process, double timeout)
359 - msg_bar_t MSG_barrier_init(unsigned int count)
360 - int MSG_barrier_wait(msg_bar_t barrier)
361 - void MSG_barrier_destroy(msg_bar_t barrier)
362 - msg_as_t MSG_environment_as_get_by_name(const char * name)
363 * New option "msg/debug_multiple_use" to help debugging when a task is used
366 - Improvements and finalization of MSG_storage, MSG_file APIs and their
368 - Increase code coverage in test suites
372 * Protect context stacks against stack overflow. The number of protected memory
373 pages allocated on the top of each stack (1 by default) can be configured
374 with the new command line option --cfg=contexts/guard_size:<value>.
375 * Simcalls are now generated by a python script that
376 - generates files included by SimGrid
377 - checks that all the functions exist, and proposes prototypes
379 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
380 host_set_data, host_get_data
382 - simcall_process_join(smx_process_t process, double timeout)
383 * Fix bug where sleeping processing could not be suspended.
386 * Translate surf models from C to C++
387 - Generic classes for all models: Model, Resource, Action
388 - A generic interface for each kind of model (CPU, Network, Storage
389 Workstation, WorkstationVM)
391 * Translate surf routings from C to C++
392 * Add callbacks using sigc++ or boost::signals2
393 - Add callback functions for resource creation/destruction
394 - Add callback functions for action state change
395 - Handle Energy as a plugin
396 * Replace swag by boost::intrusive
397 * Add new routing models for clusters. For documentation, see
398 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
399 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
400 parameters for cluster tag
401 - Fat trees, with topology="FAT_TREE" and
402 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
403 - see examples/platforms/torus_cluster.xml and
404 examples/platforms/fat_tree_cluster.xml
408 * Hostfiles support host:nb_processes construct to deploy several processes on
410 * Collective communication algorithms should not crash if used with
411 improper number of nodes and report the error.
412 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
413 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
414 MPI_Dims_create, MPI_Cart_sub are supported.
415 * New interface to use SMPI programmatically (still depends on MSG for
416 some parts, see examples/smpi/smpi_msg_masterslave) :
417 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
421 * Global variables privatization in MPI executables is now performed at runtime
422 with the option smpi/privatize_global_variables (default:no).
423 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
424 dynamic libraries loaded by the application are not privatized (static
425 linking with these libraries is advised in this case)
428 * Options defined in XML work correctly now.
431 * New cmake option, enable_lib_in_jar, to control whether native libraries are
432 copied into simgrid.jar or not (ON by default). Use this option if you want
433 to reduce the size of the installed simgrid.jar, *and* the native libraries
434 are kept installed elsewhere.
435 * Surf binding with SWIG (code generated in maintainer mode only):
436 - plugin to handle callbacks
437 - CPU model only for the moment
440 * Supernovae build mode is definitively removed. It was used to improve
441 inlining and inter-module optimizations. It is nowadays superseded by
442 link-time optimizations commonly available in compilers.
443 * Update ns-3 find lib. Bindings for ns-3 should work again now.
444 * Add boost dependency for surf++
445 * Add new macro for tests
446 - ADD_TESH(name <tesh_args>)
447 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
451 - xbt_bar_t XBT_barrier_init(unsigned int count)
452 - int XBT_barrier_wait(xbt_bar_t barrier)
453 - void XBT_barrier_destroy(xbt_bar_t barrier)
454 * Make the xbt_os_time module public
456 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
458 SimGrid (3.10) stable; urgency=low
460 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
463 * Reintegrate Java to the main archive as desynchronizing these
464 package is not acceptable anymore (Java is now considered stable)
465 * Add explicit synchronization facilities through semaphores
466 * Bug fix: Task.setDataSize() only changed the C world, not the value
467 cached in the Java world
470 * Dramatically change the way files are handled. API and internals changed, but
471 this part of MSG was not considered as production grade either.
472 * Add explicit synchronization facilities through semaphores
473 * Add a new function MSG_host_get_process_list()
474 * Preliminary DVFS support (see examples/msg/energy/ for details)
477 * SMPI is now included directly in the libsimgrid as the windows
478 linker doesn't force us on splitting it anymore.
479 * Improvements of the SMPI replay tool:
480 - Most of the collective communications are now rooted in the same process as
481 in the original application.
482 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
483 used until now). Multiple data types can now be used in a trace.
484 - The replay tool now supports traces produce either by TAU or a modified
486 - Bug Fix: the compute part of the reduce action is now taken into account.
487 - Gatherv collective is now supported
488 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
489 of smpirun outputs time independent traces for the current run. One file
490 is created per process. If too many processes are simulated, this behavior
491 can be changed to one file for all processes by using the
492 tracing/smpi/format/ti_one_file flag
493 * smpirun generates the host file if needed (with given host count and platform)
494 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
495 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
496 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
497 --cfg=smpi/coll_selector:(mpich/ompi)
498 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
500 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
501 happening inside a collective SMPI call.
502 * Fix the behavior of complex data types handling.
503 * Make MPI_Wtime another synchronization point to take computations into
505 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
506 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
507 * Add all missing Fortran bindings, SMPI should work with Fortran 90
508 (no privatization of global variables yet)
509 * Preliminary DVFS support (see examples/smpi/energy/ for details)
512 * Verification of liveness properties is now available for SMPI applications
513 (in addition to MSG applications)
514 * Bugged examples using SMPI in examples/smpi/mc/
515 * Add --cfg=model-check/visited option. Allows the verification of infinite
516 programs. Detection of loops in the execution thanks to the system state
517 comparison and reduction of the state space to explore. Can be combined with
518 DPOR for safety properties.
521 * Allow to change SimGrid configuration (see --help) within the code
522 thanks to SD_config() as it can be done in MSG.
523 * Add a new function SD_task_set_amount() upon user request.
526 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
527 * Remove rule based routing (no more PCRE dependency)
528 * Add a limiter_link option to cluster tag, to specify a maximum reachable
529 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
530 * Add a loopback_bw and loopback_lat options to cluster tag.
531 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
532 routing. Now peers are encapsulated in an AS and have their own private
533 router but this is transparent.
536 * Our own implementation of getline is renamed xbt_getline, and gets
537 used even if the OS provide a getline(). This should reduce the
538 configuration complexity by using the same code on all platforms.
539 * New type: xbt_cfg_elm_boolean.
540 * Allow to use yes/no for boolean configuration options in the command line.
541 * Allow to disable SimGrid cleanups at exit from command line option.
542 There are situations where one may want a simulation to end with an exit.
543 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
544 annoying when scripting around the simulator. Adding a
545 --cfg=clean_atexit:no allows to circumvent this issue.
548 * Lots of memory leaks were corrected in this release.
549 There is no known memory leaks anymore, in all of our 600+ tests.
550 * New command line option --version, to get SimGrid version information.
551 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
553 * Supernovae builds are deprecated, and expected to be removed in the next
556 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
558 SimGrid (3.9) stable; urgency=low
560 The Grasgory release: GRAS is really dead now.
562 * Complete overhaul of the internal host structures scheme.
565 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
566 was considered as experimental and badly maintained since then.
567 * Keeping it was thus a trap to our potential users, that could take
568 it instead of MSG or SMPI by mistake despite is pity state.
569 * GRAS seems to have very few users (if any), and no one volunteered
570 to maintain it further. It also induces a lot of XBT code (for
571 portability sake), that must be maintained too.
572 * For all these reasons, we killed GRAS. If someone wants to revive it
573 in the future, don't cry, our git history still remembers of GRAS.
576 * Major overhaul. Merge our documentation again as time proved that
577 splitting it was really not helping our users.
578 * Further improve the developer documentation to help newcomers
579 hacking on SimGrid itself. The user documentation (and in
580 particular, the beginner documentation) is still in a sorry state.
583 * Now works on Windows too!
584 * Much more extensive test suite, from MPICH
587 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
588 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
589 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
590 * Bug fix: let task be scheduled when the last dependency to be solved is
591 a control dependency.
592 * Remove SD_load_environment_script function.
593 Use the C sg_platf function if you want to declare a platform
597 * New function: MSG_process_get_number()
598 * Old function documented: MSG_config()
599 * Remove MSG_load_platform_script function
600 Use the C sg_platf function if you want to declare a platform
604 * Change the default value of the TCP_gamma constant (maximal size of TCP
605 congestion window) to a more realistic 4MiB value. If you notice changes in
606 your simulation results, you can fall back to the previous 20k tiny window
607 by adding --cfg=network/TCP_gamma:20000 on command line.
608 * (Hopefully) fix a bug wrt periodic availability and state traces
609 * Bug fix: use default values at start when first event in availability/state
610 trace is not at time 0.
613 * remove the "new_" part of function name sg_platf_new_trace_connect
614 (resulting in sg_platf_trace_connect), since it does not create
618 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
619 use case where it's really mandatory, and maintaining it was a pain in
621 * New: xbt_fifo_search(), search an item with a user-provided
622 comparison function instead of dumb pointer comparison.
625 * Fix the lua deployment:
626 Use `simgrid.init_application()` before deployment instead of
627 `simgrid.msg_register_application()` after.
630 * Transfer the tracing files into the corresponding modules.
632 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
634 SimGrid (3.8.1) stable; urgency=low
636 The "we are told that some people want to also *install* the simgrid
639 * Add missing file "tesh.1" to the archive.
641 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
643 SimGrid (3.8) stable; urgency=low
645 The Psssshiiiit release: SimGrid jumps into the Cloud.
648 * Add an experimental interface to manipulate VMs. They are mainly
649 process groups with very few intrinsic semantic, but they should
650 allow you to build the semantic you want easily.
651 * New function: MSG_host_set_property_value()
652 * New function: MSG_process_on_exit(). To clean memory in all cases.
653 * Bug fixes that made the host (and link) failures unusable.
654 * Add a way to auto-restart process when the host in which they are
655 executing comes back (ON_FAILURE="RESTART" on deployment file,
656 MSG_process_auto_restart_set).
657 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
658 please stop using the old ones, they are DEPRECATED.
660 * Deprecate functions MSG_global_init() / MSG_global_init_args()
661 Please use MSG_init() instead. (reducing the amount of entry
662 points in the library helps us).
663 * Make it impossible to link against the wrong version of the lib
664 * Deprecate MSG_clean(). No need to call it anymore.
665 * Function MSG_get_host_number() is not deprecated anymore.
668 * Split the doc into a user guide and a reference guide.
669 * Start a developper guide to help people hacking on SimGrid.
672 * Enable tracing by default. This modules rocks you should use it.
673 * Remove option custom_flags. Now use environment variables CFLAGS
675 * Use default cmake things to detect lua instead of home grown ones.
676 * New option "enable_mallocators" to disable mallocators, for debugging
677 purpose ("on" by default).
680 * Bug fixes around the resource failures: don't let the processes
681 survive the host they are running onto.
682 * Add an interface to auto-restart processes when the host in which they are
683 executing comes back.
684 * Ensures that SIMIX_clean is called automatically. It's not part of
685 the public interface anymore (bindings should be updated).
688 * Bug fix for when SD_Simulate is called with a positive value: be careful
689 when comparing doubles. Sometimes they are different for non significant
691 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
692 parallel task whose initial work is distributed among host according
693 to the Amdahl's law. Such tasks are created with a parameter alpha
694 that corresponds to the non-parallelizable part of the computation.
695 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
696 between two sets of workstations assuming a 1D block distribution (each
697 workstation owns a similar share of data) on both sides.
699 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
700 Data redistribution will be automatically scheduled once parent and child
701 are both scheduled. The filling of computation_amount and
702 communication_amount structures is now done seamlessly thanks to the chosen
704 * New function SD_workstation_dump to display various information
705 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
706 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
707 bandwidth on the route onto which the task is scheduled and the amount of
709 To divide the nominal bandwidth by 2, the rate then has to be :
710 rate = bandwidth/(2*amount)
711 * Compute tasks that have failed can now be rescheduled and executed again
712 (from their beginning)
713 * Increasing source code coverage (src/simdag is now covered at 95.8%
717 * Re-implement time-independent trace replay using SMPI (at the
718 smpi_smp_* level) instead of MSG. This should replace
719 examples/msg/actions/actions.c
720 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
721 hindexed and structs)
722 * Implement the exchange of non-contiguous data.
723 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
724 * Correct behavior of smpi/sender_gap and set its default value to 0
725 * Add option to asynchronously send small messages to allow better
726 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
727 specifies the size in bytes under which messages will be asynchronously sent.
728 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
729 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
732 * New C interface to define a platform: XML is now optional.
733 For more info, please check include/simgrid/platf.h.
734 * New interface to define random platforms from the C:
735 For more info, please check include/simgrid/platf_generator.h and
736 examples/msg/masterslave/masterslave_platfgen.c
737 * Export a sg_cmdline dynar containing all the arguments we got from
741 * Two new tracing options for adding comments to trace file so you
742 can track your experiments (see --help-tracing for details).
743 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
744 * Adding the SimGrid version that generated the trace file as a comment.
745 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
746 * Fix to avoid key clashes on Paje links
747 * Other minor fixes related to the Paje specification
750 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
751 and renamed to xbt_str_hash() and xbt_str_hash_ext().
752 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
753 * Greatly improve the robustness of mmalloc to user errors (such as
754 using an area after freeing it, or freeing it twice)
756 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
758 SimGrid-java (3.8.1) stable; urgency=low
760 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
761 Warning: all methods are visible, but only some of them are
762 implemented so far. Check the source (src/jtrace.c)
763 for further information.
764 * New module: org.simgrid.msg.File (SimGrid File management functions)
765 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
766 * Change the meaning of Process.restart: now restart the process from
767 the begining, like MSG_process_restart in C.
768 * Add Process.setAutoRestart: handling of process restart when failed
770 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
771 you to retrieve the properties of the processes/hosts
772 * Deprecate Msg.clean(): you can just forget about it now.
773 * New function Process.getCount(), that only works when compiling
774 with the not yet released version 3.9 of the C library.
776 * New context factory based on Coroutines. It mandates a modified JVM
777 but then, the simulations run about five times faster, and there is
778 no limit to the amount of processes (beside of the available memory).
780 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
782 SimGrid (3.7.1) stable; urgency=low
785 * Restore the prototype of MSG_process_create_with_environment() to
786 the pre-3.7 situation by removing the kill_time argument.
787 * Add a MSG_process_set_kill_time() function instead.
790 * Fix weird behaviors when dealing with parallel tasks.
793 * Simgrid is now built as a dll.
794 * Simgrid-java now works on Windows.
795 * Simgrid-Java is now included into Windows package.
798 * First pre-build package for MacOSX.
801 * Fix compilation when using MSG_USE_DEPRECATED.
802 * Fix some compilation issues on Macs and Windows.
803 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
804 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
806 We discovered that the Lua console is broken, but we are missing the
807 manpower to fix it right now. The problem existed in 3.7 too, so we
808 are not blocking the release for that. Sorry if you depended on this
809 feature, any help would be really welcome.
811 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
814 SimGrid-java (3.7.1) stable; urgency=low
816 The "Java aint got to be bloated and slow" release
819 * Various internal cleanups and performance improvement
820 Simulations are expected to run up to twice faster or so
821 * Make Process.kill(process) an instance method, not a static one
822 * User processes are not java.lang.Thread subclasses.
823 This breaks the compatibility (sorry), but previous API was
824 brain-dead, making it impossible to have non-trivial
825 initializations in the process constructor.
826 * Require a full constructor per Process sub-class.
827 Kinda breaks the compatibility (sorry), but this allows a much more
828 efficient way to launch the processes at simulation startup.
829 * Do not embeed our version of semaphores, java 1.5 can be considered
830 as sufficiently prevalent for us to not dupplicate its features.
834 * Add examples for almost every part of the API
835 We spotted and fixed a lot of bugs in the process
837 * New module: asynchronous communication API
838 * New function: Process.sleep()
839 It takes milliseconds as argument, just as java.lang.Thread.sleep()
840 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
841 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
843 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
845 SimGrid (3.7) stable; urgency=low
847 The "spring cleanups (before the next Big Project kicks in)" release.
850 * We can specify the SMPI latency/bandwidth factor with command line
851 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
852 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
853 You can also use the "config tag" from platform file by adding this line
854 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
855 (see "example/platforms/tag_config.xml" to use "config tag").
856 Note that the command line supersedes the platform file configuration.
857 * Change the correction factors used in LMM model, according to
858 the latest experiments described in INRIA RR-7821.
859 Accuracy should be improved this way.
860 * Use the partial invalidation optimization by default for the
861 network too. Should produce the exact same results, only faster.
862 * Major cleanup in surf to merge models and split some optimization
863 mechanisms from the core of the models. As a result you can now
864 specify which model to use (e.g., --cfg=network/model:LV08
865 --cfg=cpu/model:Cas01) and which optimization mode to use
866 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
867 Incompatible combinations should err at initialization. See
868 --help-models for the list of all models and optimization modes.
869 * The CLM03 workstation model was dropped for simplicity because it
870 used the deprecated CM02 network model. Use default instead.
871 * Rename the TCP_gamma configuration option to network/TCP_gamma
872 * Rename the coordinates configuration option to
873 network/coordinates, and document it
874 * Use now crosstraffic keyword instead of the terribly misleading
875 fullduplex keyword. It is activated by default now in the current
876 default model, use --cfg=network/crosstraffic:0 to turn it off.
877 * Ongoing refactoring the model parsing to make XML files optional
878 See include/simgrid/platf.h for details (still to be completed)
881 * Major overhaul of the documentation. Almost instructive now :/
882 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
883 functions and friends. This interface was considered as
884 deprecated since over 2 years, it's time to inform our users that it is.
885 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
886 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
887 use these (crappy) functions in your code.
888 These functions will be removed soon. Stop using them now.
889 * Deprecate MSG_get_host_{table,number}
890 Implement MSG_hosts_as_dynar() instead.
891 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
892 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
895 * Stabilize the parallel execution mode of user contexts
896 * Introduce configuration variables to control parallel execution:
897 - contexts/synchro: Synchronization mode to use when running
898 contexts in parallel (either futex, posix or busy_wait)
899 - contexts/parallel_threshold: Minimal number of user contexts
900 that must be part of a scheduling round to switch to parallel
901 execution mode (raw contexts only)
902 * Fix bugs that prevented to use suspend/resume along with
903 synchronization structures.
904 * Fix bugs in process termination that lead to invalid memory access
905 in very specific conditions.
908 * Introduce a parallel mode for the models (controlled by surf/nthreads
909 configuration item). In our tests, running the models in parallel
910 never lead to any speedups because they are so fast that the gain
911 of computing each model in parallel does not amortizes the
912 synchronization costs, even when ultra fast futexes are used.
913 This is released anyway because YMMV.
916 * Performance boost by using a swag internally to compute the set of
917 tasks that are finished and should constitute the return value of
921 * Enable it by default now that it is considered rather stable.
924 * Documentation of the tracing functions.
925 * Performance gains when tracing categorized/uncategorized resource
926 utilization by avoiding calls to get route when updating resource
927 variables. LMM constraints are being used instead.
928 * API changed to set task categories. Use MSG_task_set_category instead
929 of TRACE_msg_set_task_category, and SD_task_set_category instead
930 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
931 * Bugfix for graphicator, routes not correctly obtained, memory leaks
932 * Examples for link user variables added (see at examples/msg/tracing/)
933 * Deprecated function TRACE_msg_set_process_category completely removed
934 * Trace header updated according to the latest Paje file format
935 * Tracing network lazy updates, no longer obligate users to use full updates
936 * --cfg=tracing/platform:1 also registers power/bandwidth variables
937 * Experimental: let user code declare/set/push/pop application states for hosts
938 * API changed to allow the manual creation of graph configuration files
939 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
942 * Improve the API of Lua MSG bindings, using the Lua spirit.
943 * Each simulated process now lives in its own Lua world (globals are
944 automatically duplicated). It helps writing simulators. Will allow
945 to run Splay programs within SimGrid in the future.
946 * Add a Chord example in Lua, equivalent to the MSG one.
949 * Start the implementation of a solution to express temporal
950 properties, not only local assertions. This is still an
951 experimental work in progress, stay clear from it to be safe.
955 - Add new runtime parameters --help-logs and --help-log-categories
956 to display information about supported logging parameters and
958 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
959 don't exists anymore.
960 * Mallocators: allow value NULL for the reset function.
962 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
963 create homogeneous dictionaries, where all the elements share the
964 same free function. Non homogeneous dictionaries will be
965 deprecated in the next release.
966 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
967 - Multi-level dictionaries are deprecated.
969 - new function xbt_dynar_search_or_negative() that is useful when
970 you have less than 2 million elements in your dynar and don't
971 want of the extra complexity of catching exceptions when the
972 element is not found.
974 - Make xbt_os_thread module (for thread portability) public.
975 Documentation is still to come, sorry.
977 - Cleanups and simplifications to make it maintainable again.
978 - Exotic features (such as memalign and valloc) were removed.
979 - The metadata were extended and improved so that the
980 model-checker becomes able to explore and inspect the heaps.
981 - This may induce a performance drop when enable_model-checking is
982 ON in cmake (even if it's not used in the simulation), but it is
983 necessary at this point to get MC working.
985 Turn model-checking OFF if simulation performance matters to you.
986 Not enabling it at runtime is not enough, disable it in cmake.
988 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
991 SimGrid (3.6.2) stable; urgency=low
993 The "Not coding new stuff allows to polish old things" release.
996 * New bindings to the NS3 packet level simulator (experimental)
997 * Use the raw (efficient) execution contexts instead of the sysv
998 (portable) ones when possible.
999 * libpcre is now mandatory in any cases since not using it led to
1000 severe performance loss and possibly other issues
1001 * Update the XML platforms:
1002 - G5K: include the latest machine in Nancy
1003 - GridPP and LCG: new platforms
1004 * Documentation was partially updated, at least (more to come)
1006 Bug fixes, cosmetics and small improvements
1007 * Free terminated processes before the end of the simulation to avoid
1008 exhausting the memory of users having very dynamic amount of
1010 * Bug fix and cosmetics about canceling non-running tasks
1011 * Bug fix about the dot loader's issues when using libcgraph
1014 * Create an installer for windows with nsis (amd64 and win32)
1015 - Add an hello world project to illustrate simgrid project creation.
1016 - Embed libpcre into the Simgrid installer to avoid
1017 its compilation burden
1018 * The raw execution contexts should work on Apple now
1019 * Port to Windows 64 bits
1020 - Sysv contexts now have an implementation for this arch
1021 - GRAS communication features now support this arch
1022 * Drop support for borland compiler on windows
1023 - this code was not maintained, and we kinda depend on gcc nowadays
1024 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1025 * Fix portability issue on unstable ubuntu: linker became picky on
1028 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1030 SimGrid (3.6.1) stable; urgency=low
1032 The "Oops, we broke Macs too" release
1035 * Fixed contexts detection so that raw ones are used when possible
1036 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1037 leads to a strange error, with user code segfaulting sometimes when
1038 the generated code is not perfectly aligned (which is not
1039 controllable from the user side, depends on the amount of code)
1042 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1044 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1046 SimGrid-java (3.6) unstable; urgency=low
1049 * Split of every thing from simgrid v3.5 into a separate package.
1051 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1053 SimGrid (3.6) stable; urgency=medium
1055 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1058 * Bindings now constitute their own package, separated from the main one.
1059 Rationale: reduce our maintenance nightmare by reducing the module coupling
1060 They will soon be released on their own on gforge.
1062 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1063 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1065 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1066 * It's not quite deprecated for now because we have no replacement,
1067 but it may soon become the case.
1070 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1071 * Fortran: New user-level cache variable to store the rank of the running
1072 process. This improves performance by an order of magnitude.
1073 * C: New coccinelle script to automatically locate and modify global and
1074 local static variables.
1075 * Improved SMPI network model with a sender-side gap to account for multiple
1079 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1080 only say if a communication is finished, no matter whether it succeeded or
1081 failed. You can call MSG_comm_get_status() to know the status of a finished
1083 * New function MSG_task_dsend() to send a task and detach it. When a
1084 communication is detached, you are never notified of its success or failure
1085 and the memory is released automatically once it is finished. This function
1086 is useful when you don't care about the end nor the success of a
1088 * Change the prototypes of action replay. Sorry for inconvenience,
1089 but this is really more efficient this way (and to adapt your code,
1090 you just have to fix the initialization, that shouldn't be too long)
1091 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1092 ever managed to do anything useful with it.
1093 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1094 takes a dynar of communications. It returns immediately and gives the
1095 index of a finished communication (if any).
1096 * New example: a basic implementation of the Chord P2P algorithm.
1099 * New model for multi-core CPUs. You can now use the core attribute to
1100 precise the number of cores of a host. This is a basic model. Every
1101 process running on the host receives at most the power provided in
1102 the DTD (throughput<=power). Total throughput of process cannot exceed
1104 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1105 router linked by an up-link and a down-link (possibly asymmetrical).
1106 This kind of pattern allows to easily build last-mile model style platforms.
1107 Aggregating such patterns in a rule-based AS is thus the technique of
1108 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1109 * New model for Vivaldi routing. We transformed the Vivaldi network model
1110 into a Vivaldi routing model (based on the rule-based model). This allows to
1111 combine Vivaldi based latencies with last-mile platforms.
1114 * Added a check for NaN of IEEE754 infinite in the double entries of
1116 * Introduce a new context factory "raw", highly inspirated from the
1117 ucontext factory, but using manually crafted functions in assembly to
1118 do the work in an efficient manner.
1119 * Allow to change the used context factory at run time, not only at
1120 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1121 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1122 contexts at runtime (only with raw contexts or ucontexts).
1123 * Completely rewrote this module to allow parallel execution of user
1124 processes. Use --cfg=contexts/nthreads:N to execute user processes
1125 with N parallel threads (the default is 1, meaning no parallelism).
1126 * Allow to decide dynamically between sequential and parallel modes.
1127 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1128 processes in parallel only when their number is greater than or equal to P
1130 * Added a check for NaN of IEEE754 infinite in the double entries of
1134 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1135 won't output the state of processes when interrupted with Ctrl-C
1136 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1137 NULL-terminated array. This may solve backward compatibility issues
1138 due to the change to return type of SD_simulate. See also:
1139 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1140 * Add new macros with variable number of arguments.
1141 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1142 - in xbt/asserts.h: xbt_assert
1143 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1144 - in xbt/ex.h: THROWF and RETHROWF.
1145 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1147 * Change xbt_die() to accept a format string with arguments, just like printf.
1148 * New data structure: xbt_lib_t, like a dict but more general and with better
1152 * New configuration options
1153 Options triva/categorized and triva/uncategorized can be used to generate
1154 graph configuration files for Triva visualization tool.
1155 * Configuration option tracing/platform is renamed to tracing/categorized
1156 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1157 * New TRACE_user_link_variable interface:
1158 User provides the name of the link and the tracing variable to attach to it
1159 * the declaration of tracing categories must be done after the environment
1161 * simpler tracing interface, just one way to declare categories
1162 TRACE_category or TRACE_category_with_color, it is up to you
1163 * links in the trace file are again identified by their names
1164 * trace contains the full platform hierarchy exactly as declared using the ASes
1165 * Options tracing/msg/[task|process]:1 groups the process by hosts
1166 for both cases, tasks and processes must have names that are unique during the simulation
1167 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1168 * The experimental option tracing/msg/volume is deprecated
1169 its functionality may be reincorporated if needed
1171 The tracing generates a trace file with unordered timestamped events,
1172 because of the way the core simulator (surf) works. A script available
1173 at the tools directory (fix-paje-trace.sh) can be used to put the events
1174 in order. We have changed the tracing so it can generate ordered timestamped
1175 events in the final trace, but depending on the simulator (and how much time
1176 is simulated) that can lead to a huge memory utilization. It is deactivated
1177 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1179 Build Infrastructure
1180 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1181 - We are trying hard to keep the API stable, but it may happen that
1182 some things change (we're a research project after all, not a
1183 nuclear plant operating system). If such things should happen, you
1184 could rely on that macro to adapt.
1185 - current value: 30600 for 3.06.00, aka 3.6
1186 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1187 a number that can be compared with SIMGRID_VERSION.
1188 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1189 assertions and verbose/debug logging events are disabled at compile time.
1191 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1193 SimGrid (3.5) stable; urgency=medium
1196 * New feature to any SimGrid-based simulator: Model-Checking
1197 Check SIN#1 for more details.
1200 * New Model SMPI (three-interval linear regression for correction factors)
1201 See RR-7426, available at http://hal.inria.fr/inria-00527150
1202 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1203 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1204 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1205 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1206 * See also src/smpi/README
1211 - Tracing API changes: TRACE_start and TRACE_end should not be called
1212 by user-code. They are automatically called by simulators created
1213 with SimDAG, MSG and SMPI if the toolkit is compiled with
1214 tracing_enabled=ON. Categories declaration and utilization remain the
1215 same for MSG and SimDag.
1216 - A function was added to the tracing API to declare categories with
1218 - TRACE_category_with_color (char *category, char *color)
1219 where color must be in the following format
1220 "%f %f %f", red, green, blue
1221 and red, green, blue are float values in the interval [0, 1]
1222 - User can specify NULL as color parameter, or continue calling
1223 TRACE_category (cat)
1224 On that case, the tracing system will define random colors
1225 - The following command-line options are supported:
1226 --cfg=tracing/filename:msg.trace
1227 --cfg=tracing:1 (activate tracing, needed to use others)
1228 --cfg=tracing/platform:1 (categorized resource use)
1229 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1230 --cfg=tracing/msg/task:1 (task creation)
1231 --cfg=tracing/msg/process:1 (process creation, migration)
1232 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1233 --cfg=tracing/smpi:1 (SMPI interface tracing)
1234 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1235 - examples of examples/msg/tracing updated
1237 - DAXLoader and DOTLoader functions can generate tasks with categories
1238 - A new function to attribute a category to SD tasks:
1239 TRACE_sd_set_task_category (SD_task_t task, char *category)
1240 * Tracing the MPI interface implemented by SMPI
1241 - Collective operations are traced with states
1242 - Point-to-Point operations are traced with states/links
1243 - Tracing activated by a parameter "-trace filename" passed
1244 to smpirun during execution (considering that simgrid
1245 is compiled with tracing enabled)
1246 - To run the simulation with gdb, the simulator
1247 accepts --cfg=tracing/smpi:1 to trace SMPI
1248 - tesh files to check if smpi tracing is ok
1249 - See examples/smpi/NAS/DT-trace
1250 * GTNetS tracing re-worked
1251 - adaptation to the tracing system of GTNets to cope
1252 with modifications regarding the fullduplex mode
1253 - new tesh files to check if gtnets tracing is ok
1256 * Asynchronous communications through the functions:
1257 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1258 * New function: MSG_load_platform_script()
1259 to make possible using a lua script instead of XML files to set up platforms
1260 * New function: MSG_set_function
1261 to associate functions to processes, used when bypassing the parser
1262 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1264 Platforms: Add some more examples in examples/platforms
1265 * Grid'5000: see www.grid5000.fr
1266 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1269 * Change the XML format. This is a very important modification. SimGrid
1270 3.5 introduces a new hierarchical format based on the notion of
1271 Autonomous Systems. Compatibility with old format is ensured
1272 through the perl script provided in the install bin directory
1273 bin/simgrid_update_xml.
1274 It is now possible to build platforms with specific routing
1275 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1276 connect several platforms together. We will try to provide soon
1277 set of realistic platforms exploiting these properties (have a
1278 look at examples/platforms/ for the moment).
1279 * Take the opportunity of the XML format change to be a good XML citizen:
1280 rename link:ctn to link_ctn and similar changes (also dealt with by
1282 * Add a new routing scheme (rule-based) using regular expressions. It
1283 enables to have an extremely low memory footprint when the
1284 underlying routing is simple and can be compactly described. You
1285 need to have libpcre4-dev (perl regular expressions) installed if
1286 you want to use this routing scheme.
1287 * Revive the cluster TAG and allow to easily and efficiently (both in
1288 term of memory and speed) connect clusters together. Have a look
1289 at teshsuite/simdag/platforms/ to see how this can be done. With
1290 this tag, you can create clusters with thousands of tasks at no
1291 cost (have a look at examples/platforms/).
1292 Note: clusters are implemented as ASes, so there is no need for an
1293 enclosing AS tag if you have only one cluster in your platform.
1294 * Add new generic functions in the public interface that allows the user
1295 to call SURF 'create_resource' methods from your code (same
1296 functionality as the XML bypass mechanism but with a much lighter
1298 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1299 that uses a piecewise linear approximation to produce better
1300 results when exchanging small messages.
1301 * Add a new parameter to handle correctly full duplex link and account
1302 for interferences between uplink and downlink communications
1303 (activate with --cfg=fullduplex:1).
1306 * Rename the SD_READY (all dependencies are satisfied and task is
1307 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1308 dependencies are satisfied) state.
1309 This prevents a confusion between the notion of "ready to schedule"
1310 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1311 simulated" (SD_RUNNABLE) used by the simulation kernel.
1312 * Change the way a task is considered as ready. Instead of removing
1313 dependencies when a task is done, a counter is decreased. This way,
1314 it is always possible to reach ancestors thanks to the
1315 SD_taks_get_parents function (even after the end of the simulation.)
1316 * Change the return type of SD_Simulate from (SD_task_t*) into
1317 xbt_dynar_t. This function was in handling a dynar internally and
1318 converted it into a NULL terminated array for historical reasons.
1319 * New function SD_dotload(char*) to load a DAG described in dot
1320 format. This loader and the corresponding examples require the
1321 installation of the graphviz library.
1322 * Fix a bug in the management of tasks of size 0 in the surf network
1323 models. This problem was only visible with SIMDAG and you should
1324 thus disregard results produced with earlier versions if you
1325 relied on this feature (some tasks were blocked because of this).
1326 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1327 even though your applications were doing only point-to-point
1328 communications and sequential computations. Now you can really use any
1329 model you want (of course, if you create real parallel tasks, which are
1330 not implemented in most models beside ptaskL07, this will abort).
1331 * Add an example that schedules a DAX on an heterogeneous platform
1332 using a Min-Min strategy.
1333 * New function SD_workstation_get_current_task() that returns the kind
1334 of task currently running on a workstation in the sequential access
1336 * Raise some warnings when unexecuted tasks remains at the end of the
1337 simulation. This is usually caused by cycles in the DAG.
1340 * New function: SIMIX_process_set_function() called by MSG_set_function
1341 * Change the underlying waiting queue in semaphores so that a process
1342 can wait on several of them simultaneously (as in waitany).
1343 * Fix the way to handle tokens in semaphores so that all access patterns
1344 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1345 * kill the dirty pimple SIMIX_message_sizes_output()
1346 Please use (proper) visualization instead
1349 * New data container: setset (set of sets of elements)
1350 * New module: mmalloc (mapped malloc, allowing to have several
1351 independent segments of malloc)
1352 * New function: xbt_dict_cursor_set_data()
1353 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1354 * New function: xbt_dynar_is_empty()
1355 * New function: xbt_fifo_get_last_item()
1356 * Fix xbt_dynar_shrink(): use the right element size.
1357 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1358 uninitialized areas during expand.
1359 * Fix semaphores: previous implementation was severely broken.
1360 * Use library init/fini functions for our initialization.
1361 - you can use logs and other feature as soon as you want in your
1362 code (even before the xbt_init / MSG_init)
1363 - xbt_exit is now a no-op and produce a warning when used.
1366 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1367 benefit from the latest improvement to the simulation kernel.
1368 * Kill measurement sockets for now. If you rely on them, sorry. This
1369 release is not for you. This feature will be reintroduced in the
1370 future, but we cannot delay the release any further.
1371 * New function: gras_msgtype_get_name().
1372 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1373 * Fix (at last) the pmm example: it should not randomly fail anymore.
1375 Build chain: bug fixes and overall polishing
1376 * Cmake is now stable enough. Hence, we killed the autotools.
1377 * Port to windows ( TM :)
1378 * Fix the 'make install' target.
1379 No need to use 'make install-simgrid' anymore
1380 * Introduce a 'make dist' target compiling a *source* archive
1381 'make package' compiles a binary archive
1382 * Compile java files only on need
1383 * Add --cd and --setenv command line options to tesh
1384 * Out of source builds are not fully supported yet, but we are close
1385 * Enable supernovae and optimization flags by default for our users
1388 * Add layer to set up environment directly from lua, without XML.
1389 * The effect of gras_stub_generator can be achieved through
1390 lua too (check examples/gras/console/ping_generator.lua)
1392 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1394 SimGrid (3.4.1) stable; urgency=low
1396 The "Polishing easter eggs is probably a good idea" release.
1397 This is a bug fixes release only.
1401 * Fix a bug preventing the tasks from begin garbage collected.
1404 * Fix a bug occuring when a host involved in a communication fails.
1405 This was not detected properly by the other peer involved in the
1406 communication. Now, it's reported as a network error.
1409 * Warn the user about loop dependencies in data flow of DAX files
1410 * Obey the control-flow dependencies of DAX files
1413 * Add option "enable_smpi" allowing to not compile SMPI.
1414 Probably useful for the (Mac) users experiencing a build error here
1415 * Improve the detection of lua5.1 and ruby1.8
1417 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1419 SimGrid (3.4) stable; urgency=low
1421 The "Easter in Cargese" release. Also known as (major changes):
1423 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1424 ~> bindings were greatly improved
1425 ~> new tracing infrastructure for better visualization introduced
1427 * the "Welcome to configury modernity" release.
1428 ~> we switched from autotools to cmake, and improved our cdash
1431 A more detailled list of changes follow (full detail in svn log).
1433 Java Bindings: Various Cleanups
1434 * (install java-gcj-compat-dev on debian-like to use them)
1435 * Remove put/get: no need to export deprecated interface in Java
1436 Use send/receive instead.
1437 * Cleanup the examples and add a README per directory
1438 * Remove example autoDestination (that's the only way to go now)
1439 * Remove example explicitDestination (was a plain copy of basic)
1440 * Make JniException a runtime exception, so that there is no need to
1441 declare the fact that you may encounter such a beast. I guess that
1442 nobody will ever want to survive such error.
1443 * Create specific errors for each MSG case of failure:
1444 host failure, transfer failure, timeout, task cancelled
1445 * Cleanup the exceptions that may get thrown by each function
1446 * Other internal cleanups in Java bindings. Performance still bad :/
1447 Ruby and Lua Bindings: create them
1448 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1449 * That's new and great, you should try them out.
1450 Same functionalities than Java bindings, only even less polished
1452 * Kill the useless "rate" argument of SD_task_get_execution_time()
1453 Everyone used to provide -1 as a value, it was not used, and the
1454 semantic of a possible use wasn't even clear.
1455 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1456 either as comm costs or compute costs to mean that there is no
1457 such thing for that specific task.
1458 * Add a SD_task_set_name() function
1459 * Fix SD_task_unschedule() on typed tasks
1460 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1461 * In DAX loader, accept useless 'level' attributes to <job> since
1462 LIGO DAGs have them (seem to be to ease graphical representation).
1464 * Add an example masterslave_mailbox.c using send/receive and not
1465 the deprecated put/get interface.
1466 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1467 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1469 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1470 (declare MSG_USE_DEPRECATED to still have the old name)
1471 * Add a MSG_task_set_data() function
1472 * About trace replay (see examples/msg/actions):
1474 - Allow to work with splitted trace files for each process
1475 Give the specific trace file as argument of each process,
1476 and call MSG_action_trace_run(NULL)
1477 You can still have one merged file for all processes.
1478 - Fix implementation of collective operations
1479 * Allow task_execute() on 0-sized tasks (closes #10063)
1481 * This is the first release of SimGrid where SMPI is not considered
1482 beta anymore (even if some corners should still be improved)
1483 * Port over the new SIMIX_network submodule (internal refactoring)
1484 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1485 * Implement more missing elements of the standard:
1487 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1489 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1490 MPI_Scan() MPI_Get_processor_name()
1491 - Added implementation of missing case for Alltoall (warning: it's
1492 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1493 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1494 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1495 SMPI_MPI_Allgatherv()
1496 * Bug fixes include:
1497 - MPI_Waitsome() was broken
1498 - Allow relative includes in smpicc
1499 - Command line cfg argument 'reference_speed' was ignored...
1500 - Some functions did not properly lead to auto-benching of user code
1501 - smpicc passes -O2 by default (just like openmpi one)
1503 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1504 * Bug fixes about timeouts during communications
1505 * add SIMIX_message_sizes_output() as a pimple to write to file the
1506 amount of messages per size. Use gnuplot to get histogram.
1507 Pimple because that's the only user-visible function of simix,
1508 defined directly in xbt.h (irk, sorry)
1510 - Add a SIMIX_sem_get_capacity() function
1511 - Fix interactions with processe resume/suspende
1512 - release_forever() was stupidly broken
1513 - Fix SIMIX_display_process_status() for processes in a semaphore
1514 - Make SIMIX_sem_block_onto() user-visible
1515 * Refactoring context stuff:
1516 - Use pseudo-OOP for better modularity
1517 - reimplement SIMIX_process_kill() without process_schedule() so
1518 that the latter can take as invariant that it is called from
1520 - Merge context_start into context_new for sake of simplicity
1522 * Add a Vivaldi network model, coded live during SUD'10 ;)
1523 * Rename configuration variables to start a hierarchy:
1524 o cpu_model -> cpu/model
1525 o network_model -> network/model
1526 o workstation_model -> workstation/model
1527 * New configuration variables:
1528 o network/bandwidth_factor: correction to bandwith
1529 o network/latency_factor: correction to latency
1530 o netwotk/weight_S: correction to the weight of competing streams
1531 * Add a long description to the models, that users can see with such
1532 argument on the command line: --cfg=cpu/model:help
1533 * --help-models display the long description of all known models
1535 * config: add the ability to set a default value after registration
1536 Does not override any previously set value (e.g. from cmd line)
1537 * dict: allow to have integer key and data.
1538 When so, you need to use the following functions
1539 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1540 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1541 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1542 In contrary to regular dicts, the key is not malloced before copy.
1543 Mixing scalar and regular elements in the same dict is not tested
1545 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1546 Tracing for Visualization:
1547 * SimGrid is now instrumented in order to generate a trace file for
1548 visualization analysis: to use it, need to compile SimGrid with the
1549 "tracing" option enabled, and instrument the program using SimGrid with
1550 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1551 (among other functions).
1552 * The instrumentation only traces the platform utilization for now
1553 * Documentation to use the tracing functions and how to analyze the
1554 traces with the Triva tool is written.
1555 * More information about: SimGrid FAQ (in the section Tracing Simulations
1558 * We moved to cmake as default build system. Autotools support will
1559 be dropped soon. Check the FAQ for more info about how to use it.
1560 * Greatly improved our cdash/ctest interactions
1561 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1562 * Added memory checking tests with valgrind; lot of memleak fixing.
1563 This may be the first release of simgrid with so few memory issues
1564 * Added code coverage tests.
1565 Our coverage is still improvable, but at least we see it on cdash.
1567 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1569 SimGrid (3.3.4) stable; urgency=low
1571 The "Desktop Grid needs love too" release (also called Xmas release).
1573 Models improvements:
1574 * Major speedup in the maxmin system solving by using lazy evaluation
1575 Instead of solving completely the maxmin system at each iteration,
1576 only invalidate (and recompute) the modified parts.
1577 This new feature is enabled in default models but you can try to
1578 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1579 * Cas01 IMproved as default CPU model
1580 This CPU model is the same Cas01 model, but it uses the
1581 maxmin-selective-update flag and a heap structure to manage
1582 actions on SURF kernel.
1583 It reduces the complexity to find the next action to finish and,
1584 consequently, it's faster than the old Cas01.
1585 This is the new default CPU model (Cas01).
1586 * Rename the old Cas01 model to Cas01_fullupdate
1587 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1588 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1589 * CpuTI (CPU Trace Integration)
1590 A new CPU model whose objective is simulate faster when using
1591 availability trace files.
1592 Instead of using a full featured, over engineered maxmin system for
1593 CPU modeling, this model does the pre-integration of traces files
1594 to calculate the amount of CPU power available, and so, executes
1595 faster than the old CPU models.
1596 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1597 * Use LV08 as default network model since it gives better accuracy
1598 for small messages and shouldn't change things for big ones.
1599 Use --cfg=network_model:CM02 to get the previous behavior.
1602 ******************************************
1603 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1604 ******************************************
1605 * The new CPU model may changes simulations!
1606 The point is that events occurring at the exact same timestamp
1607 are not scheduled in the same order with the old and new
1608 version. This may be enough to completely change the execution
1609 of simulations in some cases.
1610 * The new network model will change simulations!
1611 This new model is more realistic than the previous one, so you
1612 should consider redoing your old experiments with this model.
1613 Sorry for the inconvenience.
1616 * Introduce the supernovae compilation mode
1617 When compiled that way, the whole SimGrid (or almost) is put in a
1618 single compilation unit and compiled in one shoot.
1619 This is to help gcc which has difficulties to inline stuff from one
1621 The speedup seem to be above 15%, althrough more tests are needed on
1622 amd64 to confirm that gain.
1625 * Port of MSG's mailbox on top of SIMIX network
1626 The put/get mechanism was greatly simplified on the way.
1629 * New SIMIX network module. Provides:
1630 - Mailbox: rendez-vous mechanism to find with who you want to speak
1631 - Synchronous send/recv: easier and hopefully faster since the
1632 logic is handled in the maestro process directly now
1633 - Asynchronous send/recv: you dreamt of it? It's here now
1634 Too bad that nobody cared enough to propagate the change to MSG.
1635 * Add semaphores as SIMIX synchronization mechanism.
1638 * new function SD_daxload(char*) to load a DAX file
1639 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1640 * Introduce typed tasks. Specify its kind and cost at creation.
1641 At scheduling, just give where it should be placed, and the cost
1642 for each involved resource is automatically computed.
1643 Existing constructors so far (more to come of course):
1644 - SD_task_create_comm_e2e() for end-to-end communication
1645 - SD_task_create_comp_seq() for sequential computation
1646 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1647 * new function SD_task_dump() for debuging display
1648 * new function SD_task_dotty(task,FILE*) writing to file the info
1649 about the task in dotty format
1650 * SD_task_dependency_exists() can now cope with having one of its
1651 arguments NULL. If so, it tests whether the other argument has any
1653 * Add getters on list of preceding/following tasks:
1654 SD_task_get_parents(task) and SD_task_get_children(task)
1655 * Add getters on amount of workstations and list:
1656 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1657 * Add getter on task kind: SD_task_get_kind(task)
1658 * Update the start_time and finish_time of tasks on completion/failure
1659 * Bugfix: Remove task from state swags when destroyed
1662 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1665 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1666 dynar with one separate thread per value of the dynar.
1667 * Change the prototype of xbt_thread_create(), sorry.
1668 Added a boolean parameter indicating whether we want to join this
1669 thread (used in SG only for now)
1670 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1673 * GTNetS wrappers should now be usable again (and betterly tested too)
1674 * Fix a major regression from 3.2 where the timeout provided to
1675 MSG_task_put_with_timeout() was used as absolute time before which
1676 the comm should be done.
1677 * Start to fix the <cluster> tag.
1678 - Internal links should be good now (beside of the loopback, which
1679 use the private link instead)
1680 - paths to the external world is still rather broken
1681 - the <route:multi> tag is just broken. Actually that's brain-dead.
1682 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1683 to make it less stupid
1684 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1685 * Fix a source-level compatibility glitch from 3.2: after defining
1686 MSG_USE_DEPRECATED, you can use the old name
1687 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1688 * Allow to compile from the SVN with automake 1.11
1689 * Fix some problems when using the "start_time" tag in deployment XMLs.
1690 * Fix #8569: XBT/synchro.h has redundant declarations
1691 * Fix #8563: MSG return values and exceptions
1692 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1693 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1694 Thanks to Robson Peixoto.
1695 * Fix a vicious bug in dictionaries inducing that some elements were
1696 not freed on xbt_dict_free()
1698 Portability report of this version:
1699 * Main portability targets:
1700 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1701 on (amd64/i386/ia64)
1702 - mac leopard on i386
1703 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1704 but nothing critical.
1705 * Other platforms: windows, AIX and others were not tested for this release
1707 Timing report of this version:
1708 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1709 scenario parameters). From 8h to a few seconds in desktop grid settings.
1710 * Supernovae brings about 25% speedup on i386.
1712 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1714 SimGrid (3.3.3) stable; urgency=low
1716 The "Need for Speed" release.
1718 The timings done to validate the 3.3.2 were faulty.
1719 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1721 The problem was a conversion from a manually handled vector to
1722 xbt_dynar_t on the critical path.
1723 xbt_dynar_foreach calls functions, inducing stack management crap.
1725 We inlined these functions and xbt_dynar_foreach is now breath taking.
1726 We also inlined xbt_swag_belong on the way.
1728 Here are some approximate speedup measurements (on master/slaves
1729 simulations lasting between 10s and 20s each):
1730 3.3.1 -> 3.3.2: about same performance
1731 3.3.2 -> 3.3.3: 40% speedup
1732 3.3.1 -> 3.3.3: 40% speedup
1733 3.3.1 with inline patch -> 3.3.3: 30% speedup
1735 Our reading is that the refactoring which occurred in 3.3.2 made us
1736 suffer much more from the xbt_dynar_foreach low performance, but
1737 once we solved this, this refactoring proved to be very performance
1738 effective. From the 40% speedup, somehow, 10% are due to the
1739 inlining and 30% to the refactoring.
1741 That's a pitty that gcc cannot inline functions placed in other files
1742 alone. We have to choose between:
1743 - break the encapsulation (by putting private data structures and
1744 accessors in headers files to help gcc)
1745 - live with low performance
1746 - switch to a decent compiler such as icc (not quite possible).
1748 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1750 SimGrid (3.3.2) stable; urgency=low
1752 The "Simplicity does not preceed complexity, but follows it" release.
1754 The main contributors of this release were (lexical order):
1755 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1758 * Extract the routing logic into its own object.
1759 (was dupplicated in network.c and workstation_LV07.c;
1760 Allows to implement other ways of storing that info)
1761 => kill now useless network_card concept
1762 - Use dynar to represent routes (instead of void** + int*)
1763 - kill link_set (use surf_network_model->resource_set instead)
1764 - Add a command-line option to choose the routing schema to use
1765 - Add three new models:
1766 * Floyd (shortest path computed at initialization)
1767 * Dijikstra (shortest path recomputed all the time)
1768 * Cached Dijikstra (shortest path computed on need)
1769 All these models where contributed by Silas De Munck, and are
1770 described in his ICCS09 paper.
1772 * Simplify model declaration
1773 (less redirections, less function to write when defining a model)
1774 - Factorize stuff between models:
1777 surf_model_resource_set(model)
1778 surf_model_resource_by_name(model, name)
1779 - Unify the types of models in s_surf_model_t (using an union)
1780 - Embeed fields of common_public directly into s_surf_model_t
1781 - Rename model methods:
1782 action_free ~> action_unref
1783 action_change_state ~> action_state_set
1784 action_get_state ~> action_state_get
1785 - Change model methods into functions :
1786 (model)->common_public->action_use ~> surf_action_ref
1788 * Implement a generic resource; use it as ancestor to specific ones
1789 (allows to kill duplicated code in models)
1790 Drawback: timer command don't need no name nor properties;
1791 workstation_CLM03 don't need no properties
1792 (but I guess we can live with those few bytes wasted)
1794 * Improve the action object model
1795 - implement a constructor avoiding dupplicated code about field
1796 initialization in generic_action part.
1798 * Kill the SDP model: it has an external dependency, is deprecated
1799 in flavor of modern lmm models, and didn't compile since a while
1802 * Relocation of the context module from XBT to SIMIX.
1803 (the context were decoupled from the simix processes, duplicating a lot of code)
1804 => a lot of code was factorized
1805 - less overhead is introduced during scheduling
1806 - simpler API for the context factory
1807 - the logic for process creation,destruction and manipulation was simplified
1808 * Simplification of the s_smx_process_t data structure.
1809 => accesing the simix level data associated to a process is faster now,
1810 and the code is a lot more readable.
1813 * Implement some more MPI primitives:
1814 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1815 -implementation: Bcast: flat or 2-ary tree (default),
1816 Barrier: 4-ary tree,
1818 Allreduce: Reduce then Bcast
1819 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1820 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1821 Alltoallv: flat tree, like ompi
1823 * Add support for optimized collectives (Bcast is now binomial by default)
1824 * Port smpirun and smpicc to OS X
1827 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1828 Shout out if you used it.
1831 * Display the list of still queued messages in SG mode when existing
1835 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1836 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1837 * Add xbt_str_from_file(FILE*)
1838 * Add xbt_dict_get_key achieving a linear reverse search
1839 * Remove the context module
1841 Portability report of this version:
1842 * Main portability targets:
1843 - Linux(debian)/x86/context
1844 - Linux(debian)/x86/pthreads
1845 - Linux(debian)/amd64/context
1846 - Linux(debian)/amd64/pthreads
1847 On these, we still have the eratic breakages of gras/pmm and
1848 amok/saturate_sg reported in previous version. We still think
1849 that the tests are the cause of the fault, not the tested code.
1851 - Mac OSX Leopard/x86/context
1852 Still false negative in tesh autotesting.
1853 Smpi still fails, but this time because readlink does not accept -f
1854 Everything seems to work properly beside of that.
1857 - AIX version 5.3 (only tested contexts this time)
1858 Smpi still fails there because mktemp is not installed.
1859 Everything seems to work properly beside of that.
1861 I managed to compile it for the first time, but several breakages.
1862 Won't delay the release for this exotic platform.
1864 * Windows: it's still lagging behind. If you want to help, please
1867 Timing report of this version:
1868 This version seem to be more than 5% faster than 3.3.1 (on linux
1869 64bits with contextes). The gain is less than expected, we are
1870 investigating this for next release.
1872 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1874 SimGrid (3.3.1) stable; urgency=low
1877 * Implement a --cfg-help to show existing configuration variables
1878 * Build chain do not require doxygen in maintainer mode
1881 * fix a bug on struct sizeof computation, which prevented the
1882 exchange of arrays of structs in some conditions
1883 - added a regression test about this in datadesc_usage
1884 * Allow the exchange of 0-long dynamic vectors.
1885 - for that, use -1 as indicator of dynamic size instead of 0
1886 - This implied to change any size from unsigned long to long,
1887 reducing a bit communication abilities, but I guess that with
1888 64bits being quite common, this is more than enough.
1889 - This also induce a protocol change, thus bumping network protocol
1890 version from 0 to 1 (if we have external users, we have to get
1891 clean on that point too ;)
1892 - added two regression tests about this in datadesc_usage
1893 * Be more verbose when propagating local exceptions
1894 This helps debugging.
1895 * Display the status of simulated processes when receiving SIGINT in
1899 * Allow to control the simulation from a trace file.
1900 New functions MSG_action_register() and MSG_action_trace_run()
1901 The first one allows to associate a function execution to each
1902 kind of action while the second one parses a trace file and
1903 triggers the corresponding actions within the system.
1904 For now, only a toy example is provided in examples/msg/actions
1905 * Add an exemple of process migration in examples/msg/migration
1906 * Fix a bug in task exchange which broke MSG_task_get_sender()
1907 Add a teshsuite regression test for that.
1908 [Bug: if MSG_task_get_sender() is called after sender exit,
1910 * Fix a bug which prevented suspend/resume to work properly
1911 * Display the status of simulated processes when receiving SIGINT
1912 This fixes a regression of v3.3. due to the introduction of SIMIX
1913 * Bug fixing in failure management:
1914 - trace could not start by a failure at time 0
1915 - failure during communications were not working
1918 * Add SIMIX_process_set_name() to change the name of the current
1919 process in the log messages.
1920 * Store smx_hosts in a dict since we only retrieve them by name
1921 * Move the configuration infrastructure to surf
1924 * Move the configuration infrastructure to surf
1927 * Massive internal cleanups:
1928 - Store internal structures on processes instead of hosts (allows
1929 to have more than one process per host, in addition of being more
1931 - Cleanup the initialization/finalization process
1932 - Kill a whole bunch of unneeded synchronization:
1933 processes run in exclusive manner within the simulator
1934 - Move queues from global tables to process data fields
1936 - now accept -platform and -hostfile arguments
1937 - Pass the right rank value to processes according to the hostfile
1938 * Compile the examples by default, and use them as regression tests
1939 * Implement MPI_Wtime()
1940 * Change the reference speed to a command line option
1943 * TCP_gamma can now be specified as command line option using
1944 --cfg=TCP_gamma:10000000.0
1945 * Change the --surf-path cmd line option into --cfg=path:
1948 * Also include strbuff from xbt.h public header
1949 * xbt_ex_display(): do not free the exception after displaying
1950 This allows to do more with the given exception afterward.
1951 Users should call xbt_ex_free() themselves.
1955 Portability report of this version:
1956 * Main portability targets:
1957 - Linux(debian)/x86/context
1958 - Linux(debian)/x86/pthreads
1959 - Linux(debian)/amd64/context
1960 - Linux(debian)/amd64/pthreads
1961 These targets fail about 1/10 of times on gras/pmm, but we believe
1962 that this is because of the test, not because of simgrid.
1963 amok/saturate_sg fails even more rarely, and the test may not be
1966 - Mac OSX Leopard/x86/context
1967 The test suite still spits tons of errors because some obscure
1968 force prevents us from removing the temporary directories
1969 arguing that they still contain some metadata I've never heard of.
1970 Smpi fails because seq is not installed.
1971 Everything seems to work properly beside of that.
1974 - AIX version 5.3 (both contexts and pthread)
1975 Smpi still fails there because mktemp is not installed.
1976 XML inclusions seems rosty on AIX.
1978 * Windows: it's still lagging behind. If you want to help, please
1981 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1983 SimGrid (3.3) stable; urgency=high
1987 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1988 [Malek Cherier & Mt]
1990 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1993 Until now, GRAS were using MSG as an interface to SURF. It was
1994 quite difficult because both interface have several differences
1995 (MSG channels vs GRAS sockets were the most notable point).
1997 This also opens the gate to SMPI (which should occur soon) and speed
1998 up simulations by to 40% (even if it were not the main goal).
2000 **************************************
2001 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2002 **************************************
2003 The point is that events occuring at the exact same timestamp are
2004 not scheduled in the same order with the old and new version. This
2005 may be enough to completely change the execution of simulations in
2006 some cases. Sorry for the inconvenience.
2008 * Cleanup and upgrade the XML format to push further scalability
2009 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2011 * Improve the testing infrastructure with tesh. Now a very large part of
2012 the code is tested not only by being run but also by checking that the
2013 output match an expected output [Mt].
2015 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2016 is really less memory-demanding, which should allow you to use
2017 larger files in SimGrid [AL].
2019 * Inform valgrind about our contextes, so that it becomes usable
2020 with the default (and more effecient) version of SimGrid
2021 [contributed by Sékou Diakite, many thanks]
2024 * Introduce a listener thread in charge of receiving incoming
2025 messages from the network. It allows to overlap communication and
2026 computation but most notably, it removes some stupid deadlocks due
2027 to the fact that so far, a process could not send and receive at
2028 the same time. This made most non trivial communication schema
2030 * Convert the PIDs from long int to int to match the MSG ones (and
2031 linux ones too) [Mt]
2032 * New function: gras_agent_spawn() to launch a new process on
2033 current host. Only working in simulation for now. [Mt]
2034 * New function: gras_os_hostport() returning a constant form (ie,
2035 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2038 * Make the backtrace of exceptions more human readable [Mt]
2039 * New module: xbt/str [Mt]
2040 a ton of string utility functions (split, join, printf to a newly
2041 allocated buffer, trim, etc)
2042 * New module: xbt/hash [Mt]
2043 SHA1 hashing algorithm (more to come if needed)
2044 * New module: xbt/synchro [Mt]
2045 synchronization tools (mutex and conditions) working the same way
2046 in simulation and in real life (mainly useful for GRAS, but not
2048 * New module: xbt/queue [Mt]
2049 classical producer/consumer synchronization scheme
2050 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2051 (using the classical functions will get serialized) [Mt]
2052 * Make dictionary internal table dynamic. No need to specify its size
2053 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2055 * Make sure the log channels are organized as a tree under windows
2056 (because of ANSI C compatibility issue, any channel were child of
2057 root directly) [Mt].
2060 * Cleaned many thing in surf and fixed a few bugs [AL].
2061 * Add a nice command line configuration mechanism to compose models [AL].
2062 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2063 the previous one (KCCFLN05). It relies on something that looks like
2064 a max-min sharing mechanism but cannot be written as such. A new solver
2065 was thus designed [AL].
2066 * Add a new solver to lmm. Based on Lagrange optimization and
2067 gradient-based descent, it enables to efficiently maximise systems s.a
2069 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2071 This solver enables to propose two new network models for TCP Reno and
2072 TCP Vegas based on Low's work. These models still need to be fully
2073 tested though [Pedro Velho].
2076 * Bug fix in SD_simulate. Now the time bound given as argument is
2078 * Use the new parallel task model (ptask_L07) as default.
2079 * Use the SURF command line configuration mechanism.
2080 * 0-size tasks (for synchronization) should now work.
2082 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2084 SimGrid (3.2) stable; urgency=high
2088 We still experience issues on this platform, but we believe that at
2089 least MSG is usable.
2091 GRAS API BREAKAGE (for simplification purpose, sorry):
2092 * the gras_msgtype_by_name is not used anymore. Instead of
2093 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2094 you can write (and must)
2095 gras_msg_send(toserver, "request", &request);
2096 - If you still want to pass a gras_msgtype_t to the function (to cache
2097 the type and avoid the lookup time), use the gras_msg_send_() variant.
2098 - Impacted functions:
2099 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2100 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2101 * The callbacks are now expected to return 0 when everything went well
2102 (just like the main() function)
2104 GRAS new features and improvements:
2105 * New module mechanism where user code can use per process globals [Mt]
2106 This is similar to gras_userdata_*() functions, but for libraries. It
2107 factorize some code developped over and over in the examples and AMOK.
2108 It has still to be documented and used (only amok/peermanagement is
2110 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2111 loss when they were small enough to fit into the buffer and sent quickly
2112 enough so that they can all get received in one shoot.
2113 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2114 if not found. Use the *_or_null() variant for the old semantic.
2115 * In gras_msg_handle, do not discard messages without callback.
2116 They are probably messages to be explicitly awaited later (ie, proofs of
2117 mis-synchronization in userland since they are sent before being awaited)
2118 No big deal usually.
2119 * gras_socket_meas_send/recv: semantic changed!
2120 The numerical arguments used to be (1) the total amount of data to send
2121 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2122 messages. This was need for the fool willing to send more than MAXINT
2123 bytes on quite fat pipes.
2126 * Do really rename the hostmanagement module to peermanagement. [Mt]
2127 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2128 but this is rather new and this was documented in the module
2129 documentation (poor excuses, I admit)
2130 * Bandwidth measurement semantic changed! This follows the changes to
2131 gras_socket_meas_send/recv explained above.
2134 * A sequential mode has been added to the workstations. When a workstation
2135 is in sequential mode, it can execute only one task, and the other tasks
2136 are waiting in a FIFO. [Christophe Thiery]
2139 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2140 model for SIMDAG. [Christophe Thiery]
2141 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2142 equal to 0, which caused some bad behaviors in
2143 saturated_constraint_set_update. I now use a threshold mechanism like in
2147 * When running manually src/testall, you select specific units [Mt]
2148 testall is the result of our cunit mechanism, and should replace all
2149 the scripty thingy around since bash don't run easily on billware.
2151 * A mallocator system has been added. [Christophe Thiery]
2152 Mallocators allow you to recycle your unused objects instead of freeing them
2153 and allocating new ones.
2155 Documentation update:
2156 * FAQ reworking + New FAQs:
2157 - "Valgrind spits tons of errors!" [Mt]
2158 - "How to repport bugs" [Mt]
2159 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2160 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2161 - Communication time measurement within MSG [AL]
2162 - I experience weird communication times when I change the latency [AL]
2163 * GRAS tutorial [Mt]
2165 - an introduction to the framework and to the used communication model
2166 - an initiatic tour introducing the most proheminent features:
2168 . Lesson 0: Installing GRAS
2169 . Lesson 1: Setting up your own project
2170 o Part 2: Message passing
2171 . Lesson 2: Exchanging simple messages
2172 . Lesson 3: Passing arguments to the processes (in SG)
2173 . Lesson 4: Attaching callbacks to messages
2174 . Lesson 5: Using globals in processes
2175 . Lesson 6: Logging information properly
2176 . Lesson 7: Using internal timers
2177 . Lesson 8: Handling errors through exceptions
2178 . Lesson 9: Exchanging simple data
2179 . Lesson 10: Remote Procedure Calling (RPC)
2180 . Lesson 11: Explicitely waiting for messages
2181 . Recapping of message passing features in GRAS
2182 - A HOWTO section containing:
2183 o HOWTO design a GRAS application
2184 More are due, of course. They will come latter. In the meanwhile, you can
2185 check the examples which are still here.
2187 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2189 SimGrid (3.1) stable; urgency=high
2193 There was a stack corruption somewhere, visible only when optimizing
2194 with these versions. [Vince]
2197 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2198 module that enabled to play with Directed Acyclic Graphs. It is built
2199 directly on top of SURF and provides an API rather close to the old
2200 SG. Some old codes using SG are currently under rewrite to check that
2201 all needful functions are provided. [Christophe Thiery]
2204 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2205 extension of the classical CLM03 model that gracefully handles
2206 failures. This is now the default model for MSG and GRAS. It doesn't
2207 handle parallel tasks yet though. [AL]
2208 * Bug fix: Weights were not correctly set in the network part.
2209 WARNING: This may have resulted in incorrect results with simulations
2210 where there are more than one flow on a given link. [AL]
2213 * After a (long ?) discussion on simgrid-devel, we have decided that the
2214 convention we had on units was stupid. That is why it has been decided
2215 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2216 WARNING : This means that all previous platform files will not work as
2217 such with this version! A warning is issued to ask users to update
2219 A conversion script can be found in the contrib module of the CVS, under
2220 the name contrib/platform_generation/surfxml_update.pl [MQ]
2223 * Bug fix: Processes were started in reverse order, wrt deployment file.
2224 WARNING: if your code relies on this bug, please fix it. [AL]
2225 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2226 being executed on two different locations. [AL]
2227 * Bug fix: Failures are now better supported thanks to Derrick's tests
2228 (there was many failure situations I hadn't thought of and that weren't
2229 correctly handled). [AL]
2230 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2234 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2235 Exception occurring on server-side are propagated back to client (!).
2237 API CHANGE: the callback changed their prototype. Change:
2238 int my_handler(gras_socket_t expeditor, void *payload_data) {
2240 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2241 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2243 * New! function: gras_msg_handleall to deal with all messages arriving
2244 within a given period.
2245 * New! function: gras_socket_server_range to get a server socket in a
2246 range of port numbers (ease to avoid port number conflicts) [MQ]
2247 * New! gras processes display their backtrace when they get a SIGUSR1
2248 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2249 Sweet to debug RL processes [MQ]
2253 - Do not force experiment sizes to be expressed in kb, or it becomes
2254 impossible to measure the latency this way (needs one byte-long tests)
2255 WARNING: this changes the amok_bw_* function semantic. [MQ]
2256 - Implements the link saturation stuff. [MQ]
2257 * Peer management module:
2258 New! module factorizing code that we wrote over and over [MQ].
2261 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2262 - Test units are placed directly into the library code, they get extracted
2263 automatically and placed into the src/testall binary.
2264 - Convert most of the XBT tests to this system.
2265 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2266 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2267 live in the same C code. This required to change a little bit the API
2268 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2269 * New module: sparse graph structure with basic algorithms (this is work
2270 in progress and the API is not considered to be frozen yet). [AL]
2271 * Display more information on backtraces: source line & function names are
2272 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2273 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2274 * Reworked a little bit some #include statements to load only required
2275 headers. Some user code that relied on SimGrid to include stdlib or
2276 stdio may need to include it by themselves. [AL]
2277 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2279 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2280 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2281 but I guess that almost nobody used those parts. [MQ]
2283 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2285 SimGrid (3.0.1) stable; urgency=low
2288 * Unfortunately, I had missed 5 misnamed functions:
2289 xbt_fifo_item_t xbt_fifo_newitem(void);
2290 void xbt_fifo_freeitem(xbt_fifo_item_t);
2291 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2292 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2293 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2294 They're now deprecated. Please use their new versions:
2295 xbt_fifo_item_t xbt_fifo_new_item(void);
2296 void xbt_fifo_free_item(xbt_fifo_item_t);
2297 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2298 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2299 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2301 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2302 * Documentation: xbt_log module unmercifully reworked [MQ]
2303 * Bugfix: there was a problem with the ending of contexts with
2304 the pthread backend. It caused some weird deadlock or behavior
2305 depending on the pthread implementation. [AL]
2306 * Bugfix: get the exceptions raised in the simulator repport where
2307 and why they come from when they are not catched in time [AL, MQ]
2310 * Bugfix: Do repport the error when two non-connected hosts try to
2311 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2314 * Add additionnal checkings on communications. Assert that two
2315 communicating hosts are connected by a set of links... [AL]
2318 * Add additionnal checkings on channel values in communication [AL]
2319 * New: MSG_task_get_source to see on which host a task was generated [HC]
2320 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2321 the number of tasks waiting to be received on channel and sent
2323 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2324 waits for the first task coming from a given host.. [AL]
2326 GRAS new functionnalities: [MQ]
2327 * Enhance the parsing macro to allow the size of multidimentional objects
2328 to be given thru annotations.
2329 * New example (and documentation): Matrix Multiplication a la RPC
2330 (as when I was young!) and fix a bunch of bugs found on the way.
2332 GRAS performance improvements: [MQ]
2334 * Reduce the amount of cbps creation/destruction by making it static to
2335 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2338 * Change libdata to a set so that we can search for stuff by ID (and thus
2339 reduce the insane amount of dict lookups)
2342 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2343 RL communication on top of SG since it's so uneffective.
2344 It may also be used for inter-thread communication in RL, one day.
2345 * Use gras_datadesc_copy() to exchange messages on top of SG
2347 - improve message exchange performance on top of SG
2348 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2349 * Don't exchange on the network the size of the used part of buffer,
2350 instead, specify the possible buffer size to read().
2352 - reduces the amount of read/write calls (one pair per exchange)
2353 - reduces the amount of exchanged data (the size)
2354 - allows to retrieve all arrived data on receiver side, if we don't need
2355 it right now (subsequent read will peek the buffer)
2356 - allows the receiver to proceed with the begining of the stream before
2357 everything is arrived
2358 - make it possible to build an iov transport (using readv/writev)
2360 - take care of the data with non-stable storage (like stacked data),
2362 * If possible, TCP send uses vector I/O (when writev() is here)
2363 - Don't use it for receive since we send data sizes and data on the
2364 same stream, so we wouldn't be able to chain large amount of chunks
2365 before having to flush the stuff to read the size.
2366 * Rework the transport plugin mechanism to simplify it and reduce the
2367 amount of pointer dereferencement when searching for the right function
2370 * I guess that now, we do almost as few system calls as possible while
2371 doing as few data copy as possible.
2373 To improve it further, we could try to send all the sizes first and then
2374 all the data (to use iov on receiving size), but it's only a partial
2375 solution: when you have 2 dimensional data, the sizes of the second
2376 dimension is data of the first dimension, so you need 3 streams.
2378 I'm not sure the potential performance gains justify the coding burden.
2380 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2382 SimGrid (3.00) stable; urgency=high
2385 * New! Give the possibility to hijack the surf parser and thus bypass
2386 MSG_create_environment and MSG_launch_application. Have a look at
2387 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2389 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2391 SimGrid (2.96) unstable; urgency=low
2396 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2397 This deprecates the xbt_error_t mechanisms.
2398 It modifies (simplifies) all XBT and GRAS API.
2399 MSG API keeps unchanged (exceptions raised by XBT are catched from
2400 within MSG and masked with existing error handling facilities)
2403 * New! Add a FATPIPE model. [AL]
2404 * New! Add a parallel task model. [AL]
2405 * New! Add automatically a loopback interface (in the default
2406 network model) if none was precised.
2409 * Bugfix: MSG_process_resume now works with the current running process.
2411 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2412 * Modification of MSG_task_get_compute_duration. Once a task has been
2413 processed, the value returned by this function is now equal to 0. [AL]
2414 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2415 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2416 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2417 of a m_task. For now, MSG_task_cancel only works with computation
2419 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2420 of the processor (in Mflop/s) regardless of the current load on the
2422 * API Change: use proper naming convention for MSG_getClock and
2423 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2425 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2426 This function changes the priority of a computation task. This priority
2427 doesn't affect the transfer rate. A priority of 2 will make a task
2428 receive two times more cpu power than the other ones. This function
2429 has been added to suit the needs of Nguyen The Loc and hasn't been that
2430 much tested yet. So if it fails, please report it and send me your code.
2432 * API Change: removed all functions and types that were marked "deprecated"
2433 since many months. Renamed MSG_global_init_args to MSG_global_init.
2435 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2437 SimGrid (2.95) unstable; urgency=low
2440 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2441 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2442 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2445 * Bugfix: Allow absolute paths for platform description files [MQ]
2446 * Bugfix: do free the variables after use. Leads to drastic performance
2448 * Implement max_duration (ie, timeouts) on resources [AL]
2451 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2453 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2454 the needed support (timeouts) [AL]
2457 * Implement measurement sockets. You can now get the bandwidth between two
2458 hosts thanks to AMOK (see below). [MQ]
2459 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2460 dynar over the network (yeah) [MQ]
2461 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2462 * simple/static token-ring example. [Alexandre Colucci and MQ]
2463 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2464 in place before [MQ]
2467 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2468 type parsing. "short" and "long" modifiers were also ignored; other
2469 modifier (such as reference level) are still ignored. [MQ]
2470 * Embeed the buffer size within the buffer itself on SG. [MQ]
2471 That way, send() are atomic and cannot get intermixed anymore (at least
2472 the ones which are less than 100k; bigger messages still have the issue)
2473 * Array size pushed by the field, not by the field type (or each
2474 and every long int will push stuff to the cbps) [MQ]
2475 * use select() to sleep since it allows to portably sleep less than one
2478 GRAS (minor cleanups)
2479 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2482 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2483 * type category 'ignored' killed as it was never used and were difficult
2485 * whether a type can cycle or not is now a flag, leaving room for more
2486 flags (as "ignored", if we feel the need one day ;)
2487 * Rename raw sockets to measurement sockets since "raw" has another
2488 meaning in networking community.
2491 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2492 and offers features not belonging to GRAS but that most applications
2493 need. One day, it may be a set of plugins loadable at runtime.
2494 * New module: bandwidth
2495 bandwidth measurement between arbitrary nodes running this module. [MQ]
2497 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2499 SimGrid (2.94) unstable; urgency=low
2501 The first beta release of SimGrid 3 !
2505 * Update the main page and the FAQ. Adding references to gforge.
2508 * Add a gras_os_getpid function.
2511 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2512 * Extend the logs so that they also print PID, hostname, date, ... if
2514 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2515 and kill the old version which were in testsuite/
2516 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2520 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2521 for users. Instead, each programming environment (GRAS, MSG,...) use it
2522 when needed to provide such a feature to users.
2523 Don't shortcut the mechanism or you will also shortcut the virtualization
2524 you need on the simulator.
2528 * Cleanups in configury with regard to compile optimization/warning flags.
2529 Also add -fno-loop-optimize to any powerpc since it's the optimization
2530 killing gcc (< 3.4.0).
2531 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2532 needed by MSG examples complications
2533 * Borrow configury beautifications from PHP
2536 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2537 warning (thanks loris for stumbling into this one).
2538 * Bugfix: stop loading private headers (gras_config.h) from the public
2542 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2543 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2544 RL would imply a DNS resolver, which is *hard* to do in a portable way
2545 (and therefore delayed).
2546 * Implement a real timer mechanism and use it in timing macros. This allows
2547 to avoid rounding errors and get a 0.000005 sec precision in timing
2548 macros. While I was at it, various cleanups:
2549 - allow to declare more than one timed section per file (fix a stupid bug)
2550 - move some private declaration to the right place
2551 - merge conditional execution and timing macros into emulation module
2552 - document the module
2553 - make sure the module cleanups its mess on gras_exit
2554 * Documentation improvements:
2555 - (new) how to compile applications using GRAS
2556 - (new) emulation support (timing macros)
2558 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2560 SimGrid (2.93) unstable; urgency=low
2562 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2565 - Use Paje properly where used. Still to be sanitized properly.
2566 - Portability fix: Add an implementation of the contexts using pthread
2570 - Add xbt_procname(): returns the name of the current process.
2571 Use it to show the current process's name in all logging.
2573 - fix detection of older flex version and the reaction, since we do
2574 depend on modern ones (we use lex_destroy)
2575 - Better separation of SG and RL in the libs: remove all simulation code
2576 from libgras. As a result, this lib is now only 200k when stripped.
2577 Some of the xbt modules may also be duplicated (two sets and such) and
2578 should be cleaned/killed before SG3.
2579 - Insist on using xlC on AIX because of weird problems involving gcc there.
2580 - Cleanup the make remote stuff. This is now done by scripts
2581 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2582 mainly for our private use, but we're working on changing them to user
2585 - Bugfix: flush the socket on close only if there is some *output*.
2586 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2588 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2591 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2593 SimGrid (2.92) unstable; urgency=low
2595 Alpha 3 on the path to SimGrid 3
2599 - New! First try of benchmarking macros.
2600 - New! First try so that gras_stub_generator generate deployment and
2601 remote compilation helpers.
2603 - Bugfix: Initialization fix in msg_test.
2607 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2609 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2611 - API Change: gras_os_sleep to take the amount of seconds as a double.
2612 Accepting an int was error prone since it was the only location where
2613 seconds were coded as such. It leaded to damn rounding errors.
2614 - Bugfix: Hard to belive that timers ever worked before this.
2616 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2618 SimGrid (2.91) unstable; urgency=low
2620 Alpha 2 on the path to SimGrid 3
2624 - Bug fix in the lmm_solver.
2626 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2627 through the function MSG_paje_output.
2628 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2629 - It is possible to bound the rate of a communication in MSG with
2630 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2631 to write it in the changelog).
2632 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2637 - Various cleanups to the autotools stuff
2638 - Begin to move Gras examples to examples/gras/
2639 - Let make distcheck work again (yeah!)
2641 - documentation overhauled using doxygen.
2642 gtk-doc-tools is dead in SimGrid now.
2643 - Automatically extract all existing logging categories, and add the list
2644 to the documentation (long standing one, to say the less)
2646 - Cleanup the known architecture table. Reorder the entries to group what
2647 should be, and use a more consistent naming scheme.
2648 (some of the test dataset are still to be regenerated)
2649 - New! Allow library to register globals on each process just as userdata
2651 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2652 lookup time (for now).
2653 Use it in msg and trp.
2654 This cleans a lot the internals and helps enforcing privacy of the
2655 headers between the gras components.
2656 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2657 - Bugfix: gras_os_time was delirious in RL.
2658 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2660 - Reenable GRAS now that it works.
2662 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2664 SimGrid (2.90) unstable; urgency=low
2666 Alpha 1 on the path to SimGrid 3
2668 * It is a long time since the last release of SimGrid. I'm sorry about
2669 that but as I had told you, I was rewriting a lot of things. I apologize
2670 to those who had been reporting bugs to me and that I had not answered.
2671 If your bug is still in the new version, please tell me. Here is a
2672 summary of the main changes.
2674 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2675 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2676 portability, speed, and a lot more but you'll figure it out later.
2677 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2678 I wanted to release the new MSG as soon as possible and I have
2679 broken GRAS, which is the reason why, for now, only MSG is fully
2680 functional. A laconic description of these projects is available
2681 in the documentation.
2683 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2684 kernel optimized for our needs (called SURF but only the developers
2685 should use it). Hence, MSG is now roughly 30 times faster and I think
2686 that by rewriting a little bit MSG, I could event speed it up a little
2687 bit more. Beside the gain in speed, it is also much easier to encode a
2688 new platform model with SURF than it was with SG. More to come...
2690 * REVOLUTION 3: I have tried to change a little as possible the API of
2691 MSG but a few things really had to disappear. The main differences
2692 with the previous version are :
2693 1) no more m_links_t and the corresponding functions. Platforms are
2694 directly read from a XML description and cannot be hard-coded
2695 anymore. The same format is used for application deployment
2696 description. The new format is described in the documentation.
2697 Have a look in tools/platform_generation. There is a tiny script
2698 that converts from the old platform format to the new one. Concerning
2699 the application deployment format, parsing the old one is tricky.
2700 I think most of you should however be able to convert your files. If
2701 it is really an issue, I can write a C code that does the conversion.
2703 2) the toolbox tbx does not exist anymore. We now have a library
2704 with much more data-structures but without the hash-tables (we have
2705 dictionaries that are much faster).
2707 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2709 *****************************************************************************
2710 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2711 * don't feel like converting the oldies. (Mt) *
2712 *****************************************************************************
2715 Version 2.90: "the long awaited one"
2716 - Finished rewriting and debugging MSG. Rewrote the documentation.
2717 - disable GRAS for now since it needs to be ported to the newest SG
2720 - Finish the port to windows (using mingw32 for cross-compile)
2723 - Main loop and datastructures of SURF. A cpu resource object is
2724 functional. Surf can thus be used to create cpu's with variable
2725 performance on which you can execute some actions.
2727 2004-11-15 Martin Quinson
2728 - Port to ARM. Simply added the alignment and size descriptions. Should
2729 work, but the ARM machines are so slow that I didn't had the opportunity
2730 to 'make check' over there yet.
2732 2004-11-15 Arnaud Legrand
2733 - Trace manager now written. It uses a heap structure and is therefore
2734 expected to be efficient. It may however be speeded up (particularly
2735 when many events occur at the same date) by using red and black
2736 trees. One day maybe...
2737 - Max-min linear system solver written. It uses a sparse matrix
2738 structure taking advantage of its expected use. Most operations are
2739 O(1) and free/calloc are called as few as possible. The computation of
2740 the minimum could however be improved by using a red and black tree
2743 2004-11-03 Arnaud Legrand
2744 - Rename every gras_* function that was in xbt/ to its xbt_
2746 - Add a heap and a doubly-linked list to xbt
2747 - Added a dichotomy to the dictionaries. make check works as well before
2748 so I assume that the patch is correct. I do not know however if things
2749 run effectively faster than before now. :)
2751 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2752 SimGrid, and the version number is bumped to 2.x
2754 2004-10-29 Martin Quinson
2755 - Introduction of the remote errors.
2756 They are the result of a RMI/RPC on the remote machine.
2757 ErrCodes being scalar values, you can't get the host on which those
2758 errors did happen. Extending the error mechanism as in Gnome is possible.
2759 No idea yet whether it is a good idea.
2761 2004-10-28 Martin Quinson
2762 - Interface revolution: the Starred Structure Eradication.
2763 I used to do typedef struct {} toto_t; and then handle *toto_t.
2764 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2765 - ???_t is a valid type (builded with typedef)
2766 - s_toto_t is a structure (access to fields with .)
2767 - s_toto is a structure needing 'struct' keyword to be used
2768 - e_toto_t is an enum
2769 - toto_t is an 'object' (struct*)
2771 typedef struct s_toto {} s_toto_t, *toto_t;
2772 typedef enum {} e_toto_t;
2773 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2774 s_toto_t) is private.
2776 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2777 so that it can set the variable to NULL. It was so for dicts and sets,
2778 it changed for dynars.
2780 - Fix a bunch of memleaks in dict_remove
2781 - Fix a bug in sg/server_socket opening: it failed all the time.
2783 2004-10-07 Martin Quinson
2784 - Speed up dynar lookup operation a bit.
2786 gras_dynar_get is dead.
2788 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2789 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2790 (which gives you the address of the stored data).
2792 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2793 copy of the data using an affectation to do the job and not a memcpy.
2795 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2796 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2798 It does not really speedup the dynar test because they are
2799 setting elements all the time (and look them seldom). But the dict does
2800 far more lookup than setting.
2802 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2804 2004-10-05 Martin Quinson
2805 - Allow to (en/dis)able the cycle detection at run time.
2807 Whether we should check for cycle or not is now a property of each
2808 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2809 datadesc_cycle_unset allow to remove this property when previously set.
2811 Note that the cycle detection is off by default since it impacts the
2812 performance. Watch the data you feed GRAS with ;)
2814 This property is hereditary. Any element embedded in a structure having it
2815 set have it set for the time of this data exchange.
2817 You should set it both on sender and receiver side. If you don't set it on
2818 sender side, it will enter an endless loop. If you forget on receiver
2819 side, the cycles won't be recreated after communication.
2821 - Header reorganization.
2822 Kill gras_private.h, each submodule must load the headers it needs.
2824 2004-10-04 Martin Quinson
2825 - Interface revolution: do not try to survive to malloc failure.
2827 Now, gras_malloc and friends call gras_abort() on failure.
2828 As a conclusion, malloc_error is not a valid error anymore, and all
2829 functions for which it was the only gras_error_t return value are
2830 changed. They now return void, or there result directly.
2831 This simplify the API a lot.
2833 2004-09-29 Martin Quinson
2834 - Re-enable raw sockets.
2835 Created by gras_socket_{client,server}_ext;
2836 Used with gras_raw_{send,recv}
2839 It should allow to kill the last bits of gras first version soon.
2841 This is not completely satisfactory yet (duplicate code with
2842 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2845 - Simplify transport plugin (internal) interface by not passing any
2846 argument to _server and _client, but embedding them in the socket
2849 2004-09-28 Martin Quinson
2850 - Finish the port to AIX.
2851 autoconf was my problem (segfault within the malloc replacement
2852 function. No idea why)
2854 2004-09-16 Martin Quinson
2855 - Fix some size_t madness on 64bit architectures.
2857 2004-09-08 Martin Quinson
2858 - Reduce the number of system headers loaded, overload some more system
2859 calls (such as malloc to cast the result of the system one, and work
2861 - Fix and reintroduce the config support
2863 2004-09-07 Martin Quinson
2864 - Source code reorganization to allow Arnaud to surf all over there.
2865 - Allow to document the logging categories.
2866 - Remove all uppercase from logging categories and useless cleanup in names.
2868 2004-08-18 Martin Quinson
2869 Version 0.6.2 (protocol not changed; API changed)
2870 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2871 gras_error_t), and NULL when not found. Functions expecting a msgtype
2872 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2873 hopefully usefull message.
2874 - Portability to prehistoric sparcs again
2876 2004-08-17 Martin Quinson
2877 Version 0.6.1 (protocol not changed; ABI not changed)
2878 - prealloc some buffers to speed things up
2880 2004-08-11 Martin Quinson
2881 Version 0.6 (protocol not changed; ABI expended)
2882 - The parsing macro can deal with the references, provided that you add
2883 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2885 2004-08-09 Martin Quinson
2886 Version 0.5 (protocol not changed; ABI changed)
2887 - Allow to off turn the cycle detection code in data exchange at
2888 compilation time. It should be at run time, but I'm short of time (and
2889 the config stuff is still broken). That way, we keep dict out of the
2890 critical path, which is good because the performance is poor:
2891 - search not dichotomial yet
2892 - dynar give no way to access their content and memcpy everytime
2893 - In composed data description (struct, ref and so on), stop foolness of
2894 keeping the subtype's ID, but store the type itself. This keeps sets out
2895 of the critical path, which is good since they rely on dynar and
2896 dictionnaries. The only loose of that is that we cannot detect the
2897 redeclaration of a structure/union with another content (but I'm not sure
2898 the code detected well this error before anyway). We still can detect
2899 the redefinition discrepancy for the other types.
2900 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2901 it breaks the code because of type-punning used all over the place).
2902 This breaks on all non-gcc architectures (for now).
2904 All those changes (plus the buffer of last time) allow me to gain 2 order
2905 of magnitude on cruel tests consisting of 800000 array of integers on two
2906 level of a hierarchical structure (200 secondes -> 4 secondes)
2909 - the selector of reference must now return the type it points to, not
2910 the ID of this type.
2912 2004-08-06 Martin Quinson
2913 Version 0.4 (protocol changed; ABI not changed)
2914 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2915 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2916 - Implement a buffer transport, and use it by default (it relies on tcp in
2917 real life and on sg in simulation).
2918 That's a bit hackish since I had a new field to the structure to store
2919 its data without interfering with the subtype ones. Inheritance
2920 is tricky in C. And that's a kind of reverse inheritance with one class
2921 derivating two classes. Or maybe a game with java interfaces. Anyway,
2922 that's damn hard in C (at least).
2923 Moreover, I got tired while trying to ensure plugin separation and
2924 genericity in SG mode. MSG wants me to do weird things, so let's go for
2925 cruel hacks (temporarily of course ;).
2926 See comment in transport_private.h:71
2927 - do not include all the _interface headers in private but in the files
2928 which really need them (to cut the compilation time when they are
2931 2004-07-26 Martin Quinson
2932 Version 0.3 (protocol not changed; ABI changed)
2933 - Major overhault of the datadesc interface to simplify it:
2934 - shorted the function names:
2935 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2936 - add a trivial way to push/pop integers into the cbps without malloc.
2937 This allows to make really generic sub_type description, which simply
2938 pop their size of the stack.
2939 - add a function gras_datadesc_ref_pop_arr() which does what users want
2940 most of the time: Declare a dynamic array (which pops its size of the
2941 stack) and declare a reference to it. Poor name, but anyway.
2942 - kill the post-send callback, add a post-receive one
2944 2004-07-23 Martin Quinson
2945 Version 0.2 (protocol changed; ABI changed)
2946 - add some testing for cpbs in the test cases, and fix some more bugs.
2947 This invalidate again the little64 data file, since I cannot regenerate
2949 - remove an awfull optimization in the logging stuff, allowing me to:
2950 - understand it again
2951 - learn gcc how to check that the argument match the provided format
2952 - fix all errors revealed by gcc after that
2953 - internal keys of dict are not \0 terminated. Deal with it properly in
2954 loggings instead of segfaulting when the user want to see the logs :-/
2956 2004-07-22 Martin Quinson
2957 - Fix some stupid bug preventing cbps (callback postit) from working
2959 2004-07-21 Martin Quinson
2960 - Some documentation cleanups
2961 - remove the useless last argument of msgtype_declare
2962 - rename the Virtu functions to fit into the 'os' namespace
2963 - move headers src/include -> src/include/gras/ and stop fooling with
2964 gras -> . symbolic link
2965 - make distcheck is now successful
2967 2004-07-19 Martin Quinson
2969 - Build shared library also
2970 - Install html doc to the right location
2971 - stop removing maintainer files in make clean
2972 - build tests only on make check
2974 2004-07-13 Martin Quinson
2976 - No major issue in previous version => change versioning schema
2977 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2978 corresponding dataset.
2980 2004-07-11 Martin Quinson
2982 - Get it working with any kind of structure (we can compute the padding
2983 bytes remotely for all the architectures I have access to)
2984 - Implement the structure parsing macro (still not quite robust/complete)
2985 - Improvement to the remote testing toysuite
2987 2004-07-10 Martin Quinson
2988 [autoconf mechanism]
2989 - get ride of a bunch of deprecated macros
2990 - actually run the test for two-compliment, not only compile it :-/
2991 - test whether the structures get packed (and bail out if yes. Damn.
2992 Alignment is a serious matter)
2993 - test whether the structures get compacted (but respecting the alignment
2994 constraints of each types)
2995 - test whether the array fields of structures can straddle alignment boundaries
2997 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2998 (took me 2 hours to find that bug, looking at the wrong place)
2999 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3000 before sending/receiving and are used to compute the offsets of fields
3001 - Given that padding size depend even on compiler options, keep track of
3002 alignment and aligned_size only for the current architecture. Not a big
3003 deal since we send structure fields one after the other (seems
3005 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3006 program, let it work on linux/gcc/little32. portability todo.
3008 2004-07-08 Martin Quinson
3009 - import and improve remote compilation support from FAST
3010 - make sure make check works on half a dozen of machines out there
3012 2004-07-07 Martin Quinson
3013 Let's say it's version 0.0.3 ;)
3014 - Implement conversions (yuhu!)
3015 - Let it work on solaris (beside conversion, of course)
3016 - Stupid me, using rand() to generate the conversion datatests in not wise.
3018 2004-07-06 Martin Quinson
3019 - Let make dist work, since I'm gonna need it to compile on remote hosts
3020 - Let Tests/datadesc_usage write the architecture on which the file was
3021 generated as first byte.
3022 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3023 architecture descriptions.
3024 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3025 on those architectures.
3026 - Optimization: send/recv array of scalar in one shoot
3028 2004-07-05 Martin Quinson
3029 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3031 - Plug a whole bunch of memleaks
3032 - each process now have to call gras_{init,exit}. One day, their log
3033 settings will be separated
3034 - Continue the code factorisation between SG, RL and common in Transport.
3036 2004-07-04 Martin Quinson
3038 - Redistribution between SG and RL.
3039 We wanna have to accept in SG, so move accepted related parts of RL in
3040 the common part. (more precisely, the dynar of all known sockets is no
3041 more a static in transport.c, but part of the process_data)
3043 [gras_stub_generator]
3044 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3046 2004-07-03 Martin Quinson
3047 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3048 them all up in one shot)
3050 - Fix the ugly case of reference to dynamic array.
3051 - New (semi-public) function gras_datadesc_size to allow the messaging
3052 layer to malloc the needed space for the buffer.
3054 - gras_socket_close now expect the socket to close (and not its address to
3055 put NULL in it after it). This is because the socket passed to handlers
3056 is one of their argument (=> not writable).
3058 - propagate the interface cleanup from last week in datadesc, ie remove a
3059 superfluous level of indirection. User pass adress of variable
3060 containing data (both when sending and receiving), and not of a variable
3061 being a pointer to the data. Let's say that I like it better ;)
3062 The price for that is constructs like "int msg=*(int*)payload" in
3063 handlers, but it's a fine price, IMHO.
3065 - Let it work in RL (yuhu)
3067 2004-06-21 Martin Quinson
3069 - porting SG plugin and SG select to new standards (works almost).
3070 - plug memleaks and fix bugs around.
3073 - cleanup the prototype of data recv and force users to specify when they
3074 want to handle references to objects. Test case working even for cycles.
3075 - plug memleaks. Valgrind is perfectly ok with this.
3077 2004-06-12 Martin Quinson
3079 - cleanup the separation between plugin and main code in plugin creation
3081 2004-06-11 Martin Quinson
3083 - Reput hook for raw sockets, needed for BW experiments
3084 - kill a few lines of dead code
3085 [Data description] Interface cleanup
3086 - gras_datadesc_by_name returns the searched type or NULL.
3087 That way, no variable is needed to use a type desc once, which makes
3089 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3090 two parameters were strings (field name, type name), leading to
3092 [Dicos] Interface cleanup
3093 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3094 This is consistant with the dynar API.
3096 2004-04-21 Martin Quinson
3098 - Porting to new standards.
3100 - interface cleanup.
3101 There is no bag anymore, no need to take extra provision to mask the
3102 pointers behind "ID".
3103 Better splitup of functions between files create/exchange/convert.
3104 This is still a bit artificial since convert and receive are so
3105 interleaved, but anyway.
3107 - add a queued message list to procdata (the ones not matching criteria
3109 - factorize some more code between SG and RL wrt procdata
3111 - use gras_exit in example to track memleaks
3112 - get rid of gs_example now that GS is properly integrated into gras
3113 - update run_test to integrate the latest tests (datadesc)
3115 - rename WARNINGn macros to WARNn since it prooved error-prone
3117 2004-04-19 Martin Quinson
3119 - register init/exit functions within gras module mechanism
3120 - send/receive function.
3121 Convertion is not implemented, but short-cutted if not needed.
3122 struct/array elements are sent one by one (instead of block-wise), but
3123 nobody really cares (yet). Get a prototype before optimizing.
3124 - tests (using a file socket) for DD send/receive on:
3125 - base types: int, float
3126 - array: fixed size, string (ie ref to dynamic string)
3127 - structure: homogeneous, heterogeneous
3128 - chained list, graph with cycle
3129 Believe it or not, valgrind is not too unhappy with the results. The
3130 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3134 - Bugfix when using a filename explicitely (instead of '-')
3136 2004-04-09 Martin Quinson
3138 - factorize more code between RL and SG in socket creation
3139 - Complete the implementation and tests of:
3141 o file (only in RL, and mainly for debugging)
3143 I lost 3 days to design a portable address resolver, and then decided
3144 that the prototype mainly have to run on my box.
3145 Addressing portability too early may be like optimizing too early :-/
3147 - use gras_init in the Tests instead of the crappy parse_log_opt
3148 (the latter function is removed)
3149 [Conditional execution]
3150 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3151 [Code reorganisation]
3152 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3153 Build examples against the RL library, since there is no way to disable
3154 its creation for now.
3156 For information, the beginning of coding on GRAS was back in june
3157 2003. I guess that every line has been rewritten at least twice since