1 SimGrid (3.14) UNRELEASED; urgency=low
4 * The whole documentation was reworked and reorganized. There is
5 still many room for improvement here, but we're on it.
8 * Privatization is now activated by default in smpirun
9 You can turn it off with -no-privatize if something goes wrong.
11 * Call-location tracing for SMPI.
12 You can add the exact location (filename / linenumber) of an MPI call to
13 your trace files and slow down or speed up the simulation between two
14 consecutive calls by using an adjustment file (see the documentation).
16 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
17 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
18 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
19 values were ignored in some cases. The timings of these functions can now
20 be significantly different.
22 Dropped functions and features
23 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
24 - MSG_mailbox_is_empty() -> Mailbox::empty()
25 - MSG_mailbox_front() -> Mailbox::front()
26 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
27 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
28 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
29 * Task affinity. Its intended behavior (that was very badly tested
30 and probably not really working) was deceiving what most users
31 would have hoped here.
34 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
35 EiBps, EBps, Eibps, Ebps for bandwidth.
36 They may become useful to some lucky ones.
39 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
40 * ex/app_tokenring: new example, very similar to the MSG Token Ring
41 * ex/async_waitAll: new example, on asynchronous communications
43 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
45 SimGrid (3.13) stable; urgency=low
47 The Half Release, a.k.a. the Zealous Easter Trim.
49 - We removed half of the lines, that were mostly experimental cruft.
50 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
51 (not counting blanks and comments -- according to openhub.net)
52 - The internals are now compiled in C++ (and will soon be clean C++)
53 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
54 59 klines of C, etc. We added only 29 klines of C++ in replacement.
56 * Backwards Compatibility breaks
57 - Removed Lua simulation bindings (switch to C or Java for that).
58 Lua can still be used to describe platforms
59 - Removed Java kernel plug-ins.
60 Will be reintroduced after the ongoing major internals reorg.
62 - the following functions were removed.
63 They were too specific and should be reimplemented in a generic
64 way, with filter function.
65 - MSG_task_listen_from_host
66 - MSG_mailbox_get_count_host_waiting_tasks
67 - MSG_mailbox_put_with_timeout was removed.
68 Please use MSG_task_send_with_timeout instead.
70 - the SD_application_reinit function was removed. It has been a noop for a while.
71 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
72 replaced by a more flexible mechanism.
73 - The following functions thus do not exist anymore
74 - SD_workstation_get_access_mode
75 - SD_workstation_set_access_mode
76 - SD_workstation_get_current_task
77 - Basic estimation functions have been removed but can easily be replaced
78 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
79 - SD_workstation_get_computation_time => amount / sg_host_speed()
81 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
82 Use VM.getSpeed()*load/100 for the legacy behavior.
84 - option enable_tracing was removed. It was not doing anything for a while.
85 - In the ModelChecker:
86 - the model-checker now ptraces the model-checked process which means
87 you cannot use a debugger on the latter anymore (we might make this
88 optional in the feature);
89 - removed soft-dirty page tracking;
90 - remove model-checked side snapshot management,
91 MC_snapshot() and MC_compare_snapshot();
92 - keep the MC_cut() function as a stub (it was not really working
93 in the previous release).
96 * All options are consistently in kebab-case. Old names are kept as alias.
99 * Switch to platform v4 format.
100 - Rename from 'power' to 'speed' the attributes describing the amount of
101 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
102 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
103 - In <host> and <link>, attributes availability and state are gone.
104 It was redundent with state and availability traces, and with peak values.
105 - In <cluster>, attributes availability_file and state_file are gone.
106 It was too complex and unused.
107 - Kill <gpu>. Was not doing anything.
108 - The DOCTYPE points to the right URL:
109 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
110 (the file at this address now documents the changelog since its v1)
111 - A warning is emitted for unit-less values (they are still accepted).
112 - speed. Default: 'f' or 'flops'. Also defined:
113 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
114 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
115 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
116 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
117 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
118 - latency. Default: 's' second. Also defined:
119 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
121 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
122 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
125 * s4u::Host is now the preferred public interface to the Host features.
126 sg_host_* functions are C bindings to the exact same behavior
127 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
130 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
131 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
132 They were deprecated since a few years, and probably did not even compile anymore.
135 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
136 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
137 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
138 encouraged to update their codes. A list of the modified functions can be found at the end of
139 include/simgrid/simdag.h
142 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
143 the lambda or closure passed as a parameter will run in kernel mode.
144 All the callback functions should be rewritten to that interface at some point.
147 * Reorganizing and cleaning the internals all around the place.
150 * Remove old default barrier/bcast buggy algorithms (see #18407)
151 * Various bug fixes to handle more codes
152 * Remove the need for the --foreground option of smpirun (it is still
153 accepted for backward compatibility).
156 * Kill the setset data container: MC don't use it anymore.
157 * Kill the queue data container: it made more sense with GRAS.
158 * Kill the xbt_peer_t data type: it's useless without GRAS.
159 * Kill rm_cb feature of config sets: it was never useful.
160 * Kill graphxml parsing feature. It was not used.
161 * Kill the deprecated code protected by XBT_USE_DEPRECATED
163 - xbt_dynar_sort_strings(), when the content is char*
164 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
165 They throw exceptions on invalid input;
166 * C++ support for declaring CLI flags (simgrid::config::Flag);
167 * class for abstracting different signal backends (simgrid::xbt::signal).
168 with no external dependencies (we need very simple signals).
171 * refactoring and cleanup of the code;
172 * ongoing process to cleanly separate the model-checking algorithms
173 from the code model-checking support.
175 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
177 SimGrid (3.12) stable; urgency=low
179 The Facelift Release.
182 * Require g++ v4.7 at least to not speak prehistorical C++.
183 * Require Boost 1.48 (for signal2 component).
184 * Java must be version 7 at least when activated.
185 * Builds on Windows again (including Java bindings).
186 * Tracing is now always enabled (no way to turn it out)
187 * Remove GTNetS. It was not working anyway.
188 * Various cleanups in the cmake scripts.
189 * Move headers around to sort them out on installed systems:
190 - instr/instr.h -> simgrid/instr.h
191 - instr/jedule/* -> simgrid/jedule
192 - simdag/datatypes.h was removed
193 - simdag/simdag.h -> simgrid/simdag.h
194 - msg/datatypes.h was removed
195 - msg/msg.h -> simgrid/msg.h
198 * Interface improvement:
199 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
200 - Sanitize the interface in MSG_task_ module:
201 - Merge two functions that were close enough but misleading:
202 set_compute_duration(t) -> set_flops_amount(t)
203 get_remaining_computation(t) -> get_flops_amount(t)
204 - set_data_size(t) -> set_bytes_amount(t)
205 get_data_size(t) -> get_bytes_amount(t)
206 - Massive cleanups in the functions related to the energy
207 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
208 - New: MSG_host_get_pstate()
210 - msg/energy/onoff: switching hosts on and off
212 * Interface improvement:
213 - Rename Host.isAvail() to Host.isOn()
214 - Rename Process.currentProcess() to Process.getCurrentProcess()
215 - Rename Task.setDataSize() to Task.setBytesAmount()
216 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
218 - #18874: Actually allows the GC to reclaim tasks
222 - SIMIX_process_throw: raises an exception in a remote process
223 * Refactoring: Separate sub-modules
224 - libsmx: the public interface, as libc in a real system
225 - popping: the strange dance that converts a user request into a kernel handling
226 - smx_context_*: the virtualization mechanisms that embed the user code
227 - smx_*: the handling of each simcalls
229 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
230 * Rename smx_action_t into smx_synchro_t, making explicit that these
231 things are used to synchronize processes with their environment.
232 For example, a communication is a sort of synchronization involving
233 the communicating processes (that may block until the exchange) and
234 the platform. The same can be said from computations, etc.
236 - #18888: segfault when a process ends before its kill_time
240 - 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.
241 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
242 - MPI_Type_set_name, MPI_Type_get_name
243 - MPI_*_c2f and MPI_*_f2c functions
244 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
245 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
246 - Activate a lot of new tests from the mpich 3 testsuite
248 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
249 - InfiniBand network model added : Based on the works of Jerome Vienne
250 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
251 - When smpi/display_timing is set, also display global simulation time and application times
252 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
253 * Collective communications
254 - 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
255 - 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/).
256 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
258 - Replay now uses algorithms from wanted collective selector
259 - Replay can be used with SMP-aware algorithms
260 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
261 - 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.
263 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
264 - Correctly use loopback on fat-tree clusters
265 - Asynchronous small messages shouldn't trigger deadlocks anymore
266 * Energy/DVFS cleanup and improvement
267 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
268 - new: smpi_get_host_pstate()
271 - "Full" network optimization flag was broken since Surf++
272 - Better handling of precision flags in maxmin
273 - Fix bug causing sometimes "Impossible" errors
274 - Properly pass cluster properties to included hosts
275 * Improvement of the Energy plugin.
276 - Always update the consumption before returning that value
277 - New property: watt_off to denote the dissipation when the host is off
278 - New functions getWattMinAt and getWattMaxAt to retrieve the
279 dissipation of pstates that we are not currently at.
280 * Java: class NetworkLink renamed to Link
281 * New function: simcall_process_get_kill_time()
282 * Massive rename s/workstation/host/
283 - That's intrusive, but that's good for the project consistency. Sorry.
284 - Change config option "workstation/model" into "host/model"
287 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
288 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
289 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
290 - xbt_dirname and xbt_basename for non-POSIX systems
292 * The model checker now runs as a separate process.
293 * The model checker runs must now be launched with the new simgrid-mc program.
294 * Record/Replay: the MC can display a textual representation of a path in the
295 execution graph. It can then be replayed outside of the model checker.
297 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
299 SimGrid (3.11) stable; urgency=low
304 * Normalizing pointers addresses tool for better diff between logs
307 * Add cloud examples using new VMs
308 - examples/msg/cloud/two_tasks_vm.tesh
309 - examples/msg/cloud/simple_vm.tesh
310 - examples/java/cloud/cloud.tesh
311 - examples/java/cloud/migration/migration.tesh
312 * Add java surf examples:
313 - examples/java/surfPlugin/surf_plugin.tesh
314 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
315 - examples/java/surfCpuModel/surf_cpu_model.tesh
316 * Add SMPI+MSG example:
317 - examples/smpi/smpi_msg_masterslave/
322 - msg task destroy cancel test
323 - msg_host on/off test
324 * Move all tests in testsuite to teshsuite (adding tesh files)
325 * Restructure teshsuites
326 - one folder for each kind of test
327 * Restructure AddTests.cmake
329 - structure the order of tests (with sections)
332 * Add virtual machine
333 - creation of a VM on a PM
334 - migration of a VM from a PM to another PM
336 - MSG_process_join(msg_process_t process, double timeout)
337 - msg_bar_t MSG_barrier_init(unsigned int count)
338 - int MSG_barrier_wait(msg_bar_t barrier)
339 - void MSG_barrier_destroy(msg_bar_t barrier)
340 - msg_as_t MSG_environment_as_get_by_name(const char * name)
341 * New option "msg/debug_multiple_use" to help debugging when a task is used
344 - Improvements and finalization of MSG_storage, MSG_file APIs and their
346 - Increase code coverage in test suites
350 * Protect context stacks against stack overflow. The number of protected memory
351 pages allocated on the top of each stack (1 by default) can be configured
352 with the new command line option --cfg=contexts/guard_size:<value>.
353 * Simcalls are now generated by a python script that
354 - generates files included by SimGrid
355 - checks that all the functions exist, and proposes prototypes
357 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
358 host_set_data, host_get_data
360 - simcall_process_join(smx_process_t process, double timeout)
361 * Fix bug where sleeping processing could not be suspended.
364 * Translate surf models from C to C++
365 - Generic classes for all models: Model, Resource, Action
366 - A generic interface for each kind of model (CPU, Network, Storage
367 Workstation, WorkstationVM)
369 * Translate surf routings from C to C++
370 * Add callbacks using sigc++ or boost::signals2
371 - Add callback functions for resource creation/destruction
372 - Add callback functions for action state change
373 - Handle Energy as a plugin
374 * Replace swag by boost::intrusive
375 * Add new routing models for clusters. For documentation, see
376 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
377 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
378 parameters for cluster tag
379 - Fat trees, with topology="FAT_TREE" and
380 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
381 - see examples/platforms/torus_cluster.xml and
382 examples/platforms/fat_tree_cluster.xml
386 * Hostfiles support host:nb_processes construct to deploy several processes on
388 * Collective communication algorithms should not crash if used with
389 improper number of nodes and report the error.
390 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
391 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
392 MPI_Dims_create, MPI_Cart_sub are supported.
393 * New interface to use SMPI programmatically (still depends on MSG for
394 some parts, see examples/smpi/smpi_msg_masterslave) :
395 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
399 * Global variables privatization in MPI executables is now performed at runtime
400 with the option smpi/privatize_global_variables (default:no).
401 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
402 dynamic libraries loaded by the application are not privatized (static
403 linking with these libraries is advised in this case)
406 * Options defined in XML work correctly now.
409 * New cmake option, enable_lib_in_jar, to control whether native libraries are
410 copied into simgrid.jar or not (ON by default). Use this option if you want
411 to reduce the size of the installed simgrid.jar, *and* the native libraries
412 are kept installed elsewhere.
413 * Surf binding with SWIG (code generated in maintainer mode only):
414 - plugin to handle callbacks
415 - CPU model only for the moment
418 * Supernovae build mode is definitively removed. It was used to improve
419 inlining and inter-module optimizations. It is nowadays superseded by
420 link-time optimizations commonly available in compilers.
421 * Update ns-3 find lib. Bindings for ns-3 should work again now.
422 * Add boost dependency for surf++
423 * Add new macro for tests
424 - ADD_TESH(name <tesh_args>)
425 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
429 - xbt_bar_t XBT_barrier_init(unsigned int count)
430 - int XBT_barrier_wait(xbt_bar_t barrier)
431 - void XBT_barrier_destroy(xbt_bar_t barrier)
432 * Make the xbt_os_time module public
434 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
436 SimGrid (3.10) stable; urgency=low
438 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
441 * Reintegrate Java to the main archive as desynchronizing these
442 package is not acceptable anymore (Java is now considered stable)
443 * Add explicit synchronization facilities through semaphores
444 * Bug fix: Task.setDataSize() only changed the C world, not the value
445 cached in the Java world
448 * Dramatically change the way files are handled. API and internals changed, but
449 this part of MSG was not considered as production grade either.
450 * Add explicit synchronization facilities through semaphores
451 * Add a new function MSG_host_get_process_list()
452 * Preliminary DVFS support (see examples/msg/energy/ for details)
455 * SMPI is now included directly in the libsimgrid as the windows
456 linker doesn't force us on splitting it anymore.
457 * Improvements of the SMPI replay tool:
458 - Most of the collective communications are now rooted in the same process as
459 in the original application.
460 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
461 used until now). Multiple data types can now be used in a trace.
462 - The replay tool now supports traces produce either by TAU or a modified
464 - Bug Fix: the compute part of the reduce action is now taken into account.
465 - Gatherv collective is now supported
466 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
467 of smpirun outputs time independent traces for the current run. One file
468 is created per process. If too many processes are simulated, this behavior
469 can be changed to one file for all processes by using the
470 tracing/smpi/format/ti_one_file flag
471 * smpirun generates the host file if needed (with given host count and platform)
472 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
473 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
474 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
475 --cfg=smpi/coll_selector:(mpich/ompi)
476 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
478 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
479 happening inside a collective SMPI call.
480 * Fix the behavior of complex data types handling.
481 * Make MPI_Wtime another synchronization point to take computations into
483 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
484 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
485 * Add all missing Fortran bindings, SMPI should work with Fortran 90
486 (no privatization of global variables yet)
487 * Preliminary DVFS support (see examples/smpi/energy/ for details)
490 * Verification of liveness properties is now available for SMPI applications
491 (in addition to MSG applications)
492 * Bugged examples using SMPI in examples/smpi/mc/
493 * Add --cfg=model-check/visited option. Allows the verification of infinite
494 programs. Detection of loops in the execution thanks to the system state
495 comparison and reduction of the state space to explore. Can be combined with
496 DPOR for safety properties.
499 * Allow to change SimGrid configuration (see --help) within the code
500 thanks to SD_config() as it can be done in MSG.
501 * Add a new function SD_task_set_amount() upon user request.
504 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
505 * Remove rule based routing (no more PCRE dependency)
506 * Add a limiter_link option to cluster tag, to specify a maximum reachable
507 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
508 * Add a loopback_bw and loopback_lat options to cluster tag.
509 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
510 routing. Now peers are encapsulated in an AS and have their own private
511 router but this is transparent.
514 * Our own implementation of getline is renamed xbt_getline, and gets
515 used even if the OS provide a getline(). This should reduce the
516 configuration complexity by using the same code on all platforms.
517 * New type: xbt_cfg_elm_boolean.
518 * Allow to use yes/no for boolean configuration options in the command line.
519 * Allow to disable SimGrid cleanups at exit from command line option.
520 There are situations where one may want a simulation to end with an exit.
521 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
522 annoying when scripting around the simulator. Adding a
523 --cfg=clean_atexit:no allows to circumvent this issue.
526 * Lots of memory leaks were corrected in this release.
527 There is no known memory leaks anymore, in all of our 600+ tests.
528 * New command line option --version, to get SimGrid version information.
529 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
531 * Supernovae builds are deprecated, and expected to be removed in the next
534 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
536 SimGrid (3.9) stable; urgency=low
538 The Grasgory release: GRAS is really dead now.
540 * Complete overhaul of the internal host structures scheme.
543 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
544 was considered as experimental and badly maintained since then.
545 * Keeping it was thus a trap to our potential users, that could take
546 it instead of MSG or SMPI by mistake despite is pity state.
547 * GRAS seems to have very few users (if any), and no one volunteered
548 to maintain it further. It also induces a lot of XBT code (for
549 portability sake), that must be maintained too.
550 * For all these reasons, we killed GRAS. If someone wants to revive it
551 in the future, don't cry, our git history still remembers of GRAS.
554 * Major overhaul. Merge our documentation again as time proved that
555 splitting it was really not helping our users.
556 * Further improve the developer documentation to help newcomers
557 hacking on SimGrid itself. The user documentation (and in
558 particular, the beginner documentation) is still in a sorry state.
561 * Now works on Windows too!
562 * Much more extensive test suite, from MPICH
565 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
566 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
567 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
568 * Bug fix: let task be scheduled when the last dependency to be solved is
569 a control dependency.
570 * Remove SD_load_environment_script function.
571 Use the C sg_platf function if you want to declare a platform
575 * New function: MSG_process_get_number()
576 * Old function documented: MSG_config()
577 * Remove MSG_load_platform_script function
578 Use the C sg_platf function if you want to declare a platform
582 * Change the default value of the TCP_gamma constant (maximal size of TCP
583 congestion window) to a more realistic 4MiB value. If you notice changes in
584 your simulation results, you can fall back to the previous 20k tiny window
585 by adding --cfg=network/TCP_gamma:20000 on command line.
586 * (Hopefully) fix a bug wrt periodic availability and state traces
587 * Bug fix: use default values at start when first event in availability/state
588 trace is not at time 0.
591 * remove the "new_" part of function name sg_platf_new_trace_connect
592 (resulting in sg_platf_trace_connect), since it does not create
596 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
597 use case where it's really mandatory, and maintaining it was a pain in
599 * New: xbt_fifo_search(), search an item with a user-provided
600 comparison function instead of dumb pointer comparison.
603 * Fix the lua deployment:
604 Use `simgrid.init_application()` before deployment instead of
605 `simgrid.msg_register_application()` after.
608 * Transfer the tracing files into the corresponding modules.
610 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
612 SimGrid (3.8.1) stable; urgency=low
614 The "we are told that some people want to also *install* the simgrid
617 * Add missing file "tesh.1" to the archive.
619 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
621 SimGrid (3.8) stable; urgency=low
623 The Psssshiiiit release: SimGrid jumps into the Cloud.
626 * Add an experimental interface to manipulate VMs. They are mainly
627 process groups with very few intrinsic semantic, but they should
628 allow you to build the semantic you want easily.
629 * New function: MSG_host_set_property_value()
630 * New function: MSG_process_on_exit(). To clean memory in all cases.
631 * Bug fixes that made the host (and link) failures unusable.
632 * Add a way to auto-restart process when the host in which they are
633 executing comes back (ON_FAILURE="RESTART" on deployment file,
634 MSG_process_auto_restart_set).
635 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
636 please stop using the old ones, they are DEPRECATED.
638 * Deprecate functions MSG_global_init() / MSG_global_init_args()
639 Please use MSG_init() instead. (reducing the amount of entry
640 points in the library helps us).
641 * Make it impossible to link against the wrong version of the lib
642 * Deprecate MSG_clean(). No need to call it anymore.
643 * Function MSG_get_host_number() is not deprecated anymore.
646 * Split the doc into a user guide and a reference guide.
647 * Start a developper guide to help people hacking on SimGrid.
650 * Enable tracing by default. This modules rocks you should use it.
651 * Remove option custom_flags. Now use environment variables CFLAGS
653 * Use default cmake things to detect lua instead of home grown ones.
654 * New option "enable_mallocators" to disable mallocators, for debugging
655 purpose ("on" by default).
658 * Bug fixes around the resource failures: don't let the processes
659 survive the host they are running onto.
660 * Add an interface to auto-restart processes when the host in which they are
661 executing comes back.
662 * Ensures that SIMIX_clean is called automatically. It's not part of
663 the public interface anymore (bindings should be updated).
666 * Bug fix for when SD_Simulate is called with a positive value: be careful
667 when comparing doubles. Sometimes they are different for non significant
669 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
670 parallel task whose initial work is distributed among host according
671 to the Amdahl's law. Such tasks are created with a parameter alpha
672 that corresponds to the non-parallelizable part of the computation.
673 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
674 between two sets of workstations assuming a 1D block distribution (each
675 workstation owns a similar share of data) on both sides.
677 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
678 Data redistribution will be automatically scheduled once parent and child
679 are both scheduled. The filling of computation_amount and
680 communication_amount structures is now done seamlessly thanks to the chosen
682 * New function SD_workstation_dump to display various information
683 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
684 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
685 bandwidth on the route onto which the task is scheduled and the amount of
687 To divide the nominal bandwidth by 2, the rate then has to be :
688 rate = bandwidth/(2*amount)
689 * Compute tasks that have failed can now be rescheduled and executed again
690 (from their beginning)
691 * Increasing source code coverage (src/simdag is now covered at 95.8%
695 * Re-implement time-independent trace replay using SMPI (at the
696 smpi_smp_* level) instead of MSG. This should replace
697 examples/msg/actions/actions.c
698 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
699 hindexed and structs)
700 * Implement the exchange of non-contiguous data.
701 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
702 * Correct behavior of smpi/sender_gap and set its default value to 0
703 * Add option to asynchronously send small messages to allow better
704 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
705 specifies the size in bytes under which messages will be asynchronously sent.
706 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
707 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
710 * New C interface to define a platform: XML is now optional.
711 For more info, please check include/simgrid/platf.h.
712 * New interface to define random platforms from the C:
713 For more info, please check include/simgrid/platf_generator.h and
714 examples/msg/masterslave/masterslave_platfgen.c
715 * Export a sg_cmdline dynar containing all the arguments we got from
719 * Two new tracing options for adding comments to trace file so you
720 can track your experiments (see --help-tracing for details).
721 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
722 * Adding the SimGrid version that generated the trace file as a comment.
723 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
724 * Fix to avoid key clashes on Paje links
725 * Other minor fixes related to the Paje specification
728 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
729 and renamed to xbt_str_hash() and xbt_str_hash_ext().
730 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
731 * Greatly improve the robustness of mmalloc to user errors (such as
732 using an area after freeing it, or freeing it twice)
734 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
736 SimGrid-java (3.8.1) stable; urgency=low
738 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
739 Warning: all methods are visible, but only some of them are
740 implemented so far. Check the source (src/jtrace.c)
741 for further information.
742 * New module: org.simgrid.msg.File (SimGrid File management functions)
743 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
744 * Change the meaning of Process.restart: now restart the process from
745 the begining, like MSG_process_restart in C.
746 * Add Process.setAutoRestart: handling of process restart when failed
748 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
749 you to retrieve the properties of the processes/hosts
750 * Deprecate Msg.clean(): you can just forget about it now.
751 * New function Process.getCount(), that only works when compiling
752 with the not yet released version 3.9 of the C library.
754 * New context factory based on Coroutines. It mandates a modified JVM
755 but then, the simulations run about five times faster, and there is
756 no limit to the amount of processes (beside of the available memory).
758 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
760 SimGrid (3.7.1) stable; urgency=low
763 * Restore the prototype of MSG_process_create_with_environment() to
764 the pre-3.7 situation by removing the kill_time argument.
765 * Add a MSG_process_set_kill_time() function instead.
768 * Fix weird behaviors when dealing with parallel tasks.
771 * Simgrid is now built as a dll.
772 * Simgrid-java now works on Windows.
773 * Simgrid-Java is now included into Windows package.
776 * First pre-build package for MacOSX.
779 * Fix compilation when using MSG_USE_DEPRECATED.
780 * Fix some compilation issues on Macs and Windows.
781 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
782 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
784 We discovered that the Lua console is broken, but we are missing the
785 manpower to fix it right now. The problem existed in 3.7 too, so we
786 are not blocking the release for that. Sorry if you depended on this
787 feature, any help would be really welcome.
789 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
792 SimGrid-java (3.7.1) stable; urgency=low
794 The "Java aint got to be bloated and slow" release
797 * Various internal cleanups and performance improvement
798 Simulations are expected to run up to twice faster or so
799 * Make Process.kill(process) an instance method, not a static one
800 * User processes are not java.lang.Thread subclasses.
801 This breaks the compatibility (sorry), but previous API was
802 brain-dead, making it impossible to have non-trivial
803 initializations in the process constructor.
804 * Require a full constructor per Process sub-class.
805 Kinda breaks the compatibility (sorry), but this allows a much more
806 efficient way to launch the processes at simulation startup.
807 * Do not embeed our version of semaphores, java 1.5 can be considered
808 as sufficiently prevalent for us to not dupplicate its features.
812 * Add examples for almost every part of the API
813 We spotted and fixed a lot of bugs in the process
815 * New module: asynchronous communication API
816 * New function: Process.sleep()
817 It takes milliseconds as argument, just as java.lang.Thread.sleep()
818 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
819 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
821 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
823 SimGrid (3.7) stable; urgency=low
825 The "spring cleanups (before the next Big Project kicks in)" release.
828 * We can specify the SMPI latency/bandwidth factor with command line
829 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
830 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
831 You can also use the "config tag" from platform file by adding this line
832 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
833 (see "example/platforms/tag_config.xml" to use "config tag").
834 Note that the command line supersedes the platform file configuration.
835 * Change the correction factors used in LMM model, according to
836 the latest experiments described in INRIA RR-7821.
837 Accuracy should be improved this way.
838 * Use the partial invalidation optimization by default for the
839 network too. Should produce the exact same results, only faster.
840 * Major cleanup in surf to merge models and split some optimization
841 mechanisms from the core of the models. As a result you can now
842 specify which model to use (e.g., --cfg=network/model:LV08
843 --cfg=cpu/model:Cas01) and which optimization mode to use
844 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
845 Incompatible combinations should err at initialization. See
846 --help-models for the list of all models and optimization modes.
847 * The CLM03 workstation model was dropped for simplicity because it
848 used the deprecated CM02 network model. Use default instead.
849 * Rename the TCP_gamma configuration option to network/TCP_gamma
850 * Rename the coordinates configuration option to
851 network/coordinates, and document it
852 * Use now crosstraffic keyword instead of the terribly misleading
853 fullduplex keyword. It is activated by default now in the current
854 default model, use --cfg=network/crosstraffic:0 to turn it off.
855 * Ongoing refactoring the model parsing to make XML files optional
856 See include/simgrid/platf.h for details (still to be completed)
859 * Major overhaul of the documentation. Almost instructive now :/
860 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
861 functions and friends. This interface was considered as
862 deprecated since over 2 years, it's time to inform our users that it is.
863 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
864 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
865 use these (crappy) functions in your code.
866 These functions will be removed soon. Stop using them now.
867 * Deprecate MSG_get_host_{table,number}
868 Implement MSG_hosts_as_dynar() instead.
869 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
870 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
873 * Stabilize the parallel execution mode of user contexts
874 * Introduce configuration variables to control parallel execution:
875 - contexts/synchro: Synchronization mode to use when running
876 contexts in parallel (either futex, posix or busy_wait)
877 - contexts/parallel_threshold: Minimal number of user contexts
878 that must be part of a scheduling round to switch to parallel
879 execution mode (raw contexts only)
880 * Fix bugs that prevented to use suspend/resume along with
881 synchronization structures.
882 * Fix bugs in process termination that lead to invalid memory access
883 in very specific conditions.
886 * Introduce a parallel mode for the models (controlled by surf/nthreads
887 configuration item). In our tests, running the models in parallel
888 never lead to any speedups because they are so fast that the gain
889 of computing each model in parallel does not amortizes the
890 synchronization costs, even when ultra fast futexes are used.
891 This is released anyway because YMMV.
894 * Performance boost by using a swag internally to compute the set of
895 tasks that are finished and should constitute the return value of
899 * Enable it by default now that it is considered rather stable.
902 * Documentation of the tracing functions.
903 * Performance gains when tracing categorized/uncategorized resource
904 utilization by avoiding calls to get route when updating resource
905 variables. LMM constraints are being used instead.
906 * API changed to set task categories. Use MSG_task_set_category instead
907 of TRACE_msg_set_task_category, and SD_task_set_category instead
908 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
909 * Bugfix for graphicator, routes not correctly obtained, memory leaks
910 * Examples for link user variables added (see at examples/msg/tracing/)
911 * Deprecated function TRACE_msg_set_process_category completely removed
912 * Trace header updated according to the latest Paje file format
913 * Tracing network lazy updates, no longer obligate users to use full updates
914 * --cfg=tracing/platform:1 also registers power/bandwidth variables
915 * Experimental: let user code declare/set/push/pop application states for hosts
916 * API changed to allow the manual creation of graph configuration files
917 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
920 * Improve the API of Lua MSG bindings, using the Lua spirit.
921 * Each simulated process now lives in its own Lua world (globals are
922 automatically duplicated). It helps writing simulators. Will allow
923 to run Splay programs within SimGrid in the future.
924 * Add a Chord example in Lua, equivalent to the MSG one.
927 * Start the implementation of a solution to express temporal
928 properties, not only local assertions. This is still an
929 experimental work in progress, stay clear from it to be safe.
933 - Add new runtime parameters --help-logs and --help-log-categories
934 to display information about supported logging parameters and
936 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
937 don't exists anymore.
938 * Mallocators: allow value NULL for the reset function.
940 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
941 create homogeneous dictionaries, where all the elements share the
942 same free function. Non homogeneous dictionaries will be
943 deprecated in the next release.
944 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
945 - Multi-level dictionaries are deprecated.
947 - new function xbt_dynar_search_or_negative() that is useful when
948 you have less than 2 million elements in your dynar and don't
949 want of the extra complexity of catching exceptions when the
950 element is not found.
952 - Make xbt_os_thread module (for thread portability) public.
953 Documentation is still to come, sorry.
955 - Cleanups and simplifications to make it maintainable again.
956 - Exotic features (such as memalign and valloc) were removed.
957 - The metadata were extended and improved so that the
958 model-checker becomes able to explore and inspect the heaps.
959 - This may induce a performance drop when enable_model-checking is
960 ON in cmake (even if it's not used in the simulation), but it is
961 necessary at this point to get MC working.
963 Turn model-checking OFF if simulation performance matters to you.
964 Not enabling it at runtime is not enough, disable it in cmake.
966 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
969 SimGrid (3.6.2) stable; urgency=low
971 The "Not coding new stuff allows to polish old things" release.
974 * New bindings to the NS3 packet level simulator (experimental)
975 * Use the raw (efficient) execution contexts instead of the sysv
976 (portable) ones when possible.
977 * libpcre is now mandatory in any cases since not using it led to
978 severe performance loss and possibly other issues
979 * Update the XML platforms:
980 - G5K: include the latest machine in Nancy
981 - GridPP and LCG: new platforms
982 * Documentation was partially updated, at least (more to come)
984 Bug fixes, cosmetics and small improvements
985 * Free terminated processes before the end of the simulation to avoid
986 exhausting the memory of users having very dynamic amount of
988 * Bug fix and cosmetics about canceling non-running tasks
989 * Bug fix about the dot loader's issues when using libcgraph
992 * Create an installer for windows with nsis (amd64 and win32)
993 - Add an hello world project to illustrate simgrid project creation.
994 - Embed libpcre into the Simgrid installer to avoid
995 its compilation burden
996 * The raw execution contexts should work on Apple now
997 * Port to Windows 64 bits
998 - Sysv contexts now have an implementation for this arch
999 - GRAS communication features now support this arch
1000 * Drop support for borland compiler on windows
1001 - this code was not maintained, and we kinda depend on gcc nowadays
1002 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1003 * Fix portability issue on unstable ubuntu: linker became picky on
1006 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1008 SimGrid (3.6.1) stable; urgency=low
1010 The "Oops, we broke Macs too" release
1013 * Fixed contexts detection so that raw ones are used when possible
1014 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1015 leads to a strange error, with user code segfaulting sometimes when
1016 the generated code is not perfectly aligned (which is not
1017 controllable from the user side, depends on the amount of code)
1020 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1022 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1024 SimGrid-java (3.6) unstable; urgency=low
1027 * Split of every thing from simgrid v3.5 into a separate package.
1029 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1031 SimGrid (3.6) stable; urgency=medium
1033 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1036 * Bindings now constitute their own package, separated from the main one.
1037 Rationale: reduce our maintenance nightmare by reducing the module coupling
1038 They will soon be released on their own on gforge.
1040 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1041 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1043 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1044 * It's not quite deprecated for now because we have no replacement,
1045 but it may soon become the case.
1048 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1049 * Fortran: New user-level cache variable to store the rank of the running
1050 process. This improves performance by an order of magnitude.
1051 * C: New coccinelle script to automatically locate and modify global and
1052 local static variables.
1053 * Improved SMPI network model with a sender-side gap to account for multiple
1057 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1058 only say if a communication is finished, no matter whether it succeeded or
1059 failed. You can call MSG_comm_get_status() to know the status of a finished
1061 * New function MSG_task_dsend() to send a task and detach it. When a
1062 communication is detached, you are never notified of its success or failure
1063 and the memory is released automatically once it is finished. This function
1064 is useful when you don't care about the end nor the success of a
1066 * Change the prototypes of action replay. Sorry for inconvenience,
1067 but this is really more efficient this way (and to adapt your code,
1068 you just have to fix the initialization, that shouldn't be too long)
1069 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1070 ever managed to do anything useful with it.
1071 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1072 takes a dynar of communications. It returns immediately and gives the
1073 index of a finished communication (if any).
1074 * New example: a basic implementation of the Chord P2P algorithm.
1077 * New model for multi-core CPUs. You can now use the core attribute to
1078 precise the number of cores of a host. This is a basic model. Every
1079 process running on the host receives at most the power provided in
1080 the DTD (throughput<=power). Total throughput of process cannot exceed
1082 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1083 router linked by an up-link and a down-link (possibly asymmetrical).
1084 This kind of pattern allows to easily build last-mile model style platforms.
1085 Aggregating such patterns in a rule-based AS is thus the technique of
1086 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1087 * New model for Vivaldi routing. We transformed the Vivaldi network model
1088 into a Vivaldi routing model (based on the rule-based model). This allows to
1089 combine Vivaldi based latencies with last-mile platforms.
1092 * Added a check for NaN of IEEE754 infinite in the double entries of
1094 * Introduce a new context factory "raw", highly inspirated from the
1095 ucontext factory, but using manually crafted functions in assembly to
1096 do the work in an efficient manner.
1097 * Allow to change the used context factory at run time, not only at
1098 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1099 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1100 contexts at runtime (only with raw contexts or ucontexts).
1101 * Completely rewrote this module to allow parallel execution of user
1102 processes. Use --cfg=contexts/nthreads:N to execute user processes
1103 with N parallel threads (the default is 1, meaning no parallelism).
1104 * Allow to decide dynamically between sequential and parallel modes.
1105 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1106 processes in parallel only when their number is greater than or equal to P
1108 * Added a check for NaN of IEEE754 infinite in the double entries of
1112 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1113 won't output the state of processes when interrupted with Ctrl-C
1114 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1115 NULL-terminated array. This may solve backward compatibility issues
1116 due to the change to return type of SD_simulate. See also:
1117 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1118 * Add new macros with variable number of arguments.
1119 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1120 - in xbt/asserts.h: xbt_assert
1121 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1122 - in xbt/ex.h: THROWF and RETHROWF.
1123 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1125 * Change xbt_die() to accept a format string with arguments, just like printf.
1126 * New data structure: xbt_lib_t, like a dict but more general and with better
1130 * New configuration options
1131 Options triva/categorized and triva/uncategorized can be used to generate
1132 graph configuration files for Triva visualization tool.
1133 * Configuration option tracing/platform is renamed to tracing/categorized
1134 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1135 * New TRACE_user_link_variable interface:
1136 User provides the name of the link and the tracing variable to attach to it
1137 * the declaration of tracing categories must be done after the environment
1139 * simpler tracing interface, just one way to declare categories
1140 TRACE_category or TRACE_category_with_color, it is up to you
1141 * links in the trace file are again identified by their names
1142 * trace contains the full platform hierarchy exactly as declared using the ASes
1143 * Options tracing/msg/[task|process]:1 groups the process by hosts
1144 for both cases, tasks and processes must have names that are unique during the simulation
1145 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1146 * The experimental option tracing/msg/volume is deprecated
1147 its functionality may be reincorporated if needed
1149 The tracing generates a trace file with unordered timestamped events,
1150 because of the way the core simulator (surf) works. A script available
1151 at the tools directory (fix-paje-trace.sh) can be used to put the events
1152 in order. We have changed the tracing so it can generate ordered timestamped
1153 events in the final trace, but depending on the simulator (and how much time
1154 is simulated) that can lead to a huge memory utilization. It is deactivated
1155 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1157 Build Infrastructure
1158 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1159 - We are trying hard to keep the API stable, but it may happen that
1160 some things change (we're a research project after all, not a
1161 nuclear plant operating system). If such things should happen, you
1162 could rely on that macro to adapt.
1163 - current value: 30600 for 3.06.00, aka 3.6
1164 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1165 a number that can be compared with SIMGRID_VERSION.
1166 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1167 assertions and verbose/debug logging events are disabled at compile time.
1169 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1171 SimGrid (3.5) stable; urgency=medium
1174 * New feature to any SimGrid-based simulator: Model-Checking
1175 Check SIN#1 for more details.
1178 * New Model SMPI (three-interval linear regression for correction factors)
1179 See RR-7426, available at http://hal.inria.fr/inria-00527150
1180 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1181 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1182 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1183 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1184 * See also src/smpi/README
1189 - Tracing API changes: TRACE_start and TRACE_end should not be called
1190 by user-code. They are automatically called by simulators created
1191 with SimDAG, MSG and SMPI if the toolkit is compiled with
1192 tracing_enabled=ON. Categories declaration and utilization remain the
1193 same for MSG and SimDag.
1194 - A function was added to the tracing API to declare categories with
1196 - TRACE_category_with_color (char *category, char *color)
1197 where color must be in the following format
1198 "%f %f %f", red, green, blue
1199 and red, green, blue are float values in the interval [0, 1]
1200 - User can specify NULL as color parameter, or continue calling
1201 TRACE_category (cat)
1202 On that case, the tracing system will define random colors
1203 - The following command-line options are supported:
1204 --cfg=tracing/filename:msg.trace
1205 --cfg=tracing:1 (activate tracing, needed to use others)
1206 --cfg=tracing/platform:1 (categorized resource use)
1207 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1208 --cfg=tracing/msg/task:1 (task creation)
1209 --cfg=tracing/msg/process:1 (process creation, migration)
1210 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1211 --cfg=tracing/smpi:1 (SMPI interface tracing)
1212 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1213 - examples of examples/msg/tracing updated
1215 - DAXLoader and DOTLoader functions can generate tasks with categories
1216 - A new function to attribute a category to SD tasks:
1217 TRACE_sd_set_task_category (SD_task_t task, char *category)
1218 * Tracing the MPI interface implemented by SMPI
1219 - Collective operations are traced with states
1220 - Point-to-Point operations are traced with states/links
1221 - Tracing activated by a parameter "-trace filename" passed
1222 to smpirun during execution (considering that simgrid
1223 is compiled with tracing enabled)
1224 - To run the simulation with gdb, the simulator
1225 accepts --cfg=tracing/smpi:1 to trace SMPI
1226 - tesh files to check if smpi tracing is ok
1227 - See examples/smpi/NAS/DT-trace
1228 * GTNetS tracing re-worked
1229 - adaptation to the tracing system of GTNets to cope
1230 with modifications regarding the fullduplex mode
1231 - new tesh files to check if gtnets tracing is ok
1234 * Asynchronous communications through the functions:
1235 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1236 * New function: MSG_load_platform_script()
1237 to make possible using a lua script instead of XML files to set up platforms
1238 * New function: MSG_set_function
1239 to associate functions to processes, used when bypassing the parser
1240 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1242 Platforms: Add some more examples in examples/platforms
1243 * Grid'5000: see www.grid5000.fr
1244 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1247 * Change the XML format. This is a very important modification. SimGrid
1248 3.5 introduces a new hierarchical format based on the notion of
1249 Autonomous Systems. Compatibility with old format is ensured
1250 through the perl script provided in the install bin directory
1251 bin/simgrid_update_xml.
1252 It is now possible to build platforms with specific routing
1253 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1254 connect several platforms together. We will try to provide soon
1255 set of realistic platforms exploiting these properties (have a
1256 look at examples/platforms/ for the moment).
1257 * Take the opportunity of the XML format change to be a good XML citizen:
1258 rename link:ctn to link_ctn and similar changes (also dealt with by
1260 * Add a new routing scheme (rule-based) using regular expressions. It
1261 enables to have an extremely low memory footprint when the
1262 underlying routing is simple and can be compactly described. You
1263 need to have libpcre4-dev (perl regular expressions) installed if
1264 you want to use this routing scheme.
1265 * Revive the cluster TAG and allow to easily and efficiently (both in
1266 term of memory and speed) connect clusters together. Have a look
1267 at teshsuite/simdag/platforms/ to see how this can be done. With
1268 this tag, you can create clusters with thousands of tasks at no
1269 cost (have a look at examples/platforms/).
1270 Note: clusters are implemented as ASes, so there is no need for an
1271 enclosing AS tag if you have only one cluster in your platform.
1272 * Add new generic functions in the public interface that allows the user
1273 to call SURF 'create_resource' methods from your code (same
1274 functionality as the XML bypass mechanism but with a much lighter
1276 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1277 that uses a piecewise linear approximation to produce better
1278 results when exchanging small messages.
1279 * Add a new parameter to handle correctly full duplex link and account
1280 for interferences between uplink and downlink communications
1281 (activate with --cfg=fullduplex:1).
1284 * Rename the SD_READY (all dependencies are satisfied and task is
1285 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1286 dependencies are satisfied) state.
1287 This prevents a confusion between the notion of "ready to schedule"
1288 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1289 simulated" (SD_RUNNABLE) used by the simulation kernel.
1290 * Change the way a task is considered as ready. Instead of removing
1291 dependencies when a task is done, a counter is decreased. This way,
1292 it is always possible to reach ancestors thanks to the
1293 SD_taks_get_parents function (even after the end of the simulation.)
1294 * Change the return type of SD_Simulate from (SD_task_t*) into
1295 xbt_dynar_t. This function was in handling a dynar internally and
1296 converted it into a NULL terminated array for historical reasons.
1297 * New function SD_dotload(char*) to load a DAG described in dot
1298 format. This loader and the corresponding examples require the
1299 installation of the graphviz library.
1300 * Fix a bug in the management of tasks of size 0 in the surf network
1301 models. This problem was only visible with SIMDAG and you should
1302 thus disregard results produced with earlier versions if you
1303 relied on this feature (some tasks were blocked because of this).
1304 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1305 even though your applications were doing only point-to-point
1306 communications and sequential computations. Now you can really use any
1307 model you want (of course, if you create real parallel tasks, which are
1308 not implemented in most models beside ptaskL07, this will abort).
1309 * Add an example that schedules a DAX on an heterogeneous platform
1310 using a Min-Min strategy.
1311 * New function SD_workstation_get_current_task() that returns the kind
1312 of task currently running on a workstation in the sequential access
1314 * Raise some warnings when unexecuted tasks remains at the end of the
1315 simulation. This is usually caused by cycles in the DAG.
1318 * New function: SIMIX_process_set_function() called by MSG_set_function
1319 * Change the underlying waiting queue in semaphores so that a process
1320 can wait on several of them simultaneously (as in waitany).
1321 * Fix the way to handle tokens in semaphores so that all access patterns
1322 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1323 * kill the dirty pimple SIMIX_message_sizes_output()
1324 Please use (proper) visualization instead
1327 * New data container: setset (set of sets of elements)
1328 * New module: mmalloc (mapped malloc, allowing to have several
1329 independent segments of malloc)
1330 * New function: xbt_dict_cursor_set_data()
1331 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1332 * New function: xbt_dynar_is_empty()
1333 * New function: xbt_fifo_get_last_item()
1334 * Fix xbt_dynar_shrink(): use the right element size.
1335 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1336 uninitialized areas during expand.
1337 * Fix semaphores: previous implementation was severely broken.
1338 * Use library init/fini functions for our initialization.
1339 - you can use logs and other feature as soon as you want in your
1340 code (even before the xbt_init / MSG_init)
1341 - xbt_exit is now a no-op and produce a warning when used.
1344 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1345 benefit from the latest improvement to the simulation kernel.
1346 * Kill measurement sockets for now. If you rely on them, sorry. This
1347 release is not for you. This feature will be reintroduced in the
1348 future, but we cannot delay the release any further.
1349 * New function: gras_msgtype_get_name().
1350 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1351 * Fix (at last) the pmm example: it should not randomly fail anymore.
1353 Build chain: bug fixes and overall polishing
1354 * Cmake is now stable enough. Hence, we killed the autotools.
1355 * Port to windows ( TM :)
1356 * Fix the 'make install' target.
1357 No need to use 'make install-simgrid' anymore
1358 * Introduce a 'make dist' target compiling a *source* archive
1359 'make package' compiles a binary archive
1360 * Compile java files only on need
1361 * Add --cd and --setenv command line options to tesh
1362 * Out of source builds are not fully supported yet, but we are close
1363 * Enable supernovae and optimization flags by default for our users
1366 * Add layer to set up environment directly from lua, without XML.
1367 * The effect of gras_stub_generator can be achieved through
1368 lua too (check examples/gras/console/ping_generator.lua)
1370 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1372 SimGrid (3.4.1) stable; urgency=low
1374 The "Polishing easter eggs is probably a good idea" release.
1375 This is a bug fixes release only.
1379 * Fix a bug preventing the tasks from begin garbage collected.
1382 * Fix a bug occuring when a host involved in a communication fails.
1383 This was not detected properly by the other peer involved in the
1384 communication. Now, it's reported as a network error.
1387 * Warn the user about loop dependencies in data flow of DAX files
1388 * Obey the control-flow dependencies of DAX files
1391 * Add option "enable_smpi" allowing to not compile SMPI.
1392 Probably useful for the (Mac) users experiencing a build error here
1393 * Improve the detection of lua5.1 and ruby1.8
1395 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1397 SimGrid (3.4) stable; urgency=low
1399 The "Easter in Cargese" release. Also known as (major changes):
1401 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1402 ~> bindings were greatly improved
1403 ~> new tracing infrastructure for better visualization introduced
1405 * the "Welcome to configury modernity" release.
1406 ~> we switched from autotools to cmake, and improved our cdash
1409 A more detailled list of changes follow (full detail in svn log).
1411 Java Bindings: Various Cleanups
1412 * (install java-gcj-compat-dev on debian-like to use them)
1413 * Remove put/get: no need to export deprecated interface in Java
1414 Use send/receive instead.
1415 * Cleanup the examples and add a README per directory
1416 * Remove example autoDestination (that's the only way to go now)
1417 * Remove example explicitDestination (was a plain copy of basic)
1418 * Make JniException a runtime exception, so that there is no need to
1419 declare the fact that you may encounter such a beast. I guess that
1420 nobody will ever want to survive such error.
1421 * Create specific errors for each MSG case of failure:
1422 host failure, transfer failure, timeout, task cancelled
1423 * Cleanup the exceptions that may get thrown by each function
1424 * Other internal cleanups in Java bindings. Performance still bad :/
1425 Ruby and Lua Bindings: create them
1426 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1427 * That's new and great, you should try them out.
1428 Same functionalities than Java bindings, only even less polished
1430 * Kill the useless "rate" argument of SD_task_get_execution_time()
1431 Everyone used to provide -1 as a value, it was not used, and the
1432 semantic of a possible use wasn't even clear.
1433 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1434 either as comm costs or compute costs to mean that there is no
1435 such thing for that specific task.
1436 * Add a SD_task_set_name() function
1437 * Fix SD_task_unschedule() on typed tasks
1438 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1439 * In DAX loader, accept useless 'level' attributes to <job> since
1440 LIGO DAGs have them (seem to be to ease graphical representation).
1442 * Add an example masterslave_mailbox.c using send/receive and not
1443 the deprecated put/get interface.
1444 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1445 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1447 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1448 (declare MSG_USE_DEPRECATED to still have the old name)
1449 * Add a MSG_task_set_data() function
1450 * About trace replay (see examples/msg/actions):
1452 - Allow to work with splitted trace files for each process
1453 Give the specific trace file as argument of each process,
1454 and call MSG_action_trace_run(NULL)
1455 You can still have one merged file for all processes.
1456 - Fix implementation of collective operations
1457 * Allow task_execute() on 0-sized tasks (closes #10063)
1459 * This is the first release of SimGrid where SMPI is not considered
1460 beta anymore (even if some corners should still be improved)
1461 * Port over the new SIMIX_network submodule (internal refactoring)
1462 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1463 * Implement more missing elements of the standard:
1465 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1467 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1468 MPI_Scan() MPI_Get_processor_name()
1469 - Added implementation of missing case for Alltoall (warning: it's
1470 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1471 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1472 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1473 SMPI_MPI_Allgatherv()
1474 * Bug fixes include:
1475 - MPI_Waitsome() was broken
1476 - Allow relative includes in smpicc
1477 - Command line cfg argument 'reference_speed' was ignored...
1478 - Some functions did not properly lead to auto-benching of user code
1479 - smpicc passes -O2 by default (just like openmpi one)
1481 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1482 * Bug fixes about timeouts during communications
1483 * add SIMIX_message_sizes_output() as a pimple to write to file the
1484 amount of messages per size. Use gnuplot to get histogram.
1485 Pimple because that's the only user-visible function of simix,
1486 defined directly in xbt.h (irk, sorry)
1488 - Add a SIMIX_sem_get_capacity() function
1489 - Fix interactions with processe resume/suspende
1490 - release_forever() was stupidly broken
1491 - Fix SIMIX_display_process_status() for processes in a semaphore
1492 - Make SIMIX_sem_block_onto() user-visible
1493 * Refactoring context stuff:
1494 - Use pseudo-OOP for better modularity
1495 - reimplement SIMIX_process_kill() without process_schedule() so
1496 that the latter can take as invariant that it is called from
1498 - Merge context_start into context_new for sake of simplicity
1500 * Add a Vivaldi network model, coded live during SUD'10 ;)
1501 * Rename configuration variables to start a hierarchy:
1502 o cpu_model -> cpu/model
1503 o network_model -> network/model
1504 o workstation_model -> workstation/model
1505 * New configuration variables:
1506 o network/bandwidth_factor: correction to bandwith
1507 o network/latency_factor: correction to latency
1508 o netwotk/weight_S: correction to the weight of competing streams
1509 * Add a long description to the models, that users can see with such
1510 argument on the command line: --cfg=cpu/model:help
1511 * --help-models display the long description of all known models
1513 * config: add the ability to set a default value after registration
1514 Does not override any previously set value (e.g. from cmd line)
1515 * dict: allow to have integer key and data.
1516 When so, you need to use the following functions
1517 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1518 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1519 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1520 In contrary to regular dicts, the key is not malloced before copy.
1521 Mixing scalar and regular elements in the same dict is not tested
1523 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1524 Tracing for Visualization:
1525 * SimGrid is now instrumented in order to generate a trace file for
1526 visualization analysis: to use it, need to compile SimGrid with the
1527 "tracing" option enabled, and instrument the program using SimGrid with
1528 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1529 (among other functions).
1530 * The instrumentation only traces the platform utilization for now
1531 * Documentation to use the tracing functions and how to analyze the
1532 traces with the Triva tool is written.
1533 * More information about: SimGrid FAQ (in the section Tracing Simulations
1536 * We moved to cmake as default build system. Autotools support will
1537 be dropped soon. Check the FAQ for more info about how to use it.
1538 * Greatly improved our cdash/ctest interactions
1539 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1540 * Added memory checking tests with valgrind; lot of memleak fixing.
1541 This may be the first release of simgrid with so few memory issues
1542 * Added code coverage tests.
1543 Our coverage is still improvable, but at least we see it on cdash.
1545 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1547 SimGrid (3.3.4) stable; urgency=low
1549 The "Desktop Grid needs love too" release (also called Xmas release).
1551 Models improvements:
1552 * Major speedup in the maxmin system solving by using lazy evaluation
1553 Instead of solving completely the maxmin system at each iteration,
1554 only invalidate (and recompute) the modified parts.
1555 This new feature is enabled in default models but you can try to
1556 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1557 * Cas01 IMproved as default CPU model
1558 This CPU model is the same Cas01 model, but it uses the
1559 maxmin-selective-update flag and a heap structure to manage
1560 actions on SURF kernel.
1561 It reduces the complexity to find the next action to finish and,
1562 consequently, it's faster than the old Cas01.
1563 This is the new default CPU model (Cas01).
1564 * Rename the old Cas01 model to Cas01_fullupdate
1565 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1566 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1567 * CpuTI (CPU Trace Integration)
1568 A new CPU model whose objective is simulate faster when using
1569 availability trace files.
1570 Instead of using a full featured, over engineered maxmin system for
1571 CPU modeling, this model does the pre-integration of traces files
1572 to calculate the amount of CPU power available, and so, executes
1573 faster than the old CPU models.
1574 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1575 * Use LV08 as default network model since it gives better accuracy
1576 for small messages and shouldn't change things for big ones.
1577 Use --cfg=network_model:CM02 to get the previous behavior.
1580 ******************************************
1581 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1582 ******************************************
1583 * The new CPU model may changes simulations!
1584 The point is that events occurring at the exact same timestamp
1585 are not scheduled in the same order with the old and new
1586 version. This may be enough to completely change the execution
1587 of simulations in some cases.
1588 * The new network model will change simulations!
1589 This new model is more realistic than the previous one, so you
1590 should consider redoing your old experiments with this model.
1591 Sorry for the inconvenience.
1594 * Introduce the supernovae compilation mode
1595 When compiled that way, the whole SimGrid (or almost) is put in a
1596 single compilation unit and compiled in one shoot.
1597 This is to help gcc which has difficulties to inline stuff from one
1599 The speedup seem to be above 15%, althrough more tests are needed on
1600 amd64 to confirm that gain.
1603 * Port of MSG's mailbox on top of SIMIX network
1604 The put/get mechanism was greatly simplified on the way.
1607 * New SIMIX network module. Provides:
1608 - Mailbox: rendez-vous mechanism to find with who you want to speak
1609 - Synchronous send/recv: easier and hopefully faster since the
1610 logic is handled in the maestro process directly now
1611 - Asynchronous send/recv: you dreamt of it? It's here now
1612 Too bad that nobody cared enough to propagate the change to MSG.
1613 * Add semaphores as SIMIX synchronization mechanism.
1616 * new function SD_daxload(char*) to load a DAX file
1617 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1618 * Introduce typed tasks. Specify its kind and cost at creation.
1619 At scheduling, just give where it should be placed, and the cost
1620 for each involved resource is automatically computed.
1621 Existing constructors so far (more to come of course):
1622 - SD_task_create_comm_e2e() for end-to-end communication
1623 - SD_task_create_comp_seq() for sequential computation
1624 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1625 * new function SD_task_dump() for debuging display
1626 * new function SD_task_dotty(task,FILE*) writing to file the info
1627 about the task in dotty format
1628 * SD_task_dependency_exists() can now cope with having one of its
1629 arguments NULL. If so, it tests whether the other argument has any
1631 * Add getters on list of preceding/following tasks:
1632 SD_task_get_parents(task) and SD_task_get_children(task)
1633 * Add getters on amount of workstations and list:
1634 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1635 * Add getter on task kind: SD_task_get_kind(task)
1636 * Update the start_time and finish_time of tasks on completion/failure
1637 * Bugfix: Remove task from state swags when destroyed
1640 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1643 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1644 dynar with one separate thread per value of the dynar.
1645 * Change the prototype of xbt_thread_create(), sorry.
1646 Added a boolean parameter indicating whether we want to join this
1647 thread (used in SG only for now)
1648 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1651 * GTNetS wrappers should now be usable again (and betterly tested too)
1652 * Fix a major regression from 3.2 where the timeout provided to
1653 MSG_task_put_with_timeout() was used as absolute time before which
1654 the comm should be done.
1655 * Start to fix the <cluster> tag.
1656 - Internal links should be good now (beside of the loopback, which
1657 use the private link instead)
1658 - paths to the external world is still rather broken
1659 - the <route:multi> tag is just broken. Actually that's brain-dead.
1660 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1661 to make it less stupid
1662 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1663 * Fix a source-level compatibility glitch from 3.2: after defining
1664 MSG_USE_DEPRECATED, you can use the old name
1665 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1666 * Allow to compile from the SVN with automake 1.11
1667 * Fix some problems when using the "start_time" tag in deployment XMLs.
1668 * Fix #8569: XBT/synchro.h has redundant declarations
1669 * Fix #8563: MSG return values and exceptions
1670 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1671 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1672 Thanks to Robson Peixoto.
1673 * Fix a vicious bug in dictionaries inducing that some elements were
1674 not freed on xbt_dict_free()
1676 Portability report of this version:
1677 * Main portability targets:
1678 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1679 on (amd64/i386/ia64)
1680 - mac leopard on i386
1681 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1682 but nothing critical.
1683 * Other platforms: windows, AIX and others were not tested for this release
1685 Timing report of this version:
1686 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1687 scenario parameters). From 8h to a few seconds in desktop grid settings.
1688 * Supernovae brings about 25% speedup on i386.
1690 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1692 SimGrid (3.3.3) stable; urgency=low
1694 The "Need for Speed" release.
1696 The timings done to validate the 3.3.2 were faulty.
1697 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1699 The problem was a conversion from a manually handled vector to
1700 xbt_dynar_t on the critical path.
1701 xbt_dynar_foreach calls functions, inducing stack management crap.
1703 We inlined these functions and xbt_dynar_foreach is now breath taking.
1704 We also inlined xbt_swag_belong on the way.
1706 Here are some approximate speedup measurements (on master/slaves
1707 simulations lasting between 10s and 20s each):
1708 3.3.1 -> 3.3.2: about same performance
1709 3.3.2 -> 3.3.3: 40% speedup
1710 3.3.1 -> 3.3.3: 40% speedup
1711 3.3.1 with inline patch -> 3.3.3: 30% speedup
1713 Our reading is that the refactoring which occurred in 3.3.2 made us
1714 suffer much more from the xbt_dynar_foreach low performance, but
1715 once we solved this, this refactoring proved to be very performance
1716 effective. From the 40% speedup, somehow, 10% are due to the
1717 inlining and 30% to the refactoring.
1719 That's a pitty that gcc cannot inline functions placed in other files
1720 alone. We have to choose between:
1721 - break the encapsulation (by putting private data structures and
1722 accessors in headers files to help gcc)
1723 - live with low performance
1724 - switch to a decent compiler such as icc (not quite possible).
1726 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1728 SimGrid (3.3.2) stable; urgency=low
1730 The "Simplicity does not preceed complexity, but follows it" release.
1732 The main contributors of this release were (lexical order):
1733 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1736 * Extract the routing logic into its own object.
1737 (was dupplicated in network.c and workstation_LV07.c;
1738 Allows to implement other ways of storing that info)
1739 => kill now useless network_card concept
1740 - Use dynar to represent routes (instead of void** + int*)
1741 - kill link_set (use surf_network_model->resource_set instead)
1742 - Add a command-line option to choose the routing schema to use
1743 - Add three new models:
1744 * Floyd (shortest path computed at initialization)
1745 * Dijikstra (shortest path recomputed all the time)
1746 * Cached Dijikstra (shortest path computed on need)
1747 All these models where contributed by Silas De Munck, and are
1748 described in his ICCS09 paper.
1750 * Simplify model declaration
1751 (less redirections, less function to write when defining a model)
1752 - Factorize stuff between models:
1755 surf_model_resource_set(model)
1756 surf_model_resource_by_name(model, name)
1757 - Unify the types of models in s_surf_model_t (using an union)
1758 - Embeed fields of common_public directly into s_surf_model_t
1759 - Rename model methods:
1760 action_free ~> action_unref
1761 action_change_state ~> action_state_set
1762 action_get_state ~> action_state_get
1763 - Change model methods into functions :
1764 (model)->common_public->action_use ~> surf_action_ref
1766 * Implement a generic resource; use it as ancestor to specific ones
1767 (allows to kill duplicated code in models)
1768 Drawback: timer command don't need no name nor properties;
1769 workstation_CLM03 don't need no properties
1770 (but I guess we can live with those few bytes wasted)
1772 * Improve the action object model
1773 - implement a constructor avoiding dupplicated code about field
1774 initialization in generic_action part.
1776 * Kill the SDP model: it has an external dependency, is deprecated
1777 in flavor of modern lmm models, and didn't compile since a while
1780 * Relocation of the context module from XBT to SIMIX.
1781 (the context were decoupled from the simix processes, duplicating a lot of code)
1782 => a lot of code was factorized
1783 - less overhead is introduced during scheduling
1784 - simpler API for the context factory
1785 - the logic for process creation,destruction and manipulation was simplified
1786 * Simplification of the s_smx_process_t data structure.
1787 => accesing the simix level data associated to a process is faster now,
1788 and the code is a lot more readable.
1791 * Implement some more MPI primitives:
1792 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1793 -implementation: Bcast: flat or 2-ary tree (default),
1794 Barrier: 4-ary tree,
1796 Allreduce: Reduce then Bcast
1797 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1798 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1799 Alltoallv: flat tree, like ompi
1801 * Add support for optimized collectives (Bcast is now binomial by default)
1802 * Port smpirun and smpicc to OS X
1805 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1806 Shout out if you used it.
1809 * Display the list of still queued messages in SG mode when existing
1813 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1814 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1815 * Add xbt_str_from_file(FILE*)
1816 * Add xbt_dict_get_key achieving a linear reverse search
1817 * Remove the context module
1819 Portability report of this version:
1820 * Main portability targets:
1821 - Linux(debian)/x86/context
1822 - Linux(debian)/x86/pthreads
1823 - Linux(debian)/amd64/context
1824 - Linux(debian)/amd64/pthreads
1825 On these, we still have the eratic breakages of gras/pmm and
1826 amok/saturate_sg reported in previous version. We still think
1827 that the tests are the cause of the fault, not the tested code.
1829 - Mac OSX Leopard/x86/context
1830 Still false negative in tesh autotesting.
1831 Smpi still fails, but this time because readlink does not accept -f
1832 Everything seems to work properly beside of that.
1835 - AIX version 5.3 (only tested contexts this time)
1836 Smpi still fails there because mktemp is not installed.
1837 Everything seems to work properly beside of that.
1839 I managed to compile it for the first time, but several breakages.
1840 Won't delay the release for this exotic platform.
1842 * Windows: it's still lagging behind. If you want to help, please
1845 Timing report of this version:
1846 This version seem to be more than 5% faster than 3.3.1 (on linux
1847 64bits with contextes). The gain is less than expected, we are
1848 investigating this for next release.
1850 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1852 SimGrid (3.3.1) stable; urgency=low
1855 * Implement a --cfg-help to show existing configuration variables
1856 * Build chain do not require doxygen in maintainer mode
1859 * fix a bug on struct sizeof computation, which prevented the
1860 exchange of arrays of structs in some conditions
1861 - added a regression test about this in datadesc_usage
1862 * Allow the exchange of 0-long dynamic vectors.
1863 - for that, use -1 as indicator of dynamic size instead of 0
1864 - This implied to change any size from unsigned long to long,
1865 reducing a bit communication abilities, but I guess that with
1866 64bits being quite common, this is more than enough.
1867 - This also induce a protocol change, thus bumping network protocol
1868 version from 0 to 1 (if we have external users, we have to get
1869 clean on that point too ;)
1870 - added two regression tests about this in datadesc_usage
1871 * Be more verbose when propagating local exceptions
1872 This helps debugging.
1873 * Display the status of simulated processes when receiving SIGINT in
1877 * Allow to control the simulation from a trace file.
1878 New functions MSG_action_register() and MSG_action_trace_run()
1879 The first one allows to associate a function execution to each
1880 kind of action while the second one parses a trace file and
1881 triggers the corresponding actions within the system.
1882 For now, only a toy example is provided in examples/msg/actions
1883 * Add an exemple of process migration in examples/msg/migration
1884 * Fix a bug in task exchange which broke MSG_task_get_sender()
1885 Add a teshsuite regression test for that.
1886 [Bug: if MSG_task_get_sender() is called after sender exit,
1888 * Fix a bug which prevented suspend/resume to work properly
1889 * Display the status of simulated processes when receiving SIGINT
1890 This fixes a regression of v3.3. due to the introduction of SIMIX
1891 * Bug fixing in failure management:
1892 - trace could not start by a failure at time 0
1893 - failure during communications were not working
1896 * Add SIMIX_process_set_name() to change the name of the current
1897 process in the log messages.
1898 * Store smx_hosts in a dict since we only retrieve them by name
1899 * Move the configuration infrastructure to surf
1902 * Move the configuration infrastructure to surf
1905 * Massive internal cleanups:
1906 - Store internal structures on processes instead of hosts (allows
1907 to have more than one process per host, in addition of being more
1909 - Cleanup the initialization/finalization process
1910 - Kill a whole bunch of unneeded synchronization:
1911 processes run in exclusive manner within the simulator
1912 - Move queues from global tables to process data fields
1914 - now accept -platform and -hostfile arguments
1915 - Pass the right rank value to processes according to the hostfile
1916 * Compile the examples by default, and use them as regression tests
1917 * Implement MPI_Wtime()
1918 * Change the reference speed to a command line option
1921 * TCP_gamma can now be specified as command line option using
1922 --cfg=TCP_gamma:10000000.0
1923 * Change the --surf-path cmd line option into --cfg=path:
1926 * Also include strbuff from xbt.h public header
1927 * xbt_ex_display(): do not free the exception after displaying
1928 This allows to do more with the given exception afterward.
1929 Users should call xbt_ex_free() themselves.
1933 Portability report of this version:
1934 * Main portability targets:
1935 - Linux(debian)/x86/context
1936 - Linux(debian)/x86/pthreads
1937 - Linux(debian)/amd64/context
1938 - Linux(debian)/amd64/pthreads
1939 These targets fail about 1/10 of times on gras/pmm, but we believe
1940 that this is because of the test, not because of simgrid.
1941 amok/saturate_sg fails even more rarely, and the test may not be
1944 - Mac OSX Leopard/x86/context
1945 The test suite still spits tons of errors because some obscure
1946 force prevents us from removing the temporary directories
1947 arguing that they still contain some metadata I've never heard of.
1948 Smpi fails because seq is not installed.
1949 Everything seems to work properly beside of that.
1952 - AIX version 5.3 (both contexts and pthread)
1953 Smpi still fails there because mktemp is not installed.
1954 XML inclusions seems rosty on AIX.
1956 * Windows: it's still lagging behind. If you want to help, please
1959 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1961 SimGrid (3.3) stable; urgency=high
1965 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1966 [Malek Cherier & Mt]
1968 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1971 Until now, GRAS were using MSG as an interface to SURF. It was
1972 quite difficult because both interface have several differences
1973 (MSG channels vs GRAS sockets were the most notable point).
1975 This also opens the gate to SMPI (which should occur soon) and speed
1976 up simulations by to 40% (even if it were not the main goal).
1978 **************************************
1979 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1980 **************************************
1981 The point is that events occuring at the exact same timestamp are
1982 not scheduled in the same order with the old and new version. This
1983 may be enough to completely change the execution of simulations in
1984 some cases. Sorry for the inconvenience.
1986 * Cleanup and upgrade the XML format to push further scalability
1987 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1989 * Improve the testing infrastructure with tesh. Now a very large part of
1990 the code is tested not only by being run but also by checking that the
1991 output match an expected output [Mt].
1993 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1994 is really less memory-demanding, which should allow you to use
1995 larger files in SimGrid [AL].
1997 * Inform valgrind about our contextes, so that it becomes usable
1998 with the default (and more effecient) version of SimGrid
1999 [contributed by Sékou Diakite, many thanks]
2002 * Introduce a listener thread in charge of receiving incoming
2003 messages from the network. It allows to overlap communication and
2004 computation but most notably, it removes some stupid deadlocks due
2005 to the fact that so far, a process could not send and receive at
2006 the same time. This made most non trivial communication schema
2008 * Convert the PIDs from long int to int to match the MSG ones (and
2009 linux ones too) [Mt]
2010 * New function: gras_agent_spawn() to launch a new process on
2011 current host. Only working in simulation for now. [Mt]
2012 * New function: gras_os_hostport() returning a constant form (ie,
2013 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2016 * Make the backtrace of exceptions more human readable [Mt]
2017 * New module: xbt/str [Mt]
2018 a ton of string utility functions (split, join, printf to a newly
2019 allocated buffer, trim, etc)
2020 * New module: xbt/hash [Mt]
2021 SHA1 hashing algorithm (more to come if needed)
2022 * New module: xbt/synchro [Mt]
2023 synchronization tools (mutex and conditions) working the same way
2024 in simulation and in real life (mainly useful for GRAS, but not
2026 * New module: xbt/queue [Mt]
2027 classical producer/consumer synchronization scheme
2028 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2029 (using the classical functions will get serialized) [Mt]
2030 * Make dictionary internal table dynamic. No need to specify its size
2031 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2033 * Make sure the log channels are organized as a tree under windows
2034 (because of ANSI C compatibility issue, any channel were child of
2035 root directly) [Mt].
2038 * Cleaned many thing in surf and fixed a few bugs [AL].
2039 * Add a nice command line configuration mechanism to compose models [AL].
2040 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2041 the previous one (KCCFLN05). It relies on something that looks like
2042 a max-min sharing mechanism but cannot be written as such. A new solver
2043 was thus designed [AL].
2044 * Add a new solver to lmm. Based on Lagrange optimization and
2045 gradient-based descent, it enables to efficiently maximise systems s.a
2047 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2049 This solver enables to propose two new network models for TCP Reno and
2050 TCP Vegas based on Low's work. These models still need to be fully
2051 tested though [Pedro Velho].
2054 * Bug fix in SD_simulate. Now the time bound given as argument is
2056 * Use the new parallel task model (ptask_L07) as default.
2057 * Use the SURF command line configuration mechanism.
2058 * 0-size tasks (for synchronization) should now work.
2060 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2062 SimGrid (3.2) stable; urgency=high
2066 We still experience issues on this platform, but we believe that at
2067 least MSG is usable.
2069 GRAS API BREAKAGE (for simplification purpose, sorry):
2070 * the gras_msgtype_by_name is not used anymore. Instead of
2071 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2072 you can write (and must)
2073 gras_msg_send(toserver, "request", &request);
2074 - If you still want to pass a gras_msgtype_t to the function (to cache
2075 the type and avoid the lookup time), use the gras_msg_send_() variant.
2076 - Impacted functions:
2077 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2078 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2079 * The callbacks are now expected to return 0 when everything went well
2080 (just like the main() function)
2082 GRAS new features and improvements:
2083 * New module mechanism where user code can use per process globals [Mt]
2084 This is similar to gras_userdata_*() functions, but for libraries. It
2085 factorize some code developped over and over in the examples and AMOK.
2086 It has still to be documented and used (only amok/peermanagement is
2088 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2089 loss when they were small enough to fit into the buffer and sent quickly
2090 enough so that they can all get received in one shoot.
2091 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2092 if not found. Use the *_or_null() variant for the old semantic.
2093 * In gras_msg_handle, do not discard messages without callback.
2094 They are probably messages to be explicitly awaited later (ie, proofs of
2095 mis-synchronization in userland since they are sent before being awaited)
2096 No big deal usually.
2097 * gras_socket_meas_send/recv: semantic changed!
2098 The numerical arguments used to be (1) the total amount of data to send
2099 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2100 messages. This was need for the fool willing to send more than MAXINT
2101 bytes on quite fat pipes.
2104 * Do really rename the hostmanagement module to peermanagement. [Mt]
2105 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2106 but this is rather new and this was documented in the module
2107 documentation (poor excuses, I admit)
2108 * Bandwidth measurement semantic changed! This follows the changes to
2109 gras_socket_meas_send/recv explained above.
2112 * A sequential mode has been added to the workstations. When a workstation
2113 is in sequential mode, it can execute only one task, and the other tasks
2114 are waiting in a FIFO. [Christophe Thiery]
2117 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2118 model for SIMDAG. [Christophe Thiery]
2119 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2120 equal to 0, which caused some bad behaviors in
2121 saturated_constraint_set_update. I now use a threshold mechanism like in
2125 * When running manually src/testall, you select specific units [Mt]
2126 testall is the result of our cunit mechanism, and should replace all
2127 the scripty thingy around since bash don't run easily on billware.
2129 * A mallocator system has been added. [Christophe Thiery]
2130 Mallocators allow you to recycle your unused objects instead of freeing them
2131 and allocating new ones.
2133 Documentation update:
2134 * FAQ reworking + New FAQs:
2135 - "Valgrind spits tons of errors!" [Mt]
2136 - "How to repport bugs" [Mt]
2137 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2138 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2139 - Communication time measurement within MSG [AL]
2140 - I experience weird communication times when I change the latency [AL]
2141 * GRAS tutorial [Mt]
2143 - an introduction to the framework and to the used communication model
2144 - an initiatic tour introducing the most proheminent features:
2146 . Lesson 0: Installing GRAS
2147 . Lesson 1: Setting up your own project
2148 o Part 2: Message passing
2149 . Lesson 2: Exchanging simple messages
2150 . Lesson 3: Passing arguments to the processes (in SG)
2151 . Lesson 4: Attaching callbacks to messages
2152 . Lesson 5: Using globals in processes
2153 . Lesson 6: Logging information properly
2154 . Lesson 7: Using internal timers
2155 . Lesson 8: Handling errors through exceptions
2156 . Lesson 9: Exchanging simple data
2157 . Lesson 10: Remote Procedure Calling (RPC)
2158 . Lesson 11: Explicitely waiting for messages
2159 . Recapping of message passing features in GRAS
2160 - A HOWTO section containing:
2161 o HOWTO design a GRAS application
2162 More are due, of course. They will come latter. In the meanwhile, you can
2163 check the examples which are still here.
2165 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2167 SimGrid (3.1) stable; urgency=high
2171 There was a stack corruption somewhere, visible only when optimizing
2172 with these versions. [Vince]
2175 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2176 module that enabled to play with Directed Acyclic Graphs. It is built
2177 directly on top of SURF and provides an API rather close to the old
2178 SG. Some old codes using SG are currently under rewrite to check that
2179 all needful functions are provided. [Christophe Thiery]
2182 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2183 extension of the classical CLM03 model that gracefully handles
2184 failures. This is now the default model for MSG and GRAS. It doesn't
2185 handle parallel tasks yet though. [AL]
2186 * Bug fix: Weights were not correctly set in the network part.
2187 WARNING: This may have resulted in incorrect results with simulations
2188 where there are more than one flow on a given link. [AL]
2191 * After a (long ?) discussion on simgrid-devel, we have decided that the
2192 convention we had on units was stupid. That is why it has been decided
2193 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2194 WARNING : This means that all previous platform files will not work as
2195 such with this version! A warning is issued to ask users to update
2197 A conversion script can be found in the contrib module of the CVS, under
2198 the name contrib/platform_generation/surfxml_update.pl [MQ]
2201 * Bug fix: Processes were started in reverse order, wrt deployment file.
2202 WARNING: if your code relies on this bug, please fix it. [AL]
2203 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2204 being executed on two different locations. [AL]
2205 * Bug fix: Failures are now better supported thanks to Derrick's tests
2206 (there was many failure situations I hadn't thought of and that weren't
2207 correctly handled). [AL]
2208 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2212 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2213 Exception occurring on server-side are propagated back to client (!).
2215 API CHANGE: the callback changed their prototype. Change:
2216 int my_handler(gras_socket_t expeditor, void *payload_data) {
2218 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2219 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2221 * New! function: gras_msg_handleall to deal with all messages arriving
2222 within a given period.
2223 * New! function: gras_socket_server_range to get a server socket in a
2224 range of port numbers (ease to avoid port number conflicts) [MQ]
2225 * New! gras processes display their backtrace when they get a SIGUSR1
2226 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2227 Sweet to debug RL processes [MQ]
2231 - Do not force experiment sizes to be expressed in kb, or it becomes
2232 impossible to measure the latency this way (needs one byte-long tests)
2233 WARNING: this changes the amok_bw_* function semantic. [MQ]
2234 - Implements the link saturation stuff. [MQ]
2235 * Peer management module:
2236 New! module factorizing code that we wrote over and over [MQ].
2239 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2240 - Test units are placed directly into the library code, they get extracted
2241 automatically and placed into the src/testall binary.
2242 - Convert most of the XBT tests to this system.
2243 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2244 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2245 live in the same C code. This required to change a little bit the API
2246 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2247 * New module: sparse graph structure with basic algorithms (this is work
2248 in progress and the API is not considered to be frozen yet). [AL]
2249 * Display more information on backtraces: source line & function names are
2250 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2251 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2252 * Reworked a little bit some #include statements to load only required
2253 headers. Some user code that relied on SimGrid to include stdlib or
2254 stdio may need to include it by themselves. [AL]
2255 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2257 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2258 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2259 but I guess that almost nobody used those parts. [MQ]
2261 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2263 SimGrid (3.0.1) stable; urgency=low
2266 * Unfortunately, I had missed 5 misnamed functions:
2267 xbt_fifo_item_t xbt_fifo_newitem(void);
2268 void xbt_fifo_freeitem(xbt_fifo_item_t);
2269 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2270 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2271 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2272 They're now deprecated. Please use their new versions:
2273 xbt_fifo_item_t xbt_fifo_new_item(void);
2274 void xbt_fifo_free_item(xbt_fifo_item_t);
2275 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2276 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2277 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2279 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2280 * Documentation: xbt_log module unmercifully reworked [MQ]
2281 * Bugfix: there was a problem with the ending of contexts with
2282 the pthread backend. It caused some weird deadlock or behavior
2283 depending on the pthread implementation. [AL]
2284 * Bugfix: get the exceptions raised in the simulator repport where
2285 and why they come from when they are not catched in time [AL, MQ]
2288 * Bugfix: Do repport the error when two non-connected hosts try to
2289 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2292 * Add additionnal checkings on communications. Assert that two
2293 communicating hosts are connected by a set of links... [AL]
2296 * Add additionnal checkings on channel values in communication [AL]
2297 * New: MSG_task_get_source to see on which host a task was generated [HC]
2298 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2299 the number of tasks waiting to be received on channel and sent
2301 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2302 waits for the first task coming from a given host.. [AL]
2304 GRAS new functionnalities: [MQ]
2305 * Enhance the parsing macro to allow the size of multidimentional objects
2306 to be given thru annotations.
2307 * New example (and documentation): Matrix Multiplication a la RPC
2308 (as when I was young!) and fix a bunch of bugs found on the way.
2310 GRAS performance improvements: [MQ]
2312 * Reduce the amount of cbps creation/destruction by making it static to
2313 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2316 * Change libdata to a set so that we can search for stuff by ID (and thus
2317 reduce the insane amount of dict lookups)
2320 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2321 RL communication on top of SG since it's so uneffective.
2322 It may also be used for inter-thread communication in RL, one day.
2323 * Use gras_datadesc_copy() to exchange messages on top of SG
2325 - improve message exchange performance on top of SG
2326 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2327 * Don't exchange on the network the size of the used part of buffer,
2328 instead, specify the possible buffer size to read().
2330 - reduces the amount of read/write calls (one pair per exchange)
2331 - reduces the amount of exchanged data (the size)
2332 - allows to retrieve all arrived data on receiver side, if we don't need
2333 it right now (subsequent read will peek the buffer)
2334 - allows the receiver to proceed with the begining of the stream before
2335 everything is arrived
2336 - make it possible to build an iov transport (using readv/writev)
2338 - take care of the data with non-stable storage (like stacked data),
2340 * If possible, TCP send uses vector I/O (when writev() is here)
2341 - Don't use it for receive since we send data sizes and data on the
2342 same stream, so we wouldn't be able to chain large amount of chunks
2343 before having to flush the stuff to read the size.
2344 * Rework the transport plugin mechanism to simplify it and reduce the
2345 amount of pointer dereferencement when searching for the right function
2348 * I guess that now, we do almost as few system calls as possible while
2349 doing as few data copy as possible.
2351 To improve it further, we could try to send all the sizes first and then
2352 all the data (to use iov on receiving size), but it's only a partial
2353 solution: when you have 2 dimensional data, the sizes of the second
2354 dimension is data of the first dimension, so you need 3 streams.
2356 I'm not sure the potential performance gains justify the coding burden.
2358 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2360 SimGrid (3.00) stable; urgency=high
2363 * New! Give the possibility to hijack the surf parser and thus bypass
2364 MSG_create_environment and MSG_launch_application. Have a look at
2365 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2367 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2369 SimGrid (2.96) unstable; urgency=low
2374 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2375 This deprecates the xbt_error_t mechanisms.
2376 It modifies (simplifies) all XBT and GRAS API.
2377 MSG API keeps unchanged (exceptions raised by XBT are catched from
2378 within MSG and masked with existing error handling facilities)
2381 * New! Add a FATPIPE model. [AL]
2382 * New! Add a parallel task model. [AL]
2383 * New! Add automatically a loopback interface (in the default
2384 network model) if none was precised.
2387 * Bugfix: MSG_process_resume now works with the current running process.
2389 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2390 * Modification of MSG_task_get_compute_duration. Once a task has been
2391 processed, the value returned by this function is now equal to 0. [AL]
2392 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2393 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2394 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2395 of a m_task. For now, MSG_task_cancel only works with computation
2397 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2398 of the processor (in Mflop/s) regardless of the current load on the
2400 * API Change: use proper naming convention for MSG_getClock and
2401 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2403 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2404 This function changes the priority of a computation task. This priority
2405 doesn't affect the transfer rate. A priority of 2 will make a task
2406 receive two times more cpu power than the other ones. This function
2407 has been added to suit the needs of Nguyen The Loc and hasn't been that
2408 much tested yet. So if it fails, please report it and send me your code.
2410 * API Change: removed all functions and types that were marked "deprecated"
2411 since many months. Renamed MSG_global_init_args to MSG_global_init.
2413 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2415 SimGrid (2.95) unstable; urgency=low
2418 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2419 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2420 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2423 * Bugfix: Allow absolute paths for platform description files [MQ]
2424 * Bugfix: do free the variables after use. Leads to drastic performance
2426 * Implement max_duration (ie, timeouts) on resources [AL]
2429 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2431 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2432 the needed support (timeouts) [AL]
2435 * Implement measurement sockets. You can now get the bandwidth between two
2436 hosts thanks to AMOK (see below). [MQ]
2437 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2438 dynar over the network (yeah) [MQ]
2439 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2440 * simple/static token-ring example. [Alexandre Colucci and MQ]
2441 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2442 in place before [MQ]
2445 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2446 type parsing. "short" and "long" modifiers were also ignored; other
2447 modifier (such as reference level) are still ignored. [MQ]
2448 * Embeed the buffer size within the buffer itself on SG. [MQ]
2449 That way, send() are atomic and cannot get intermixed anymore (at least
2450 the ones which are less than 100k; bigger messages still have the issue)
2451 * Array size pushed by the field, not by the field type (or each
2452 and every long int will push stuff to the cbps) [MQ]
2453 * use select() to sleep since it allows to portably sleep less than one
2456 GRAS (minor cleanups)
2457 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2460 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2461 * type category 'ignored' killed as it was never used and were difficult
2463 * whether a type can cycle or not is now a flag, leaving room for more
2464 flags (as "ignored", if we feel the need one day ;)
2465 * Rename raw sockets to measurement sockets since "raw" has another
2466 meaning in networking community.
2469 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2470 and offers features not belonging to GRAS but that most applications
2471 need. One day, it may be a set of plugins loadable at runtime.
2472 * New module: bandwidth
2473 bandwidth measurement between arbitrary nodes running this module. [MQ]
2475 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2477 SimGrid (2.94) unstable; urgency=low
2479 The first beta release of SimGrid 3 !
2483 * Update the main page and the FAQ. Adding references to gforge.
2486 * Add a gras_os_getpid function.
2489 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2490 * Extend the logs so that they also print PID, hostname, date, ... if
2492 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2493 and kill the old version which were in testsuite/
2494 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2498 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2499 for users. Instead, each programming environment (GRAS, MSG,...) use it
2500 when needed to provide such a feature to users.
2501 Don't shortcut the mechanism or you will also shortcut the virtualization
2502 you need on the simulator.
2506 * Cleanups in configury with regard to compile optimization/warning flags.
2507 Also add -fno-loop-optimize to any powerpc since it's the optimization
2508 killing gcc (< 3.4.0).
2509 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2510 needed by MSG examples complications
2511 * Borrow configury beautifications from PHP
2514 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2515 warning (thanks loris for stumbling into this one).
2516 * Bugfix: stop loading private headers (gras_config.h) from the public
2520 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2521 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2522 RL would imply a DNS resolver, which is *hard* to do in a portable way
2523 (and therefore delayed).
2524 * Implement a real timer mechanism and use it in timing macros. This allows
2525 to avoid rounding errors and get a 0.000005 sec precision in timing
2526 macros. While I was at it, various cleanups:
2527 - allow to declare more than one timed section per file (fix a stupid bug)
2528 - move some private declaration to the right place
2529 - merge conditional execution and timing macros into emulation module
2530 - document the module
2531 - make sure the module cleanups its mess on gras_exit
2532 * Documentation improvements:
2533 - (new) how to compile applications using GRAS
2534 - (new) emulation support (timing macros)
2536 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2538 SimGrid (2.93) unstable; urgency=low
2540 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2543 - Use Paje properly where used. Still to be sanitized properly.
2544 - Portability fix: Add an implementation of the contexts using pthread
2548 - Add xbt_procname(): returns the name of the current process.
2549 Use it to show the current process's name in all logging.
2551 - fix detection of older flex version and the reaction, since we do
2552 depend on modern ones (we use lex_destroy)
2553 - Better separation of SG and RL in the libs: remove all simulation code
2554 from libgras. As a result, this lib is now only 200k when stripped.
2555 Some of the xbt modules may also be duplicated (two sets and such) and
2556 should be cleaned/killed before SG3.
2557 - Insist on using xlC on AIX because of weird problems involving gcc there.
2558 - Cleanup the make remote stuff. This is now done by scripts
2559 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2560 mainly for our private use, but we're working on changing them to user
2563 - Bugfix: flush the socket on close only if there is some *output*.
2564 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2566 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2569 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2571 SimGrid (2.92) unstable; urgency=low
2573 Alpha 3 on the path to SimGrid 3
2577 - New! First try of benchmarking macros.
2578 - New! First try so that gras_stub_generator generate deployment and
2579 remote compilation helpers.
2581 - Bugfix: Initialization fix in msg_test.
2585 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2587 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2589 - API Change: gras_os_sleep to take the amount of seconds as a double.
2590 Accepting an int was error prone since it was the only location where
2591 seconds were coded as such. It leaded to damn rounding errors.
2592 - Bugfix: Hard to belive that timers ever worked before this.
2594 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2596 SimGrid (2.91) unstable; urgency=low
2598 Alpha 2 on the path to SimGrid 3
2602 - Bug fix in the lmm_solver.
2604 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2605 through the function MSG_paje_output.
2606 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2607 - It is possible to bound the rate of a communication in MSG with
2608 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2609 to write it in the changelog).
2610 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2615 - Various cleanups to the autotools stuff
2616 - Begin to move Gras examples to examples/gras/
2617 - Let make distcheck work again (yeah!)
2619 - documentation overhauled using doxygen.
2620 gtk-doc-tools is dead in SimGrid now.
2621 - Automatically extract all existing logging categories, and add the list
2622 to the documentation (long standing one, to say the less)
2624 - Cleanup the known architecture table. Reorder the entries to group what
2625 should be, and use a more consistent naming scheme.
2626 (some of the test dataset are still to be regenerated)
2627 - New! Allow library to register globals on each process just as userdata
2629 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2630 lookup time (for now).
2631 Use it in msg and trp.
2632 This cleans a lot the internals and helps enforcing privacy of the
2633 headers between the gras components.
2634 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2635 - Bugfix: gras_os_time was delirious in RL.
2636 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2638 - Reenable GRAS now that it works.
2640 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2642 SimGrid (2.90) unstable; urgency=low
2644 Alpha 1 on the path to SimGrid 3
2646 * It is a long time since the last release of SimGrid. I'm sorry about
2647 that but as I had told you, I was rewriting a lot of things. I apologize
2648 to those who had been reporting bugs to me and that I had not answered.
2649 If your bug is still in the new version, please tell me. Here is a
2650 summary of the main changes.
2652 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2653 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2654 portability, speed, and a lot more but you'll figure it out later.
2655 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2656 I wanted to release the new MSG as soon as possible and I have
2657 broken GRAS, which is the reason why, for now, only MSG is fully
2658 functional. A laconic description of these projects is available
2659 in the documentation.
2661 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2662 kernel optimized for our needs (called SURF but only the developers
2663 should use it). Hence, MSG is now roughly 30 times faster and I think
2664 that by rewriting a little bit MSG, I could event speed it up a little
2665 bit more. Beside the gain in speed, it is also much easier to encode a
2666 new platform model with SURF than it was with SG. More to come...
2668 * REVOLUTION 3: I have tried to change a little as possible the API of
2669 MSG but a few things really had to disappear. The main differences
2670 with the previous version are :
2671 1) no more m_links_t and the corresponding functions. Platforms are
2672 directly read from a XML description and cannot be hard-coded
2673 anymore. The same format is used for application deployment
2674 description. The new format is described in the documentation.
2675 Have a look in tools/platform_generation. There is a tiny script
2676 that converts from the old platform format to the new one. Concerning
2677 the application deployment format, parsing the old one is tricky.
2678 I think most of you should however be able to convert your files. If
2679 it is really an issue, I can write a C code that does the conversion.
2681 2) the toolbox tbx does not exist anymore. We now have a library
2682 with much more data-structures but without the hash-tables (we have
2683 dictionaries that are much faster).
2685 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2687 *****************************************************************************
2688 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2689 * don't feel like converting the oldies. (Mt) *
2690 *****************************************************************************
2693 Version 2.90: "the long awaited one"
2694 - Finished rewriting and debugging MSG. Rewrote the documentation.
2695 - disable GRAS for now since it needs to be ported to the newest SG
2698 - Finish the port to windows (using mingw32 for cross-compile)
2701 - Main loop and datastructures of SURF. A cpu resource object is
2702 functional. Surf can thus be used to create cpu's with variable
2703 performance on which you can execute some actions.
2705 2004-11-15 Martin Quinson
2706 - Port to ARM. Simply added the alignment and size descriptions. Should
2707 work, but the ARM machines are so slow that I didn't had the opportunity
2708 to 'make check' over there yet.
2710 2004-11-15 Arnaud Legrand
2711 - Trace manager now written. It uses a heap structure and is therefore
2712 expected to be efficient. It may however be speeded up (particularly
2713 when many events occur at the same date) by using red and black
2714 trees. One day maybe...
2715 - Max-min linear system solver written. It uses a sparse matrix
2716 structure taking advantage of its expected use. Most operations are
2717 O(1) and free/calloc are called as few as possible. The computation of
2718 the minimum could however be improved by using a red and black tree
2721 2004-11-03 Arnaud Legrand
2722 - Rename every gras_* function that was in xbt/ to its xbt_
2724 - Add a heap and a doubly-linked list to xbt
2725 - Added a dichotomy to the dictionaries. make check works as well before
2726 so I assume that the patch is correct. I do not know however if things
2727 run effectively faster than before now. :)
2729 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2730 SimGrid, and the version number is bumped to 2.x
2732 2004-10-29 Martin Quinson
2733 - Introduction of the remote errors.
2734 They are the result of a RMI/RPC on the remote machine.
2735 ErrCodes being scalar values, you can't get the host on which those
2736 errors did happen. Extending the error mechanism as in Gnome is possible.
2737 No idea yet whether it is a good idea.
2739 2004-10-28 Martin Quinson
2740 - Interface revolution: the Starred Structure Eradication.
2741 I used to do typedef struct {} toto_t; and then handle *toto_t.
2742 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2743 - ???_t is a valid type (builded with typedef)
2744 - s_toto_t is a structure (access to fields with .)
2745 - s_toto is a structure needing 'struct' keyword to be used
2746 - e_toto_t is an enum
2747 - toto_t is an 'object' (struct*)
2749 typedef struct s_toto {} s_toto_t, *toto_t;
2750 typedef enum {} e_toto_t;
2751 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2752 s_toto_t) is private.
2754 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2755 so that it can set the variable to NULL. It was so for dicts and sets,
2756 it changed for dynars.
2758 - Fix a bunch of memleaks in dict_remove
2759 - Fix a bug in sg/server_socket opening: it failed all the time.
2761 2004-10-07 Martin Quinson
2762 - Speed up dynar lookup operation a bit.
2764 gras_dynar_get is dead.
2766 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2767 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2768 (which gives you the address of the stored data).
2770 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2771 copy of the data using an affectation to do the job and not a memcpy.
2773 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2774 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2776 It does not really speedup the dynar test because they are
2777 setting elements all the time (and look them seldom). But the dict does
2778 far more lookup than setting.
2780 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2782 2004-10-05 Martin Quinson
2783 - Allow to (en/dis)able the cycle detection at run time.
2785 Whether we should check for cycle or not is now a property of each
2786 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2787 datadesc_cycle_unset allow to remove this property when previously set.
2789 Note that the cycle detection is off by default since it impacts the
2790 performance. Watch the data you feed GRAS with ;)
2792 This property is hereditary. Any element embedded in a structure having it
2793 set have it set for the time of this data exchange.
2795 You should set it both on sender and receiver side. If you don't set it on
2796 sender side, it will enter an endless loop. If you forget on receiver
2797 side, the cycles won't be recreated after communication.
2799 - Header reorganization.
2800 Kill gras_private.h, each submodule must load the headers it needs.
2802 2004-10-04 Martin Quinson
2803 - Interface revolution: do not try to survive to malloc failure.
2805 Now, gras_malloc and friends call gras_abort() on failure.
2806 As a conclusion, malloc_error is not a valid error anymore, and all
2807 functions for which it was the only gras_error_t return value are
2808 changed. They now return void, or there result directly.
2809 This simplify the API a lot.
2811 2004-09-29 Martin Quinson
2812 - Re-enable raw sockets.
2813 Created by gras_socket_{client,server}_ext;
2814 Used with gras_raw_{send,recv}
2817 It should allow to kill the last bits of gras first version soon.
2819 This is not completely satisfactory yet (duplicate code with
2820 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2823 - Simplify transport plugin (internal) interface by not passing any
2824 argument to _server and _client, but embedding them in the socket
2827 2004-09-28 Martin Quinson
2828 - Finish the port to AIX.
2829 autoconf was my problem (segfault within the malloc replacement
2830 function. No idea why)
2832 2004-09-16 Martin Quinson
2833 - Fix some size_t madness on 64bit architectures.
2835 2004-09-08 Martin Quinson
2836 - Reduce the number of system headers loaded, overload some more system
2837 calls (such as malloc to cast the result of the system one, and work
2839 - Fix and reintroduce the config support
2841 2004-09-07 Martin Quinson
2842 - Source code reorganization to allow Arnaud to surf all over there.
2843 - Allow to document the logging categories.
2844 - Remove all uppercase from logging categories and useless cleanup in names.
2846 2004-08-18 Martin Quinson
2847 Version 0.6.2 (protocol not changed; API changed)
2848 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2849 gras_error_t), and NULL when not found. Functions expecting a msgtype
2850 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2851 hopefully usefull message.
2852 - Portability to prehistoric sparcs again
2854 2004-08-17 Martin Quinson
2855 Version 0.6.1 (protocol not changed; ABI not changed)
2856 - prealloc some buffers to speed things up
2858 2004-08-11 Martin Quinson
2859 Version 0.6 (protocol not changed; ABI expended)
2860 - The parsing macro can deal with the references, provided that you add
2861 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2863 2004-08-09 Martin Quinson
2864 Version 0.5 (protocol not changed; ABI changed)
2865 - Allow to off turn the cycle detection code in data exchange at
2866 compilation time. It should be at run time, but I'm short of time (and
2867 the config stuff is still broken). That way, we keep dict out of the
2868 critical path, which is good because the performance is poor:
2869 - search not dichotomial yet
2870 - dynar give no way to access their content and memcpy everytime
2871 - In composed data description (struct, ref and so on), stop foolness of
2872 keeping the subtype's ID, but store the type itself. This keeps sets out
2873 of the critical path, which is good since they rely on dynar and
2874 dictionnaries. The only loose of that is that we cannot detect the
2875 redeclaration of a structure/union with another content (but I'm not sure
2876 the code detected well this error before anyway). We still can detect
2877 the redefinition discrepancy for the other types.
2878 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2879 it breaks the code because of type-punning used all over the place).
2880 This breaks on all non-gcc architectures (for now).
2882 All those changes (plus the buffer of last time) allow me to gain 2 order
2883 of magnitude on cruel tests consisting of 800000 array of integers on two
2884 level of a hierarchical structure (200 secondes -> 4 secondes)
2887 - the selector of reference must now return the type it points to, not
2888 the ID of this type.
2890 2004-08-06 Martin Quinson
2891 Version 0.4 (protocol changed; ABI not changed)
2892 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2893 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2894 - Implement a buffer transport, and use it by default (it relies on tcp in
2895 real life and on sg in simulation).
2896 That's a bit hackish since I had a new field to the structure to store
2897 its data without interfering with the subtype ones. Inheritance
2898 is tricky in C. And that's a kind of reverse inheritance with one class
2899 derivating two classes. Or maybe a game with java interfaces. Anyway,
2900 that's damn hard in C (at least).
2901 Moreover, I got tired while trying to ensure plugin separation and
2902 genericity in SG mode. MSG wants me to do weird things, so let's go for
2903 cruel hacks (temporarily of course ;).
2904 See comment in transport_private.h:71
2905 - do not include all the _interface headers in private but in the files
2906 which really need them (to cut the compilation time when they are
2909 2004-07-26 Martin Quinson
2910 Version 0.3 (protocol not changed; ABI changed)
2911 - Major overhault of the datadesc interface to simplify it:
2912 - shorted the function names:
2913 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2914 - add a trivial way to push/pop integers into the cbps without malloc.
2915 This allows to make really generic sub_type description, which simply
2916 pop their size of the stack.
2917 - add a function gras_datadesc_ref_pop_arr() which does what users want
2918 most of the time: Declare a dynamic array (which pops its size of the
2919 stack) and declare a reference to it. Poor name, but anyway.
2920 - kill the post-send callback, add a post-receive one
2922 2004-07-23 Martin Quinson
2923 Version 0.2 (protocol changed; ABI changed)
2924 - add some testing for cpbs in the test cases, and fix some more bugs.
2925 This invalidate again the little64 data file, since I cannot regenerate
2927 - remove an awfull optimization in the logging stuff, allowing me to:
2928 - understand it again
2929 - learn gcc how to check that the argument match the provided format
2930 - fix all errors revealed by gcc after that
2931 - internal keys of dict are not \0 terminated. Deal with it properly in
2932 loggings instead of segfaulting when the user want to see the logs :-/
2934 2004-07-22 Martin Quinson
2935 - Fix some stupid bug preventing cbps (callback postit) from working
2937 2004-07-21 Martin Quinson
2938 - Some documentation cleanups
2939 - remove the useless last argument of msgtype_declare
2940 - rename the Virtu functions to fit into the 'os' namespace
2941 - move headers src/include -> src/include/gras/ and stop fooling with
2942 gras -> . symbolic link
2943 - make distcheck is now successful
2945 2004-07-19 Martin Quinson
2947 - Build shared library also
2948 - Install html doc to the right location
2949 - stop removing maintainer files in make clean
2950 - build tests only on make check
2952 2004-07-13 Martin Quinson
2954 - No major issue in previous version => change versioning schema
2955 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2956 corresponding dataset.
2958 2004-07-11 Martin Quinson
2960 - Get it working with any kind of structure (we can compute the padding
2961 bytes remotely for all the architectures I have access to)
2962 - Implement the structure parsing macro (still not quite robust/complete)
2963 - Improvement to the remote testing toysuite
2965 2004-07-10 Martin Quinson
2966 [autoconf mechanism]
2967 - get ride of a bunch of deprecated macros
2968 - actually run the test for two-compliment, not only compile it :-/
2969 - test whether the structures get packed (and bail out if yes. Damn.
2970 Alignment is a serious matter)
2971 - test whether the structures get compacted (but respecting the alignment
2972 constraints of each types)
2973 - test whether the array fields of structures can straddle alignment boundaries
2975 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2976 (took me 2 hours to find that bug, looking at the wrong place)
2977 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2978 before sending/receiving and are used to compute the offsets of fields
2979 - Given that padding size depend even on compiler options, keep track of
2980 alignment and aligned_size only for the current architecture. Not a big
2981 deal since we send structure fields one after the other (seems
2983 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2984 program, let it work on linux/gcc/little32. portability todo.
2986 2004-07-08 Martin Quinson
2987 - import and improve remote compilation support from FAST
2988 - make sure make check works on half a dozen of machines out there
2990 2004-07-07 Martin Quinson
2991 Let's say it's version 0.0.3 ;)
2992 - Implement conversions (yuhu!)
2993 - Let it work on solaris (beside conversion, of course)
2994 - Stupid me, using rand() to generate the conversion datatests in not wise.
2996 2004-07-06 Martin Quinson
2997 - Let make dist work, since I'm gonna need it to compile on remote hosts
2998 - Let Tests/datadesc_usage write the architecture on which the file was
2999 generated as first byte.
3000 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3001 architecture descriptions.
3002 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3003 on those architectures.
3004 - Optimization: send/recv array of scalar in one shoot
3006 2004-07-05 Martin Quinson
3007 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3009 - Plug a whole bunch of memleaks
3010 - each process now have to call gras_{init,exit}. One day, their log
3011 settings will be separated
3012 - Continue the code factorisation between SG, RL and common in Transport.
3014 2004-07-04 Martin Quinson
3016 - Redistribution between SG and RL.
3017 We wanna have to accept in SG, so move accepted related parts of RL in
3018 the common part. (more precisely, the dynar of all known sockets is no
3019 more a static in transport.c, but part of the process_data)
3021 [gras_stub_generator]
3022 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3024 2004-07-03 Martin Quinson
3025 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3026 them all up in one shot)
3028 - Fix the ugly case of reference to dynamic array.
3029 - New (semi-public) function gras_datadesc_size to allow the messaging
3030 layer to malloc the needed space for the buffer.
3032 - gras_socket_close now expect the socket to close (and not its address to
3033 put NULL in it after it). This is because the socket passed to handlers
3034 is one of their argument (=> not writable).
3036 - propagate the interface cleanup from last week in datadesc, ie remove a
3037 superfluous level of indirection. User pass adress of variable
3038 containing data (both when sending and receiving), and not of a variable
3039 being a pointer to the data. Let's say that I like it better ;)
3040 The price for that is constructs like "int msg=*(int*)payload" in
3041 handlers, but it's a fine price, IMHO.
3043 - Let it work in RL (yuhu)
3045 2004-06-21 Martin Quinson
3047 - porting SG plugin and SG select to new standards (works almost).
3048 - plug memleaks and fix bugs around.
3051 - cleanup the prototype of data recv and force users to specify when they
3052 want to handle references to objects. Test case working even for cycles.
3053 - plug memleaks. Valgrind is perfectly ok with this.
3055 2004-06-12 Martin Quinson
3057 - cleanup the separation between plugin and main code in plugin creation
3059 2004-06-11 Martin Quinson
3061 - Reput hook for raw sockets, needed for BW experiments
3062 - kill a few lines of dead code
3063 [Data description] Interface cleanup
3064 - gras_datadesc_by_name returns the searched type or NULL.
3065 That way, no variable is needed to use a type desc once, which makes
3067 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3068 two parameters were strings (field name, type name), leading to
3070 [Dicos] Interface cleanup
3071 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3072 This is consistant with the dynar API.
3074 2004-04-21 Martin Quinson
3076 - Porting to new standards.
3078 - interface cleanup.
3079 There is no bag anymore, no need to take extra provision to mask the
3080 pointers behind "ID".
3081 Better splitup of functions between files create/exchange/convert.
3082 This is still a bit artificial since convert and receive are so
3083 interleaved, but anyway.
3085 - add a queued message list to procdata (the ones not matching criteria
3087 - factorize some more code between SG and RL wrt procdata
3089 - use gras_exit in example to track memleaks
3090 - get rid of gs_example now that GS is properly integrated into gras
3091 - update run_test to integrate the latest tests (datadesc)
3093 - rename WARNINGn macros to WARNn since it prooved error-prone
3095 2004-04-19 Martin Quinson
3097 - register init/exit functions within gras module mechanism
3098 - send/receive function.
3099 Convertion is not implemented, but short-cutted if not needed.
3100 struct/array elements are sent one by one (instead of block-wise), but
3101 nobody really cares (yet). Get a prototype before optimizing.
3102 - tests (using a file socket) for DD send/receive on:
3103 - base types: int, float
3104 - array: fixed size, string (ie ref to dynamic string)
3105 - structure: homogeneous, heterogeneous
3106 - chained list, graph with cycle
3107 Believe it or not, valgrind is not too unhappy with the results. The
3108 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3112 - Bugfix when using a filename explicitely (instead of '-')
3114 2004-04-09 Martin Quinson
3116 - factorize more code between RL and SG in socket creation
3117 - Complete the implementation and tests of:
3119 o file (only in RL, and mainly for debugging)
3121 I lost 3 days to design a portable address resolver, and then decided
3122 that the prototype mainly have to run on my box.
3123 Addressing portability too early may be like optimizing too early :-/
3125 - use gras_init in the Tests instead of the crappy parse_log_opt
3126 (the latter function is removed)
3127 [Conditional execution]
3128 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3129 [Code reorganisation]
3130 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3131 Build examples against the RL library, since there is no way to disable
3132 its creation for now.
3134 For information, the beginning of coding on GRAS was back in june
3135 2003. I guess that every line has been rewritten at least twice since