1 SimGrid (3.14) UNRELEASED; urgency=low
4 * Call-location tracing for SMPI.
5 You can add the exact location (filename / linenumber) of an MPI call to
6 your trace files and slow down or speed up the simulation between two
7 consecutive calls by using an adjustment file. (See the documentation.)
10 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
12 SimGrid (3.13) stable; urgency=low
14 The Half Release, a.k.a. the Zealous Easter Trim.
16 - We removed half of the lines, that were mostly experimental cruft.
17 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
18 (not counting blanks and comments -- according to openhub.net)
19 - The internals are now compiled in C++ (and will soon be clean C++)
20 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
21 59 klines of C, etc. We added only 29 klines of C++ in replacement.
23 * Backwards Compatibility breaks
24 - Removed Lua simulation bindings (switch to C or Java for that).
25 Lua can still be used to describe platforms
26 - Removed Java kernel plug-ins.
27 Will be reintroduced after the ongoing major internals reorg.
29 - the following functions were removed.
30 They were too specific and should be reimplemented in a generic
31 way, with filter function.
32 - MSG_task_listen_from_host
33 - MSG_mailbox_get_count_host_waiting_tasks
34 - MSG_mailbox_put_with_timeout was removed.
35 Please use MSG_task_send_with_timeout instead.
37 - the SD_application_reinit function was removed. It has been a noop for a while.
38 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
39 replaced by a more flexible mechanism.
40 - The following functions thus do not exist anymore
41 - SD_workstation_get_access_mode
42 - SD_workstation_set_access_mode
43 - SD_workstation_get_current_task
44 - Basic estimation functions have been removed but can easily be replaced
45 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
46 - SD_workstation_get_computation_time => amount / sg_host_speed()
48 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
49 Use VM.getSpeed()*load/100 for the legacy behavior.
51 - option enable_tracing was removed. It was not doing anything for a while.
52 - In the ModelChecker:
53 - the model-checker now ptraces the model-checked process which means
54 you cannot use a debugger on the latter anymore (we might make this
55 optional in the feature);
56 - removed soft-dirty page tracking;
57 - remove model-checked side snapshot management,
58 MC_snapshot() and MC_compare_snapshot();
59 - keep the MC_cut() function as a stub (it was not really working
60 in the previous release).
63 * All options are consistently in kebab-case. Old names are kept as alias.
66 * Switch to platform v4 format.
67 - Rename from 'power' to 'speed' the attributes describing the amount of
68 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
69 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
70 - In <host> and <link>, attributes availability and state are gone.
71 It was redundent with state and availability traces, and with peak values.
72 - In <cluster>, attributes availability_file and state_file are gone.
73 It was too complex and unused.
74 - Kill <gpu>. Was not doing anything.
75 - The DOCTYPE points to the right URL:
76 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
77 (the file at this address now documents the changelog since its v1)
78 - A warning is emitted for unit-less values (they are still accepted).
79 - speed. Default: 'f' or 'flops'. Also defined:
80 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
81 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
82 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
83 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
84 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
85 - latency. Default: 's' second. Also defined:
86 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
88 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
89 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
92 * s4u::Host is now the preferred public interface to the Host features.
93 sg_host_* functions are C bindings to the exact same behavior
94 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
97 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
98 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
99 They were deprecated since a few years, and probably did not even compile anymore.
102 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
103 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
104 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
105 encouraged to update their codes. A list of the modified functions can be found at the end of
106 include/simgrid/simdag.h
109 * simgrid::simix::kernel() is the closure callback. It ensures that
110 the lambda or closure passed as a parameter will run in kernel mode.
111 All the callback functions should be rewritten to that interface at some point.
114 * Reorganizing and cleaning the internals all around the place.
117 * Remove old default barrier/bcast buggy algorithms (see #18407)
118 * Various bug fixes to handle more codes
119 * Remove the need for the --foreground option of smpirun (it is still
120 accepted for backward compatibility).
123 * Kill the setset data container: MC don't use it anymore.
124 * Kill the queue data container: it made more sense with GRAS.
125 * Kill the xbt_peer_t data type: it's useless without GRAS.
126 * Kill rm_cb feature of config sets: it was never useful.
127 * Kill graphxml parsing feature. It was not used.
128 * Kill the deprecated code protected by XBT_USE_DEPRECATED
130 - xbt_dynar_sort_strings(), when the content is char*
131 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
132 They throw exceptions on invalid input;
133 * C++ support for declaring CLI flags (simgrid::config::Flag);
134 * class for abstracting different signal backends (simgrid::xbt::signal).
135 with no external dependencies (we need very simple signals).
138 * refactoring and cleanup of the code;
139 * ongoing process to cleanly separate the model-checking algorithms
140 from the code model-checking support.
142 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
144 SimGrid (3.12) stable; urgency=low
146 The Facelift Release.
149 * Require g++ v4.7 at least to not speak prehistorical C++.
150 * Require Boost 1.48 (for signal2 component).
151 * Java must be version 7 at least when activated.
152 * Builds on Windows again (including Java bindings).
153 * Tracing is now always enabled (no way to turn it out)
154 * Remove GTNetS. It was not working anyway.
155 * Various cleanups in the cmake scripts.
156 * Move headers around to sort them out on installed systems:
157 - instr/instr.h -> simgrid/instr.h
158 - instr/jedule/* -> simgrid/jedule
159 - simdag/datatypes.h was removed
160 - simdag/simdag.h -> simgrid/simdag.h
161 - msg/datatypes.h was removed
162 - msg/msg.h -> simgrid/msg.h
165 * Interface improvement:
166 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
167 - Sanitize the interface in MSG_task_ module:
168 - Merge two functions that were close enough but misleading:
169 set_compute_duration(t) -> set_flops_amount(t)
170 get_remaining_computation(t) -> get_flops_amount(t)
171 - set_data_size(t) -> set_bytes_amount(t)
172 get_data_size(t) -> get_bytes_amount(t)
173 - Massive cleanups in the functions related to the energy
174 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
175 - New: MSG_host_get_pstate()
177 - msg/energy/onoff: switching hosts on and off
179 * Interface improvement:
180 - Rename Host.isAvail() to Host.isOn()
181 - Rename Process.currentProcess() to Process.getCurrentProcess()
182 - Rename Task.setDataSize() to Task.setBytesAmount()
183 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
185 - #18874: Actually allows the GC to reclaim tasks
189 - SIMIX_process_throw: raises an exception in a remote process
190 * Refactoring: Separate sub-modules
191 - libsmx: the public interface, as libc in a real system
192 - popping: the strange dance that converts a user request into a kernel handling
193 - smx_context_*: the virtualization mechanisms that embed the user code
194 - smx_*: the handling of each simcalls
196 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
197 * Rename smx_action_t into smx_synchro_t, making explicit that these
198 things are used to synchronize processes with their environment.
199 For example, a communication is a sort of synchronization involving
200 the communicating processes (that may block until the exchange) and
201 the platform. The same can be said from computations, etc.
203 - #18888: segfault when a process ends before its kill_time
207 - 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.
208 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
209 - MPI_Type_set_name, MPI_Type_get_name
210 - MPI_*_c2f and MPI_*_f2c functions
211 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
212 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
213 - Activate a lot of new tests from the mpich 3 testsuite
215 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
216 - InfiniBand network model added : Based on the works of Jerome Vienne
217 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
218 - When smpi/display_timing is set, also display global simulation time and application times
219 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
220 * Collective communications
221 - 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
222 - 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/).
223 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
225 - Replay now uses algorithms from wanted collective selector
226 - Replay can be used with SMP-aware algorithms
227 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
228 - 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.
230 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
231 - Correctly use loopback on fat-tree clusters
232 - Asynchronous small messages shouldn't trigger deadlocks anymore
233 * Energy/DVFS cleanup and improvement
234 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
235 - new: smpi_get_host_pstate()
238 - "Full" network optimization flag was broken since Surf++
239 - Better handling of precision flags in maxmin
240 - Fix bug causing sometimes "Impossible" errors
241 - Properly pass cluster properties to included hosts
242 * Improvement of the Energy plugin.
243 - Always update the consumption before returning that value
244 - New property: watt_off to denote the dissipation when the host is off
245 - New functions getWattMinAt and getWattMaxAt to retrieve the
246 dissipation of pstates that we are not currently at.
247 * Java: class NetworkLink renamed to Link
248 * New function: simcall_process_get_kill_time()
249 * Massive rename s/workstation/host/
250 - That's intrusive, but that's good for the project consistency. Sorry.
251 - Change config option "workstation/model" into "host/model"
254 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
255 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
256 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
257 - xbt_dirname and xbt_basename for non-POSIX systems
259 * The model checker now runs as a separate process.
260 * The model checker runs must now be launched with the new simgrid-mc program.
261 * Record/Replay: the MC can display a textual representation of a path in the
262 execution graph. It can then be replayed outside of the model checker.
264 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
266 SimGrid (3.11) stable; urgency=low
271 * Normalizing pointers addresses tool for better diff between logs
274 * Add cloud examples using new VMs
275 - examples/msg/cloud/two_tasks_vm.tesh
276 - examples/msg/cloud/simple_vm.tesh
277 - examples/java/cloud/cloud.tesh
278 - examples/java/cloud/migration/migration.tesh
279 * Add java surf examples:
280 - examples/java/surfPlugin/surf_plugin.tesh
281 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
282 - examples/java/surfCpuModel/surf_cpu_model.tesh
283 * Add SMPI+MSG example:
284 - examples/smpi/smpi_msg_masterslave/
289 - msg task destroy cancel test
290 - msg_host on/off test
291 * Move all tests in testsuite to teshsuite (adding tesh files)
292 * Restructure teshsuites
293 - one folder for each kind of test
294 * Restructure AddTests.cmake
296 - structure the order of tests (with sections)
299 * Add virtual machine
300 - creation of a VM on a PM
301 - migration of a VM from a PM to another PM
303 - MSG_process_join(msg_process_t process, double timeout)
304 - msg_bar_t MSG_barrier_init(unsigned int count)
305 - int MSG_barrier_wait(msg_bar_t barrier)
306 - void MSG_barrier_destroy(msg_bar_t barrier)
307 - msg_as_t MSG_environment_as_get_by_name(const char * name)
308 * New option "msg/debug_multiple_use" to help debugging when a task is used
311 - Improvements and finalization of MSG_storage, MSG_file APIs and their
313 - Increase code coverage in test suites
317 * Protect context stacks against stack overflow. The number of protected memory
318 pages allocated on the top of each stack (1 by default) can be configured
319 with the new command line option --cfg=contexts/guard_size:<value>.
320 * Simcalls are now generated by a python script that
321 - generates files included by SimGrid
322 - checks that all the functions exist, and proposes prototypes
324 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
325 host_set_data, host_get_data
327 - simcall_process_join(smx_process_t process, double timeout)
328 * Fix bug where sleeping processing could not be suspended.
331 * Translate surf models from C to C++
332 - Generic classes for all models: Model, Resource, Action
333 - A generic interface for each kind of model (CPU, Network, Storage
334 Workstation, WorkstationVM)
336 * Translate surf routings from C to C++
337 * Add callbacks using sigc++ or boost::signals2
338 - Add callback functions for resource creation/destruction
339 - Add callback functions for action state change
340 - Handle Energy as a plugin
341 * Replace swag by boost::intrusive
342 * Add new routing models for clusters. For documentation, see
343 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
344 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
345 parameters for cluster tag
346 - Fat trees, with topology="FAT_TREE" and
347 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
348 - see examples/platforms/torus_cluster.xml and
349 examples/platforms/fat_tree_cluster.xml
353 * Hostfiles support host:nb_processes construct to deploy several processes on
355 * Collective communication algorithms should not crash if used with
356 improper number of nodes and report the error.
357 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
358 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
359 MPI_Dims_create, MPI_Cart_sub are supported.
360 * New interface to use SMPI programmatically (still depends on MSG for
361 some parts, see examples/smpi/smpi_msg_masterslave) :
362 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
366 * Global variables privatization in MPI executables is now performed at runtime
367 with the option smpi/privatize_global_variables (default:no).
368 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
369 dynamic libraries loaded by the application are not privatized (static
370 linking with these libraries is advised in this case)
373 * Options defined in XML work correctly now.
376 * New cmake option, enable_lib_in_jar, to control whether native libraries are
377 copied into simgrid.jar or not (ON by default). Use this option if you want
378 to reduce the size of the installed simgrid.jar, *and* the native libraries
379 are kept installed elsewhere.
380 * Surf binding with SWIG (code generated in maintainer mode only):
381 - plugin to handle callbacks
382 - CPU model only for the moment
385 * Supernovae build mode is definitively removed. It was used to improve
386 inlining and inter-module optimizations. It is nowadays superseded by
387 link-time optimizations commonly available in compilers.
388 * Update ns-3 find lib. Bindings for ns-3 should work again now.
389 * Add boost dependency for surf++
390 * Add new macro for tests
391 - ADD_TESH(name <tesh_args>)
392 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
396 - xbt_bar_t XBT_barrier_init(unsigned int count)
397 - int XBT_barrier_wait(xbt_bar_t barrier)
398 - void XBT_barrier_destroy(xbt_bar_t barrier)
399 * Make the xbt_os_time module public
401 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
403 SimGrid (3.10) stable; urgency=low
405 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
408 * Reintegrate Java to the main archive as desynchronizing these
409 package is not acceptable anymore (Java is now considered stable)
410 * Add explicit synchronization facilities through semaphores
411 * Bug fix: Task.setDataSize() only changed the C world, not the value
412 cached in the Java world
415 * Dramatically change the way files are handled. API and internals changed, but
416 this part of MSG was not considered as production grade either.
417 * Add explicit synchronization facilities through semaphores
418 * Add a new function MSG_host_get_process_list()
419 * Preliminary DVFS support (see examples/msg/energy/ for details)
422 * SMPI is now included directly in the libsimgrid as the windows
423 linker doesn't force us on splitting it anymore.
424 * Improvements of the SMPI replay tool:
425 - Most of the collective communications are now rooted in the same process as
426 in the original application.
427 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
428 used until now). Multiple data types can now be used in a trace.
429 - The replay tool now supports traces produce either by TAU or a modified
431 - Bug Fix: the compute part of the reduce action is now taken into account.
432 - Gatherv collective is now supported
433 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
434 of smpirun outputs time independent traces for the current run. One file
435 is created per process. If too many processes are simulated, this behavior
436 can be changed to one file for all processes by using the
437 tracing/smpi/format/ti_one_file flag
438 * smpirun generates the host file if needed (with given host count and platform)
439 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
440 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
441 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
442 --cfg=smpi/coll_selector:(mpich/ompi)
443 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
445 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
446 happening inside a collective SMPI call.
447 * Fix the behavior of complex data types handling.
448 * Make MPI_Wtime another synchronization point to take computations into
450 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
451 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
452 * Add all missing Fortran bindings, SMPI should work with Fortran 90
453 (no privatization of global variables yet)
454 * Preliminary DVFS support (see examples/smpi/energy/ for details)
457 * Verification of liveness properties is now available for SMPI applications
458 (in addition to MSG applications)
459 * Bugged examples using SMPI in examples/smpi/mc/
460 * Add --cfg=model-check/visited option. Allows the verification of infinite
461 programs. Detection of loops in the execution thanks to the system state
462 comparison and reduction of the state space to explore. Can be combined with
463 DPOR for safety properties.
466 * Allow to change SimGrid configuration (see --help) within the code
467 thanks to SD_config() as it can be done in MSG.
468 * Add a new function SD_task_set_amount() upon user request.
471 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
472 * Remove rule based routing (no more PCRE dependency)
473 * Add a limiter_link option to cluster tag, to specify a maximum reachable
474 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
475 * Add a loopback_bw and loopback_lat options to cluster tag.
476 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
477 routing. Now peers are encapsulated in an AS and have their own private
478 router but this is transparent.
481 * Our own implementation of getline is renamed xbt_getline, and gets
482 used even if the OS provide a getline(). This should reduce the
483 configuration complexity by using the same code on all platforms.
484 * New type: xbt_cfg_elm_boolean.
485 * Allow to use yes/no for boolean configuration options in the command line.
486 * Allow to disable SimGrid cleanups at exit from command line option.
487 There are situations where one may want a simulation to end with an exit.
488 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
489 annoying when scripting around the simulator. Adding a
490 --cfg=clean_atexit:no allows to circumvent this issue.
493 * Lots of memory leaks were corrected in this release.
494 There is no known memory leaks anymore, in all of our 600+ tests.
495 * New command line option --version, to get SimGrid version information.
496 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
498 * Supernovae builds are deprecated, and expected to be removed in the next
501 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
503 SimGrid (3.9) stable; urgency=low
505 The Grasgory release: GRAS is really dead now.
507 * Complete overhaul of the internal host structures scheme.
510 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
511 was considered as experimental and badly maintained since then.
512 * Keeping it was thus a trap to our potential users, that could take
513 it instead of MSG or SMPI by mistake despite is pity state.
514 * GRAS seems to have very few users (if any), and no one volunteered
515 to maintain it further. It also induces a lot of XBT code (for
516 portability sake), that must be maintained too.
517 * For all these reasons, we killed GRAS. If someone wants to revive it
518 in the future, don't cry, our git history still remembers of GRAS.
521 * Major overhaul. Merge our documentation again as time proved that
522 splitting it was really not helping our users.
523 * Further improve the developer documentation to help newcomers
524 hacking on SimGrid itself. The user documentation (and in
525 particular, the beginner documentation) is still in a sorry state.
528 * Now works on Windows too!
529 * Much more extensive test suite, from MPICH
532 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
533 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
534 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
535 * Bug fix: let task be scheduled when the last dependency to be solved is
536 a control dependency.
537 * Remove SD_load_environment_script function.
538 Use the C sg_platf function if you want to declare a platform
542 * New function: MSG_process_get_number()
543 * Old function documented: MSG_config()
544 * Remove MSG_load_platform_script function
545 Use the C sg_platf function if you want to declare a platform
549 * Change the default value of the TCP_gamma constant (maximal size of TCP
550 congestion window) to a more realistic 4MiB value. If you notice changes in
551 your simulation results, you can fall back to the previous 20k tiny window
552 by adding --cfg=network/TCP_gamma:20000 on command line.
553 * (Hopefully) fix a bug wrt periodic availability and state traces
554 * Bug fix: use default values at start when first event in availability/state
555 trace is not at time 0.
558 * remove the "new_" part of function name sg_platf_new_trace_connect
559 (resulting in sg_platf_trace_connect), since it does not create
563 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
564 use case where it's really mandatory, and maintaining it was a pain in
566 * New: xbt_fifo_search(), search an item with a user-provided
567 comparison function instead of dumb pointer comparison.
570 * Fix the lua deployment:
571 Use `simgrid.init_application()` before deployment instead of
572 `simgrid.msg_register_application()` after.
575 * Transfer the tracing files into the corresponding modules.
577 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
579 SimGrid (3.8.1) stable; urgency=low
581 The "we are told that some people want to also *install* the simgrid
584 * Add missing file "tesh.1" to the archive.
586 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
588 SimGrid (3.8) stable; urgency=low
590 The Psssshiiiit release: SimGrid jumps into the Cloud.
593 * Add an experimental interface to manipulate VMs. They are mainly
594 process groups with very few intrinsic semantic, but they should
595 allow you to build the semantic you want easily.
596 * New function: MSG_host_set_property_value()
597 * New function: MSG_process_on_exit(). To clean memory in all cases.
598 * Bug fixes that made the host (and link) failures unusable.
599 * Add a way to auto-restart process when the host in which they are
600 executing comes back (ON_FAILURE="RESTART" on deployment file,
601 MSG_process_auto_restart_set).
602 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
603 please stop using the old ones, they are DEPRECATED.
605 * Deprecate functions MSG_global_init() / MSG_global_init_args()
606 Please use MSG_init() instead. (reducing the amount of entry
607 points in the library helps us).
608 * Make it impossible to link against the wrong version of the lib
609 * Deprecate MSG_clean(). No need to call it anymore.
610 * Function MSG_get_host_number() is not deprecated anymore.
613 * Split the doc into a user guide and a reference guide.
614 * Start a developper guide to help people hacking on SimGrid.
617 * Enable tracing by default. This modules rocks you should use it.
618 * Remove option custom_flags. Now use environment variables CFLAGS
620 * Use default cmake things to detect lua instead of home grown ones.
621 * New option "enable_mallocators" to disable mallocators, for debugging
622 purpose ("on" by default).
625 * Bug fixes around the resource failures: don't let the processes
626 survive the host they are running onto.
627 * Add an interface to auto-restart processes when the host in which they are
628 executing comes back.
629 * Ensures that SIMIX_clean is called automatically. It's not part of
630 the public interface anymore (bindings should be updated).
633 * Bug fix for when SD_Simulate is called with a positive value: be careful
634 when comparing doubles. Sometimes they are different for non significant
636 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
637 parallel task whose initial work is distributed among host according
638 to the Amdahl's law. Such tasks are created with a parameter alpha
639 that corresponds to the non-parallelizable part of the computation.
640 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
641 between two sets of workstations assuming a 1D block distribution (each
642 workstation owns a similar share of data) on both sides.
644 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
645 Data redistribution will be automatically scheduled once parent and child
646 are both scheduled. The filling of computation_amount and
647 communication_amount structures is now done seamlessly thanks to the chosen
649 * New function SD_workstation_dump to display various information
650 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
651 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
652 bandwidth on the route onto which the task is scheduled and the amount of
654 To divide the nominal bandwidth by 2, the rate then has to be :
655 rate = bandwidth/(2*amount)
656 * Compute tasks that have failed can now be rescheduled and executed again
657 (from their beginning)
658 * Increasing source code coverage (src/simdag is now covered at 95.8%
662 * Re-implement time-independent trace replay using SMPI (at the
663 smpi_smp_* level) instead of MSG. This should replace
664 examples/msg/actions/actions.c
665 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
666 hindexed and structs)
667 * Implement the exchange of non-contiguous data.
668 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
669 * Correct behavior of smpi/sender_gap and set its default value to 0
670 * Add option to asynchronously send small messages to allow better
671 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
672 specifies the size in bytes under which messages will be asynchronously sent.
673 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
674 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
677 * New C interface to define a platform: XML is now optional.
678 For more info, please check include/simgrid/platf.h.
679 * New interface to define random platforms from the C:
680 For more info, please check include/simgrid/platf_generator.h and
681 examples/msg/masterslave/masterslave_platfgen.c
682 * Export a sg_cmdline dynar containing all the arguments we got from
686 * Two new tracing options for adding comments to trace file so you
687 can track your experiments (see --help-tracing for details).
688 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
689 * Adding the SimGrid version that generated the trace file as a comment.
690 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
691 * Fix to avoid key clashes on Paje links
692 * Other minor fixes related to the Paje specification
695 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
696 and renamed to xbt_str_hash() and xbt_str_hash_ext().
697 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
698 * Greatly improve the robustness of mmalloc to user errors (such as
699 using an area after freeing it, or freeing it twice)
701 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
703 SimGrid-java (3.8.1) stable; urgency=low
705 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
706 Warning: all methods are visible, but only some of them are
707 implemented so far. Check the source (src/jtrace.c)
708 for further information.
709 * New module: org.simgrid.msg.File (SimGrid File management functions)
710 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
711 * Change the meaning of Process.restart: now restart the process from
712 the begining, like MSG_process_restart in C.
713 * Add Process.setAutoRestart: handling of process restart when failed
715 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
716 you to retrieve the properties of the processes/hosts
717 * Deprecate Msg.clean(): you can just forget about it now.
718 * New function Process.getCount(), that only works when compiling
719 with the not yet released version 3.9 of the C library.
721 * New context factory based on Coroutines. It mandates a modified JVM
722 but then, the simulations run about five times faster, and there is
723 no limit to the amount of processes (beside of the available memory).
725 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
727 SimGrid (3.7.1) stable; urgency=low
730 * Restore the prototype of MSG_process_create_with_environment() to
731 the pre-3.7 situation by removing the kill_time argument.
732 * Add a MSG_process_set_kill_time() function instead.
735 * Fix weird behaviors when dealing with parallel tasks.
738 * Simgrid is now built as a dll.
739 * Simgrid-java now works on Windows.
740 * Simgrid-Java is now included into Windows package.
743 * First pre-build package for MacOSX.
746 * Fix compilation when using MSG_USE_DEPRECATED.
747 * Fix some compilation issues on Macs and Windows.
748 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
749 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
751 We discovered that the Lua console is broken, but we are missing the
752 manpower to fix it right now. The problem existed in 3.7 too, so we
753 are not blocking the release for that. Sorry if you depended on this
754 feature, any help would be really welcome.
756 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
759 SimGrid-java (3.7.1) stable; urgency=low
761 The "Java aint got to be bloated and slow" release
764 * Various internal cleanups and performance improvement
765 Simulations are expected to run up to twice faster or so
766 * Make Process.kill(process) an instance method, not a static one
767 * User processes are not java.lang.Thread subclasses.
768 This breaks the compatibility (sorry), but previous API was
769 brain-dead, making it impossible to have non-trivial
770 initializations in the process constructor.
771 * Require a full constructor per Process sub-class.
772 Kinda breaks the compatibility (sorry), but this allows a much more
773 efficient way to launch the processes at simulation startup.
774 * Do not embeed our version of semaphores, java 1.5 can be considered
775 as sufficiently prevalent for us to not dupplicate its features.
779 * Add examples for almost every part of the API
780 We spotted and fixed a lot of bugs in the process
782 * New module: asynchronous communication API
783 * New function: Process.sleep()
784 It takes milliseconds as argument, just as java.lang.Thread.sleep()
785 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
786 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
788 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
790 SimGrid (3.7) stable; urgency=low
792 The "spring cleanups (before the next Big Project kicks in)" release.
795 * We can specify the SMPI latency/bandwidth factor with command line
796 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
797 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
798 You can also use the "config tag" from platform file by adding this line
799 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
800 (see "example/platforms/tag_config.xml" to use "config tag").
801 Note that the command line supersedes the platform file configuration.
802 * Change the correction factors used in LMM model, according to
803 the latest experiments described in INRIA RR-7821.
804 Accuracy should be improved this way.
805 * Use the partial invalidation optimization by default for the
806 network too. Should produce the exact same results, only faster.
807 * Major cleanup in surf to merge models and split some optimization
808 mechanisms from the core of the models. As a result you can now
809 specify which model to use (e.g., --cfg=network/model:LV08
810 --cfg=cpu/model:Cas01) and which optimization mode to use
811 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
812 Incompatible combinations should err at initialization. See
813 --help-models for the list of all models and optimization modes.
814 * The CLM03 workstation model was dropped for simplicity because it
815 used the deprecated CM02 network model. Use default instead.
816 * Rename the TCP_gamma configuration option to network/TCP_gamma
817 * Rename the coordinates configuration option to
818 network/coordinates, and document it
819 * Use now crosstraffic keyword instead of the terribly misleading
820 fullduplex keyword. It is activated by default now in the current
821 default model, use --cfg=network/crosstraffic:0 to turn it off.
822 * Ongoing refactoring the model parsing to make XML files optional
823 See include/simgrid/platf.h for details (still to be completed)
826 * Major overhaul of the documentation. Almost instructive now :/
827 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
828 functions and friends. This interface was considered as
829 deprecated since over 2 years, it's time to inform our users that it is.
830 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
831 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
832 use these (crappy) functions in your code.
833 These functions will be removed soon. Stop using them now.
834 * Deprecate MSG_get_host_{table,number}
835 Implement MSG_hosts_as_dynar() instead.
836 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
837 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
840 * Stabilize the parallel execution mode of user contexts
841 * Introduce configuration variables to control parallel execution:
842 - contexts/synchro: Synchronization mode to use when running
843 contexts in parallel (either futex, posix or busy_wait)
844 - contexts/parallel_threshold: Minimal number of user contexts
845 that must be part of a scheduling round to switch to parallel
846 execution mode (raw contexts only)
847 * Fix bugs that prevented to use suspend/resume along with
848 synchronization structures.
849 * Fix bugs in process termination that lead to invalid memory access
850 in very specific conditions.
853 * Introduce a parallel mode for the models (controlled by surf/nthreads
854 configuration item). In our tests, running the models in parallel
855 never lead to any speedups because they are so fast that the gain
856 of computing each model in parallel does not amortizes the
857 synchronization costs, even when ultra fast futexes are used.
858 This is released anyway because YMMV.
861 * Performance boost by using a swag internally to compute the set of
862 tasks that are finished and should constitute the return value of
866 * Enable it by default now that it is considered rather stable.
869 * Documentation of the tracing functions.
870 * Performance gains when tracing categorized/uncategorized resource
871 utilization by avoiding calls to get route when updating resource
872 variables. LMM constraints are being used instead.
873 * API changed to set task categories. Use MSG_task_set_category instead
874 of TRACE_msg_set_task_category, and SD_task_set_category instead
875 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
876 * Bugfix for graphicator, routes not correctly obtained, memory leaks
877 * Examples for link user variables added (see at examples/msg/tracing/)
878 * Deprecated function TRACE_msg_set_process_category completely removed
879 * Trace header updated according to the latest Paje file format
880 * Tracing network lazy updates, no longer obligate users to use full updates
881 * --cfg=tracing/platform:1 also registers power/bandwidth variables
882 * Experimental: let user code declare/set/push/pop application states for hosts
883 * API changed to allow the manual creation of graph configuration files
884 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
887 * Improve the API of Lua MSG bindings, using the Lua spirit.
888 * Each simulated process now lives in its own Lua world (globals are
889 automatically duplicated). It helps writing simulators. Will allow
890 to run Splay programs within SimGrid in the future.
891 * Add a Chord example in Lua, equivalent to the MSG one.
894 * Start the implementation of a solution to express temporal
895 properties, not only local assertions. This is still an
896 experimental work in progress, stay clear from it to be safe.
900 - Add new runtime parameters --help-logs and --help-log-categories
901 to display information about supported logging parameters and
903 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
904 don't exists anymore.
905 * Mallocators: allow value NULL for the reset function.
907 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
908 create homogeneous dictionaries, where all the elements share the
909 same free function. Non homogeneous dictionaries will be
910 deprecated in the next release.
911 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
912 - Multi-level dictionaries are deprecated.
914 - new function xbt_dynar_search_or_negative() that is useful when
915 you have less than 2 million elements in your dynar and don't
916 want of the extra complexity of catching exceptions when the
917 element is not found.
919 - Make xbt_os_thread module (for thread portability) public.
920 Documentation is still to come, sorry.
922 - Cleanups and simplifications to make it maintainable again.
923 - Exotic features (such as memalign and valloc) were removed.
924 - The metadata were extended and improved so that the
925 model-checker becomes able to explore and inspect the heaps.
926 - This may induce a performance drop when enable_model-checking is
927 ON in cmake (even if it's not used in the simulation), but it is
928 necessary at this point to get MC working.
930 Turn model-checking OFF if simulation performance matters to you.
931 Not enabling it at runtime is not enough, disable it in cmake.
933 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
936 SimGrid (3.6.2) stable; urgency=low
938 The "Not coding new stuff allows to polish old things" release.
941 * New bindings to the NS3 packet level simulator (experimental)
942 * Use the raw (efficient) execution contexts instead of the sysv
943 (portable) ones when possible.
944 * libpcre is now mandatory in any cases since not using it led to
945 severe performance loss and possibly other issues
946 * Update the XML platforms:
947 - G5K: include the latest machine in Nancy
948 - GridPP and LCG: new platforms
949 * Documentation was partially updated, at least (more to come)
951 Bug fixes, cosmetics and small improvements
952 * Free terminated processes before the end of the simulation to avoid
953 exhausting the memory of users having very dynamic amount of
955 * Bug fix and cosmetics about canceling non-running tasks
956 * Bug fix about the dot loader's issues when using libcgraph
959 * Create an installer for windows with nsis (amd64 and win32)
960 - Add an hello world project to illustrate simgrid project creation.
961 - Embed libpcre into the Simgrid installer to avoid
962 its compilation burden
963 * The raw execution contexts should work on Apple now
964 * Port to Windows 64 bits
965 - Sysv contexts now have an implementation for this arch
966 - GRAS communication features now support this arch
967 * Drop support for borland compiler on windows
968 - this code was not maintained, and we kinda depend on gcc nowadays
969 * Fix portability issues on kfreebsd/gnu: build error about semaphores
970 * Fix portability issue on unstable ubuntu: linker became picky on
973 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
975 SimGrid (3.6.1) stable; urgency=low
977 The "Oops, we broke Macs too" release
980 * Fixed contexts detection so that raw ones are used when possible
981 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
982 leads to a strange error, with user code segfaulting sometimes when
983 the generated code is not perfectly aligned (which is not
984 controllable from the user side, depends on the amount of code)
987 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
989 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
991 SimGrid-java (3.6) unstable; urgency=low
994 * Split of every thing from simgrid v3.5 into a separate package.
996 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
998 SimGrid (3.6) stable; urgency=medium
1000 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1003 * Bindings now constitute their own package, separated from the main one.
1004 Rationale: reduce our maintenance nightmare by reducing the module coupling
1005 They will soon be released on their own on gforge.
1007 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1008 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1010 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1011 * It's not quite deprecated for now because we have no replacement,
1012 but it may soon become the case.
1015 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1016 * Fortran: New user-level cache variable to store the rank of the running
1017 process. This improves performance by an order of magnitude.
1018 * C: New coccinelle script to automatically locate and modify global and
1019 local static variables.
1020 * Improved SMPI network model with a sender-side gap to account for multiple
1024 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1025 only say if a communication is finished, no matter whether it succeeded or
1026 failed. You can call MSG_comm_get_status() to know the status of a finished
1028 * New function MSG_task_dsend() to send a task and detach it. When a
1029 communication is detached, you are never notified of its success or failure
1030 and the memory is released automatically once it is finished. This function
1031 is useful when you don't care about the end nor the success of a
1033 * Change the prototypes of action replay. Sorry for inconvenience,
1034 but this is really more efficient this way (and to adapt your code,
1035 you just have to fix the initialization, that shouldn't be too long)
1036 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1037 ever managed to do anything useful with it.
1038 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1039 takes a dynar of communications. It returns immediately and gives the
1040 index of a finished communication (if any).
1041 * New example: a basic implementation of the Chord P2P algorithm.
1044 * New model for multi-core CPUs. You can now use the core attribute to
1045 precise the number of cores of a host. This is a basic model. Every
1046 process running on the host receives at most the power provided in
1047 the DTD (throughput<=power). Total throughput of process cannot exceed
1049 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1050 router linked by an up-link and a down-link (possibly asymmetrical).
1051 This kind of pattern allows to easily build last-mile model style platforms.
1052 Aggregating such patterns in a rule-based AS is thus the technique of
1053 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1054 * New model for Vivaldi routing. We transformed the Vivaldi network model
1055 into a Vivaldi routing model (based on the rule-based model). This allows to
1056 combine Vivaldi based latencies with last-mile platforms.
1059 * Added a check for NaN of IEEE754 infinite in the double entries of
1061 * Introduce a new context factory "raw", highly inspirated from the
1062 ucontext factory, but using manually crafted functions in assembly to
1063 do the work in an efficient manner.
1064 * Allow to change the used context factory at run time, not only at
1065 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1066 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1067 contexts at runtime (only with raw contexts or ucontexts).
1068 * Completely rewrote this module to allow parallel execution of user
1069 processes. Use --cfg=contexts/nthreads:N to execute user processes
1070 with N parallel threads (the default is 1, meaning no parallelism).
1071 * Allow to decide dynamically between sequential and parallel modes.
1072 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1073 processes in parallel only when their number is greater than or equal to P
1075 * Added a check for NaN of IEEE754 infinite in the double entries of
1079 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1080 won't output the state of processes when interrupted with Ctrl-C
1081 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1082 NULL-terminated array. This may solve backward compatibility issues
1083 due to the change to return type of SD_simulate. See also:
1084 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1085 * Add new macros with variable number of arguments.
1086 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1087 - in xbt/asserts.h: xbt_assert
1088 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1089 - in xbt/ex.h: THROWF and RETHROWF.
1090 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1092 * Change xbt_die() to accept a format string with arguments, just like printf.
1093 * New data structure: xbt_lib_t, like a dict but more general and with better
1097 * New configuration options
1098 Options triva/categorized and triva/uncategorized can be used to generate
1099 graph configuration files for Triva visualization tool.
1100 * Configuration option tracing/platform is renamed to tracing/categorized
1101 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1102 * New TRACE_user_link_variable interface:
1103 User provides the name of the link and the tracing variable to attach to it
1104 * the declaration of tracing categories must be done after the environment
1106 * simpler tracing interface, just one way to declare categories
1107 TRACE_category or TRACE_category_with_color, it is up to you
1108 * links in the trace file are again identified by their names
1109 * trace contains the full platform hierarchy exactly as declared using the ASes
1110 * Options tracing/msg/[task|process]:1 groups the process by hosts
1111 for both cases, tasks and processes must have names that are unique during the simulation
1112 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1113 * The experimental option tracing/msg/volume is deprecated
1114 its functionality may be reincorporated if needed
1116 The tracing generates a trace file with unordered timestamped events,
1117 because of the way the core simulator (surf) works. A script available
1118 at the tools directory (fix-paje-trace.sh) can be used to put the events
1119 in order. We have changed the tracing so it can generate ordered timestamped
1120 events in the final trace, but depending on the simulator (and how much time
1121 is simulated) that can lead to a huge memory utilization. It is deactivated
1122 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1124 Build Infrastructure
1125 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1126 - We are trying hard to keep the API stable, but it may happen that
1127 some things change (we're a research project after all, not a
1128 nuclear plant operating system). If such things should happen, you
1129 could rely on that macro to adapt.
1130 - current value: 30600 for 3.06.00, aka 3.6
1131 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1132 a number that can be compared with SIMGRID_VERSION.
1133 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1134 assertions and verbose/debug logging events are disabled at compile time.
1136 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1138 SimGrid (3.5) stable; urgency=medium
1141 * New feature to any SimGrid-based simulator: Model-Checking
1142 Check SIN#1 for more details.
1145 * New Model SMPI (three-interval linear regression for correction factors)
1146 See RR-7426, available at http://hal.inria.fr/inria-00527150
1147 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1148 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1149 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1150 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1151 * See also src/smpi/README
1156 - Tracing API changes: TRACE_start and TRACE_end should not be called
1157 by user-code. They are automatically called by simulators created
1158 with SimDAG, MSG and SMPI if the toolkit is compiled with
1159 tracing_enabled=ON. Categories declaration and utilization remain the
1160 same for MSG and SimDag.
1161 - A function was added to the tracing API to declare categories with
1163 - TRACE_category_with_color (char *category, char *color)
1164 where color must be in the following format
1165 "%f %f %f", red, green, blue
1166 and red, green, blue are float values in the interval [0, 1]
1167 - User can specify NULL as color parameter, or continue calling
1168 TRACE_category (cat)
1169 On that case, the tracing system will define random colors
1170 - The following command-line options are supported:
1171 --cfg=tracing/filename:msg.trace
1172 --cfg=tracing:1 (activate tracing, needed to use others)
1173 --cfg=tracing/platform:1 (categorized resource use)
1174 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1175 --cfg=tracing/msg/task:1 (task creation)
1176 --cfg=tracing/msg/process:1 (process creation, migration)
1177 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1178 --cfg=tracing/smpi:1 (SMPI interface tracing)
1179 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1180 - examples of examples/msg/tracing updated
1182 - DAXLoader and DOTLoader functions can generate tasks with categories
1183 - A new function to attribute a category to SD tasks:
1184 TRACE_sd_set_task_category (SD_task_t task, char *category)
1185 * Tracing the MPI interface implemented by SMPI
1186 - Collective operations are traced with states
1187 - Point-to-Point operations are traced with states/links
1188 - Tracing activated by a parameter "-trace filename" passed
1189 to smpirun during execution (considering that simgrid
1190 is compiled with tracing enabled)
1191 - To run the simulation with gdb, the simulator
1192 accepts --cfg=tracing/smpi:1 to trace SMPI
1193 - tesh files to check if smpi tracing is ok
1194 - See examples/smpi/NAS/DT-trace
1195 * GTNetS tracing re-worked
1196 - adaptation to the tracing system of GTNets to cope
1197 with modifications regarding the fullduplex mode
1198 - new tesh files to check if gtnets tracing is ok
1201 * Asynchronous communications through the functions:
1202 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1203 * New function: MSG_load_platform_script()
1204 to make possible using a lua script instead of XML files to set up platforms
1205 * New function: MSG_set_function
1206 to associate functions to processes, used when bypassing the parser
1207 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1209 Platforms: Add some more examples in examples/platforms
1210 * Grid'5000: see www.grid5000.fr
1211 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1214 * Change the XML format. This is a very important modification. SimGrid
1215 3.5 introduces a new hierarchical format based on the notion of
1216 Autonomous Systems. Compatibility with old format is ensured
1217 through the perl script provided in the install bin directory
1218 bin/simgrid_update_xml.
1219 It is now possible to build platforms with specific routing
1220 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1221 connect several platforms together. We will try to provide soon
1222 set of realistic platforms exploiting these properties (have a
1223 look at examples/platforms/ for the moment).
1224 * Take the opportunity of the XML format change to be a good XML citizen:
1225 rename link:ctn to link_ctn and similar changes (also dealt with by
1227 * Add a new routing scheme (rule-based) using regular expressions. It
1228 enables to have an extremely low memory footprint when the
1229 underlying routing is simple and can be compactly described. You
1230 need to have libpcre4-dev (perl regular expressions) installed if
1231 you want to use this routing scheme.
1232 * Revive the cluster TAG and allow to easily and efficiently (both in
1233 term of memory and speed) connect clusters together. Have a look
1234 at teshsuite/simdag/platforms/ to see how this can be done. With
1235 this tag, you can create clusters with thousands of tasks at no
1236 cost (have a look at examples/platforms/).
1237 Note: clusters are implemented as ASes, so there is no need for an
1238 enclosing AS tag if you have only one cluster in your platform.
1239 * Add new generic functions in the public interface that allows the user
1240 to call SURF 'create_resource' methods from your code (same
1241 functionality as the XML bypass mechanism but with a much lighter
1243 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1244 that uses a piecewise linear approximation to produce better
1245 results when exchanging small messages.
1246 * Add a new parameter to handle correctly full duplex link and account
1247 for interferences between uplink and downlink communications
1248 (activate with --cfg=fullduplex:1).
1251 * Rename the SD_READY (all dependencies are satisfied and task is
1252 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1253 dependencies are satisfied) state.
1254 This prevents a confusion between the notion of "ready to schedule"
1255 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1256 simulated" (SD_RUNNABLE) used by the simulation kernel.
1257 * Change the way a task is considered as ready. Instead of removing
1258 dependencies when a task is done, a counter is decreased. This way,
1259 it is always possible to reach ancestors thanks to the
1260 SD_taks_get_parents function (even after the end of the simulation.)
1261 * Change the return type of SD_Simulate from (SD_task_t*) into
1262 xbt_dynar_t. This function was in handling a dynar internally and
1263 converted it into a NULL terminated array for historical reasons.
1264 * New function SD_dotload(char*) to load a DAG described in dot
1265 format. This loader and the corresponding examples require the
1266 installation of the graphviz library.
1267 * Fix a bug in the management of tasks of size 0 in the surf network
1268 models. This problem was only visible with SIMDAG and you should
1269 thus disregard results produced with earlier versions if you
1270 relied on this feature (some tasks were blocked because of this).
1271 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1272 even though your applications were doing only point-to-point
1273 communications and sequential computations. Now you can really use any
1274 model you want (of course, if you create real parallel tasks, which are
1275 not implemented in most models beside ptaskL07, this will abort).
1276 * Add an example that schedules a DAX on an heterogeneous platform
1277 using a Min-Min strategy.
1278 * New function SD_workstation_get_current_task() that returns the kind
1279 of task currently running on a workstation in the sequential access
1281 * Raise some warnings when unexecuted tasks remains at the end of the
1282 simulation. This is usually caused by cycles in the DAG.
1285 * New function: SIMIX_process_set_function() called by MSG_set_function
1286 * Change the underlying waiting queue in semaphores so that a process
1287 can wait on several of them simultaneously (as in waitany).
1288 * Fix the way to handle tokens in semaphores so that all access patterns
1289 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1290 * kill the dirty pimple SIMIX_message_sizes_output()
1291 Please use (proper) visualization instead
1294 * New data container: setset (set of sets of elements)
1295 * New module: mmalloc (mapped malloc, allowing to have several
1296 independent segments of malloc)
1297 * New function: xbt_dict_cursor_set_data()
1298 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1299 * New function: xbt_dynar_is_empty()
1300 * New function: xbt_fifo_get_last_item()
1301 * Fix xbt_dynar_shrink(): use the right element size.
1302 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1303 uninitialized areas during expand.
1304 * Fix semaphores: previous implementation was severely broken.
1305 * Use library init/fini functions for our initialization.
1306 - you can use logs and other feature as soon as you want in your
1307 code (even before the xbt_init / MSG_init)
1308 - xbt_exit is now a no-op and produce a warning when used.
1311 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1312 benefit from the latest improvement to the simulation kernel.
1313 * Kill measurement sockets for now. If you rely on them, sorry. This
1314 release is not for you. This feature will be reintroduced in the
1315 future, but we cannot delay the release any further.
1316 * New function: gras_msgtype_get_name().
1317 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1318 * Fix (at last) the pmm example: it should not randomly fail anymore.
1320 Build chain: bug fixes and overall polishing
1321 * Cmake is now stable enough. Hence, we killed the autotools.
1322 * Port to windows ( TM :)
1323 * Fix the 'make install' target.
1324 No need to use 'make install-simgrid' anymore
1325 * Introduce a 'make dist' target compiling a *source* archive
1326 'make package' compiles a binary archive
1327 * Compile java files only on need
1328 * Add --cd and --setenv command line options to tesh
1329 * Out of source builds are not fully supported yet, but we are close
1330 * Enable supernovae and optimization flags by default for our users
1333 * Add layer to set up environment directly from lua, without XML.
1334 * The effect of gras_stub_generator can be achieved through
1335 lua too (check examples/gras/console/ping_generator.lua)
1337 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1339 SimGrid (3.4.1) stable; urgency=low
1341 The "Polishing easter eggs is probably a good idea" release.
1342 This is a bug fixes release only.
1346 * Fix a bug preventing the tasks from begin garbage collected.
1349 * Fix a bug occuring when a host involved in a communication fails.
1350 This was not detected properly by the other peer involved in the
1351 communication. Now, it's reported as a network error.
1354 * Warn the user about loop dependencies in data flow of DAX files
1355 * Obey the control-flow dependencies of DAX files
1358 * Add option "enable_smpi" allowing to not compile SMPI.
1359 Probably useful for the (Mac) users experiencing a build error here
1360 * Improve the detection of lua5.1 and ruby1.8
1362 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1364 SimGrid (3.4) stable; urgency=low
1366 The "Easter in Cargese" release. Also known as (major changes):
1368 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1369 ~> bindings were greatly improved
1370 ~> new tracing infrastructure for better visualization introduced
1372 * the "Welcome to configury modernity" release.
1373 ~> we switched from autotools to cmake, and improved our cdash
1376 A more detailled list of changes follow (full detail in svn log).
1378 Java Bindings: Various Cleanups
1379 * (install java-gcj-compat-dev on debian-like to use them)
1380 * Remove put/get: no need to export deprecated interface in Java
1381 Use send/receive instead.
1382 * Cleanup the examples and add a README per directory
1383 * Remove example autoDestination (that's the only way to go now)
1384 * Remove example explicitDestination (was a plain copy of basic)
1385 * Make JniException a runtime exception, so that there is no need to
1386 declare the fact that you may encounter such a beast. I guess that
1387 nobody will ever want to survive such error.
1388 * Create specific errors for each MSG case of failure:
1389 host failure, transfer failure, timeout, task cancelled
1390 * Cleanup the exceptions that may get thrown by each function
1391 * Other internal cleanups in Java bindings. Performance still bad :/
1392 Ruby and Lua Bindings: create them
1393 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1394 * That's new and great, you should try them out.
1395 Same functionalities than Java bindings, only even less polished
1397 * Kill the useless "rate" argument of SD_task_get_execution_time()
1398 Everyone used to provide -1 as a value, it was not used, and the
1399 semantic of a possible use wasn't even clear.
1400 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1401 either as comm costs or compute costs to mean that there is no
1402 such thing for that specific task.
1403 * Add a SD_task_set_name() function
1404 * Fix SD_task_unschedule() on typed tasks
1405 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1406 * In DAX loader, accept useless 'level' attributes to <job> since
1407 LIGO DAGs have them (seem to be to ease graphical representation).
1409 * Add an example masterslave_mailbox.c using send/receive and not
1410 the deprecated put/get interface.
1411 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1412 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1414 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1415 (declare MSG_USE_DEPRECATED to still have the old name)
1416 * Add a MSG_task_set_data() function
1417 * About trace replay (see examples/msg/actions):
1419 - Allow to work with splitted trace files for each process
1420 Give the specific trace file as argument of each process,
1421 and call MSG_action_trace_run(NULL)
1422 You can still have one merged file for all processes.
1423 - Fix implementation of collective operations
1424 * Allow task_execute() on 0-sized tasks (closes #10063)
1426 * This is the first release of SimGrid where SMPI is not considered
1427 beta anymore (even if some corners should still be improved)
1428 * Port over the new SIMIX_network submodule (internal refactoring)
1429 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1430 * Implement more missing elements of the standard:
1432 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1434 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1435 MPI_Scan() MPI_Get_processor_name()
1436 - Added implementation of missing case for Alltoall (warning: it's
1437 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1438 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1439 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1440 SMPI_MPI_Allgatherv()
1441 * Bug fixes include:
1442 - MPI_Waitsome() was broken
1443 - Allow relative includes in smpicc
1444 - Command line cfg argument 'reference_speed' was ignored...
1445 - Some functions did not properly lead to auto-benching of user code
1446 - smpicc passes -O2 by default (just like openmpi one)
1448 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1449 * Bug fixes about timeouts during communications
1450 * add SIMIX_message_sizes_output() as a pimple to write to file the
1451 amount of messages per size. Use gnuplot to get histogram.
1452 Pimple because that's the only user-visible function of simix,
1453 defined directly in xbt.h (irk, sorry)
1455 - Add a SIMIX_sem_get_capacity() function
1456 - Fix interactions with processe resume/suspende
1457 - release_forever() was stupidly broken
1458 - Fix SIMIX_display_process_status() for processes in a semaphore
1459 - Make SIMIX_sem_block_onto() user-visible
1460 * Refactoring context stuff:
1461 - Use pseudo-OOP for better modularity
1462 - reimplement SIMIX_process_kill() without process_schedule() so
1463 that the latter can take as invariant that it is called from
1465 - Merge context_start into context_new for sake of simplicity
1467 * Add a Vivaldi network model, coded live during SUD'10 ;)
1468 * Rename configuration variables to start a hierarchy:
1469 o cpu_model -> cpu/model
1470 o network_model -> network/model
1471 o workstation_model -> workstation/model
1472 * New configuration variables:
1473 o network/bandwidth_factor: correction to bandwith
1474 o network/latency_factor: correction to latency
1475 o netwotk/weight_S: correction to the weight of competing streams
1476 * Add a long description to the models, that users can see with such
1477 argument on the command line: --cfg=cpu/model:help
1478 * --help-models display the long description of all known models
1480 * config: add the ability to set a default value after registration
1481 Does not override any previously set value (e.g. from cmd line)
1482 * dict: allow to have integer key and data.
1483 When so, you need to use the following functions
1484 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1485 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1486 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1487 In contrary to regular dicts, the key is not malloced before copy.
1488 Mixing scalar and regular elements in the same dict is not tested
1490 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1491 Tracing for Visualization:
1492 * SimGrid is now instrumented in order to generate a trace file for
1493 visualization analysis: to use it, need to compile SimGrid with the
1494 "tracing" option enabled, and instrument the program using SimGrid with
1495 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1496 (among other functions).
1497 * The instrumentation only traces the platform utilization for now
1498 * Documentation to use the tracing functions and how to analyze the
1499 traces with the Triva tool is written.
1500 * More information about: SimGrid FAQ (in the section Tracing Simulations
1503 * We moved to cmake as default build system. Autotools support will
1504 be dropped soon. Check the FAQ for more info about how to use it.
1505 * Greatly improved our cdash/ctest interactions
1506 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1507 * Added memory checking tests with valgrind; lot of memleak fixing.
1508 This may be the first release of simgrid with so few memory issues
1509 * Added code coverage tests.
1510 Our coverage is still improvable, but at least we see it on cdash.
1512 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1514 SimGrid (3.3.4) stable; urgency=low
1516 The "Desktop Grid needs love too" release (also called Xmas release).
1518 Models improvements:
1519 * Major speedup in the maxmin system solving by using lazy evaluation
1520 Instead of solving completely the maxmin system at each iteration,
1521 only invalidate (and recompute) the modified parts.
1522 This new feature is enabled in default models but you can try to
1523 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1524 * Cas01 IMproved as default CPU model
1525 This CPU model is the same Cas01 model, but it uses the
1526 maxmin-selective-update flag and a heap structure to manage
1527 actions on SURF kernel.
1528 It reduces the complexity to find the next action to finish and,
1529 consequently, it's faster than the old Cas01.
1530 This is the new default CPU model (Cas01).
1531 * Rename the old Cas01 model to Cas01_fullupdate
1532 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1533 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1534 * CpuTI (CPU Trace Integration)
1535 A new CPU model whose objective is simulate faster when using
1536 availability trace files.
1537 Instead of using a full featured, over engineered maxmin system for
1538 CPU modeling, this model does the pre-integration of traces files
1539 to calculate the amount of CPU power available, and so, executes
1540 faster than the old CPU models.
1541 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1542 * Use LV08 as default network model since it gives better accuracy
1543 for small messages and shouldn't change things for big ones.
1544 Use --cfg=network_model:CM02 to get the previous behavior.
1547 ******************************************
1548 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1549 ******************************************
1550 * The new CPU model may changes simulations!
1551 The point is that events occurring at the exact same timestamp
1552 are not scheduled in the same order with the old and new
1553 version. This may be enough to completely change the execution
1554 of simulations in some cases.
1555 * The new network model will change simulations!
1556 This new model is more realistic than the previous one, so you
1557 should consider redoing your old experiments with this model.
1558 Sorry for the inconvenience.
1561 * Introduce the supernovae compilation mode
1562 When compiled that way, the whole SimGrid (or almost) is put in a
1563 single compilation unit and compiled in one shoot.
1564 This is to help gcc which has difficulties to inline stuff from one
1566 The speedup seem to be above 15%, althrough more tests are needed on
1567 amd64 to confirm that gain.
1570 * Port of MSG's mailbox on top of SIMIX network
1571 The put/get mechanism was greatly simplified on the way.
1574 * New SIMIX network module. Provides:
1575 - Mailbox: rendez-vous mechanism to find with who you want to speak
1576 - Synchronous send/recv: easier and hopefully faster since the
1577 logic is handled in the maestro process directly now
1578 - Asynchronous send/recv: you dreamt of it? It's here now
1579 Too bad that nobody cared enough to propagate the change to MSG.
1580 * Add semaphores as SIMIX synchronization mechanism.
1583 * new function SD_daxload(char*) to load a DAX file
1584 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1585 * Introduce typed tasks. Specify its kind and cost at creation.
1586 At scheduling, just give where it should be placed, and the cost
1587 for each involved resource is automatically computed.
1588 Existing constructors so far (more to come of course):
1589 - SD_task_create_comm_e2e() for end-to-end communication
1590 - SD_task_create_comp_seq() for sequential computation
1591 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1592 * new function SD_task_dump() for debuging display
1593 * new function SD_task_dotty(task,FILE*) writing to file the info
1594 about the task in dotty format
1595 * SD_task_dependency_exists() can now cope with having one of its
1596 arguments NULL. If so, it tests whether the other argument has any
1598 * Add getters on list of preceding/following tasks:
1599 SD_task_get_parents(task) and SD_task_get_children(task)
1600 * Add getters on amount of workstations and list:
1601 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1602 * Add getter on task kind: SD_task_get_kind(task)
1603 * Update the start_time and finish_time of tasks on completion/failure
1604 * Bugfix: Remove task from state swags when destroyed
1607 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1610 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1611 dynar with one separate thread per value of the dynar.
1612 * Change the prototype of xbt_thread_create(), sorry.
1613 Added a boolean parameter indicating whether we want to join this
1614 thread (used in SG only for now)
1615 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1618 * GTNetS wrappers should now be usable again (and betterly tested too)
1619 * Fix a major regression from 3.2 where the timeout provided to
1620 MSG_task_put_with_timeout() was used as absolute time before which
1621 the comm should be done.
1622 * Start to fix the <cluster> tag.
1623 - Internal links should be good now (beside of the loopback, which
1624 use the private link instead)
1625 - paths to the external world is still rather broken
1626 - the <route:multi> tag is just broken. Actually that's brain-dead.
1627 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1628 to make it less stupid
1629 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1630 * Fix a source-level compatibility glitch from 3.2: after defining
1631 MSG_USE_DEPRECATED, you can use the old name
1632 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1633 * Allow to compile from the SVN with automake 1.11
1634 * Fix some problems when using the "start_time" tag in deployment XMLs.
1635 * Fix #8569: XBT/synchro.h has redundant declarations
1636 * Fix #8563: MSG return values and exceptions
1637 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1638 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1639 Thanks to Robson Peixoto.
1640 * Fix a vicious bug in dictionaries inducing that some elements were
1641 not freed on xbt_dict_free()
1643 Portability report of this version:
1644 * Main portability targets:
1645 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1646 on (amd64/i386/ia64)
1647 - mac leopard on i386
1648 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1649 but nothing critical.
1650 * Other platforms: windows, AIX and others were not tested for this release
1652 Timing report of this version:
1653 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1654 scenario parameters). From 8h to a few seconds in desktop grid settings.
1655 * Supernovae brings about 25% speedup on i386.
1657 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1659 SimGrid (3.3.3) stable; urgency=low
1661 The "Need for Speed" release.
1663 The timings done to validate the 3.3.2 were faulty.
1664 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1666 The problem was a conversion from a manually handled vector to
1667 xbt_dynar_t on the critical path.
1668 xbt_dynar_foreach calls functions, inducing stack management crap.
1670 We inlined these functions and xbt_dynar_foreach is now breath taking.
1671 We also inlined xbt_swag_belong on the way.
1673 Here are some approximate speedup measurements (on master/slaves
1674 simulations lasting between 10s and 20s each):
1675 3.3.1 -> 3.3.2: about same performance
1676 3.3.2 -> 3.3.3: 40% speedup
1677 3.3.1 -> 3.3.3: 40% speedup
1678 3.3.1 with inline patch -> 3.3.3: 30% speedup
1680 Our reading is that the refactoring which occurred in 3.3.2 made us
1681 suffer much more from the xbt_dynar_foreach low performance, but
1682 once we solved this, this refactoring proved to be very performance
1683 effective. From the 40% speedup, somehow, 10% are due to the
1684 inlining and 30% to the refactoring.
1686 That's a pitty that gcc cannot inline functions placed in other files
1687 alone. We have to choose between:
1688 - break the encapsulation (by putting private data structures and
1689 accessors in headers files to help gcc)
1690 - live with low performance
1691 - switch to a decent compiler such as icc (not quite possible).
1693 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1695 SimGrid (3.3.2) stable; urgency=low
1697 The "Simplicity does not preceed complexity, but follows it" release.
1699 The main contributors of this release were (lexical order):
1700 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1703 * Extract the routing logic into its own object.
1704 (was dupplicated in network.c and workstation_LV07.c;
1705 Allows to implement other ways of storing that info)
1706 => kill now useless network_card concept
1707 - Use dynar to represent routes (instead of void** + int*)
1708 - kill link_set (use surf_network_model->resource_set instead)
1709 - Add a command-line option to choose the routing schema to use
1710 - Add three new models:
1711 * Floyd (shortest path computed at initialization)
1712 * Dijikstra (shortest path recomputed all the time)
1713 * Cached Dijikstra (shortest path computed on need)
1714 All these models where contributed by Silas De Munck, and are
1715 described in his ICCS09 paper.
1717 * Simplify model declaration
1718 (less redirections, less function to write when defining a model)
1719 - Factorize stuff between models:
1722 surf_model_resource_set(model)
1723 surf_model_resource_by_name(model, name)
1724 - Unify the types of models in s_surf_model_t (using an union)
1725 - Embeed fields of common_public directly into s_surf_model_t
1726 - Rename model methods:
1727 action_free ~> action_unref
1728 action_change_state ~> action_state_set
1729 action_get_state ~> action_state_get
1730 - Change model methods into functions :
1731 (model)->common_public->action_use ~> surf_action_ref
1733 * Implement a generic resource; use it as ancestor to specific ones
1734 (allows to kill duplicated code in models)
1735 Drawback: timer command don't need no name nor properties;
1736 workstation_CLM03 don't need no properties
1737 (but I guess we can live with those few bytes wasted)
1739 * Improve the action object model
1740 - implement a constructor avoiding dupplicated code about field
1741 initialization in generic_action part.
1743 * Kill the SDP model: it has an external dependency, is deprecated
1744 in flavor of modern lmm models, and didn't compile since a while
1747 * Relocation of the context module from XBT to SIMIX.
1748 (the context were decoupled from the simix processes, duplicating a lot of code)
1749 => a lot of code was factorized
1750 - less overhead is introduced during scheduling
1751 - simpler API for the context factory
1752 - the logic for process creation,destruction and manipulation was simplified
1753 * Simplification of the s_smx_process_t data structure.
1754 => accesing the simix level data associated to a process is faster now,
1755 and the code is a lot more readable.
1758 * Implement some more MPI primitives:
1759 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1760 -implementation: Bcast: flat or 2-ary tree (default),
1761 Barrier: 4-ary tree,
1763 Allreduce: Reduce then Bcast
1764 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1765 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1766 Alltoallv: flat tree, like ompi
1768 * Add support for optimized collectives (Bcast is now binomial by default)
1769 * Port smpirun and smpicc to OS X
1772 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1773 Shout out if you used it.
1776 * Display the list of still queued messages in SG mode when existing
1780 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1781 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1782 * Add xbt_str_from_file(FILE*)
1783 * Add xbt_dict_get_key achieving a linear reverse search
1784 * Remove the context module
1786 Portability report of this version:
1787 * Main portability targets:
1788 - Linux(debian)/x86/context
1789 - Linux(debian)/x86/pthreads
1790 - Linux(debian)/amd64/context
1791 - Linux(debian)/amd64/pthreads
1792 On these, we still have the eratic breakages of gras/pmm and
1793 amok/saturate_sg reported in previous version. We still think
1794 that the tests are the cause of the fault, not the tested code.
1796 - Mac OSX Leopard/x86/context
1797 Still false negative in tesh autotesting.
1798 Smpi still fails, but this time because readlink does not accept -f
1799 Everything seems to work properly beside of that.
1802 - AIX version 5.3 (only tested contexts this time)
1803 Smpi still fails there because mktemp is not installed.
1804 Everything seems to work properly beside of that.
1806 I managed to compile it for the first time, but several breakages.
1807 Won't delay the release for this exotic platform.
1809 * Windows: it's still lagging behind. If you want to help, please
1812 Timing report of this version:
1813 This version seem to be more than 5% faster than 3.3.1 (on linux
1814 64bits with contextes). The gain is less than expected, we are
1815 investigating this for next release.
1817 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1819 SimGrid (3.3.1) stable; urgency=low
1822 * Implement a --cfg-help to show existing configuration variables
1823 * Build chain do not require doxygen in maintainer mode
1826 * fix a bug on struct sizeof computation, which prevented the
1827 exchange of arrays of structs in some conditions
1828 - added a regression test about this in datadesc_usage
1829 * Allow the exchange of 0-long dynamic vectors.
1830 - for that, use -1 as indicator of dynamic size instead of 0
1831 - This implied to change any size from unsigned long to long,
1832 reducing a bit communication abilities, but I guess that with
1833 64bits being quite common, this is more than enough.
1834 - This also induce a protocol change, thus bumping network protocol
1835 version from 0 to 1 (if we have external users, we have to get
1836 clean on that point too ;)
1837 - added two regression tests about this in datadesc_usage
1838 * Be more verbose when propagating local exceptions
1839 This helps debugging.
1840 * Display the status of simulated processes when receiving SIGINT in
1844 * Allow to control the simulation from a trace file.
1845 New functions MSG_action_register() and MSG_action_trace_run()
1846 The first one allows to associate a function execution to each
1847 kind of action while the second one parses a trace file and
1848 triggers the corresponding actions within the system.
1849 For now, only a toy example is provided in examples/msg/actions
1850 * Add an exemple of process migration in examples/msg/migration
1851 * Fix a bug in task exchange which broke MSG_task_get_sender()
1852 Add a teshsuite regression test for that.
1853 [Bug: if MSG_task_get_sender() is called after sender exit,
1855 * Fix a bug which prevented suspend/resume to work properly
1856 * Display the status of simulated processes when receiving SIGINT
1857 This fixes a regression of v3.3. due to the introduction of SIMIX
1858 * Bug fixing in failure management:
1859 - trace could not start by a failure at time 0
1860 - failure during communications were not working
1863 * Add SIMIX_process_set_name() to change the name of the current
1864 process in the log messages.
1865 * Store smx_hosts in a dict since we only retrieve them by name
1866 * Move the configuration infrastructure to surf
1869 * Move the configuration infrastructure to surf
1872 * Massive internal cleanups:
1873 - Store internal structures on processes instead of hosts (allows
1874 to have more than one process per host, in addition of being more
1876 - Cleanup the initialization/finalization process
1877 - Kill a whole bunch of unneeded synchronization:
1878 processes run in exclusive manner within the simulator
1879 - Move queues from global tables to process data fields
1881 - now accept -platform and -hostfile arguments
1882 - Pass the right rank value to processes according to the hostfile
1883 * Compile the examples by default, and use them as regression tests
1884 * Implement MPI_Wtime()
1885 * Change the reference speed to a command line option
1888 * TCP_gamma can now be specified as command line option using
1889 --cfg=TCP_gamma:10000000.0
1890 * Change the --surf-path cmd line option into --cfg=path:
1893 * Also include strbuff from xbt.h public header
1894 * xbt_ex_display(): do not free the exception after displaying
1895 This allows to do more with the given exception afterward.
1896 Users should call xbt_ex_free() themselves.
1900 Portability report of this version:
1901 * Main portability targets:
1902 - Linux(debian)/x86/context
1903 - Linux(debian)/x86/pthreads
1904 - Linux(debian)/amd64/context
1905 - Linux(debian)/amd64/pthreads
1906 These targets fail about 1/10 of times on gras/pmm, but we believe
1907 that this is because of the test, not because of simgrid.
1908 amok/saturate_sg fails even more rarely, and the test may not be
1911 - Mac OSX Leopard/x86/context
1912 The test suite still spits tons of errors because some obscure
1913 force prevents us from removing the temporary directories
1914 arguing that they still contain some metadata I've never heard of.
1915 Smpi fails because seq is not installed.
1916 Everything seems to work properly beside of that.
1919 - AIX version 5.3 (both contexts and pthread)
1920 Smpi still fails there because mktemp is not installed.
1921 XML inclusions seems rosty on AIX.
1923 * Windows: it's still lagging behind. If you want to help, please
1926 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1928 SimGrid (3.3) stable; urgency=high
1932 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1933 [Malek Cherier & Mt]
1935 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1938 Until now, GRAS were using MSG as an interface to SURF. It was
1939 quite difficult because both interface have several differences
1940 (MSG channels vs GRAS sockets were the most notable point).
1942 This also opens the gate to SMPI (which should occur soon) and speed
1943 up simulations by to 40% (even if it were not the main goal).
1945 **************************************
1946 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1947 **************************************
1948 The point is that events occuring at the exact same timestamp are
1949 not scheduled in the same order with the old and new version. This
1950 may be enough to completely change the execution of simulations in
1951 some cases. Sorry for the inconvenience.
1953 * Cleanup and upgrade the XML format to push further scalability
1954 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1956 * Improve the testing infrastructure with tesh. Now a very large part of
1957 the code is tested not only by being run but also by checking that the
1958 output match an expected output [Mt].
1960 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1961 is really less memory-demanding, which should allow you to use
1962 larger files in SimGrid [AL].
1964 * Inform valgrind about our contextes, so that it becomes usable
1965 with the default (and more effecient) version of SimGrid
1966 [contributed by Sékou Diakite, many thanks]
1969 * Introduce a listener thread in charge of receiving incoming
1970 messages from the network. It allows to overlap communication and
1971 computation but most notably, it removes some stupid deadlocks due
1972 to the fact that so far, a process could not send and receive at
1973 the same time. This made most non trivial communication schema
1975 * Convert the PIDs from long int to int to match the MSG ones (and
1976 linux ones too) [Mt]
1977 * New function: gras_agent_spawn() to launch a new process on
1978 current host. Only working in simulation for now. [Mt]
1979 * New function: gras_os_hostport() returning a constant form (ie,
1980 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1983 * Make the backtrace of exceptions more human readable [Mt]
1984 * New module: xbt/str [Mt]
1985 a ton of string utility functions (split, join, printf to a newly
1986 allocated buffer, trim, etc)
1987 * New module: xbt/hash [Mt]
1988 SHA1 hashing algorithm (more to come if needed)
1989 * New module: xbt/synchro [Mt]
1990 synchronization tools (mutex and conditions) working the same way
1991 in simulation and in real life (mainly useful for GRAS, but not
1993 * New module: xbt/queue [Mt]
1994 classical producer/consumer synchronization scheme
1995 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1996 (using the classical functions will get serialized) [Mt]
1997 * Make dictionary internal table dynamic. No need to specify its size
1998 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2000 * Make sure the log channels are organized as a tree under windows
2001 (because of ANSI C compatibility issue, any channel were child of
2002 root directly) [Mt].
2005 * Cleaned many thing in surf and fixed a few bugs [AL].
2006 * Add a nice command line configuration mechanism to compose models [AL].
2007 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2008 the previous one (KCCFLN05). It relies on something that looks like
2009 a max-min sharing mechanism but cannot be written as such. A new solver
2010 was thus designed [AL].
2011 * Add a new solver to lmm. Based on Lagrange optimization and
2012 gradient-based descent, it enables to efficiently maximise systems s.a
2014 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2016 This solver enables to propose two new network models for TCP Reno and
2017 TCP Vegas based on Low's work. These models still need to be fully
2018 tested though [Pedro Velho].
2021 * Bug fix in SD_simulate. Now the time bound given as argument is
2023 * Use the new parallel task model (ptask_L07) as default.
2024 * Use the SURF command line configuration mechanism.
2025 * 0-size tasks (for synchronization) should now work.
2027 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2029 SimGrid (3.2) stable; urgency=high
2033 We still experience issues on this platform, but we believe that at
2034 least MSG is usable.
2036 GRAS API BREAKAGE (for simplification purpose, sorry):
2037 * the gras_msgtype_by_name is not used anymore. Instead of
2038 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2039 you can write (and must)
2040 gras_msg_send(toserver, "request", &request);
2041 - If you still want to pass a gras_msgtype_t to the function (to cache
2042 the type and avoid the lookup time), use the gras_msg_send_() variant.
2043 - Impacted functions:
2044 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2045 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2046 * The callbacks are now expected to return 0 when everything went well
2047 (just like the main() function)
2049 GRAS new features and improvements:
2050 * New module mechanism where user code can use per process globals [Mt]
2051 This is similar to gras_userdata_*() functions, but for libraries. It
2052 factorize some code developped over and over in the examples and AMOK.
2053 It has still to be documented and used (only amok/peermanagement is
2055 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2056 loss when they were small enough to fit into the buffer and sent quickly
2057 enough so that they can all get received in one shoot.
2058 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2059 if not found. Use the *_or_null() variant for the old semantic.
2060 * In gras_msg_handle, do not discard messages without callback.
2061 They are probably messages to be explicitly awaited later (ie, proofs of
2062 mis-synchronization in userland since they are sent before being awaited)
2063 No big deal usually.
2064 * gras_socket_meas_send/recv: semantic changed!
2065 The numerical arguments used to be (1) the total amount of data to send
2066 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2067 messages. This was need for the fool willing to send more than MAXINT
2068 bytes on quite fat pipes.
2071 * Do really rename the hostmanagement module to peermanagement. [Mt]
2072 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2073 but this is rather new and this was documented in the module
2074 documentation (poor excuses, I admit)
2075 * Bandwidth measurement semantic changed! This follows the changes to
2076 gras_socket_meas_send/recv explained above.
2079 * A sequential mode has been added to the workstations. When a workstation
2080 is in sequential mode, it can execute only one task, and the other tasks
2081 are waiting in a FIFO. [Christophe Thiery]
2084 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2085 model for SIMDAG. [Christophe Thiery]
2086 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2087 equal to 0, which caused some bad behaviors in
2088 saturated_constraint_set_update. I now use a threshold mechanism like in
2092 * When running manually src/testall, you select specific units [Mt]
2093 testall is the result of our cunit mechanism, and should replace all
2094 the scripty thingy around since bash don't run easily on billware.
2096 * A mallocator system has been added. [Christophe Thiery]
2097 Mallocators allow you to recycle your unused objects instead of freeing them
2098 and allocating new ones.
2100 Documentation update:
2101 * FAQ reworking + New FAQs:
2102 - "Valgrind spits tons of errors!" [Mt]
2103 - "How to repport bugs" [Mt]
2104 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2105 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2106 - Communication time measurement within MSG [AL]
2107 - I experience weird communication times when I change the latency [AL]
2108 * GRAS tutorial [Mt]
2110 - an introduction to the framework and to the used communication model
2111 - an initiatic tour introducing the most proheminent features:
2113 . Lesson 0: Installing GRAS
2114 . Lesson 1: Setting up your own project
2115 o Part 2: Message passing
2116 . Lesson 2: Exchanging simple messages
2117 . Lesson 3: Passing arguments to the processes (in SG)
2118 . Lesson 4: Attaching callbacks to messages
2119 . Lesson 5: Using globals in processes
2120 . Lesson 6: Logging information properly
2121 . Lesson 7: Using internal timers
2122 . Lesson 8: Handling errors through exceptions
2123 . Lesson 9: Exchanging simple data
2124 . Lesson 10: Remote Procedure Calling (RPC)
2125 . Lesson 11: Explicitely waiting for messages
2126 . Recapping of message passing features in GRAS
2127 - A HOWTO section containing:
2128 o HOWTO design a GRAS application
2129 More are due, of course. They will come latter. In the meanwhile, you can
2130 check the examples which are still here.
2132 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2134 SimGrid (3.1) stable; urgency=high
2138 There was a stack corruption somewhere, visible only when optimizing
2139 with these versions. [Vince]
2142 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2143 module that enabled to play with Directed Acyclic Graphs. It is built
2144 directly on top of SURF and provides an API rather close to the old
2145 SG. Some old codes using SG are currently under rewrite to check that
2146 all needful functions are provided. [Christophe Thiery]
2149 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2150 extension of the classical CLM03 model that gracefully handles
2151 failures. This is now the default model for MSG and GRAS. It doesn't
2152 handle parallel tasks yet though. [AL]
2153 * Bug fix: Weights were not correctly set in the network part.
2154 WARNING: This may have resulted in incorrect results with simulations
2155 where there are more than one flow on a given link. [AL]
2158 * After a (long ?) discussion on simgrid-devel, we have decided that the
2159 convention we had on units was stupid. That is why it has been decided
2160 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2161 WARNING : This means that all previous platform files will not work as
2162 such with this version! A warning is issued to ask users to update
2164 A conversion script can be found in the contrib module of the CVS, under
2165 the name contrib/platform_generation/surfxml_update.pl [MQ]
2168 * Bug fix: Processes were started in reverse order, wrt deployment file.
2169 WARNING: if your code relies on this bug, please fix it. [AL]
2170 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2171 being executed on two different locations. [AL]
2172 * Bug fix: Failures are now better supported thanks to Derrick's tests
2173 (there was many failure situations I hadn't thought of and that weren't
2174 correctly handled). [AL]
2175 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2179 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2180 Exception occurring on server-side are propagated back to client (!).
2182 API CHANGE: the callback changed their prototype. Change:
2183 int my_handler(gras_socket_t expeditor, void *payload_data) {
2185 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2186 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2188 * New! function: gras_msg_handleall to deal with all messages arriving
2189 within a given period.
2190 * New! function: gras_socket_server_range to get a server socket in a
2191 range of port numbers (ease to avoid port number conflicts) [MQ]
2192 * New! gras processes display their backtrace when they get a SIGUSR1
2193 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2194 Sweet to debug RL processes [MQ]
2198 - Do not force experiment sizes to be expressed in kb, or it becomes
2199 impossible to measure the latency this way (needs one byte-long tests)
2200 WARNING: this changes the amok_bw_* function semantic. [MQ]
2201 - Implements the link saturation stuff. [MQ]
2202 * Peer management module:
2203 New! module factorizing code that we wrote over and over [MQ].
2206 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2207 - Test units are placed directly into the library code, they get extracted
2208 automatically and placed into the src/testall binary.
2209 - Convert most of the XBT tests to this system.
2210 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2211 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2212 live in the same C code. This required to change a little bit the API
2213 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2214 * New module: sparse graph structure with basic algorithms (this is work
2215 in progress and the API is not considered to be frozen yet). [AL]
2216 * Display more information on backtraces: source line & function names are
2217 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2218 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2219 * Reworked a little bit some #include statements to load only required
2220 headers. Some user code that relied on SimGrid to include stdlib or
2221 stdio may need to include it by themselves. [AL]
2222 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2224 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2225 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2226 but I guess that almost nobody used those parts. [MQ]
2228 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2230 SimGrid (3.0.1) stable; urgency=low
2233 * Unfortunately, I had missed 5 misnamed functions:
2234 xbt_fifo_item_t xbt_fifo_newitem(void);
2235 void xbt_fifo_freeitem(xbt_fifo_item_t);
2236 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2237 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2238 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2239 They're now deprecated. Please use their new versions:
2240 xbt_fifo_item_t xbt_fifo_new_item(void);
2241 void xbt_fifo_free_item(xbt_fifo_item_t);
2242 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2243 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2244 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2246 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2247 * Documentation: xbt_log module unmercifully reworked [MQ]
2248 * Bugfix: there was a problem with the ending of contexts with
2249 the pthread backend. It caused some weird deadlock or behavior
2250 depending on the pthread implementation. [AL]
2251 * Bugfix: get the exceptions raised in the simulator repport where
2252 and why they come from when they are not catched in time [AL, MQ]
2255 * Bugfix: Do repport the error when two non-connected hosts try to
2256 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2259 * Add additionnal checkings on communications. Assert that two
2260 communicating hosts are connected by a set of links... [AL]
2263 * Add additionnal checkings on channel values in communication [AL]
2264 * New: MSG_task_get_source to see on which host a task was generated [HC]
2265 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2266 the number of tasks waiting to be received on channel and sent
2268 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2269 waits for the first task coming from a given host.. [AL]
2271 GRAS new functionnalities: [MQ]
2272 * Enhance the parsing macro to allow the size of multidimentional objects
2273 to be given thru annotations.
2274 * New example (and documentation): Matrix Multiplication a la RPC
2275 (as when I was young!) and fix a bunch of bugs found on the way.
2277 GRAS performance improvements: [MQ]
2279 * Reduce the amount of cbps creation/destruction by making it static to
2280 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2283 * Change libdata to a set so that we can search for stuff by ID (and thus
2284 reduce the insane amount of dict lookups)
2287 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2288 RL communication on top of SG since it's so uneffective.
2289 It may also be used for inter-thread communication in RL, one day.
2290 * Use gras_datadesc_copy() to exchange messages on top of SG
2292 - improve message exchange performance on top of SG
2293 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2294 * Don't exchange on the network the size of the used part of buffer,
2295 instead, specify the possible buffer size to read().
2297 - reduces the amount of read/write calls (one pair per exchange)
2298 - reduces the amount of exchanged data (the size)
2299 - allows to retrieve all arrived data on receiver side, if we don't need
2300 it right now (subsequent read will peek the buffer)
2301 - allows the receiver to proceed with the begining of the stream before
2302 everything is arrived
2303 - make it possible to build an iov transport (using readv/writev)
2305 - take care of the data with non-stable storage (like stacked data),
2307 * If possible, TCP send uses vector I/O (when writev() is here)
2308 - Don't use it for receive since we send data sizes and data on the
2309 same stream, so we wouldn't be able to chain large amount of chunks
2310 before having to flush the stuff to read the size.
2311 * Rework the transport plugin mechanism to simplify it and reduce the
2312 amount of pointer dereferencement when searching for the right function
2315 * I guess that now, we do almost as few system calls as possible while
2316 doing as few data copy as possible.
2318 To improve it further, we could try to send all the sizes first and then
2319 all the data (to use iov on receiving size), but it's only a partial
2320 solution: when you have 2 dimensional data, the sizes of the second
2321 dimension is data of the first dimension, so you need 3 streams.
2323 I'm not sure the potential performance gains justify the coding burden.
2325 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2327 SimGrid (3.00) stable; urgency=high
2330 * New! Give the possibility to hijack the surf parser and thus bypass
2331 MSG_create_environment and MSG_launch_application. Have a look at
2332 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2334 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2336 SimGrid (2.96) unstable; urgency=low
2341 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2342 This deprecates the xbt_error_t mechanisms.
2343 It modifies (simplifies) all XBT and GRAS API.
2344 MSG API keeps unchanged (exceptions raised by XBT are catched from
2345 within MSG and masked with existing error handling facilities)
2348 * New! Add a FATPIPE model. [AL]
2349 * New! Add a parallel task model. [AL]
2350 * New! Add automatically a loopback interface (in the default
2351 network model) if none was precised.
2354 * Bugfix: MSG_process_resume now works with the current running process.
2356 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2357 * Modification of MSG_task_get_compute_duration. Once a task has been
2358 processed, the value returned by this function is now equal to 0. [AL]
2359 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2360 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2361 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2362 of a m_task. For now, MSG_task_cancel only works with computation
2364 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2365 of the processor (in Mflop/s) regardless of the current load on the
2367 * API Change: use proper naming convention for MSG_getClock and
2368 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2370 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2371 This function changes the priority of a computation task. This priority
2372 doesn't affect the transfer rate. A priority of 2 will make a task
2373 receive two times more cpu power than the other ones. This function
2374 has been added to suit the needs of Nguyen The Loc and hasn't been that
2375 much tested yet. So if it fails, please report it and send me your code.
2377 * API Change: removed all functions and types that were marked "deprecated"
2378 since many months. Renamed MSG_global_init_args to MSG_global_init.
2380 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2382 SimGrid (2.95) unstable; urgency=low
2385 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2386 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2387 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2390 * Bugfix: Allow absolute paths for platform description files [MQ]
2391 * Bugfix: do free the variables after use. Leads to drastic performance
2393 * Implement max_duration (ie, timeouts) on resources [AL]
2396 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2398 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2399 the needed support (timeouts) [AL]
2402 * Implement measurement sockets. You can now get the bandwidth between two
2403 hosts thanks to AMOK (see below). [MQ]
2404 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2405 dynar over the network (yeah) [MQ]
2406 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2407 * simple/static token-ring example. [Alexandre Colucci and MQ]
2408 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2409 in place before [MQ]
2412 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2413 type parsing. "short" and "long" modifiers were also ignored; other
2414 modifier (such as reference level) are still ignored. [MQ]
2415 * Embeed the buffer size within the buffer itself on SG. [MQ]
2416 That way, send() are atomic and cannot get intermixed anymore (at least
2417 the ones which are less than 100k; bigger messages still have the issue)
2418 * Array size pushed by the field, not by the field type (or each
2419 and every long int will push stuff to the cbps) [MQ]
2420 * use select() to sleep since it allows to portably sleep less than one
2423 GRAS (minor cleanups)
2424 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2427 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2428 * type category 'ignored' killed as it was never used and were difficult
2430 * whether a type can cycle or not is now a flag, leaving room for more
2431 flags (as "ignored", if we feel the need one day ;)
2432 * Rename raw sockets to measurement sockets since "raw" has another
2433 meaning in networking community.
2436 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2437 and offers features not belonging to GRAS but that most applications
2438 need. One day, it may be a set of plugins loadable at runtime.
2439 * New module: bandwidth
2440 bandwidth measurement between arbitrary nodes running this module. [MQ]
2442 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2444 SimGrid (2.94) unstable; urgency=low
2446 The first beta release of SimGrid 3 !
2450 * Update the main page and the FAQ. Adding references to gforge.
2453 * Add a gras_os_getpid function.
2456 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2457 * Extend the logs so that they also print PID, hostname, date, ... if
2459 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2460 and kill the old version which were in testsuite/
2461 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2465 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2466 for users. Instead, each programming environment (GRAS, MSG,...) use it
2467 when needed to provide such a feature to users.
2468 Don't shortcut the mechanism or you will also shortcut the virtualization
2469 you need on the simulator.
2473 * Cleanups in configury with regard to compile optimization/warning flags.
2474 Also add -fno-loop-optimize to any powerpc since it's the optimization
2475 killing gcc (< 3.4.0).
2476 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2477 needed by MSG examples complications
2478 * Borrow configury beautifications from PHP
2481 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2482 warning (thanks loris for stumbling into this one).
2483 * Bugfix: stop loading private headers (gras_config.h) from the public
2487 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2488 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2489 RL would imply a DNS resolver, which is *hard* to do in a portable way
2490 (and therefore delayed).
2491 * Implement a real timer mechanism and use it in timing macros. This allows
2492 to avoid rounding errors and get a 0.000005 sec precision in timing
2493 macros. While I was at it, various cleanups:
2494 - allow to declare more than one timed section per file (fix a stupid bug)
2495 - move some private declaration to the right place
2496 - merge conditional execution and timing macros into emulation module
2497 - document the module
2498 - make sure the module cleanups its mess on gras_exit
2499 * Documentation improvements:
2500 - (new) how to compile applications using GRAS
2501 - (new) emulation support (timing macros)
2503 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2505 SimGrid (2.93) unstable; urgency=low
2507 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2510 - Use Paje properly where used. Still to be sanitized properly.
2511 - Portability fix: Add an implementation of the contexts using pthread
2515 - Add xbt_procname(): returns the name of the current process.
2516 Use it to show the current process's name in all logging.
2518 - fix detection of older flex version and the reaction, since we do
2519 depend on modern ones (we use lex_destroy)
2520 - Better separation of SG and RL in the libs: remove all simulation code
2521 from libgras. As a result, this lib is now only 200k when stripped.
2522 Some of the xbt modules may also be duplicated (two sets and such) and
2523 should be cleaned/killed before SG3.
2524 - Insist on using xlC on AIX because of weird problems involving gcc there.
2525 - Cleanup the make remote stuff. This is now done by scripts
2526 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2527 mainly for our private use, but we're working on changing them to user
2530 - Bugfix: flush the socket on close only if there is some *output*.
2531 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2533 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2536 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2538 SimGrid (2.92) unstable; urgency=low
2540 Alpha 3 on the path to SimGrid 3
2544 - New! First try of benchmarking macros.
2545 - New! First try so that gras_stub_generator generate deployment and
2546 remote compilation helpers.
2548 - Bugfix: Initialization fix in msg_test.
2552 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2554 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2556 - API Change: gras_os_sleep to take the amount of seconds as a double.
2557 Accepting an int was error prone since it was the only location where
2558 seconds were coded as such. It leaded to damn rounding errors.
2559 - Bugfix: Hard to belive that timers ever worked before this.
2561 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2563 SimGrid (2.91) unstable; urgency=low
2565 Alpha 2 on the path to SimGrid 3
2569 - Bug fix in the lmm_solver.
2571 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2572 through the function MSG_paje_output.
2573 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2574 - It is possible to bound the rate of a communication in MSG with
2575 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2576 to write it in the changelog).
2577 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2582 - Various cleanups to the autotools stuff
2583 - Begin to move Gras examples to examples/gras/
2584 - Let make distcheck work again (yeah!)
2586 - documentation overhauled using doxygen.
2587 gtk-doc-tools is dead in SimGrid now.
2588 - Automatically extract all existing logging categories, and add the list
2589 to the documentation (long standing one, to say the less)
2591 - Cleanup the known architecture table. Reorder the entries to group what
2592 should be, and use a more consistent naming scheme.
2593 (some of the test dataset are still to be regenerated)
2594 - New! Allow library to register globals on each process just as userdata
2596 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2597 lookup time (for now).
2598 Use it in msg and trp.
2599 This cleans a lot the internals and helps enforcing privacy of the
2600 headers between the gras components.
2601 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2602 - Bugfix: gras_os_time was delirious in RL.
2603 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2605 - Reenable GRAS now that it works.
2607 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2609 SimGrid (2.90) unstable; urgency=low
2611 Alpha 1 on the path to SimGrid 3
2613 * It is a long time since the last release of SimGrid. I'm sorry about
2614 that but as I had told you, I was rewriting a lot of things. I apologize
2615 to those who had been reporting bugs to me and that I had not answered.
2616 If your bug is still in the new version, please tell me. Here is a
2617 summary of the main changes.
2619 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2620 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2621 portability, speed, and a lot more but you'll figure it out later.
2622 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2623 I wanted to release the new MSG as soon as possible and I have
2624 broken GRAS, which is the reason why, for now, only MSG is fully
2625 functional. A laconic description of these projects is available
2626 in the documentation.
2628 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2629 kernel optimized for our needs (called SURF but only the developers
2630 should use it). Hence, MSG is now roughly 30 times faster and I think
2631 that by rewriting a little bit MSG, I could event speed it up a little
2632 bit more. Beside the gain in speed, it is also much easier to encode a
2633 new platform model with SURF than it was with SG. More to come...
2635 * REVOLUTION 3: I have tried to change a little as possible the API of
2636 MSG but a few things really had to disappear. The main differences
2637 with the previous version are :
2638 1) no more m_links_t and the corresponding functions. Platforms are
2639 directly read from a XML description and cannot be hard-coded
2640 anymore. The same format is used for application deployment
2641 description. The new format is described in the documentation.
2642 Have a look in tools/platform_generation. There is a tiny script
2643 that converts from the old platform format to the new one. Concerning
2644 the application deployment format, parsing the old one is tricky.
2645 I think most of you should however be able to convert your files. If
2646 it is really an issue, I can write a C code that does the conversion.
2648 2) the toolbox tbx does not exist anymore. We now have a library
2649 with much more data-structures but without the hash-tables (we have
2650 dictionaries that are much faster).
2652 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2654 *****************************************************************************
2655 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2656 * don't feel like converting the oldies. (Mt) *
2657 *****************************************************************************
2660 Version 2.90: "the long awaited one"
2661 - Finished rewriting and debugging MSG. Rewrote the documentation.
2662 - disable GRAS for now since it needs to be ported to the newest SG
2665 - Finish the port to windows (using mingw32 for cross-compile)
2668 - Main loop and datastructures of SURF. A cpu resource object is
2669 functional. Surf can thus be used to create cpu's with variable
2670 performance on which you can execute some actions.
2672 2004-11-15 Martin Quinson
2673 - Port to ARM. Simply added the alignment and size descriptions. Should
2674 work, but the ARM machines are so slow that I didn't had the opportunity
2675 to 'make check' over there yet.
2677 2004-11-15 Arnaud Legrand
2678 - Trace manager now written. It uses a heap structure and is therefore
2679 expected to be efficient. It may however be speeded up (particularly
2680 when many events occur at the same date) by using red and black
2681 trees. One day maybe...
2682 - Max-min linear system solver written. It uses a sparse matrix
2683 structure taking advantage of its expected use. Most operations are
2684 O(1) and free/calloc are called as few as possible. The computation of
2685 the minimum could however be improved by using a red and black tree
2688 2004-11-03 Arnaud Legrand
2689 - Rename every gras_* function that was in xbt/ to its xbt_
2691 - Add a heap and a doubly-linked list to xbt
2692 - Added a dichotomy to the dictionaries. make check works as well before
2693 so I assume that the patch is correct. I do not know however if things
2694 run effectively faster than before now. :)
2696 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2697 SimGrid, and the version number is bumped to 2.x
2699 2004-10-29 Martin Quinson
2700 - Introduction of the remote errors.
2701 They are the result of a RMI/RPC on the remote machine.
2702 ErrCodes being scalar values, you can't get the host on which those
2703 errors did happen. Extending the error mechanism as in Gnome is possible.
2704 No idea yet whether it is a good idea.
2706 2004-10-28 Martin Quinson
2707 - Interface revolution: the Starred Structure Eradication.
2708 I used to do typedef struct {} toto_t; and then handle *toto_t.
2709 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2710 - ???_t is a valid type (builded with typedef)
2711 - s_toto_t is a structure (access to fields with .)
2712 - s_toto is a structure needing 'struct' keyword to be used
2713 - e_toto_t is an enum
2714 - toto_t is an 'object' (struct*)
2716 typedef struct s_toto {} s_toto_t, *toto_t;
2717 typedef enum {} e_toto_t;
2718 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2719 s_toto_t) is private.
2721 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2722 so that it can set the variable to NULL. It was so for dicts and sets,
2723 it changed for dynars.
2725 - Fix a bunch of memleaks in dict_remove
2726 - Fix a bug in sg/server_socket opening: it failed all the time.
2728 2004-10-07 Martin Quinson
2729 - Speed up dynar lookup operation a bit.
2731 gras_dynar_get is dead.
2733 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2734 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2735 (which gives you the address of the stored data).
2737 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2738 copy of the data using an affectation to do the job and not a memcpy.
2740 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2741 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2743 It does not really speedup the dynar test because they are
2744 setting elements all the time (and look them seldom). But the dict does
2745 far more lookup than setting.
2747 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2749 2004-10-05 Martin Quinson
2750 - Allow to (en/dis)able the cycle detection at run time.
2752 Whether we should check for cycle or not is now a property of each
2753 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2754 datadesc_cycle_unset allow to remove this property when previously set.
2756 Note that the cycle detection is off by default since it impacts the
2757 performance. Watch the data you feed GRAS with ;)
2759 This property is hereditary. Any element embedded in a structure having it
2760 set have it set for the time of this data exchange.
2762 You should set it both on sender and receiver side. If you don't set it on
2763 sender side, it will enter an endless loop. If you forget on receiver
2764 side, the cycles won't be recreated after communication.
2766 - Header reorganization.
2767 Kill gras_private.h, each submodule must load the headers it needs.
2769 2004-10-04 Martin Quinson
2770 - Interface revolution: do not try to survive to malloc failure.
2772 Now, gras_malloc and friends call gras_abort() on failure.
2773 As a conclusion, malloc_error is not a valid error anymore, and all
2774 functions for which it was the only gras_error_t return value are
2775 changed. They now return void, or there result directly.
2776 This simplify the API a lot.
2778 2004-09-29 Martin Quinson
2779 - Re-enable raw sockets.
2780 Created by gras_socket_{client,server}_ext;
2781 Used with gras_raw_{send,recv}
2784 It should allow to kill the last bits of gras first version soon.
2786 This is not completely satisfactory yet (duplicate code with
2787 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2790 - Simplify transport plugin (internal) interface by not passing any
2791 argument to _server and _client, but embedding them in the socket
2794 2004-09-28 Martin Quinson
2795 - Finish the port to AIX.
2796 autoconf was my problem (segfault within the malloc replacement
2797 function. No idea why)
2799 2004-09-16 Martin Quinson
2800 - Fix some size_t madness on 64bit architectures.
2802 2004-09-08 Martin Quinson
2803 - Reduce the number of system headers loaded, overload some more system
2804 calls (such as malloc to cast the result of the system one, and work
2806 - Fix and reintroduce the config support
2808 2004-09-07 Martin Quinson
2809 - Source code reorganization to allow Arnaud to surf all over there.
2810 - Allow to document the logging categories.
2811 - Remove all uppercase from logging categories and useless cleanup in names.
2813 2004-08-18 Martin Quinson
2814 Version 0.6.2 (protocol not changed; API changed)
2815 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2816 gras_error_t), and NULL when not found. Functions expecting a msgtype
2817 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2818 hopefully usefull message.
2819 - Portability to prehistoric sparcs again
2821 2004-08-17 Martin Quinson
2822 Version 0.6.1 (protocol not changed; ABI not changed)
2823 - prealloc some buffers to speed things up
2825 2004-08-11 Martin Quinson
2826 Version 0.6 (protocol not changed; ABI expended)
2827 - The parsing macro can deal with the references, provided that you add
2828 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2830 2004-08-09 Martin Quinson
2831 Version 0.5 (protocol not changed; ABI changed)
2832 - Allow to off turn the cycle detection code in data exchange at
2833 compilation time. It should be at run time, but I'm short of time (and
2834 the config stuff is still broken). That way, we keep dict out of the
2835 critical path, which is good because the performance is poor:
2836 - search not dichotomial yet
2837 - dynar give no way to access their content and memcpy everytime
2838 - In composed data description (struct, ref and so on), stop foolness of
2839 keeping the subtype's ID, but store the type itself. This keeps sets out
2840 of the critical path, which is good since they rely on dynar and
2841 dictionnaries. The only loose of that is that we cannot detect the
2842 redeclaration of a structure/union with another content (but I'm not sure
2843 the code detected well this error before anyway). We still can detect
2844 the redefinition discrepancy for the other types.
2845 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2846 it breaks the code because of type-punning used all over the place).
2847 This breaks on all non-gcc architectures (for now).
2849 All those changes (plus the buffer of last time) allow me to gain 2 order
2850 of magnitude on cruel tests consisting of 800000 array of integers on two
2851 level of a hierarchical structure (200 secondes -> 4 secondes)
2854 - the selector of reference must now return the type it points to, not
2855 the ID of this type.
2857 2004-08-06 Martin Quinson
2858 Version 0.4 (protocol changed; ABI not changed)
2859 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2860 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2861 - Implement a buffer transport, and use it by default (it relies on tcp in
2862 real life and on sg in simulation).
2863 That's a bit hackish since I had a new field to the structure to store
2864 its data without interfering with the subtype ones. Inheritance
2865 is tricky in C. And that's a kind of reverse inheritance with one class
2866 derivating two classes. Or maybe a game with java interfaces. Anyway,
2867 that's damn hard in C (at least).
2868 Moreover, I got tired while trying to ensure plugin separation and
2869 genericity in SG mode. MSG wants me to do weird things, so let's go for
2870 cruel hacks (temporarily of course ;).
2871 See comment in transport_private.h:71
2872 - do not include all the _interface headers in private but in the files
2873 which really need them (to cut the compilation time when they are
2876 2004-07-26 Martin Quinson
2877 Version 0.3 (protocol not changed; ABI changed)
2878 - Major overhault of the datadesc interface to simplify it:
2879 - shorted the function names:
2880 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2881 - add a trivial way to push/pop integers into the cbps without malloc.
2882 This allows to make really generic sub_type description, which simply
2883 pop their size of the stack.
2884 - add a function gras_datadesc_ref_pop_arr() which does what users want
2885 most of the time: Declare a dynamic array (which pops its size of the
2886 stack) and declare a reference to it. Poor name, but anyway.
2887 - kill the post-send callback, add a post-receive one
2889 2004-07-23 Martin Quinson
2890 Version 0.2 (protocol changed; ABI changed)
2891 - add some testing for cpbs in the test cases, and fix some more bugs.
2892 This invalidate again the little64 data file, since I cannot regenerate
2894 - remove an awfull optimization in the logging stuff, allowing me to:
2895 - understand it again
2896 - learn gcc how to check that the argument match the provided format
2897 - fix all errors revealed by gcc after that
2898 - internal keys of dict are not \0 terminated. Deal with it properly in
2899 loggings instead of segfaulting when the user want to see the logs :-/
2901 2004-07-22 Martin Quinson
2902 - Fix some stupid bug preventing cbps (callback postit) from working
2904 2004-07-21 Martin Quinson
2905 - Some documentation cleanups
2906 - remove the useless last argument of msgtype_declare
2907 - rename the Virtu functions to fit into the 'os' namespace
2908 - move headers src/include -> src/include/gras/ and stop fooling with
2909 gras -> . symbolic link
2910 - make distcheck is now successful
2912 2004-07-19 Martin Quinson
2914 - Build shared library also
2915 - Install html doc to the right location
2916 - stop removing maintainer files in make clean
2917 - build tests only on make check
2919 2004-07-13 Martin Quinson
2921 - No major issue in previous version => change versioning schema
2922 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2923 corresponding dataset.
2925 2004-07-11 Martin Quinson
2927 - Get it working with any kind of structure (we can compute the padding
2928 bytes remotely for all the architectures I have access to)
2929 - Implement the structure parsing macro (still not quite robust/complete)
2930 - Improvement to the remote testing toysuite
2932 2004-07-10 Martin Quinson
2933 [autoconf mechanism]
2934 - get ride of a bunch of deprecated macros
2935 - actually run the test for two-compliment, not only compile it :-/
2936 - test whether the structures get packed (and bail out if yes. Damn.
2937 Alignment is a serious matter)
2938 - test whether the structures get compacted (but respecting the alignment
2939 constraints of each types)
2940 - test whether the array fields of structures can straddle alignment boundaries
2942 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2943 (took me 2 hours to find that bug, looking at the wrong place)
2944 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2945 before sending/receiving and are used to compute the offsets of fields
2946 - Given that padding size depend even on compiler options, keep track of
2947 alignment and aligned_size only for the current architecture. Not a big
2948 deal since we send structure fields one after the other (seems
2950 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2951 program, let it work on linux/gcc/little32. portability todo.
2953 2004-07-08 Martin Quinson
2954 - import and improve remote compilation support from FAST
2955 - make sure make check works on half a dozen of machines out there
2957 2004-07-07 Martin Quinson
2958 Let's say it's version 0.0.3 ;)
2959 - Implement conversions (yuhu!)
2960 - Let it work on solaris (beside conversion, of course)
2961 - Stupid me, using rand() to generate the conversion datatests in not wise.
2963 2004-07-06 Martin Quinson
2964 - Let make dist work, since I'm gonna need it to compile on remote hosts
2965 - Let Tests/datadesc_usage write the architecture on which the file was
2966 generated as first byte.
2967 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2968 architecture descriptions.
2969 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2970 on those architectures.
2971 - Optimization: send/recv array of scalar in one shoot
2973 2004-07-05 Martin Quinson
2974 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2976 - Plug a whole bunch of memleaks
2977 - each process now have to call gras_{init,exit}. One day, their log
2978 settings will be separated
2979 - Continue the code factorisation between SG, RL and common in Transport.
2981 2004-07-04 Martin Quinson
2983 - Redistribution between SG and RL.
2984 We wanna have to accept in SG, so move accepted related parts of RL in
2985 the common part. (more precisely, the dynar of all known sockets is no
2986 more a static in transport.c, but part of the process_data)
2988 [gras_stub_generator]
2989 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2991 2004-07-03 Martin Quinson
2992 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2993 them all up in one shot)
2995 - Fix the ugly case of reference to dynamic array.
2996 - New (semi-public) function gras_datadesc_size to allow the messaging
2997 layer to malloc the needed space for the buffer.
2999 - gras_socket_close now expect the socket to close (and not its address to
3000 put NULL in it after it). This is because the socket passed to handlers
3001 is one of their argument (=> not writable).
3003 - propagate the interface cleanup from last week in datadesc, ie remove a
3004 superfluous level of indirection. User pass adress of variable
3005 containing data (both when sending and receiving), and not of a variable
3006 being a pointer to the data. Let's say that I like it better ;)
3007 The price for that is constructs like "int msg=*(int*)payload" in
3008 handlers, but it's a fine price, IMHO.
3010 - Let it work in RL (yuhu)
3012 2004-06-21 Martin Quinson
3014 - porting SG plugin and SG select to new standards (works almost).
3015 - plug memleaks and fix bugs around.
3018 - cleanup the prototype of data recv and force users to specify when they
3019 want to handle references to objects. Test case working even for cycles.
3020 - plug memleaks. Valgrind is perfectly ok with this.
3022 2004-06-12 Martin Quinson
3024 - cleanup the separation between plugin and main code in plugin creation
3026 2004-06-11 Martin Quinson
3028 - Reput hook for raw sockets, needed for BW experiments
3029 - kill a few lines of dead code
3030 [Data description] Interface cleanup
3031 - gras_datadesc_by_name returns the searched type or NULL.
3032 That way, no variable is needed to use a type desc once, which makes
3034 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3035 two parameters were strings (field name, type name), leading to
3037 [Dicos] Interface cleanup
3038 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3039 This is consistant with the dynar API.
3041 2004-04-21 Martin Quinson
3043 - Porting to new standards.
3045 - interface cleanup.
3046 There is no bag anymore, no need to take extra provision to mask the
3047 pointers behind "ID".
3048 Better splitup of functions between files create/exchange/convert.
3049 This is still a bit artificial since convert and receive are so
3050 interleaved, but anyway.
3052 - add a queued message list to procdata (the ones not matching criteria
3054 - factorize some more code between SG and RL wrt procdata
3056 - use gras_exit in example to track memleaks
3057 - get rid of gs_example now that GS is properly integrated into gras
3058 - update run_test to integrate the latest tests (datadesc)
3060 - rename WARNINGn macros to WARNn since it prooved error-prone
3062 2004-04-19 Martin Quinson
3064 - register init/exit functions within gras module mechanism
3065 - send/receive function.
3066 Convertion is not implemented, but short-cutted if not needed.
3067 struct/array elements are sent one by one (instead of block-wise), but
3068 nobody really cares (yet). Get a prototype before optimizing.
3069 - tests (using a file socket) for DD send/receive on:
3070 - base types: int, float
3071 - array: fixed size, string (ie ref to dynamic string)
3072 - structure: homogeneous, heterogeneous
3073 - chained list, graph with cycle
3074 Believe it or not, valgrind is not too unhappy with the results. The
3075 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3079 - Bugfix when using a filename explicitely (instead of '-')
3081 2004-04-09 Martin Quinson
3083 - factorize more code between RL and SG in socket creation
3084 - Complete the implementation and tests of:
3086 o file (only in RL, and mainly for debugging)
3088 I lost 3 days to design a portable address resolver, and then decided
3089 that the prototype mainly have to run on my box.
3090 Addressing portability too early may be like optimizing too early :-/
3092 - use gras_init in the Tests instead of the crappy parse_log_opt
3093 (the latter function is removed)
3094 [Conditional execution]
3095 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3096 [Code reorganisation]
3097 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3098 Build examples against the RL library, since there is no way to disable
3099 its creation for now.
3101 For information, the beginning of coding on GRAS was back in june
3102 2003. I guess that every line has been rewritten at least twice since