1 SimGrid (3.12) NOT RELEASED; urgency=low
4 * Interface improvement:
5 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
6 - Sanitize the interface in MSG_task_ module:
7 - Merge two functions that were close enough but misleading:
8 set_compute_duration(t) -> set_flops_amount(t)
9 get_remaining_computation(t) -> set_flops_amount(t)
10 - set_data_size(t) -> set_bytes_amount(t)
11 get_data_size(t) -> get_bytes_amount(t)
12 - Massive cleanups in the functions related to the energy
13 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
14 - New: MSG_host_get_pstate()
16 * Interface improvement:
17 - Rename Host.isAvail() to Host.isOn()
18 - Rename Process.currentProcess() to Process.getCurrentProcess()
19 - Rename Task.setDataSize() to Task.setBytesAmount()
20 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
22 - #18874: Actually allows the GC to reclaim tasks
26 - SIMIX_process_throw: raises an exception in a remote process
27 * Refactoring: Separate submodules
28 - libsmx: the public interface, as libc in a real system
29 - popping: the strange dance that converts a user request into a kernel handling
30 - smx_context_*: the virtualization mechanisms that embeed the user code
31 - smx_*: the handling of each simcalls
33 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
34 * Rename smx_action_t into smx_synchro_t, making explicit that these
35 things are used to synchronize processes with their environment.
36 For example, a communication is a sort of synchronization involving
37 the communicating processes (that may block until the exchange) and
38 the platform. The same can be said from computations, etc.
40 - #18888: segfault when a process ends before its kill_time
44 - 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.
45 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
46 - MPI_Type_set_name, MPI_Type_get_name
47 - MPI_*_c2f and MPI_*_f2c functions
48 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
49 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
50 - Activate a lot of new tests from the mpich 3 testsuite
52 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
53 - InfiniBand network model added : Based on the works of Jerome Vienne (http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf )
54 - When smpi/display_timing is set, also display global simulation time and application times
55 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
56 * Collective communications
57 - 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
58 - 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/).
59 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
61 - Replay now uses algorithms from wanted collective selector
62 - Replay can be used with SMP-aware algorithms
63 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
64 - 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.
66 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
67 - Correctly use loopback on fat-tree clusters
68 - Asynchronous small messages shouldn't trigger deadlocks anymore
69 * Energy/DVFS cleanup and improvement
70 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
71 - new: smpi_get_host_pstate()
74 - "Full" network optimization flag was broken since Surf++
75 - Better handling of precision flags in maxmin
76 - Fix bug causing sometimes "Impossible" errors
77 - Properly pass cluster properties to included hosts
78 * Improvement of the Energy plugin.
79 - Always update the consumption before returning that value
80 - New property: watt_off to denote the disipation when the host is off
83 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
84 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
85 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
87 * Tracing is now always enabled (no way to turn it out)
88 * Move headers around to sort them out on installed systems:
89 - instr/instr.h -> simgrid/instr.h
90 - instr/jedule/* -> simgrid/jedule
91 - simdag/datatypes.h was removed
92 - simdag/simdag.h -> simgrid/simdag.h
93 - msg/datatypes.h was removed
94 - msg/msg.h -> simgrid/msg.h
96 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
98 SimGrid (3.11) stable; urgency=low
103 * Normalizing pointers addresses tool for better diff between logs
106 * Add cloud examples using new VMs
107 - examples/msg/cloud/two_tasks_vm.tesh
108 - examples/msg/cloud/simple_vm.tesh
109 - examples/java/cloud/cloud.tesh
110 - examples/java/cloud/migration/migration.tesh
111 * Add java surf examples:
112 - examples/java/surfPlugin/surf_plugin.tesh
113 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
114 - examples/java/surfCpuModel/surf_cpu_model.tesh
115 * Add SMPI+MSG example:
116 - examples/smpi/smpi_msg_masterslave/
121 - msg task destroy cancel test
122 - msg_host on/off test
123 * Move all tests in testsuite to teshsuite (adding tesh files)
124 * Restructure teshsuites
125 - one folder for each kind of test
126 * Restructure AddTests.cmake
128 - structure the order of tests (with sections)
131 * Add virtual machine
132 - creation of a VM on a PM
133 - migration of a VM from a PM to another PM
135 - MSG_process_join(msg_process_t process, double timeout)
136 - msg_bar_t MSG_barrier_init(unsigned int count)
137 - int MSG_barrier_wait(msg_bar_t barrier)
138 - void MSG_barrier_destroy(msg_bar_t barrier)
139 - msg_as_t MSG_environment_as_get_by_name(const char * name)
140 * New option "msg/debug_multiple_use" to help debugging when a task is used
143 - Improvements and finalization of MSG_storage, MSG_file APIs and their
145 - Increase code coverage in test suites
149 * Protect context stacks against stack overflow. The number of protected memory
150 pages allocated on the top of each stack (1 by default) can be configured
151 with the new command line option --cfg=contexts/guard_size:<value>.
152 * Simcalls are now generated by a python script that
153 - generates files included by SimGrid
154 - checks that all the functions exist, and proposes prototypes
156 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
157 host_set_data, host_get_data
159 - simcall_process_join(smx_process_t process, double timeout)
160 * Fix bug where sleeping processing could not be suspended.
163 * Translate surf models from C to C++
164 - Generic classes for all models: Model, Resource, Action
165 - A generic interface for each kind of model (CPU, Network, Storage
166 Workstation, WorkstationVM)
168 * Translate surf routings from C to C++
169 * Add callbacks using sigc++ or boost::signals2
170 - Add callback functions for resource creation/destruction
171 - Add callback functions for action state change
172 - Handle Energy as a plugin
173 * Replace swag by boost::intrusive
174 * Add new routing models for clusters. For documentation, see
175 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
176 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
177 parameters for cluster tag
178 - Fat trees, with topology="FAT_TREE" and
179 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
180 - see examples/platforms/torus_cluster.xml and
181 examples/platforms/fat_tree_cluster.xml
185 * Hostfiles support host:nb_processes construct to deploy several processes on
187 * Collective communication algorithms should not crash if used with
188 improper number of nodes and report the error.
189 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
190 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
191 MPI_Dims_create, MPI_Cart_sub are supported.
192 * New interface to use SMPI programmatically (still depends on MSG for
193 some parts, see examples/smpi/smpi_msg_masterslave) :
194 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
198 * Global variables privatization in MPI executables is now performed at runtime
199 with the option smpi/privatize_global_variables (default:no).
200 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
201 dynamic libraries loaded by the application are not privatized (static
202 linking with these libraries is advised in this case)
205 * Options defined in XML work correctly now.
208 * New cmake option, enable_lib_in_jar, to control whether native libraries are
209 copied into simgrid.jar or not (ON by default). Use this option if you want
210 to reduce the size of the installed simgrid.jar, *and* the native libraries
211 are kept installed elsewhere.
212 * Surf binding with SWIG (code generated in maintainer mode only):
213 - plugin to handle callbacks
214 - CPU model only for the moment
217 * Supernovae build mode is definitively removed. It was used to improve
218 inlining and inter-module optimizations. It is nowadays superseded by
219 link-time optimizations commonly available in compilers.
220 * Update ns-3 find lib. Bindings for ns-3 should work again now.
221 * Add boost dependency for surf++
222 * Add new macro for tests
223 - ADD_TESH(name <tesh_args>)
224 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
228 - xbt_bar_t XBT_barrier_init(unsigned int count)
229 - int XBT_barrier_wait(xbt_bar_t barrier)
230 - void XBT_barrier_destroy(xbt_bar_t barrier)
231 * Make the xbt_os_time module public
233 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
235 SimGrid (3.10) stable; urgency=low
237 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
240 * Reintegrate Java to the main archive as desynchronizing these
241 package is not acceptable anymore (Java is now considered stable)
242 * Add explicit synchronization facilities through semaphores
243 * Bug fix: Task.setDataSize() only changed the C world, not the value
244 cached in the Java world
247 * Dramatically change the way files are handled. API and internals changed, but
248 this part of MSG was not considered as production grade either.
249 * Add explicit synchronization facilities through semaphores
250 * Add a new function MSG_host_get_process_list()
251 * Preliminary DVFS support (see examples/msg/energy/ for details)
254 * SMPI is now included directly in the libsimgrid as the windows
255 linker doesn't force us on splitting it anymore.
256 * Improvements of the SMPI replay tool:
257 - Most of the collective communications are now rooted in the same process as
258 in the original application.
259 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
260 used until now). Multiple data types can now be used in a trace.
261 - The replay tool now supports traces produce either by TAU or a modified
263 - Bug Fix: the compute part of the reduce action is now taken into account.
264 - Gatherv collective is now supported
265 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
266 of smpirun outputs time independent traces for the current run. One file
267 is created per process. If too many processes are simulated, this behavior
268 can be changed to one file for all processes by using the
269 tracing/smpi/format/ti_one_file flag
270 * smpirun generates the host file if needed (with given host count and platform)
271 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
272 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
273 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
274 --cfg=smpi/coll_selector:(mpich/ompi)
275 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
277 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
278 happening inside a collective SMPI call.
279 * Fix the behavior of complex data types handling.
280 * Make MPI_Wtime another synchronization point to take computations into
282 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
283 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
284 * Add all missing Fortran bindings, SMPI should work with Fortran 90
285 (no privatization of global variables yet)
286 * Preliminary DVFS support (see examples/smpi/energy/ for details)
289 * Verification of liveness properties is now available for SMPI applications
290 (in addition to MSG applications)
291 * Bugged examples using SMPI in examples/smpi/mc/
292 * Add --cfg=model-check/visited option. Allows the verification of infinite
293 programs. Detection of loops in the execution thanks to the system state
294 comparison and reduction of the state space to explore. Can be combined with
295 DPOR for safety properties.
298 * Allow to change SimGrid configuration (see --help) within the code
299 thanks to SD_config() as it can be done in MSG.
300 * Add a new function SD_task_set_amount() upon user request.
303 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
304 * Remove rule based routing (no more PCRE dependency)
305 * Add a limiter_link option to cluster tag, to specify a maximum reachable
306 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
307 * Add a loopback_bw and loopback_lat options to cluster tag.
308 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
309 routing. Now peers are encapsulated in an AS and have their own private
310 router but this is transparent.
313 * Our own implementation of getline is renamed xbt_getline, and gets
314 used even if the OS provide a getline(). This should reduce the
315 configuration complexity by using the same code on all platforms.
316 * New type: xbt_cfg_elm_boolean.
317 * Allow to use yes/no for boolean configuration options in the command line.
318 * Allow to disable SimGrid cleanups at exit from command line option.
319 There are situations where one may want a simulation to end with an exit.
320 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
321 annoying when scripting around the simulator. Adding a
322 --cfg=clean_atexit:no allows to circumvent this issue.
325 * Lots of memory leaks were corrected in this release.
326 There is no known memory leaks anymore, in all of our 600+ tests.
327 * New command line option --version, to get SimGrid version information.
328 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
330 * Supernovae builds are deprecated, and expected to be removed in the next
333 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
335 SimGrid (3.9) stable; urgency=low
337 The Grasgory release: GRAS is really dead now.
339 * Complete overhaul of the internal host structures scheme.
342 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
343 was considered as experimental and badly maintained since then.
344 * Keeping it was thus a trap to our potential users, that could take
345 it instead of MSG or SMPI by mistake despite is pity state.
346 * GRAS seems to have very few users (if any), and no one volunteered
347 to maintain it further. It also induces a lot of XBT code (for
348 portability sake), that must be maintained too.
349 * For all these reasons, we killed GRAS. If someone wants to revive it
350 in the future, don't cry, our git history still remembers of GRAS.
353 * Major overhaul. Merge our documentation again as time proved that
354 splitting it was really not helping our users.
355 * Further improve the developer documentation to help newcomers
356 hacking on SimGrid itself. The user documentation (and in
357 particular, the beginner documentation) is still in a sorry state.
360 * Now works on Windows too!
361 * Much more extensive test suite, from MPICH
364 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
365 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
366 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
367 * Bug fix: let task be scheduled when the last dependency to be solved is
368 a control dependency.
369 * Remove SD_load_environment_script function.
370 Use the C sg_platf function if you want to declare a platform
374 * New function: MSG_process_get_number()
375 * Old function documented: MSG_config()
376 * Remove MSG_load_platform_script function
377 Use the C sg_platf function if you want to declare a platform
381 * Change the default value of the TCP_gamma constant (maximal size of TCP
382 congestion window) to a more realistic 4MiB value. If you notice changes in
383 your simulation results, you can fall back to the previous 20k tiny window
384 by adding --cfg=network/TCP_gamma:20000 on command line.
385 * (Hopefully) fix a bug wrt periodic availability and state traces
386 * Bug fix: use default values at start when first event in availability/state
387 trace is not at time 0.
390 * remove the "new_" part of function name sg_platf_new_trace_connect
391 (resulting in sg_platf_trace_connect), since it does not create
395 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
396 use case where it's really mandatory, and maintaining it was a pain in
398 * New: xbt_fifo_search(), search an item with a user-provided
399 comparison function instead of dumb pointer comparison.
402 * Fix the lua deployment:
403 Use `simgrid.init_application()` before deployment instead of
404 `simgrid.msg_register_application()` after.
407 * Transfer the tracing files into the corresponding modules.
409 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
411 SimGrid (3.8.1) stable; urgency=low
413 The "we are told that some people want to also *install* the simgrid
416 * Add missing file "tesh.1" to the archive.
418 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
420 SimGrid (3.8) stable; urgency=low
422 The Psssshiiiit release: SimGrid jumps into the Cloud.
425 * Add an experimental interface to manipulate VMs. They are mainly
426 process groups with very few intrinsic semantic, but they should
427 allow you to build the semantic you want easily.
428 * New function: MSG_host_set_property_value()
429 * New function: MSG_process_on_exit(). To clean memory in all cases.
430 * Bug fixes that made the host (and link) failures unusable.
431 * Add a way to auto-restart process when the host in which they are
432 executing comes back (ON_FAILURE="RESTART" on deployment file,
433 MSG_process_auto_restart_set).
434 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
435 please stop using the old ones, they are DEPRECATED.
437 * Deprecate functions MSG_global_init() / MSG_global_init_args()
438 Please use MSG_init() instead. (reducing the amount of entry
439 points in the library helps us).
440 * Make it impossible to link against the wrong version of the lib
441 * Deprecate MSG_clean(). No need to call it anymore.
442 * Function MSG_get_host_number() is not deprecated anymore.
445 * Split the doc into a user guide and a reference guide.
446 * Start a developper guide to help people hacking on SimGrid.
449 * Enable tracing by default. This modules rocks you should use it.
450 * Remove option custom_flags. Now use environment variables CFLAGS
452 * Use default cmake things to detect lua instead of home grown ones.
453 * New option "enable_mallocators" to disable mallocators, for debugging
454 purpose ("on" by default).
457 * Bug fixes around the resource failures: don't let the processes
458 survive the host they are running onto.
459 * Add an interface to auto-restart processes when the host in which they are
460 executing comes back.
461 * Ensures that SIMIX_clean is called automatically. It's not part of
462 the public interface anymore (bindings should be updated).
465 * Bug fix for when SD_Simulate is called with a positive value: be careful
466 when comparing doubles. Sometimes they are different for non significant
468 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
469 parallel task whose initial work is distributed among host according
470 to the Amdahl's law. Such tasks are created with a parameter alpha
471 that corresponds to the non-parallelizable part of the computation.
472 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
473 between two sets of workstations assuming a 1D block distribution (each
474 workstation owns a similar share of data) on both sides.
476 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
477 Data redistribution will be automatically scheduled once parent and child
478 are both scheduled. The filling of computation_amount and
479 communication_amount structures is now done seamlessly thanks to the chosen
481 * New function SD_workstation_dump to display various information
482 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
483 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
484 bandwidth on the route onto which the task is scheduled and the amount of
486 To divide the nominal bandwidth by 2, the rate then has to be :
487 rate = bandwidth/(2*amount)
488 * Compute tasks that have failed can now be rescheduled and executed again
489 (from their beginning)
490 * Increasing source code coverage (src/simdag is now covered at 95.8%
494 * Re-implement time-independent trace replay using SMPI (at the
495 smpi_smp_* level) instead of MSG. This should replace
496 examples/msg/actions/actions.c
497 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
498 hindexed and structs)
499 * Implement the exchange of non-contiguous data.
500 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
501 * Correct behavior of smpi/sender_gap and set its default value to 0
502 * Add option to asynchronously send small messages to allow better
503 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
504 specifies the size in bytes under which messages will be asynchronously sent.
505 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
506 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
509 * New C interface to define a platform: XML is now optional.
510 For more info, please check include/simgrid/platf.h.
511 * New interface to define random platforms from the C:
512 For more info, please check include/simgrid/platf_generator.h and
513 examples/msg/masterslave/masterslave_platfgen.c
514 * Export a sg_cmdline dynar containing all the arguments we got from
518 * Two new tracing options for adding comments to trace file so you
519 can track your experiments (see --help-tracing for details).
520 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
521 * Adding the SimGrid version that generated the trace file as a comment.
522 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
523 * Fix to avoid key clashes on Paje links
524 * Other minor fixes related to the Paje specification
527 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
528 and renamed to xbt_str_hash() and xbt_str_hash_ext().
529 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
530 * Greatly improve the robustness of mmalloc to user errors (such as
531 using an area after freeing it, or freeing it twice)
533 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
535 SimGrid (3.7.1) stable; urgency=low
538 * Restore the prototype of MSG_process_create_with_environment() to
539 the pre-3.7 situation by removing the kill_time argument.
540 * Add a MSG_process_set_kill_time() function instead.
543 * Fix weird behaviors when dealing with parallel tasks.
546 * Simgrid is now built as a dll.
547 * Simgrid-java now works on Windows.
548 * Simgrid-Java is now included into Windows package.
551 * First pre-build package for MacOSX.
554 * Fix compilation when using MSG_USE_DEPRECATED.
555 * Fix some compilation issues on Macs and Windows.
556 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
557 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
559 We discovered that the Lua console is broken, but we are missing the
560 manpower to fix it right now. The problem existed in 3.7 too, so we
561 are not blocking the release for that. Sorry if you depended on this
562 feature, any help would be really welcome.
564 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
566 SimGrid (3.7) stable; urgency=low
568 The "spring cleanups (before the next Big Project kicks in)" release.
571 * We can specify the SMPI latency/bandwidth factor with command line
572 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
573 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
574 You can also use the "config tag" from platform file by adding this line
575 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
576 (see "example/platforms/tag_config.xml" to use "config tag").
577 Note that the command line supersedes the platform file configuration.
578 * Change the correction factors used in LMM model, according to
579 the latest experiments described in INRIA RR-7821.
580 Accuracy should be improved this way.
581 * Use the partial invalidation optimization by default for the
582 network too. Should produce the exact same results, only faster.
583 * Major cleanup in surf to merge models and split some optimization
584 mechanisms from the core of the models. As a result you can now
585 specify which model to use (e.g., --cfg=network/model:LV08
586 --cfg=cpu/model:Cas01) and which optimization mode to use
587 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
588 Incompatible combinations should err at initialization. See
589 --help-models for the list of all models and optimization modes.
590 * The CLM03 workstation model was dropped for simplicity because it
591 used the deprecated CM02 network model. Use default instead.
592 * Rename the TCP_gamma configuration option to network/TCP_gamma
593 * Rename the coordinates configuration option to
594 network/coordinates, and document it
595 * Use now crosstraffic keyword instead of the terribly misleading
596 fullduplex keyword. It is activated by default now in the current
597 default model, use --cfg=network/crosstraffic:0 to turn it off.
598 * Ongoing refactoring the model parsing to make XML files optional
599 See include/simgrid/platf.h for details (still to be completed)
602 * Major overhaul of the documentation. Almost instructive now :/
603 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
604 functions and friends. This interface was considered as
605 deprecated since over 2 years, it's time to inform our users that it is.
606 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
607 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
608 use these (crappy) functions in your code.
609 These functions will be removed soon. Stop using them now.
610 * Deprecate MSG_get_host_{table,number}
611 Implement MSG_hosts_as_dynar() instead.
612 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
613 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
616 * Stabilize the parallel execution mode of user contexts
617 * Introduce configuration variables to control parallel execution:
618 - contexts/synchro: Synchronization mode to use when running
619 contexts in parallel (either futex, posix or busy_wait)
620 - contexts/parallel_threshold: Minimal number of user contexts
621 that must be part of a scheduling round to switch to parallel
622 execution mode (raw contexts only)
623 * Fix bugs that prevented to use suspend/resume along with
624 synchronization structures.
625 * Fix bugs in process termination that lead to invalid memory access
626 in very specific conditions.
629 * Introduce a parallel mode for the models (controlled by surf/nthreads
630 configuration item). In our tests, running the models in parallel
631 never lead to any speedups because they are so fast that the gain
632 of computing each model in parallel does not amortizes the
633 synchronization costs, even when ultra fast futexes are used.
634 This is released anyway because YMMV.
637 * Performance boost by using a swag internally to compute the set of
638 tasks that are finished and should constitute the return value of
642 * Enable it by default now that it is considered rather stable.
645 * Documentation of the tracing functions.
646 * Performance gains when tracing categorized/uncategorized resource
647 utilization by avoiding calls to get route when updating resource
648 variables. LMM constraints are being used instead.
649 * API changed to set task categories. Use MSG_task_set_category instead
650 of TRACE_msg_set_task_category, and SD_task_set_category instead
651 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
652 * Bugfix for graphicator, routes not correctly obtained, memory leaks
653 * Examples for link user variables added (see at examples/msg/tracing/)
654 * Deprecated function TRACE_msg_set_process_category completely removed
655 * Trace header updated according to the latest Paje file format
656 * Tracing network lazy updates, no longer obligate users to use full updates
657 * --cfg=tracing/platform:1 also registers power/bandwidth variables
658 * Experimental: let user code declare/set/push/pop application states for hosts
659 * API changed to allow the manual creation of graph configuration files
660 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
663 * Improve the API of Lua MSG bindings, using the Lua spirit.
664 * Each simulated process now lives in its own Lua world (globals are
665 automatically duplicated). It helps writing simulators. Will allow
666 to run Splay programs within SimGrid in the future.
667 * Add a Chord example in Lua, equivalent to the MSG one.
670 * Start the implementation of a solution to express temporal
671 properties, not only local assertions. This is still an
672 experimental work in progress, stay clear from it to be safe.
676 - Add new runtime parameters --help-logs and --help-log-categories
677 to display informations about supported logging parameters and
679 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
680 don't exists anymore.
681 * Mallocators: allow value NULL for the reset function.
683 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
684 create homogeneous dictionaries, where all the elements share the
685 same free function. Non homogeneous dictionaries will be
686 deprecated in the next release.
687 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
688 - Multi-level dictionaries are deprecated.
690 - new function xbt_dynar_search_or_negative() that is useful when
691 you have less than 2 million elements in your dynar and don't
692 want of the extra complexity of catching exceptions when the
693 element is not found.
695 - Make xbt_os_thread module (for thread portability) public.
696 Documentation is still to come, sorry.
698 - Cleanups and simplifications to make it maintainable again.
699 - Exotic features (such as memalign and valloc) were removed.
700 - The metadata were extended and improved so that the
701 model-checker becomes able to explore and inspect the heaps.
702 - This may induce a performance drop when enable_model-checking is
703 ON in cmake (even if it's not used in the simulation), but it is
704 necessary at this point to get MC working.
706 Turn model-checking OFF if simulation performance matters to you.
707 Not enabling it at runtime is not enough, disable it in cmake.
709 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
711 SimGrid (3.6.2) stable; urgency=low
713 The "Not coding new stuff allows to polish old things" release.
716 * New bindings to the NS3 packet level simulator (experimental)
717 * Use the raw (efficient) execution contexts instead of the sysv
718 (portable) ones when possible.
719 * libpcre is now mandatory in any cases since not using it led to
720 severe performance loss and possibly other issues
721 * Update the XML platforms:
722 - G5K: include the latest machine in Nancy
723 - GridPP and LCG: new platforms
724 * Documentation was partially updated, at least (more to come)
726 Bug fixes, cosmetics and small improvements
727 * Free terminated processes before the end of the simulation to avoid
728 exhausting the memory of users having very dynamic amount of
730 * Bug fix and cosmetics about canceling non-running tasks
731 * Bug fix about the dot loader's issues when using libcgraph
734 * Create an installer for windows with nsis (amd64 and win32)
735 - Add an hello world project to illustrate simgrid project creation.
736 - Embed libpcre into the Simgrid installer to avoid
737 its compilation burden
738 * The raw execution contexts should work on Apple now
739 * Port to Windows 64 bits
740 - Sysv contexts now have an implementation for this arch
741 - GRAS communication features now support this arch
742 * Drop support for borland compiler on windows
743 - this code was not maintained, and we kinda depend on gcc nowadays
744 * Fix portability issues on kfreebsd/gnu: build error about semaphores
745 * Fix portability issue on unstable ubuntu: linker became picky on
748 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
750 SimGrid (3.6.1) stable; urgency=low
752 The "Oops, we broke Macs too" release
755 * Fixed contexts detection so that raw ones are used when possible
756 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
757 leads to a strange error, with user code segfaulting sometimes when
758 the generated code is not perfectly aligned (which is not
759 controllable from the user side, depends on the amount of code)
762 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
764 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
766 SimGrid (3.6) stable; urgency=medium
768 The Summer Release, also known as the "OMG! They Killed Kenny!" version
771 * Bindings now constitute their own package, separated from the main one.
772 Rationale: reduce our maintenance nightmare by reducing the module coupling
773 They will soon be released on their own on gforge.
775 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
776 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
778 GRAS: It is not considered as stable anymore, but experimental. Sorry.
779 * It's not quite deprecated for now because we have no replacement,
780 but it may soon become the case.
783 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
784 * Fortran: New user-level cache variable to store the rank of the running
785 process. This improves performance by an order of magnitude.
786 * C: New coccinelle script to automatically locate and modify global and
787 local static variables.
788 * Improved SMPI network model with a sender-side gap to account for multiple
792 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
793 only say if a communication is finished, no matter whether it succeeded or
794 failed. You can call MSG_comm_get_status() to know the status of a finished
796 * New function MSG_task_dsend() to send a task and detach it. When a
797 communication is detached, you are never notified of its success or failure
798 and the memory is released automatically once it is finished. This function
799 is useful when you don't care about the end nor the success of a
801 * Change the prototypes of action replay. Sorry for inconvenience,
802 but this is really more efficient this way (and to adapt your code,
803 you just have to fix the initialization, that shouldn't be too long)
804 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
805 ever managed to do anything useful with it.
806 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
807 takes a dynar of communications. It returns immediately and gives the
808 index of a finished communication (if any).
809 * New example: a basic implementation of the Chord P2P algorithm.
812 * New model for multi-core CPUs. You can now use the core attribute to
813 precise the number of cores of a host. This is a basic model. Every
814 process running on the host receives at most the power provided in
815 the DTD (throughput<=power). Total throughput of process cannot exceed
817 * New peer tag. This peer tag creates a tiny AS comprising a host and a
818 router linked by an up-link and a down-link (possibly asymmetrical).
819 This kind of pattern allows to easily build last-mile model style platforms.
820 Aggregating such patterns in a rule-based AS is thus the technique of
821 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
822 * New model for Vivaldi routing. We transformed the Vivaldi network model
823 into a Vivaldi routing model (based on the rule-based model). This allows to
824 combine Vivaldi based latencies with last-mile platforms.
827 * Added a check for NaN of IEEE754 infinite in the double entries of
829 * Introduce a new context factory "raw", highly inspirated from the
830 ucontext factory, but using manually crafted functions in assembly to
831 do the work in an efficient manner.
832 * Allow to change the used context factory at run time, not only at
833 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
834 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
835 contexts at runtime (only with raw contexts or ucontexts).
836 * Completely rewrote this module to allow parallel execution of user
837 processes. Use --cfg=contexts/nthreads:N to execute user processes
838 with N parallel threads (the default is 1, meaning no parallelism).
839 * Allow to decide dynamically between sequential and parallel modes.
840 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
841 processes in parallel only when their number is greater than or equal to P
843 * Added a check for NaN of IEEE754 infinite in the double entries of
847 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
848 won't output the state of processes when interrupted with Ctrl-C
849 * Add a new function xbt_dynar_to_array that transforms a dynar into a
850 NULL-terminated array. This may solve backward compatibility issues
851 due to the change to return type of SD_simulate. See also:
852 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
853 * Add new macros with variable number of arguments.
854 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
855 - in xbt/asserts.h: xbt_assert
856 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
857 - in xbt/ex.h: THROWF and RETHROWF.
858 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
860 * Change xbt_die() to accept a format string with arguments, just like printf.
861 * New data structure: xbt_lib_t, like a dict but more general and with better
865 * New configuration options
866 Options triva/categorized and triva/uncategorized can be used to generate
867 graph configuration files for Triva visualization tool.
868 * Configuration option tracing/platform is renamed to tracing/categorized
869 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
870 * New TRACE_user_link_variable interface:
871 User provides the name of the link and the tracing variable to attach to it
872 * the declaration of tracing categories must be done after the environment
874 * simpler tracing interface, just one way to declare categories
875 TRACE_category or TRACE_category_with_color, it is up to you
876 * links in the trace file are again identified by their names
877 * trace contains the full platform hierarchy exactly as declared using the ASes
878 * Options tracing/msg/[task|process]:1 groups the process by hosts
879 for both cases, tasks and processes must have names that are unique during the simulation
880 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
881 * The experimental option tracing/msg/volume is deprecated
882 its functionality may be reincorporated if needed
884 The tracing generates a trace file with unordered timestamped events,
885 because of the way the core simulator (surf) works. A script available
886 at the tools directory (fix-paje-trace.sh) can be used to put the events
887 in order. We have changed the tracing so it can generate ordered timestamped
888 events in the final trace, but depending on the simulator (and how much time
889 is simulated) that can lead to a huge memory utilization. It is deactivated
890 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
893 * Define a SIMGRID_VERSION macro in simgrid_config.h.
894 - We are trying hard to keep the API stable, but it may happen that
895 some things change (we're a research project after all, not a
896 nuclear plant operating system). If such things should happen, you
897 could rely on that macro to adapt.
898 - current value: 30600 for 3.06.00, aka 3.6
899 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
900 a number that can be compared with SIMGRID_VERSION.
901 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
902 assertions and verbose/debug logging events are disabled at compile time.
904 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
906 SimGrid (3.5) stable; urgency=medium
909 * New feature to any SimGrid-based simulator: Model-Checking
910 Check SIN#1 for more details.
913 * New Model SMPI (three-interval linear regression for correction factors)
914 See RR-7426, available at http://hal.inria.fr/inria-00527150
915 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
916 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
917 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
918 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
919 * See also src/smpi/README
924 - Tracing API changes: TRACE_start and TRACE_end should not be called
925 by user-code. They are automatically called by simulators created
926 with SimDAG, MSG and SMPI if the toolkit is compiled with
927 tracing_enabled=ON. Categories declaration and utilization remain the
928 same for MSG and SimDag.
929 - A function was added to the tracing API to declare categories with
931 - TRACE_category_with_color (char *category, char *color)
932 where color must be in the following format
933 "%f %f %f", red, green, blue
934 and red, green, blue are float values in the interval [0, 1]
935 - User can specify NULL as color parameter, or continue calling
937 On that case, the tracing system will define random colors
938 - The following command-line options are supported:
939 --cfg=tracing/filename:msg.trace
940 --cfg=tracing:1 (activate tracing, needed to use others)
941 --cfg=tracing/platform:1 (categorized resource use)
942 --cfg=tracing/uncategorized:1 (uncategorized resource use)
943 --cfg=tracing/msg/task:1 (task creation)
944 --cfg=tracing/msg/process:1 (process creation, migration)
945 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
946 --cfg=tracing/smpi:1 (SMPI interface tracing)
947 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
948 - examples of examples/msg/tracing updated
950 - DAXLoader and DOTLoader functions can generate tasks with categories
951 - A new function to attribute a category to SD tasks:
952 TRACE_sd_set_task_category (SD_task_t task, char *category)
953 * Tracing the MPI interface implemented by SMPI
954 - Collective operations are traced with states
955 - Point-to-Point operations are traced with states/links
956 - Tracing activated by a parameter "-trace filename" passed
957 to smpirun during execution (considering that simgrid
958 is compiled with tracing enabled)
959 - To run the simulation with gdb, the simulator
960 accepts --cfg=tracing/smpi:1 to trace SMPI
961 - tesh files to check if smpi tracing is ok
962 - See examples/smpi/NAS/DT-trace
963 * GTNetS tracing re-worked
964 - adaptation to the tracing system of GTNets to cope
965 with modifications regarding the fullduplex mode
966 - new tesh files to check if gtnets tracing is ok
969 * Asynchronous communications through the functions:
970 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
971 * New function: MSG_load_platform_script()
972 to make possible using a lua script instead of XML files to set up platforms
973 * New function: MSG_set_function
974 to associate functions to processes, used when bypassing the parser
975 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
977 Platforms: Add some more examples in examples/platforms
978 * Grid'5000: see www.grid5000.fr
979 * *_30000_hosts.xml: various huge files [mainly scalability testing]
982 * Change the XML format. This is a very important modification. SimGrid
983 3.5 introduces a new hierarchical format based on the notion of
984 Autonomous Systems. Compatibility with old format is ensured
985 through the perl script provided in the install bin directory
986 bin/simgrid_update_xml.
987 It is now possible to build platforms with specific routing
988 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
989 connect several platforms together. We will try to provide soon
990 set of realistic platforms exploiting these properties (have a
991 look at examples/platforms/ for the moment).
992 * Take the opportunity of the XML format change to be a good XML citizen:
993 rename link:ctn to link_ctn and similar changes (also dealt with by
995 * Add a new routing scheme (rule-based) using regular expressions. It
996 enables to have an extremely low memory footprint when the
997 underlying routing is simple and can be compactly described. You
998 need to have libpcre4-dev (perl regular expressions) installed if
999 you want to use this routing scheme.
1000 * Revive the cluster TAG and allow to easily and efficiently (both in
1001 term of memory and speed) connect clusters together. Have a look
1002 at teshsuite/simdag/platforms/ to see how this can be done. With
1003 this tag, you can create clusters with thousands of tasks at no
1004 cost (have a look at examples/platforms/).
1005 Note: clusters are implemented as ASes, so there is no need for an
1006 enclosing AS tag if you have only one cluster in your platform.
1007 * Add new generic functions in the public interface that allows the user
1008 to call SURF 'create_resource' methods from your code (same
1009 functionality as the XML bypass mechanism but with a much lighter
1011 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1012 that uses a piecewise linear approximation to produce better
1013 results when exchanging small messages.
1014 * Add a new parameter to handle correctly full duplex link and account
1015 for interferences between uplink and downlink communications
1016 (activate with --cfg=fullduplex:1).
1019 * Rename the SD_READY (all dependencies are satisfied and task is
1020 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1021 dependencies are satisfied) state.
1022 This prevents a confusion between the notion of "ready to schedule"
1023 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1024 simulated" (SD_RUNNABLE) used by the simulation kernel.
1025 * Change the way a task is considered as ready. Instead of removing
1026 dependencies when a task is done, a counter is decreased. This way,
1027 it is always possible to reach ancestors thanks to the
1028 SD_taks_get_parents function (even after the end of the simulation.)
1029 * Change the return type of SD_Simulate from (SD_task_t*) into
1030 xbt_dynar_t. This function was in handling a dynar internally and
1031 converted it into a NULL terminated array for historical reasons.
1032 * New function SD_dotload(char*) to load a DAG described in dot
1033 format. This loader and the corresponding examples require the
1034 installation of the graphviz library.
1035 * Fix a bug in the management of tasks of size 0 in the surf network
1036 models. This problem was only visible with SIMDAG and you should
1037 thus disregard results produced with earlier versions if you
1038 relied on this feature (some tasks were blocked because of this).
1039 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1040 even though your applications were doing only point-to-point
1041 communications and sequential computations. Now you can really use any
1042 model you want (of course, if you create real parallel tasks, which are
1043 not implemented in most models beside ptaskL07, this will abort).
1044 * Add an example that schedules a DAX on an heterogeneous platform
1045 using a Min-Min strategy.
1046 * New function SD_workstation_get_current_task() that returns the kind
1047 of task currently running on a workstation in the sequential access
1049 * Raise some warnings when unexecuted tasks remains at the end of the
1050 simulation. This is usually caused by cycles in the DAG.
1053 * New function: SIMIX_process_set_function() called by MSG_set_function
1054 * Change the underlying waiting queue in semaphores so that a process
1055 can wait on several of them simultaneously (as in waitany).
1056 * Fix the way to handle tokens in semaphores so that all access patterns
1057 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1058 * kill the dirty pimple SIMIX_message_sizes_output()
1059 Please use (proper) visualization instead
1062 * New data container: setset (set of sets of elements)
1063 * New module: mmalloc (mapped malloc, allowing to have several
1064 independent segments of malloc)
1065 * New function: xbt_dict_cursor_set_data()
1066 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1067 * New function: xbt_dynar_is_empty()
1068 * New function: xbt_fifo_get_last_item()
1069 * Fix xbt_dynar_shrink(): use the right element size.
1070 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1071 uninitialized areas during expand.
1072 * Fix semaphores: previous implementation was severely broken.
1073 * Use library init/fini functions for our initialization.
1074 - you can use logs and other feature as soon as you want in your
1075 code (even before the xbt_init / MSG_init)
1076 - xbt_exit is now a no-op and produce a warning when used.
1079 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1080 benefit from the latest improvement to the simulation kernel.
1081 * Kill measurement sockets for now. If you rely on them, sorry. This
1082 release is not for you. This feature will be reintroduced in the
1083 future, but we cannot delay the release any further.
1084 * New function: gras_msgtype_get_name().
1085 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1086 * Fix (at last) the pmm example: it should not randomly fail anymore.
1088 Build chain: bug fixes and overall polishing
1089 * Cmake is now stable enough. Hence, we killed the autotools.
1090 * Port to windows ( TM :)
1091 * Fix the 'make install' target.
1092 No need to use 'make install-simgrid' anymore
1093 * Introduce a 'make dist' target compiling a *source* archive
1094 'make package' compiles a binary archive
1095 * Compile java files only on need
1096 * Add --cd and --setenv command line options to tesh
1097 * Out of source builds are not fully supported yet, but we are close
1098 * Enable supernovae and optimization flags by default for our users
1101 * Add layer to set up environment directly from lua, without XML.
1102 * The effect of gras_stub_generator can be achieved through
1103 lua too (check examples/gras/console/ping_generator.lua)
1105 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1107 SimGrid (3.4.1) stable; urgency=low
1109 The "Polishing easter eggs is probably a good idea" release.
1110 This is a bug fixes release only.
1114 * Fix a bug preventing the tasks from begin garbage collected.
1117 * Fix a bug occuring when a host involved in a communication fails.
1118 This was not detected properly by the other peer involved in the
1119 communication. Now, it's reported as a network error.
1122 * Warn the user about loop dependencies in data flow of DAX files
1123 * Obey the control-flow dependencies of DAX files
1126 * Add option "enable_smpi" allowing to not compile SMPI.
1127 Probably useful for the (Mac) users experiencing a build error here
1128 * Improve the detection of lua5.1 and ruby1.8
1130 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1132 SimGrid (3.4) stable; urgency=low
1134 The "Easter in Cargese" release. Also known as (major changes):
1136 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1137 ~> bindings were greatly improved
1138 ~> new tracing infrastructure for better visualization introduced
1140 * the "Welcome to configury modernity" release.
1141 ~> we switched from autotools to cmake, and improved our cdash
1144 A more detailled list of changes follow (full detail in svn log).
1146 Java Bindings: Various Cleanups
1147 * (install java-gcj-compat-dev on debian-like to use them)
1148 * Remove put/get: no need to export deprecated interface in Java
1149 Use send/receive instead.
1150 * Cleanup the examples and add a README per directory
1151 * Remove example autoDestination (that's the only way to go now)
1152 * Remove example explicitDestination (was a plain copy of basic)
1153 * Make JniException a runtime exception, so that there is no need to
1154 declare the fact that you may encounter such a beast. I guess that
1155 nobody will ever want to survive such error.
1156 * Create specific errors for each MSG case of failure:
1157 host failure, transfer failure, timeout, task cancelled
1158 * Cleanup the exceptions that may get thrown by each function
1159 * Other internal cleanups in Java bindings. Performance still bad :/
1160 Ruby and Lua Bindings: create them
1161 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1162 * That's new and great, you should try them out.
1163 Same functionalities than Java bindings, only even less polished
1165 * Kill the useless "rate" argument of SD_task_get_execution_time()
1166 Everyone used to provide -1 as a value, it was not used, and the
1167 semantic of a possible use wasn't even clear.
1168 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1169 either as comm costs or compute costs to mean that there is no
1170 such thing for that specific task.
1171 * Add a SD_task_set_name() function
1172 * Fix SD_task_unschedule() on typed tasks
1173 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1174 * In DAX loader, accept useless 'level' attributes to <job> since
1175 LIGO DAGs have them (seem to be to ease graphical representation).
1177 * Add an example masterslave_mailbox.c using send/receive and not
1178 the deprecated put/get interface.
1179 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1180 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1182 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1183 (declare MSG_USE_DEPRECATED to still have the old name)
1184 * Add a MSG_task_set_data() function
1185 * About trace replay (see examples/msg/actions):
1187 - Allow to work with splitted trace files for each process
1188 Give the specific trace file as argument of each process,
1189 and call MSG_action_trace_run(NULL)
1190 You can still have one merged file for all processes.
1191 - Fix implementation of collective operations
1192 * Allow task_execute() on 0-sized tasks (closes #10063)
1194 * This is the first release of SimGrid where SMPI is not considered
1195 beta anymore (even if some corners should still be improved)
1196 * Port over the new SIMIX_network submodule (internal refactoring)
1197 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1198 * Implement more missing elements of the standard:
1200 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1202 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1203 MPI_Scan() MPI_Get_processor_name()
1204 - Added implementation of missing case for Alltoall (warning: it's
1205 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1206 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1207 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1208 SMPI_MPI_Allgatherv()
1209 * Bug fixes include:
1210 - MPI_Waitsome() was broken
1211 - Allow relative includes in smpicc
1212 - Command line cfg argument 'reference_speed' was ignored...
1213 - Some functions did not properly lead to auto-benching of user code
1214 - smpicc passes -O2 by default (just like openmpi one)
1216 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1217 * Bug fixes about timeouts during communications
1218 * add SIMIX_message_sizes_output() as a pimple to write to file the
1219 amount of messages per size. Use gnuplot to get histogram.
1220 Pimple because that's the only user-visible function of simix,
1221 defined directly in xbt.h (irk, sorry)
1223 - Add a SIMIX_sem_get_capacity() function
1224 - Fix interactions with processe resume/suspende
1225 - release_forever() was stupidly broken
1226 - Fix SIMIX_display_process_status() for processes in a semaphore
1227 - Make SIMIX_sem_block_onto() user-visible
1228 * Refactoring context stuff:
1229 - Use pseudo-OOP for better modularity
1230 - reimplement SIMIX_process_kill() without process_schedule() so
1231 that the latter can take as invariant that it is called from
1233 - Merge context_start into context_new for sake of simplicity
1235 * Add a Vivaldi network model, coded live during SUD'10 ;)
1236 * Rename configuration variables to start a hierarchy:
1237 o cpu_model -> cpu/model
1238 o network_model -> network/model
1239 o workstation_model -> workstation/model
1240 * New configuration variables:
1241 o network/bandwidth_factor: correction to bandwith
1242 o network/latency_factor: correction to latency
1243 o netwotk/weight_S: correction to the weight of competing streams
1244 * Add a long description to the models, that users can see with such
1245 argument on the command line: --cfg=cpu/model:help
1246 * --help-models display the long description of all known models
1248 * config: add the ability to set a default value after registration
1249 Does not override any previously set value (e.g. from cmd line)
1250 * dict: allow to have integer key and data.
1251 When so, you need to use the following functions
1252 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1253 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1254 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1255 In contrary to regular dicts, the key is not malloced before copy.
1256 Mixing scalar and regular elements in the same dict is not tested
1258 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1259 Tracing for Visualization:
1260 * SimGrid is now instrumented in order to generate a trace file for
1261 visualization analysis: to use it, need to compile SimGrid with the
1262 "tracing" option enabled, and instrument the program using SimGrid with
1263 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1264 (among other functions).
1265 * The instrumentation only traces the platform utilization for now
1266 * Documentation to use the tracing functions and how to analyze the
1267 traces with the Triva tool is written.
1268 * More information about: SimGrid FAQ (in the section Tracing Simulations
1271 * We moved to cmake as default build system. Autotools support will
1272 be dropped soon. Check the FAQ for more info about how to use it.
1273 * Greatly improved our cdash/ctest interactions
1274 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1275 * Added memory checking tests with valgrind; lot of memleak fixing.
1276 This may be the first release of simgrid with so few memory issues
1277 * Added code coverage tests.
1278 Our coverage is still improvable, but at least we see it on cdash.
1280 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1282 SimGrid (3.3.4) stable; urgency=low
1284 The "Desktop Grid needs love too" release (also called Xmas release).
1286 Models improvements:
1287 * Major speedup in the maxmin system solving by using lazy evaluation
1288 Instead of solving completely the maxmin system at each iteration,
1289 only invalidate (and recompute) the modified parts.
1290 This new feature is enabled in default models but you can try to
1291 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1292 * Cas01 IMproved as default CPU model
1293 This CPU model is the same Cas01 model, but it uses the
1294 maxmin-selective-update flag and a heap structure to manage
1295 actions on SURF kernel.
1296 It reduces the complexity to find the next action to finish and,
1297 consequently, it's faster than the old Cas01.
1298 This is the new default CPU model (Cas01).
1299 * Rename the old Cas01 model to Cas01_fullupdate
1300 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1301 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1302 * CpuTI (CPU Trace Integration)
1303 A new CPU model whose objective is simulate faster when using
1304 availability trace files.
1305 Instead of using a full featured, over engineered maxmin system for
1306 CPU modeling, this model does the pre-integration of traces files
1307 to calculate the amount of CPU power available, and so, executes
1308 faster than the old CPU models.
1309 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1310 * Use LV08 as default network model since it gives better accuracy
1311 for small messages and shouldn't change things for big ones.
1312 Use --cfg=network_model:CM02 to get the previous behavior.
1315 ******************************************
1316 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1317 ******************************************
1318 * The new CPU model may changes simulations!
1319 The point is that events occurring at the exact same timestamp
1320 are not scheduled in the same order with the old and new
1321 version. This may be enough to completely change the execution
1322 of simulations in some cases.
1323 * The new network model will change simulations!
1324 This new model is more realistic than the previous one, so you
1325 should consider redoing your old experiments with this model.
1326 Sorry for the inconvenience.
1329 * Introduce the supernovae compilation mode
1330 When compiled that way, the whole SimGrid (or almost) is put in a
1331 single compilation unit and compiled in one shoot.
1332 This is to help gcc which has difficulties to inline stuff from one
1334 The speedup seem to be above 15%, althrough more tests are needed on
1335 amd64 to confirm that gain.
1338 * Port of MSG's mailbox on top of SIMIX network
1339 The put/get mechanism was greatly simplified on the way.
1342 * New SIMIX network module. Provides:
1343 - Mailbox: rendez-vous mecanism to find with who you want to speak
1344 - Synchronous send/recv: easier and hopefully faster since the
1345 logic is handled in the maestro process directly now
1346 - Asynchronous send/recv: you dreamt of it? It's here now
1347 Too bad that nobody cared enough to propagate the change to MSG.
1348 * Add semaphores as SIMIX synchronization mechanism.
1351 * new function SD_daxload(char*) to load a DAX file
1352 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1353 * Introduce typed tasks. Specify its kind and cost at creation.
1354 At scheduling, just give where it should be placed, and the cost
1355 for each involved resource is automatically computed.
1356 Existing constructors so far (more to come of course):
1357 - SD_task_create_comm_e2e() for end-to-end communication
1358 - SD_task_create_comp_seq() for sequential computation
1359 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1360 * new function SD_task_dump() for debuging display
1361 * new function SD_task_dotty(task,FILE*) writing to file the info
1362 about the task in dotty format
1363 * SD_task_dependency_exists() can now cope with having one of its
1364 arguments NULL. If so, it tests whether the other argument has any
1366 * Add getters on list of preceding/following tasks:
1367 SD_task_get_parents(task) and SD_task_get_children(task)
1368 * Add getters on amount of workstations and list:
1369 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1370 * Add getter on task kind: SD_task_get_kind(task)
1371 * Update the start_time and finish_time of tasks on completion/failure
1372 * Bugfix: Remove task from state swags when destroyed
1375 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1378 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1379 dynar with one separate thread per value of the dynar.
1380 * Change the prototype of xbt_thread_create(), sorry.
1381 Added a boolean parameter indicating whether we want to join this
1382 thread (used in SG only for now)
1383 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1386 * GTNetS wrappers should now be usable again (and betterly tested too)
1387 * Fix a major regression from 3.2 where the timeout provided to
1388 MSG_task_put_with_timeout() was used as absolute time before which
1389 the comm should be done.
1390 * Start to fix the <cluster> tag.
1391 - Internal links should be good now (beside of the loopback, which
1392 use the private link instead)
1393 - paths to the external world is still rather broken
1394 - the <route:multi> tag is just broken. Actually that's brain-dead.
1395 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1396 to make it less stupid
1397 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1398 * Fix a source-level compatibility glitch from 3.2: after defining
1399 MSG_USE_DEPRECATED, you can use the old name
1400 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1401 * Allow to compile from the SVN with automake 1.11
1402 * Fix some problems when using the "start_time" tag in deployment XMLs.
1403 * Fix #8569: XBT/synchro.h has redundant declarations
1404 * Fix #8563: MSG return values and exceptions
1405 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1406 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1407 Thanks to Robson Peixoto.
1408 * Fix a vicious bug in dictionaries inducing that some elements were
1409 not freed on xbt_dict_free()
1411 Portability report of this version:
1412 * Main portability targets:
1413 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1414 on (amd64/i386/ia64)
1415 - mac leopard on i386
1416 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1417 but nothing critical.
1418 * Other platforms: windows, AIX and others were not tested for this release
1420 Timing report of this version:
1421 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1422 scenario parameters). From 8h to a few seconds in desktop grid settings.
1423 * Supernovae brings about 25% speedup on i386.
1425 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1427 SimGrid (3.3.3) stable; urgency=low
1429 The "Need for Speed" release.
1431 The timings done to validate the 3.3.2 were faulty.
1432 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1434 The problem was a conversion from a manually handled vector to
1435 xbt_dynar_t on the critical path.
1436 xbt_dynar_foreach calls functions, inducing stack management crap.
1438 We inlined these functions and xbt_dynar_foreach is now breath taking.
1439 We also inlined xbt_swag_belong on the way.
1441 Here are some approximate speedup measurements (on master/slaves
1442 simulations lasting between 10s and 20s each):
1443 3.3.1 -> 3.3.2: about same performance
1444 3.3.2 -> 3.3.3: 40% speedup
1445 3.3.1 -> 3.3.3: 40% speedup
1446 3.3.1 with inline patch -> 3.3.3: 30% speedup
1448 Our reading is that the refactoring which occurred in 3.3.2 made us
1449 suffer much more from the xbt_dynar_foreach low performance, but
1450 once we solved this, this refactoring proved to be very performance
1451 effective. From the 40% speedup, somehow, 10% are due to the
1452 inlining and 30% to the refactoring.
1454 That's a pitty that gcc cannot inline functions placed in other files
1455 alone. We have to choose between:
1456 - break the encapsulation (by putting private data structures and
1457 accessors in headers files to help gcc)
1458 - live with low performance
1459 - switch to a decent compiler such as icc (not quite possible).
1461 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1463 SimGrid (3.3.2) stable; urgency=low
1465 The "Simplicity does not preceed complexity, but follows it" release.
1467 The main contributors of this release were (lexical order):
1468 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1471 * Extract the routing logic into its own object.
1472 (was dupplicated in network.c and workstation_LV07.c;
1473 Allows to implement other ways of storing that info)
1474 => kill now useless network_card concept
1475 - Use dynar to represent routes (instead of void** + int*)
1476 - kill link_set (use surf_network_model->resource_set instead)
1477 - Add a command-line option to choose the routing schema to use
1478 - Add three new models:
1479 * Floyd (shortest path computed at initialization)
1480 * Dijikstra (shortest path recomputed all the time)
1481 * Cached Dijikstra (shortest path computed on need)
1482 All these models where contributed by Silas De Munck, and are
1483 described in his ICCS09 paper.
1485 * Simplify model declaration
1486 (less redirections, less function to write when defining a model)
1487 - Factorize stuff between models:
1490 surf_model_resource_set(model)
1491 surf_model_resource_by_name(model, name)
1492 - Unify the types of models in s_surf_model_t (using an union)
1493 - Embeed fields of common_public directly into s_surf_model_t
1494 - Rename model methods:
1495 action_free ~> action_unref
1496 action_change_state ~> action_state_set
1497 action_get_state ~> action_state_get
1498 - Change model methods into functions :
1499 (model)->common_public->action_use ~> surf_action_ref
1501 * Implement a generic resource; use it as ancestor to specific ones
1502 (allows to kill duplicated code in models)
1503 Drawback: timer command don't need no name nor properties;
1504 workstation_CLM03 don't need no properties
1505 (but I guess we can live with those few bytes wasted)
1507 * Improve the action object model
1508 - implement a constructor avoiding dupplicated code about field
1509 initialization in generic_action part.
1511 * Kill the SDP model: it has an external dependency, is deprecated
1512 in flavor of modern lmm models, and didn't compile since a while
1515 * Relocation of the context module from XBT to SIMIX.
1516 (the context were decoupled from the simix processes, duplicating a lot of code)
1517 => a lot of code was factorized
1518 - less overhead is introduced during scheduling
1519 - simpler API for the context factory
1520 - the logic for process creation,destruction and manipulation was simplified
1521 * Simplification of the s_smx_process_t data structure.
1522 => accesing the simix level data associated to a process is faster now,
1523 and the code is a lot more readable.
1526 * Implement some more MPI primitives:
1527 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1528 -implementation: Bcast: flat or 2-ary tree (default),
1529 Barrier: 4-ary tree,
1531 Allreduce: Reduce then Bcast
1532 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1533 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1534 Alltoallv: flat tree, like ompi
1536 * Add support for optimized collectives (Bcast is now binomial by default)
1537 * Port smpirun and smpicc to OS X
1540 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1541 Shout out if you used it.
1544 * Display the list of still queued messages in SG mode when existing
1548 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1549 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1550 * Add xbt_str_from_file(FILE*)
1551 * Add xbt_dict_get_key achieving a linear reverse search
1552 * Remove the context module
1554 Portability report of this version:
1555 * Main portability targets:
1556 - Linux(debian)/x86/context
1557 - Linux(debian)/x86/pthreads
1558 - Linux(debian)/amd64/context
1559 - Linux(debian)/amd64/pthreads
1560 On these, we still have the eratic breakages of gras/pmm and
1561 amok/saturate_sg reported in previous version. We still think
1562 that the tests are the cause of the fault, not the tested code.
1564 - Mac OSX Leopard/x86/context
1565 Still false negative in tesh autotesting.
1566 Smpi still fails, but this time because readlink does not accept -f
1567 Everything seems to work properly beside of that.
1570 - AIX version 5.3 (only tested contexts this time)
1571 Smpi still fails there because mktemp is not installed.
1572 Everything seems to work properly beside of that.
1574 I managed to compile it for the first time, but several breakages.
1575 Won't delay the release for this exotic platform.
1577 * Windows: it's still lagging behind. If you want to help, please
1580 Timing report of this version:
1581 This version seem to be more than 5% faster than 3.3.1 (on linux
1582 64bits with contextes). The gain is less than expected, we are
1583 investigating this for next release.
1585 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1587 SimGrid (3.3.1) stable; urgency=low
1590 * Implement a --cfg-help to show existing configuration variables
1591 * Build chain do not require doxygen in maintainer mode
1594 * fix a bug on struct sizeof computation, which prevented the
1595 exchange of arrays of structs in some conditions
1596 - added a regression test about this in datadesc_usage
1597 * Allow the exchange of 0-long dynamic vectors.
1598 - for that, use -1 as indicator of dynamic size instead of 0
1599 - This implied to change any size from unsigned long to long,
1600 reducing a bit communication abilities, but I guess that with
1601 64bits being quite common, this is more than enough.
1602 - This also induce a protocol change, thus bumping network protocol
1603 version from 0 to 1 (if we have external users, we have to get
1604 clean on that point too ;)
1605 - added two regression tests about this in datadesc_usage
1606 * Be more verbose when propagating local exceptions
1607 This helps debugging.
1608 * Display the status of simulated processes when receiving SIGINT in
1612 * Allow to control the simulation from a trace file.
1613 New functions MSG_action_register() and MSG_action_trace_run()
1614 The first one allows to associate a function execution to each
1615 kind of action while the second one parses a trace file and
1616 triggers the corresponding actions within the system.
1617 For now, only a toy example is provided in examples/msg/actions
1618 * Add an exemple of process migration in examples/msg/migration
1619 * Fix a bug in task exchange which broke MSG_task_get_sender()
1620 Add a teshsuite regression test for that.
1621 [Bug: if MSG_task_get_sender() is called after sender exit,
1623 * Fix a bug which prevented suspend/resume to work properly
1624 * Display the status of simulated processes when receiving SIGINT
1625 This fixes a regression of v3.3. due to the introduction of SIMIX
1626 * Bug fixing in failure management:
1627 - trace could not start by a failure at time 0
1628 - failure during communications were not working
1631 * Add SIMIX_process_set_name() to change the name of the current
1632 process in the log messages.
1633 * Store smx_hosts in a dict since we only retrieve them by name
1634 * Move the configuration infrastructure to surf
1637 * Move the configuration infrastructure to surf
1640 * Massive internal cleanups:
1641 - Store internal structures on processes instead of hosts (allows
1642 to have more than one process per host, in addition of being more
1644 - Cleanup the initialization/finalization process
1645 - Kill a whole bunch of unneeded synchronization:
1646 processes run in exclusive manner within the simulator
1647 - Move queues from global tables to process data fields
1649 - now accept -platform and -hostfile arguments
1650 - Pass the right rank value to processes according to the hostfile
1651 * Compile the examples by default, and use them as regression tests
1652 * Implement MPI_Wtime()
1653 * Change the reference speed to a command line option
1656 * TCP_gamma can now be specified as command line option using
1657 --cfg=TCP_gamma:10000000.0
1658 * Change the --surf-path cmd line option into --cfg=path:
1661 * Also include strbuff from xbt.h public header
1662 * xbt_ex_display(): do not free the exception after displaying
1663 This allows to do more with the given exception afterward.
1664 Users should call xbt_ex_free() themselves.
1668 Portability report of this version:
1669 * Main portability targets:
1670 - Linux(debian)/x86/context
1671 - Linux(debian)/x86/pthreads
1672 - Linux(debian)/amd64/context
1673 - Linux(debian)/amd64/pthreads
1674 These targets fail about 1/10 of times on gras/pmm, but we believe
1675 that this is because of the test, not because of simgrid.
1676 amok/saturate_sg fails even more rarely, and the test may not be
1679 - Mac OSX Leopard/x86/context
1680 The test suite still spits tons of errors because some obscure
1681 force prevents us from removing the temporary directories
1682 arguing that they still contain some metadata I've never heard of.
1683 Smpi fails because seq is not installed.
1684 Everything seems to work properly beside of that.
1687 - AIX version 5.3 (both contexts and pthread)
1688 Smpi still fails there because mktemp is not installed.
1689 XML inclusions seems rosty on AIX.
1691 * Windows: it's still lagging behind. If you want to help, please
1694 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1696 SimGrid (3.3) stable; urgency=high
1700 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1701 [Malek Cherier & Mt]
1703 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1706 Until now, GRAS were using MSG as an interface to SURF. It was
1707 quite difficult because both interface have several differences
1708 (MSG channels vs GRAS sockets were the most notable point).
1710 This also opens the gate to SMPI (which should occur soon) and speed
1711 up simulations by to 40% (even if it were not the main goal).
1713 **************************************
1714 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1715 **************************************
1716 The point is that events occuring at the exact same timestamp are
1717 not scheduled in the same order with the old and new version. This
1718 may be enough to completely change the execution of simulations in
1719 some cases. Sorry for the inconvenience.
1721 * Cleanup and upgrade the XML format to push further scalability
1722 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1724 * Improve the testing infrastructure with tesh. Now a very large part of
1725 the code is tested not only by being run but also by checking that the
1726 output match an expected output [Mt].
1728 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1729 is really less memory-demanding, which should allow you to use
1730 larger files in SimGrid [AL].
1732 * Inform valgrind about our contextes, so that it becomes usable
1733 with the default (and more effecient) version of SimGrid
1734 [contributed by Sékou Diakite, many thanks]
1737 * Introduce a listener thread in charge of receiving incoming
1738 messages from the network. It allows to overlap communication and
1739 computation but most notably, it removes some stupid deadlocks due
1740 to the fact that so far, a process could not send and receive at
1741 the same time. This made most non trivial communication schema
1743 * Convert the PIDs from long int to int to match the MSG ones (and
1744 linux ones too) [Mt]
1745 * New function: gras_agent_spawn() to launch a new process on
1746 current host. Only working in simulation for now. [Mt]
1747 * New function: gras_os_hostport() returning a constant form (ie,
1748 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1751 * Make the backtrace of exceptions more human readable [Mt]
1752 * New module: xbt/str [Mt]
1753 a ton of string utility functions (split, join, printf to a newly
1754 allocated buffer, trim, etc)
1755 * New module: xbt/hash [Mt]
1756 SHA1 hashing algorithm (more to come if needed)
1757 * New module: xbt/synchro [Mt]
1758 synchronization tools (mutex and conditions) working the same way
1759 in simulation and in real life (mainly useful for GRAS, but not
1761 * New module: xbt/queue [Mt]
1762 classical producer/consumer synchronization scheme
1763 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1764 (using the classical functions will get serialized) [Mt]
1765 * Make dictionary internal table dynamic. No need to specify its size
1766 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1768 * Make sure the log channels are organized as a tree under windows
1769 (because of ANSI C compatibility issue, any channel were child of
1770 root directly) [Mt].
1773 * Cleaned many thing in surf and fixed a few bugs [AL].
1774 * Add a nice command line configuration mechanism to compose models [AL].
1775 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1776 the previous one (KCCFLN05). It relies on something that looks like
1777 a max-min sharing mechanism but cannot be written as such. A new solver
1778 was thus designed [AL].
1779 * Add a new solver to lmm. Based on Lagrange optimization and
1780 gradient-based descent, it enables to efficiently maximise systems s.a
1782 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1784 This solver enables to propose two new network models for TCP Reno and
1785 TCP Vegas based on Low's work. These models still need to be fully
1786 tested though [Pedro Velho].
1789 * Bug fix in SD_simulate. Now the time bound given as argument is
1791 * Use the new parallel task model (ptask_L07) as default.
1792 * Use the SURF command line configuration mechanism.
1793 * 0-size tasks (for synchronization) should now work.
1795 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1797 SimGrid (3.2) stable; urgency=high
1801 We still experience issues on this platform, but we believe that at
1802 least MSG is usable.
1804 GRAS API BREAKAGE (for simplification purpose, sorry):
1805 * the gras_msgtype_by_name is not used anymore. Instead of
1806 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1807 you can write (and must)
1808 gras_msg_send(toserver, "request", &request);
1809 - If you still want to pass a gras_msgtype_t to the function (to cache
1810 the type and avoid the lookup time), use the gras_msg_send_() variant.
1811 - Impacted functions:
1812 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1813 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1814 * The callbacks are now expected to return 0 when everything went well
1815 (just like the main() function)
1817 GRAS new features and improvements:
1818 * New module mecanism where user code can use per process globals [Mt]
1819 This is similar to gras_userdata_*() functions, but for libraries. It
1820 factorize some code developped over and over in the examples and AMOK.
1821 It has still to be documented and used (only amok/peermanagement is
1823 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1824 loss when they were small enough to fit into the buffer and sent quickly
1825 enough so that they can all get received in one shoot.
1826 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1827 if not found. Use the *_or_null() variant for the old semantic.
1828 * In gras_msg_handle, do not discard messages without callback.
1829 They are probably messages to be explicitly awaited later (ie, proofs of
1830 mis-synchronization in userland since they are sent before being awaited)
1831 No big deal usually.
1832 * gras_socket_meas_send/recv: semantic changed!
1833 The numerical arguments used to be (1) the total amount of data to send
1834 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1835 messages. This was need for the fool willing to send more than MAXINT
1836 bytes on quite fat pipes.
1839 * Do really rename the hostmanagement module to peermanagement. [Mt]
1840 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1841 but this is rather new and this was documented in the module
1842 documentation (poor excuses, I admit)
1843 * Bandwidth measurement semantic changed! This follows the changes to
1844 gras_socket_meas_send/recv explained above.
1847 * A sequential mode has been added to the workstations. When a workstation
1848 is in sequential mode, it can execute only one task, and the other tasks
1849 are waiting in a FIFO. [Christophe Thiery]
1852 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1853 model for SIMDAG. [Christophe Thiery]
1854 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1855 equal to 0, which caused some bad behaviors in
1856 saturated_constraint_set_update. I now use a threshold mechanism like in
1860 * When running manually src/testall, you select specific units [Mt]
1861 testall is the result of our cunit mecanism, and should replace all
1862 the scripty thingy around since bash don't run easily on billware.
1864 * A mallocator system has been added. [Christophe Thiery]
1865 Mallocators allow you to recycle your unused objects instead of freeing them
1866 and allocating new ones.
1868 Documentation update:
1869 * FAQ reworking + New FAQs:
1870 - "Valgrind spits tons of errors!" [Mt]
1871 - "How to repport bugs" [Mt]
1872 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1873 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1874 - Communication time measurement within MSG [AL]
1875 - I experience weird communication times when I change the latency [AL]
1876 * GRAS tutorial [Mt]
1878 - an introduction to the framework and to the used communication model
1879 - an initiatic tour introducing the most proheminent features:
1881 . Lesson 0: Installing GRAS
1882 . Lesson 1: Setting up your own project
1883 o Part 2: Message passing
1884 . Lesson 2: Exchanging simple messages
1885 . Lesson 3: Passing arguments to the processes (in SG)
1886 . Lesson 4: Attaching callbacks to messages
1887 . Lesson 5: Using globals in processes
1888 . Lesson 6: Logging informations properly
1889 . Lesson 7: Using internal timers
1890 . Lesson 8: Handling errors through exceptions
1891 . Lesson 9: Exchanging simple data
1892 . Lesson 10: Remote Procedure Calling (RPC)
1893 . Lesson 11: Explicitely waiting for messages
1894 . Recapping of message passing features in GRAS
1895 - A HOWTO section containing:
1896 o HOWTO design a GRAS application
1897 More are due, of course. They will come latter. In the meanwhile, you can
1898 check the examples which are still here.
1900 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1902 SimGrid (3.1) stable; urgency=high
1906 There was a stack corruption somewhere, visible only when optimizing
1907 with these versions. [Vince]
1910 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1911 module that enabled to play with Directed Acyclic Graphs. It is built
1912 directly on top of SURF and provides an API rather close to the old
1913 SG. Some old codes using SG are currently under rewrite to check that
1914 all needful functions are provided. [Christophe Thiery]
1917 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1918 extension of the classical CLM03 model that gracefully handles
1919 failures. This is now the default model for MSG and GRAS. It doesn't
1920 handle parallel tasks yet though. [AL]
1921 * Bug fix: Weights were not correctly set in the network part.
1922 WARNING: This may have resulted in incorrect results with simulations
1923 where there are more than one flow on a given link. [AL]
1926 * After a (long ?) discussion on simgrid-devel, we have decided that the
1927 convention we had on units was stupid. That is why it has been decided
1928 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1929 WARNING : This means that all previous platform files will not work as
1930 such with this version! A warning is issued to ask users to update
1932 A conversion script can be found in the contrib module of the CVS, under
1933 the name contrib/platform_generation/surfxml_update.pl [MQ]
1936 * Bug fix: Processes were started in reverse order, wrt deployment file.
1937 WARNING: if your code relies on this bug, please fix it. [AL]
1938 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1939 being executed on two different locations. [AL]
1940 * Bug fix: Failures are now better supported thanks to Derrick's tests
1941 (there was many failure situations I hadn't thought of and that weren't
1942 correctly handled). [AL]
1943 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1947 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1948 Exception occurring on server-side are propagated back to client (!).
1950 API CHANGE: the callback changed their prototype. Change:
1951 int my_handler(gras_socket_t expeditor, void *payload_data) {
1953 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1954 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1956 * New! function: gras_msg_handleall to deal with all messages arriving
1957 within a given period.
1958 * New! function: gras_socket_server_range to get a server socket in a
1959 range of port numbers (ease to avoid port number conflicts) [MQ]
1960 * New! gras processes display their backtrace when they get a SIGUSR1
1961 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1962 Sweet to debug RL processes [MQ]
1966 - Do not force experiment sizes to be expressed in kb, or it becomes
1967 impossible to measure the latency this way (needs one byte-long tests)
1968 WARNING: this changes the amok_bw_* function semantic. [MQ]
1969 - Implements the link saturation stuff. [MQ]
1970 * Peer management module:
1971 New! module factorizing code that we wrote over and over [MQ].
1974 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1975 - Test units are placed directly into the library code, they get extracted
1976 automatically and placed into the src/testall binary.
1977 - Convert most of the XBT tests to this system.
1978 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1979 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1980 live in the same C code. This required to change a little bit the API
1981 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1982 * New module: sparse graph structure with basic algorithms (this is work
1983 in progress and the API is not considered to be frozen yet). [AL]
1984 * Display more information on backtraces: source line & function names are
1985 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1986 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1987 * Reworked a little bit some #include statements to load only required
1988 headers. Some user code that relied on SimGrid to include stdlib or
1989 stdio may need to include it by themselves. [AL]
1990 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1992 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1993 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1994 but I guess that almost nobody used those parts. [MQ]
1996 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1998 SimGrid (3.0.1) stable; urgency=low
2001 * Unfortunately, I had missed 5 misnamed functions:
2002 xbt_fifo_item_t xbt_fifo_newitem(void);
2003 void xbt_fifo_freeitem(xbt_fifo_item_t);
2004 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2005 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2006 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2007 They're now deprecated. Please use their new versions:
2008 xbt_fifo_item_t xbt_fifo_new_item(void);
2009 void xbt_fifo_free_item(xbt_fifo_item_t);
2010 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2011 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2012 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2014 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2015 * Documentation: xbt_log module unmercifully reworked [MQ]
2016 * Bugfix: there was a problem with the ending of contexts with
2017 the pthread backend. It caused some weird deadlock or behavior
2018 depending on the pthread implementation. [AL]
2019 * Bugfix: get the exceptions raised in the simulator repport where
2020 and why they come from when they are not catched in time [AL, MQ]
2023 * Bugfix: Do repport the error when two non-connected hosts try to
2024 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2027 * Add additionnal checkings on communications. Assert that two
2028 communicating hosts are connected by a set of links... [AL]
2031 * Add additionnal checkings on channel values in communication [AL]
2032 * New: MSG_task_get_source to see on which host a task was generated [HC]
2033 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2034 the number of tasks waiting to be received on channel and sent
2036 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2037 waits for the first task coming from a given host.. [AL]
2039 GRAS new functionnalities: [MQ]
2040 * Enhance the parsing macro to allow the size of multidimentional objects
2041 to be given thru annotations.
2042 * New example (and documentation): Matrix Multiplication a la RPC
2043 (as when I was young!) and fix a bunch of bugs found on the way.
2045 GRAS performance improvements: [MQ]
2047 * Reduce the amount of cbps creation/destruction by making it static to
2048 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2051 * Change libdata to a set so that we can search for stuff by ID (and thus
2052 reduce the insane amount of dict lookups)
2055 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2056 RL communication on top of SG since it's so uneffective.
2057 It may also be used for inter-thread communication in RL, one day.
2058 * Use gras_datadesc_copy() to exchange messages on top of SG
2060 - improve message exchange performance on top of SG
2061 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2062 * Don't exchange on the network the size of the used part of buffer,
2063 instead, specify the possible buffer size to read().
2065 - reduces the amount of read/write calls (one pair per exchange)
2066 - reduces the amount of exchanged data (the size)
2067 - allows to retrieve all arrived data on receiver side, if we don't need
2068 it right now (subsequent read will peek the buffer)
2069 - allows the receiver to proceed with the begining of the stream before
2070 everything is arrived
2071 - make it possible to build an iov transport (using readv/writev)
2073 - take care of the data with non-stable storage (like stacked data),
2075 * If possible, TCP send uses vector I/O (when writev() is here)
2076 - Don't use it for receive since we send data sizes and data on the
2077 same stream, so we wouldn't be able to chain large amount of chunks
2078 before having to flush the stuff to read the size.
2079 * Rework the transport plugin mecanism to simplify it and reduce the
2080 amount of pointer dereferencement when searching for the right function
2083 * I guess that now, we do almost as few system calls as possible while
2084 doing as few data copy as possible.
2086 To improve it further, we could try to send all the sizes first and then
2087 all the data (to use iov on receiving size), but it's only a partial
2088 solution: when you have 2 dimensional data, the sizes of the second
2089 dimension is data of the first dimension, so you need 3 streams.
2091 I'm not sure the potential performance gains justify the coding burden.
2093 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2095 SimGrid (3.00) stable; urgency=high
2098 * New! Give the possibility to hijack the surf parser and thus bypass
2099 MSG_create_environment and MSG_launch_application. Have a look at
2100 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2102 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2104 SimGrid (2.96) unstable; urgency=low
2109 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2110 This deprecates the xbt_error_t mecanisms.
2111 It modifies (simplifies) all XBT and GRAS API.
2112 MSG API keeps unchanged (exceptions raised by XBT are catched from
2113 within MSG and masked with existing error handling facilities)
2116 * New! Add a FATPIPE model. [AL]
2117 * New! Add a parallel task model. [AL]
2118 * New! Add automatically a loopback interface (in the default
2119 network model) if none was precised.
2122 * Bugfix: MSG_process_resume now works with the current running process.
2124 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2125 * Modification of MSG_task_get_compute_duration. Once a task has been
2126 processed, the value returned by this function is now equal to 0. [AL]
2127 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2128 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2129 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2130 of a m_task. For now, MSG_task_cancel only works with computation
2132 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2133 of the processor (in Mflop/s) regardless of the current load on the
2135 * API Change: use proper naming convention for MSG_getClock and
2136 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2138 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2139 This function changes the priority of a computation task. This priority
2140 doesn't affect the transfer rate. A priority of 2 will make a task
2141 receive two times more cpu power than the other ones. This function
2142 has been added to suit the needs of Nguyen The Loc and hasn't been that
2143 much tested yet. So if it fails, please report it and send me your code.
2145 * API Change: removed all functions and types that were marked "deprecated"
2146 since many months. Renamed MSG_global_init_args to MSG_global_init.
2148 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2150 SimGrid (2.95) unstable; urgency=low
2153 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2154 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2155 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2158 * Bugfix: Allow absolute paths for platform description files [MQ]
2159 * Bugfix: do free the variables after use. Leads to drastic performance
2161 * Implement max_duration (ie, timeouts) on resources [AL]
2164 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2166 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2167 the needed support (timeouts) [AL]
2170 * Implement measurement sockets. You can now get the bandwidth between two
2171 hosts thanks to AMOK (see below). [MQ]
2172 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2173 dynar over the network (yeah) [MQ]
2174 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2175 * simple/static token-ring example. [Alexandre Colucci and MQ]
2176 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2177 in place before [MQ]
2180 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2181 type parsing. "short" and "long" modifiers were also ignored; other
2182 modifier (such as reference level) are still ignored. [MQ]
2183 * Embeed the buffer size within the buffer itself on SG. [MQ]
2184 That way, send() are atomic and cannot get intermixed anymore (at least
2185 the ones which are less than 100k; bigger messages still have the issue)
2186 * Array size pushed by the field, not by the field type (or each
2187 and every long int will push stuff to the cbps) [MQ]
2188 * use select() to sleep since it allows to portably sleep less than one
2191 GRAS (minor cleanups)
2192 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2195 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2196 * type category 'ignored' killed as it was never used and were difficult
2198 * whether a type can cycle or not is now a flag, leaving room for more
2199 flags (as "ignored", if we feel the need one day ;)
2200 * Rename raw sockets to measurement sockets since "raw" has another
2201 meaning in networking community.
2204 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2205 and offers features not belonging to GRAS but that most applications
2206 need. One day, it may be a set of plugins loadable at runtime.
2207 * New module: bandwidth
2208 bandwidth measurement between arbitrary nodes running this module. [MQ]
2210 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2212 SimGrid (2.94) unstable; urgency=low
2214 The first beta release of SimGrid 3 !
2218 * Update the main page and the FAQ. Adding references to gforge.
2221 * Add a gras_os_getpid function.
2224 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2225 * Extend the logs so that they also print PID, hostname, date, ... if
2227 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2228 and kill the old version which were in testsuite/
2229 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2233 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2234 for users. Instead, each programming environment (GRAS, MSG,...) use it
2235 when needed to provide such a feature to users.
2236 Don't shortcut the mecanism or you will also shortcut the virtualization
2237 you need on the simulator.
2241 * Cleanups in configury with regard to compile optimization/warning flags.
2242 Also add -fno-loop-optimize to any powerpc since it's the optimization
2243 killing gcc (< 3.4.0).
2244 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2245 needed by MSG examples complications
2246 * Borrow configury beautifications from PHP
2249 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2250 warning (thanks loris for stumbling into this one).
2251 * Bugfix: stop loading private headers (gras_config.h) from the public
2255 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2256 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2257 RL would imply a DNS resolver, which is *hard* to do in a portable way
2258 (and therefore delayed).
2259 * Implement a real timer mecanism and use it in timing macros. This allows
2260 to avoid rounding errors and get a 0.000005 sec precision in timing
2261 macros. While I was at it, various cleanups:
2262 - allow to declare more than one timed section per file (fix a stupid bug)
2263 - move some private declaration to the right place
2264 - merge conditional execution and timing macros into emulation module
2265 - document the module
2266 - make sure the module cleanups its mess on gras_exit
2267 * Documentation improvements:
2268 - (new) how to compile applications using GRAS
2269 - (new) emulation support (timing macros)
2271 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2273 SimGrid (2.93) unstable; urgency=low
2275 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2278 - Use Paje properly where used. Still to be sanitized properly.
2279 - Portability fix: Add an implementation of the contexts using pthread
2283 - Add xbt_procname(): returns the name of the current process.
2284 Use it to show the current process's name in all logging.
2286 - fix detection of older flex version and the reaction, since we do
2287 depend on modern ones (we use lex_destroy)
2288 - Better separation of SG and RL in the libs: remove all simulation code
2289 from libgras. As a result, this lib is now only 200k when stripped.
2290 Some of the xbt modules may also be duplicated (two sets and such) and
2291 should be cleaned/killed before SG3.
2292 - Insist on using xlC on AIX because of weird problems involving gcc there.
2293 - Cleanup the make remote stuff. This is now done by scripts
2294 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2295 mainly for our private use, but we're working on changing them to user
2298 - Bugfix: flush the socket on close only if there is some *output*.
2299 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2301 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2304 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2306 SimGrid (2.92) unstable; urgency=low
2308 Alpha 3 on the path to SimGrid 3
2312 - New! First try of benchmarking macros.
2313 - New! First try so that gras_stub_generator generate deployment and
2314 remote compilation helpers.
2316 - Bugfix: Initialization fix in msg_test.
2320 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2322 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2324 - API Change: gras_os_sleep to take the amount of seconds as a double.
2325 Accepting an int was error prone since it was the only location where
2326 seconds were coded as such. It leaded to damn rounding errors.
2327 - Bugfix: Hard to belive that timers ever worked before this.
2329 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2331 SimGrid (2.91) unstable; urgency=low
2333 Alpha 2 on the path to SimGrid 3
2337 - Bug fix in the lmm_solver.
2339 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2340 through the function MSG_paje_output.
2341 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2342 - It is possible to bound the rate of a communication in MSG with
2343 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2344 to write it in the changelog).
2345 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2350 - Various cleanups to the autotools stuff
2351 - Begin to move Gras examples to examples/gras/
2352 - Let make distcheck work again (yeah!)
2354 - documentation overhauled using doxygen.
2355 gtk-doc-tools is dead in SimGrid now.
2356 - Automatically extract all existing logging categories, and add the list
2357 to the documentation (long standing one, to say the less)
2359 - Cleanup the known architecture table. Reorder the entries to group what
2360 should be, and use a more consistent naming scheme.
2361 (some of the test dataset are still to be regenerated)
2362 - New! Allow library to register globals on each process just as userdata
2364 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2365 lookup time (for now).
2366 Use it in msg and trp.
2367 This cleans a lot the internals and helps enforcing privacy of the
2368 headers between the gras components.
2369 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2370 - Bugfix: gras_os_time was delirious in RL.
2371 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2373 - Reenable GRAS now that it works.
2375 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2377 SimGrid (2.90) unstable; urgency=low
2379 Alpha 1 on the path to SimGrid 3
2381 * It is a long time since the last release of SimGrid. I'm sorry about
2382 that but as I had told you, I was rewriting a lot of things. I apologize
2383 to those who had been reporting bugs to me and that I had not answered.
2384 If your bug is still in the new version, please tell me. Here is a
2385 summary of the main changes.
2387 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2388 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2389 portability, speed, and a lot more but you'll figure it out later.
2390 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2391 I wanted to release the new MSG as soon as possible and I have
2392 broken GRAS, which is the reason why, for now, only MSG is fully
2393 functional. A laconic description of these projects is available
2394 in the documentation.
2396 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2397 kernel optimized for our needs (called SURF but only the developers
2398 should use it). Hence, MSG is now roughly 30 times faster and I think
2399 that by rewriting a little bit MSG, I could event speed it up a little
2400 bit more. Beside the gain in speed, it is also much easier to encode a
2401 new platform model with SURF than it was with SG. More to come...
2403 * REVOLUTION 3: I have tried to change a little as possible the API of
2404 MSG but a few things really had to disappear. The main differences
2405 with the previous version are :
2406 1) no more m_links_t and the corresponding functions. Platforms are
2407 directly read from a XML description and cannot be hard-coded
2408 anymore. The same format is used for application deployment
2409 description. The new format is described in the documentation.
2410 Have a look in tools/platform_generation. There is a tiny script
2411 that converts from the old platform format to the new one. Concerning
2412 the application deployment format, parsing the old one is tricky.
2413 I think most of you should however be able to convert your files. If
2414 it is really an issue, I can write a C code that does the conversion.
2416 2) the toolbox tbx does not exist anymore. We now have a library
2417 with much more data-structures but without the hash-tables (we have
2418 dictionaries that are much faster).
2420 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2422 *****************************************************************************
2423 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2424 * don't feel like converting the oldies. (Mt) *
2425 *****************************************************************************
2428 Version 2.90: "the long awaited one"
2429 - Finished rewriting and debugging MSG. Rewrote the documentation.
2430 - disable GRAS for now since it needs to be ported to the newest SG
2433 - Finish the port to windows (using mingw32 for cross-compile)
2436 - Main loop and datastructures of SURF. A cpu resource object is
2437 functional. Surf can thus be used to create cpu's with variable
2438 performance on which you can execute some actions.
2440 2004-11-15 Martin Quinson
2441 - Port to ARM. Simply added the alignment and size descriptions. Should
2442 work, but the ARM machines are so slow that I didn't had the opportunity
2443 to 'make check' over there yet.
2445 2004-11-15 Arnaud Legrand
2446 - Trace manager now written. It uses a heap structure and is therefore
2447 expected to be efficient. It may however be speeded up (particularly
2448 when many events occur at the same date) by using red and black
2449 trees. One day maybe...
2450 - Max-min linear system solver written. It uses a sparse matrix
2451 structure taking advantage of its expected use. Most operations are
2452 O(1) and free/calloc are called as few as possible. The computation of
2453 the minimum could however be improved by using a red and black tree
2456 2004-11-03 Arnaud Legrand
2457 - Rename every gras_* function that was in xbt/ to its xbt_
2459 - Add a heap and a doubly-linked list to xbt
2460 - Added a dichotomy to the dictionaries. make check works as well before
2461 so I assume that the patch is correct. I do not know however if things
2462 run effectively faster than before now. :)
2464 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2465 SimGrid, and the version number is bumped to 2.x
2467 2004-10-29 Martin Quinson
2468 - Introduction of the remote errors.
2469 They are the result of a RMI/RPC on the remote machine.
2470 ErrCodes being scalar values, you can't get the host on which those
2471 errors did happen. Extending the error mechanism as in Gnome is possible.
2472 No idea yet whether it is a good idea.
2474 2004-10-28 Martin Quinson
2475 - Interface revolution: the Starred Structure Eradication.
2476 I used to do typedef struct {} toto_t; and then handle *toto_t.
2477 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2478 - ???_t is a valid type (builded with typedef)
2479 - s_toto_t is a structure (access to fields with .)
2480 - s_toto is a structure needing 'struct' keyword to be used
2481 - e_toto_t is an enum
2482 - toto_t is an 'object' (struct*)
2484 typedef struct s_toto {} s_toto_t, *toto_t;
2485 typedef enum {} e_toto_t;
2486 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2487 s_toto_t) is private.
2489 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2490 so that it can set the variable to NULL. It was so for dicts and sets,
2491 it changed for dynars.
2493 - Fix a bunch of memleaks in dict_remove
2494 - Fix a bug in sg/server_socket opening: it failed all the time.
2496 2004-10-07 Martin Quinson
2497 - Speed up dynar lookup operation a bit.
2499 gras_dynar_get is dead.
2501 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2502 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2503 (which gives you the address of the stored data).
2505 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2506 copy of the data using an affectation to do the job and not a memcpy.
2508 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2509 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2511 It does not really speedup the dynar test because they are
2512 setting elements all the time (and look them seldom). But the dict does
2513 far more lookup than setting.
2515 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2517 2004-10-05 Martin Quinson
2518 - Allow to (en/dis)able the cycle detection at run time.
2520 Whether we should check for cycle or not is now a property of each
2521 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2522 datadesc_cycle_unset allow to remove this property when previously set.
2524 Note that the cycle detection is off by default since it impacts the
2525 performance. Watch the data you feed GRAS with ;)
2527 This property is hereditary. Any element embedded in a structure having it
2528 set have it set for the time of this data exchange.
2530 You should set it both on sender and receiver side. If you don't set it on
2531 sender side, it will enter an endless loop. If you forget on receiver
2532 side, the cycles won't be recreated after communication.
2534 - Header reorganization.
2535 Kill gras_private.h, each submodule must load the headers it needs.
2537 2004-10-04 Martin Quinson
2538 - Interface revolution: do not try to survive to malloc failure.
2540 Now, gras_malloc and friends call gras_abort() on failure.
2541 As a conclusion, malloc_error is not a valid error anymore, and all
2542 functions for which it was the only gras_error_t return value are
2543 changed. They now return void, or there result directly.
2544 This simplify the API a lot.
2546 2004-09-29 Martin Quinson
2547 - Re-enable raw sockets.
2548 Created by gras_socket_{client,server}_ext;
2549 Used with gras_raw_{send,recv}
2552 It should allow to kill the last bits of gras first version soon.
2554 This is not completely satisfactory yet (duplicate code with
2555 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2558 - Simplify transport plugin (internal) interface by not passing any
2559 argument to _server and _client, but embedding them in the socket
2562 2004-09-28 Martin Quinson
2563 - Finish the port to AIX.
2564 autoconf was my problem (segfault within the malloc replacement
2565 function. No idea why)
2567 2004-09-16 Martin Quinson
2568 - Fix some size_t madness on 64bit architectures.
2570 2004-09-08 Martin Quinson
2571 - Reduce the number of system headers loaded, overload some more system
2572 calls (such as malloc to cast the result of the system one, and work
2574 - Fix and reintroduce the config support
2576 2004-09-07 Martin Quinson
2577 - Source code reorganization to allow Arnaud to surf all over there.
2578 - Allow to document the logging categories.
2579 - Remove all uppercase from logging categories and useless cleanup in names.
2581 2004-08-18 Martin Quinson
2582 Version 0.6.2 (protocol not changed; API changed)
2583 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2584 gras_error_t), and NULL when not found. Functions expecting a msgtype
2585 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2586 hopefully usefull message.
2587 - Portability to prehistoric sparcs again
2589 2004-08-17 Martin Quinson
2590 Version 0.6.1 (protocol not changed; ABI not changed)
2591 - prealloc some buffers to speed things up
2593 2004-08-11 Martin Quinson
2594 Version 0.6 (protocol not changed; ABI expended)
2595 - The parsing macro can deal with the references, provided that you add
2596 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2598 2004-08-09 Martin Quinson
2599 Version 0.5 (protocol not changed; ABI changed)
2600 - Allow to off turn the cycle detection code in data exchange at
2601 compilation time. It should be at run time, but I'm short of time (and
2602 the config stuff is still broken). That way, we keep dict out of the
2603 critical path, which is good because the performance is poor:
2604 - search not dichotomial yet
2605 - dynar give no way to access their content and memcpy everytime
2606 - In composed data description (struct, ref and so on), stop foolness of
2607 keeping the subtype's ID, but store the type itself. This keeps sets out
2608 of the critical path, which is good since they rely on dynar and
2609 dictionnaries. The only loose of that is that we cannot detect the
2610 redeclaration of a structure/union with another content (but I'm not sure
2611 the code detected well this error before anyway). We still can detect
2612 the redefinition discrepancy for the other types.
2613 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2614 it breaks the code because of type-punning used all over the place).
2615 This breaks on all non-gcc architectures (for now).
2617 All those changes (plus the buffer of last time) allow me to gain 2 order
2618 of magnitude on cruel tests consisting of 800000 array of integers on two
2619 level of a hierarchical structure (200 secondes -> 4 secondes)
2622 - the selector of reference must now return the type it points to, not
2623 the ID of this type.
2625 2004-08-06 Martin Quinson
2626 Version 0.4 (protocol changed; ABI not changed)
2627 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2628 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2629 - Implement a buffer transport, and use it by default (it relies on tcp in
2630 real life and on sg in simulation).
2631 That's a bit hackish since I had a new field to the structure to store
2632 its data without interfering with the subtype ones. Inheritance
2633 is tricky in C. And that's a kind of reverse inheritance with one class
2634 derivating two classes. Or maybe a game with java interfaces. Anyway,
2635 that's damn hard in C (at least).
2636 Moreover, I got tired while trying to ensure plugin separation and
2637 genericity in SG mode. MSG wants me to do weird things, so let's go for
2638 cruel hacks (temporarily of course ;).
2639 See comment in transport_private.h:71
2640 - do not include all the _interface headers in private but in the files
2641 which really need them (to cut the compilation time when they are
2644 2004-07-26 Martin Quinson
2645 Version 0.3 (protocol not changed; ABI changed)
2646 - Major overhault of the datadesc interface to simplify it:
2647 - shorted the function names:
2648 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2649 - add a trivial way to push/pop integers into the cbps without malloc.
2650 This allows to make really generic sub_type description, which simply
2651 pop their size of the stack.
2652 - add a function gras_datadesc_ref_pop_arr() which does what users want
2653 most of the time: Declare a dynamic array (which pops its size of the
2654 stack) and declare a reference to it. Poor name, but anyway.
2655 - kill the post-send callback, add a post-receive one
2657 2004-07-23 Martin Quinson
2658 Version 0.2 (protocol changed; ABI changed)
2659 - add some testing for cpbs in the test cases, and fix some more bugs.
2660 This invalidate again the little64 data file, since I cannot regenerate
2662 - remove an awfull optimization in the logging stuff, allowing me to:
2663 - understand it again
2664 - learn gcc how to check that the argument match the provided format
2665 - fix all errors revealed by gcc after that
2666 - internal keys of dict are not \0 terminated. Deal with it properly in
2667 loggings instead of segfaulting when the user want to see the logs :-/
2669 2004-07-22 Martin Quinson
2670 - Fix some stupid bug preventing cbps (callback postit) from working
2672 2004-07-21 Martin Quinson
2673 - Some documentation cleanups
2674 - remove the useless last argument of msgtype_declare
2675 - rename the Virtu functions to fit into the 'os' namespace
2676 - move headers src/include -> src/include/gras/ and stop fooling with
2677 gras -> . symbolic link
2678 - make distcheck is now successful
2680 2004-07-19 Martin Quinson
2682 - Build shared library also
2683 - Install html doc to the right location
2684 - stop removing maintainer files in make clean
2685 - build tests only on make check
2687 2004-07-13 Martin Quinson
2689 - No major issue in previous version => change versionning schema
2690 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2691 corresponding dataset.
2693 2004-07-11 Martin Quinson
2695 - Get it working with any kind of structure (we can compute the padding
2696 bytes remotely for all the architectures I have access to)
2697 - Implement the structure parsing macro (still not quite robust/complete)
2698 - Improvement to the remote testing toysuite
2700 2004-07-10 Martin Quinson
2701 [autoconf mechanism]
2702 - get ride of a bunch of deprecated macros
2703 - actually run the test for two-compliment, not only compile it :-/
2704 - test whether the structures get packed (and bail out if yes. Damn.
2705 Alignment is a serious matter)
2706 - test whether the structures get compacted (but respecting the alignment
2707 constraints of each types)
2708 - test whether the array fields of structures can straddle alignment boundaries
2710 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2711 (took me 2 hours to find that bug, looking at the wrong place)
2712 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2713 before sending/receiving and are used to compute the offsets of fields
2714 - Given that padding size depend even on compiler options, keep track of
2715 alignment and aligned_size only for the current architecture. Not a big
2716 deal since we send structure fields one after the other (seems
2718 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2719 program, let it work on linux/gcc/little32. portability todo.
2721 2004-07-08 Martin Quinson
2722 - import and improve remote compilation support from FAST
2723 - make sure make check works on half a dozen of machines out there
2725 2004-07-07 Martin Quinson
2726 Let's say it's version 0.0.3 ;)
2727 - Implement conversions (yuhu!)
2728 - Let it work on solaris (beside conversion, of course)
2729 - Stupid me, using rand() to generate the conversion datatests in not wise.
2731 2004-07-06 Martin Quinson
2732 - Let make dist work, since I'm gonna need it to compile on remote hosts
2733 - Let Tests/datadesc_usage write the architecture on which the file was
2734 generated as first byte.
2735 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2736 architecture descriptions.
2737 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2738 on those architectures.
2739 - Optimization: send/recv array of scalar in one shoot
2741 2004-07-05 Martin Quinson
2742 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2744 - Plug a whole bunch of memleaks
2745 - each process now have to call gras_{init,exit}. One day, their log
2746 settings will be separated
2747 - Continue the code factorisation between SG, RL and common in Transport.
2749 2004-07-04 Martin Quinson
2751 - Redistribution between SG and RL.
2752 We wanna have to accept in SG, so move accepted related parts of RL in
2753 the common part. (more precisely, the dynar of all known sockets is no
2754 more a static in transport.c, but part of the process_data)
2756 [gras_stub_generator]
2757 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2759 2004-07-03 Martin Quinson
2760 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2761 them all up in one shot)
2763 - Fix the ugly case of reference to dynamic array.
2764 - New (semi-public) function gras_datadesc_size to allow the messaging
2765 layer to malloc the needed space for the buffer.
2767 - gras_socket_close now expect the socket to close (and not its address to
2768 put NULL in it after it). This is because the socket passed to handlers
2769 is one of their argument (=> not writable).
2771 - propagate the interface cleanup from last week in datadesc, ie remove a
2772 superfluous level of indirection. User pass adress of variable
2773 containing data (both when sending and receiving), and not of a variable
2774 being a pointer to the data. Let's say that I like it better ;)
2775 The price for that is constructs like "int msg=*(int*)payload" in
2776 handlers, but it's a fine price, IMHO.
2778 - Let it work in RL (yuhu)
2780 2004-06-21 Martin Quinson
2782 - porting SG plugin and SG select to new standards (works almost).
2783 - plug memleaks and fix bugs around.
2786 - cleanup the prototype of data recv and force users to specify when they
2787 want to handle references to objects. Test case working even for cycles.
2788 - plug memleaks. Valgrind is perfectly ok with this.
2790 2004-06-12 Martin Quinson
2792 - cleanup the separation between plugin and main code in plugin creation
2794 2004-06-11 Martin Quinson
2796 - Reput hook for raw sockets, needed for BW experiments
2797 - kill a few lines of dead code
2798 [Data description] Interface cleanup
2799 - gras_datadesc_by_name returns the searched type or NULL.
2800 That way, no variable is needed to use a type desc once, which makes
2802 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2803 two parameters were strings (field name, type name), leading to
2805 [Dicos] Interface cleanup
2806 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2807 This is consistant with the dynar API.
2809 2004-04-21 Martin Quinson
2811 - Porting to new standards.
2813 - interface cleanup.
2814 There is no bag anymore, no need to take extra provision to mask the
2815 pointers behind "ID".
2816 Better splitup of functions between files create/exchange/convert.
2817 This is still a bit artificial since convert and receive are so
2818 interleaved, but anyway.
2820 - add a queued message list to procdata (the ones not matching criteria
2822 - factorize some more code between SG and RL wrt procdata
2824 - use gras_exit in example to track memleaks
2825 - get rid of gs_example now that GS is properly integrated into gras
2826 - update run_test to integrate the latest tests (datadesc)
2828 - rename WARNINGn macros to WARNn since it prooved error-prone
2830 2004-04-19 Martin Quinson
2832 - register init/exit functions within gras module mechanism
2833 - send/receive function.
2834 Convertion is not implemented, but short-cutted if not needed.
2835 struct/array elements are sent one by one (instead of block-wise), but
2836 nobody really cares (yet). Get a prototype before optimizing.
2837 - tests (using a file socket) for DD send/receive on:
2838 - base types: int, float
2839 - array: fixed size, string (ie ref to dynamic string)
2840 - structure: homogeneous, heterogeneous
2841 - chained list, graph with cycle
2842 Believe it or not, valgrind is not too unhappy with the results. The
2843 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2847 - Bugfix when using a filename explicitely (instead of '-')
2849 2004-04-09 Martin Quinson
2851 - factorize more code between RL and SG in socket creation
2852 - Complete the implementation and tests of:
2854 o file (only in RL, and mainly for debugging)
2856 I lost 3 days to design a portable address resolver, and then decided
2857 that the prototype mainly have to run on my box.
2858 Addressing portability too early may be like optimizing too early :-/
2860 - use gras_init in the Tests instead of the crappy parse_log_opt
2861 (the latter function is removed)
2862 [Conditional execution]
2863 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2864 [Code reorganisation]
2865 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2866 Build examples against the RL library, since there is no way to disable
2867 its creation for now.
2869 For information, the beginning of coding on GRAS was back in june
2870 2003. I guess that every line has been rewritten at least twice since