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 - Rename MSG_task_set_compute_duration(t) to MSG_task_set_flops_amount(t)
7 - Rename MSG_task_set_data_size(t) to MSG_task_set_bytes_amount(t)
8 Rename MSG_task_get_data_size(t) to MSG_task_get_bytes_amount(t)
9 - Massive cleanups in the functions related to the energy
11 * Interface improvement:
12 - Rename Host.isAvail() to Host.isOn()
13 - Rename Process.currentProcess() to Process.getCurrentProcess()
17 - SIMIX_process_throw: raises an exception in a remote process
18 * Refactoring: Separate submodules
19 - libsmx: the public interface, as libc in a real system
20 - popping: the strange dance that converts a user request into a kernel handling
21 - smx_context_*: the virtualization mechanisms that embeed the user code
22 - smx_*: the handling of each simcalls
23 * Rename smx_action_t into smx_synchro_t, making explicit that these
24 things are used to synchronize processes with their environment.
25 For example, a communication is a sort of synchronization involving
26 the communicating processes (that may block until the exchange) and
27 the platform. The same can be said from computations, etc.
31 - 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.
32 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
33 - MPI_Type_set_name, MPI_Type_get_name
34 - MPI_*_c2f and MPI_*_f2c functions
35 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
36 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
37 - Activate a lot of new tests from the mpich 3 testsuite
39 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
40 - InfiniBand network model added : Based on the works of Jerome Vienne (http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf )
41 - When smpi/display_timing is set, also display global simulation time and application times
42 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
43 * Collective communications
44 - 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
45 - 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/).
46 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
48 - Replay now uses algorithms from wanted collective selector
49 - Replay can be used with SMP-aware algorithms
50 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
51 - 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.
53 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
54 - Correctly use loopback on fat-tree clusters
55 - Asynchronous small messages shouldn't trigger deadlocks anymore
58 - "Full" network optimization flag was broken since Surf++
59 - Better handling of precision flags in maxmin
60 - Fix bug causing sometimes "Impossible" errors
61 - Properly pass cluster properties to included hosts
62 * Improvement of the Energy plugin.
63 - Always update the consumption before returning that value
64 - New property: watt_off to denote the disipation when the host is off
67 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
68 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
69 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
71 * Tracing is now always enabled (no way to turn it out)
72 * Move headers around to sort them out on installed systems:
73 - instr/instr.h -> simgrid/instr.h
74 - instr/jedule/* -> simgrid/jedule
75 - simdag/datatypes.h was removed
76 - simdag/simdag.h -> simgrid/simdag.h
77 - msg/datatypes.h was removed
78 - msg/msg.h -> simgrid/msg.h
80 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
82 SimGrid (3.11) stable; urgency=low
87 * Normalizing pointers addresses tool for better diff between logs
90 * Add cloud examples using new VMs
91 - examples/msg/cloud/two_tasks_vm.tesh
92 - examples/msg/cloud/simple_vm.tesh
93 - examples/java/cloud/cloud.tesh
94 - examples/java/cloud/migration/migration.tesh
95 * Add java surf examples:
96 - examples/java/surfPlugin/surf_plugin.tesh
97 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
98 - examples/java/surfCpuModel/surf_cpu_model.tesh
99 * Add SMPI+MSG example:
100 - examples/smpi/smpi_msg_masterslave/
105 - msg task destroy cancel test
106 - msg_host on/off test
107 * Move all tests in testsuite to teshsuite (adding tesh files)
108 * Restructure teshsuites
109 - one folder for each kind of test
110 * Restructure AddTests.cmake
112 - structure the order of tests (with sections)
115 * Add virtual machine
116 - creation of a VM on a PM
117 - migration of a VM from a PM to another PM
119 - MSG_process_join(msg_process_t process, double timeout)
120 - msg_bar_t MSG_barrier_init(unsigned int count)
121 - int MSG_barrier_wait(msg_bar_t barrier)
122 - void MSG_barrier_destroy(msg_bar_t barrier)
123 - msg_as_t MSG_environment_as_get_by_name(const char * name)
124 * New option "msg/debug_multiple_use" to help debugging when a task is used
127 - Improvements and finalization of MSG_storage, MSG_file APIs and their
129 - Increase code coverage in test suites
133 * Protect context stacks against stack overflow. The number of protected memory
134 pages allocated on the top of each stack (1 by default) can be configured
135 with the new command line option --cfg=contexts/guard_size:<value>.
136 * Simcalls are now generated by a python script that
137 - generates files included by SimGrid
138 - checks that all the functions exist, and proposes prototypes
140 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
141 host_set_data, host_get_data
143 - simcall_process_join(smx_process_t process, double timeout)
144 * Fix bug where sleeping processing could not be suspended.
147 * Translate surf models from C to C++
148 - Generic classes for all models: Model, Resource, Action
149 - A generic interface for each kind of model (CPU, Network, Storage
150 Workstation, WorkstationVM)
152 * Translate surf routings from C to C++
153 * Add callbacks using sigc++ or boost::signals2
154 - Add callback functions for resource creation/destruction
155 - Add callback functions for action state change
156 - Handle Energy as a plugin
157 * Replace swag by boost::intrusive
158 * Add new routing models for clusters. For documentation, see
159 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
160 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
161 parameters for cluster tag
162 - Fat trees, with topology="FAT_TREE" and
163 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
164 - see examples/platforms/torus_cluster.xml and
165 examples/platforms/fat_tree_cluster.xml
169 * Hostfiles support host:nb_processes construct to deploy several processes on
171 * Collective communication algorithms should not crash if used with
172 improper number of nodes and report the error.
173 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
174 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
175 MPI_Dims_create, MPI_Cart_sub are supported.
176 * New interface to use SMPI programmatically (still depends on MSG for
177 some parts, see examples/smpi/smpi_msg_masterslave) :
178 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
182 * Global variables privatization in MPI executables is now performed at runtime
183 with the option smpi/privatize_global_variables (default:no).
184 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
185 dynamic libraries loaded by the application are not privatized (static
186 linking with these libraries is advised in this case)
189 * Options defined in XML work correctly now.
192 * New cmake option, enable_lib_in_jar, to control whether native libraries are
193 copied into simgrid.jar or not (ON by default). Use this option if you want
194 to reduce the size of the installed simgrid.jar, *and* the native libraries
195 are kept installed elsewhere.
196 * Surf binding with SWIG (code generated in maintainer mode only):
197 - plugin to handle callbacks
198 - CPU model only for the moment
201 * Supernovae build mode is definitively removed. It was used to improve
202 inlining and inter-module optimizations. It is nowadays superseded by
203 link-time optimizations commonly available in compilers.
204 * Update ns-3 find lib. Bindings for ns-3 should work again now.
205 * Add boost dependency for surf++
206 * Add new macro for tests
207 - ADD_TESH(name <tesh_args>)
208 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
212 - xbt_bar_t XBT_barrier_init(unsigned int count)
213 - int XBT_barrier_wait(xbt_bar_t barrier)
214 - void XBT_barrier_destroy(xbt_bar_t barrier)
215 * Make the xbt_os_time module public
217 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
219 SimGrid (3.10) stable; urgency=low
221 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
224 * Reintegrate Java to the main archive as desynchronizing these
225 package is not acceptable anymore (Java is now considered stable)
226 * Add explicit synchronization facilities through semaphores
227 * Bug fix: Task.setDataSize() only changed the C world, not the value
228 cached in the Java world
231 * Dramatically change the way files are handled. API and internals changed, but
232 this part of MSG was not considered as production grade either.
233 * Add explicit synchronization facilities through semaphores
234 * Add a new function MSG_host_get_process_list()
235 * Preliminary DVFS support (see examples/msg/energy/ for details)
238 * SMPI is now included directly in the libsimgrid as the windows
239 linker doesn't force us on splitting it anymore.
240 * Improvements of the SMPI replay tool:
241 - Most of the collective communications are now rooted in the same process as
242 in the original application.
243 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
244 used until now). Multiple data types can now be used in a trace.
245 - The replay tool now supports traces produce either by TAU or a modified
247 - Bug Fix: the compute part of the reduce action is now taken into account.
248 - Gatherv collective is now supported
249 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
250 of smpirun outputs time independent traces for the current run. One file
251 is created per process. If too many processes are simulated, this behavior
252 can be changed to one file for all processes by using the
253 tracing/smpi/format/ti_one_file flag
254 * smpirun generates the host file if needed (with given host count and platform)
255 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
256 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
257 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
258 --cfg=smpi/coll_selector:(mpich/ompi)
259 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
261 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
262 happening inside a collective SMPI call.
263 * Fix the behavior of complex data types handling.
264 * Make MPI_Wtime another synchronization point to take computations into
266 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
267 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
268 * Add all missing Fortran bindings, SMPI should work with Fortran 90
269 (no privatization of global variables yet)
270 * Preliminary DVFS support (see examples/smpi/energy/ for details)
273 * Verification of liveness properties is now available for SMPI applications
274 (in addition to MSG applications)
275 * Bugged examples using SMPI in examples/smpi/mc/
276 * Add --cfg=model-check/visited option. Allows the verification of infinite
277 programs. Detection of loops in the execution thanks to the system state
278 comparison and reduction of the state space to explore. Can be combined with
279 DPOR for safety properties.
282 * Allow to change SimGrid configuration (see --help) within the code
283 thanks to SD_config() as it can be done in MSG.
284 * Add a new function SD_task_set_amount() upon user request.
287 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
288 * Remove rule based routing (no more PCRE dependency)
289 * Add a limiter_link option to cluster tag, to specify a maximum reachable
290 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
291 * Add a loopback_bw and loopback_lat options to cluster tag.
292 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
293 routing. Now peers are encapsulated in an AS and have their own private
294 router but this is transparent.
297 * Our own implementation of getline is renamed xbt_getline, and gets
298 used even if the OS provide a getline(). This should reduce the
299 configuration complexity by using the same code on all platforms.
300 * New type: xbt_cfg_elm_boolean.
301 * Allow to use yes/no for boolean configuration options in the command line.
302 * Allow to disable SimGrid cleanups at exit from command line option.
303 There are situations where one may want a simulation to end with an exit.
304 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
305 annoying when scripting around the simulator. Adding a
306 --cfg=clean_atexit:no allows to circumvent this issue.
309 * Lots of memory leaks were corrected in this release.
310 There is no known memory leaks anymore, in all of our 600+ tests.
311 * New command line option --version, to get SimGrid version information.
312 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
314 * Supernovae builds are deprecated, and expected to be removed in the next
317 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
319 SimGrid (3.9) stable; urgency=low
321 The Grasgory release: GRAS is really dead now.
323 * Complete overhaul of the internal host structures scheme.
326 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
327 was considered as experimental and badly maintained since then.
328 * Keeping it was thus a trap to our potential users, that could take
329 it instead of MSG or SMPI by mistake despite is pity state.
330 * GRAS seems to have very few users (if any), and no one volunteered
331 to maintain it further. It also induces a lot of XBT code (for
332 portability sake), that must be maintained too.
333 * For all these reasons, we killed GRAS. If someone wants to revive it
334 in the future, don't cry, our git history still remembers of GRAS.
337 * Major overhaul. Merge our documentation again as time proved that
338 splitting it was really not helping our users.
339 * Further improve the developer documentation to help newcomers
340 hacking on SimGrid itself. The user documentation (and in
341 particular, the beginner documentation) is still in a sorry state.
344 * Now works on Windows too!
345 * Much more extensive test suite, from MPICH
348 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
349 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
350 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
351 * Bug fix: let task be scheduled when the last dependency to be solved is
352 a control dependency.
353 * Remove SD_load_environment_script function.
354 Use the C sg_platf function if you want to declare a platform
358 * New function: MSG_process_get_number()
359 * Old function documented: MSG_config()
360 * Remove MSG_load_platform_script function
361 Use the C sg_platf function if you want to declare a platform
365 * Change the default value of the TCP_gamma constant (maximal size of TCP
366 congestion window) to a more realistic 4MiB value. If you notice changes in
367 your simulation results, you can fall back to the previous 20k tiny window
368 by adding --cfg=network/TCP_gamma:20000 on command line.
369 * (Hopefully) fix a bug wrt periodic availability and state traces
370 * Bug fix: use default values at start when first event in availability/state
371 trace is not at time 0.
374 * remove the "new_" part of function name sg_platf_new_trace_connect
375 (resulting in sg_platf_trace_connect), since it does not create
379 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
380 use case where it's really mandatory, and maintaining it was a pain in
382 * New: xbt_fifo_search(), search an item with a user-provided
383 comparison function instead of dumb pointer comparison.
386 * Fix the lua deployment:
387 Use `simgrid.init_application()` before deployment instead of
388 `simgrid.msg_register_application()` after.
391 * Transfer the tracing files into the corresponding modules.
393 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
395 SimGrid (3.8.1) stable; urgency=low
397 The "we are told that some people want to also *install* the simgrid
400 * Add missing file "tesh.1" to the archive.
402 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
404 SimGrid (3.8) stable; urgency=low
406 The Psssshiiiit release: SimGrid jumps into the Cloud.
409 * Add an experimental interface to manipulate VMs. They are mainly
410 process groups with very few intrinsic semantic, but they should
411 allow you to build the semantic you want easily.
412 * New function: MSG_host_set_property_value()
413 * New function: MSG_process_on_exit(). To clean memory in all cases.
414 * Bug fixes that made the host (and link) failures unusable.
415 * Add a way to auto-restart process when the host in which they are
416 executing comes back (ON_FAILURE="RESTART" on deployment file,
417 MSG_process_auto_restart_set).
418 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
419 please stop using the old ones, they are DEPRECATED.
421 * Deprecate functions MSG_global_init() / MSG_global_init_args()
422 Please use MSG_init() instead. (reducing the amount of entry
423 points in the library helps us).
424 * Make it impossible to link against the wrong version of the lib
425 * Deprecate MSG_clean(). No need to call it anymore.
426 * Function MSG_get_host_number() is not deprecated anymore.
429 * Split the doc into a user guide and a reference guide.
430 * Start a developper guide to help people hacking on SimGrid.
433 * Enable tracing by default. This modules rocks you should use it.
434 * Remove option custom_flags. Now use environment variables CFLAGS
436 * Use default cmake things to detect lua instead of home grown ones.
437 * New option "enable_mallocators" to disable mallocators, for debugging
438 purpose ("on" by default).
441 * Bug fixes around the resource failures: don't let the processes
442 survive the host they are running onto.
443 * Add an interface to auto-restart processes when the host in which they are
444 executing comes back.
445 * Ensures that SIMIX_clean is called automatically. It's not part of
446 the public interface anymore (bindings should be updated).
449 * Bug fix for when SD_Simulate is called with a positive value: be careful
450 when comparing doubles. Sometimes they are different for non significant
452 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
453 parallel task whose initial work is distributed among host according
454 to the Amdahl's law. Such tasks are created with a parameter alpha
455 that corresponds to the non-parallelizable part of the computation.
456 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
457 between two sets of workstations assuming a 1D block distribution (each
458 workstation owns a similar share of data) on both sides.
460 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
461 Data redistribution will be automatically scheduled once parent and child
462 are both scheduled. The filling of computation_amount and
463 communication_amount structures is now done seamlessly thanks to the chosen
465 * New function SD_workstation_dump to display various information
466 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
467 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
468 bandwidth on the route onto which the task is scheduled and the amount of
470 To divide the nominal bandwidth by 2, the rate then has to be :
471 rate = bandwidth/(2*amount)
472 * Compute tasks that have failed can now be rescheduled and executed again
473 (from their beginning)
474 * Increasing source code coverage (src/simdag is now covered at 95.8%
478 * Re-implement time-independent trace replay using SMPI (at the
479 smpi_smp_* level) instead of MSG. This should replace
480 examples/msg/actions/actions.c
481 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
482 hindexed and structs)
483 * Implement the exchange of non-contiguous data.
484 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
485 * Correct behavior of smpi/sender_gap and set its default value to 0
486 * Add option to asynchronously send small messages to allow better
487 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
488 specifies the size in bytes under which messages will be asynchronously sent.
489 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
490 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
493 * New C interface to define a platform: XML is now optional.
494 For more info, please check include/simgrid/platf.h.
495 * New interface to define random platforms from the C:
496 For more info, please check include/simgrid/platf_generator.h and
497 examples/msg/masterslave/masterslave_platfgen.c
498 * Export a sg_cmdline dynar containing all the arguments we got from
502 * Two new tracing options for adding comments to trace file so you
503 can track your experiments (see --help-tracing for details).
504 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
505 * Adding the SimGrid version that generated the trace file as a comment.
506 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
507 * Fix to avoid key clashes on Paje links
508 * Other minor fixes related to the Paje specification
511 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
512 and renamed to xbt_str_hash() and xbt_str_hash_ext().
513 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
514 * Greatly improve the robustness of mmalloc to user errors (such as
515 using an area after freeing it, or freeing it twice)
517 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
519 SimGrid (3.7.1) stable; urgency=low
522 * Restore the prototype of MSG_process_create_with_environment() to
523 the pre-3.7 situation by removing the kill_time argument.
524 * Add a MSG_process_set_kill_time() function instead.
527 * Fix weird behaviors when dealing with parallel tasks.
530 * Simgrid is now built as a dll.
531 * Simgrid-java now works on Windows.
532 * Simgrid-Java is now included into Windows package.
535 * First pre-build package for MacOSX.
538 * Fix compilation when using MSG_USE_DEPRECATED.
539 * Fix some compilation issues on Macs and Windows.
540 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
541 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
543 We discovered that the Lua console is broken, but we are missing the
544 manpower to fix it right now. The problem existed in 3.7 too, so we
545 are not blocking the release for that. Sorry if you depended on this
546 feature, any help would be really welcome.
548 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
550 SimGrid (3.7) stable; urgency=low
552 The "spring cleanups (before the next Big Project kicks in)" release.
555 * We can specify the SMPI latency/bandwidth factor with command line
556 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
557 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
558 You can also use the "config tag" from platform file by adding this line
559 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
560 (see "example/platforms/tag_config.xml" to use "config tag").
561 Note that the command line supersedes the platform file configuration.
562 * Change the correction factors used in LMM model, according to
563 the latest experiments described in INRIA RR-7821.
564 Accuracy should be improved this way.
565 * Use the partial invalidation optimization by default for the
566 network too. Should produce the exact same results, only faster.
567 * Major cleanup in surf to merge models and split some optimization
568 mechanisms from the core of the models. As a result you can now
569 specify which model to use (e.g., --cfg=network/model:LV08
570 --cfg=cpu/model:Cas01) and which optimization mode to use
571 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
572 Incompatible combinations should err at initialization. See
573 --help-models for the list of all models and optimization modes.
574 * The CLM03 workstation model was dropped for simplicity because it
575 used the deprecated CM02 network model. Use default instead.
576 * Rename the TCP_gamma configuration option to network/TCP_gamma
577 * Rename the coordinates configuration option to
578 network/coordinates, and document it
579 * Use now crosstraffic keyword instead of the terribly misleading
580 fullduplex keyword. It is activated by default now in the current
581 default model, use --cfg=network/crosstraffic:0 to turn it off.
582 * Ongoing refactoring the model parsing to make XML files optional
583 See include/simgrid/platf.h for details (still to be completed)
586 * Major overhaul of the documentation. Almost instructive now :/
587 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
588 functions and friends. This interface was considered as
589 deprecated since over 2 years, it's time to inform our users that it is.
590 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
591 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
592 use these (crappy) functions in your code.
593 These functions will be removed soon. Stop using them now.
594 * Deprecate MSG_get_host_{table,number}
595 Implement MSG_hosts_as_dynar() instead.
596 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
597 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
600 * Stabilize the parallel execution mode of user contexts
601 * Introduce configuration variables to control parallel execution:
602 - contexts/synchro: Synchronization mode to use when running
603 contexts in parallel (either futex, posix or busy_wait)
604 - contexts/parallel_threshold: Minimal number of user contexts
605 that must be part of a scheduling round to switch to parallel
606 execution mode (raw contexts only)
607 * Fix bugs that prevented to use suspend/resume along with
608 synchronization structures.
609 * Fix bugs in process termination that lead to invalid memory access
610 in very specific conditions.
613 * Introduce a parallel mode for the models (controlled by surf/nthreads
614 configuration item). In our tests, running the models in parallel
615 never lead to any speedups because they are so fast that the gain
616 of computing each model in parallel does not amortizes the
617 synchronization costs, even when ultra fast futexes are used.
618 This is released anyway because YMMV.
621 * Performance boost by using a swag internally to compute the set of
622 tasks that are finished and should constitute the return value of
626 * Enable it by default now that it is considered rather stable.
629 * Documentation of the tracing functions.
630 * Performance gains when tracing categorized/uncategorized resource
631 utilization by avoiding calls to get route when updating resource
632 variables. LMM constraints are being used instead.
633 * API changed to set task categories. Use MSG_task_set_category instead
634 of TRACE_msg_set_task_category, and SD_task_set_category instead
635 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
636 * Bugfix for graphicator, routes not correctly obtained, memory leaks
637 * Examples for link user variables added (see at examples/msg/tracing/)
638 * Deprecated function TRACE_msg_set_process_category completely removed
639 * Trace header updated according to the latest Paje file format
640 * Tracing network lazy updates, no longer obligate users to use full updates
641 * --cfg=tracing/platform:1 also registers power/bandwidth variables
642 * Experimental: let user code declare/set/push/pop application states for hosts
643 * API changed to allow the manual creation of graph configuration files
644 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
647 * Improve the API of Lua MSG bindings, using the Lua spirit.
648 * Each simulated process now lives in its own Lua world (globals are
649 automatically duplicated). It helps writing simulators. Will allow
650 to run Splay programs within SimGrid in the future.
651 * Add a Chord example in Lua, equivalent to the MSG one.
654 * Start the implementation of a solution to express temporal
655 properties, not only local assertions. This is still an
656 experimental work in progress, stay clear from it to be safe.
660 - Add new runtime parameters --help-logs and --help-log-categories
661 to display informations about supported logging parameters and
663 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
664 don't exists anymore.
665 * Mallocators: allow value NULL for the reset function.
667 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
668 create homogeneous dictionaries, where all the elements share the
669 same free function. Non homogeneous dictionaries will be
670 deprecated in the next release.
671 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
672 - Multi-level dictionaries are deprecated.
674 - new function xbt_dynar_search_or_negative() that is useful when
675 you have less than 2 million elements in your dynar and don't
676 want of the extra complexity of catching exceptions when the
677 element is not found.
679 - Make xbt_os_thread module (for thread portability) public.
680 Documentation is still to come, sorry.
682 - Cleanups and simplifications to make it maintainable again.
683 - Exotic features (such as memalign and valloc) were removed.
684 - The metadata were extended and improved so that the
685 model-checker becomes able to explore and inspect the heaps.
686 - This may induce a performance drop when enable_model-checking is
687 ON in cmake (even if it's not used in the simulation), but it is
688 necessary at this point to get MC working.
690 Turn model-checking OFF if simulation performance matters to you.
691 Not enabling it at runtime is not enough, disable it in cmake.
693 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
695 SimGrid (3.6.2) stable; urgency=low
697 The "Not coding new stuff allows to polish old things" release.
700 * New bindings to the NS3 packet level simulator (experimental)
701 * Use the raw (efficient) execution contexts instead of the sysv
702 (portable) ones when possible.
703 * libpcre is now mandatory in any cases since not using it led to
704 severe performance loss and possibly other issues
705 * Update the XML platforms:
706 - G5K: include the latest machine in Nancy
707 - GridPP and LCG: new platforms
708 * Documentation was partially updated, at least (more to come)
710 Bug fixes, cosmetics and small improvements
711 * Free terminated processes before the end of the simulation to avoid
712 exhausting the memory of users having very dynamic amount of
714 * Bug fix and cosmetics about canceling non-running tasks
715 * Bug fix about the dot loader's issues when using libcgraph
718 * Create an installer for windows with nsis (amd64 and win32)
719 - Add an hello world project to illustrate simgrid project creation.
720 - Embed libpcre into the Simgrid installer to avoid
721 its compilation burden
722 * The raw execution contexts should work on Apple now
723 * Port to Windows 64 bits
724 - Sysv contexts now have an implementation for this arch
725 - GRAS communication features now support this arch
726 * Drop support for borland compiler on windows
727 - this code was not maintained, and we kinda depend on gcc nowadays
728 * Fix portability issues on kfreebsd/gnu: build error about semaphores
729 * Fix portability issue on unstable ubuntu: linker became picky on
732 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
734 SimGrid (3.6.1) stable; urgency=low
736 The "Oops, we broke Macs too" release
739 * Fixed contexts detection so that raw ones are used when possible
740 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
741 leads to a strange error, with user code segfaulting sometimes when
742 the generated code is not perfectly aligned (which is not
743 controllable from the user side, depends on the amount of code)
746 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
748 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
750 SimGrid (3.6) stable; urgency=medium
752 The Summer Release, also known as the "OMG! They Killed Kenny!" version
755 * Bindings now constitute their own package, separated from the main one.
756 Rationale: reduce our maintenance nightmare by reducing the module coupling
757 They will soon be released on their own on gforge.
759 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
760 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
762 GRAS: It is not considered as stable anymore, but experimental. Sorry.
763 * It's not quite deprecated for now because we have no replacement,
764 but it may soon become the case.
767 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
768 * Fortran: New user-level cache variable to store the rank of the running
769 process. This improves performance by an order of magnitude.
770 * C: New coccinelle script to automatically locate and modify global and
771 local static variables.
772 * Improved SMPI network model with a sender-side gap to account for multiple
776 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
777 only say if a communication is finished, no matter whether it succeeded or
778 failed. You can call MSG_comm_get_status() to know the status of a finished
780 * New function MSG_task_dsend() to send a task and detach it. When a
781 communication is detached, you are never notified of its success or failure
782 and the memory is released automatically once it is finished. This function
783 is useful when you don't care about the end nor the success of a
785 * Change the prototypes of action replay. Sorry for inconvenience,
786 but this is really more efficient this way (and to adapt your code,
787 you just have to fix the initialization, that shouldn't be too long)
788 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
789 ever managed to do anything useful with it.
790 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
791 takes a dynar of communications. It returns immediately and gives the
792 index of a finished communication (if any).
793 * New example: a basic implementation of the Chord P2P algorithm.
796 * New model for multi-core CPUs. You can now use the core attribute to
797 precise the number of cores of a host. This is a basic model. Every
798 process running on the host receives at most the power provided in
799 the DTD (throughput<=power). Total throughput of process cannot exceed
801 * New peer tag. This peer tag creates a tiny AS comprising a host and a
802 router linked by an up-link and a down-link (possibly asymmetrical).
803 This kind of pattern allows to easily build last-mile model style platforms.
804 Aggregating such patterns in a rule-based AS is thus the technique of
805 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
806 * New model for Vivaldi routing. We transformed the Vivaldi network model
807 into a Vivaldi routing model (based on the rule-based model). This allows to
808 combine Vivaldi based latencies with last-mile platforms.
811 * Added a check for NaN of IEEE754 infinite in the double entries of
813 * Introduce a new context factory "raw", highly inspirated from the
814 ucontext factory, but using manually crafted functions in assembly to
815 do the work in an efficient manner.
816 * Allow to change the used context factory at run time, not only at
817 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
818 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
819 contexts at runtime (only with raw contexts or ucontexts).
820 * Completely rewrote this module to allow parallel execution of user
821 processes. Use --cfg=contexts/nthreads:N to execute user processes
822 with N parallel threads (the default is 1, meaning no parallelism).
823 * Allow to decide dynamically between sequential and parallel modes.
824 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
825 processes in parallel only when their number is greater than or equal to P
827 * Added a check for NaN of IEEE754 infinite in the double entries of
831 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
832 won't output the state of processes when interrupted with Ctrl-C
833 * Add a new function xbt_dynar_to_array that transforms a dynar into a
834 NULL-terminated array. This may solve backward compatibility issues
835 due to the change to return type of SD_simulate. See also:
836 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
837 * Add new macros with variable number of arguments.
838 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
839 - in xbt/asserts.h: xbt_assert
840 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
841 - in xbt/ex.h: THROWF and RETHROWF.
842 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
844 * Change xbt_die() to accept a format string with arguments, just like printf.
845 * New data structure: xbt_lib_t, like a dict but more general and with better
849 * New configuration options
850 Options triva/categorized and triva/uncategorized can be used to generate
851 graph configuration files for Triva visualization tool.
852 * Configuration option tracing/platform is renamed to tracing/categorized
853 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
854 * New TRACE_user_link_variable interface:
855 User provides the name of the link and the tracing variable to attach to it
856 * the declaration of tracing categories must be done after the environment
858 * simpler tracing interface, just one way to declare categories
859 TRACE_category or TRACE_category_with_color, it is up to you
860 * links in the trace file are again identified by their names
861 * trace contains the full platform hierarchy exactly as declared using the ASes
862 * Options tracing/msg/[task|process]:1 groups the process by hosts
863 for both cases, tasks and processes must have names that are unique during the simulation
864 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
865 * The experimental option tracing/msg/volume is deprecated
866 its functionality may be reincorporated if needed
868 The tracing generates a trace file with unordered timestamped events,
869 because of the way the core simulator (surf) works. A script available
870 at the tools directory (fix-paje-trace.sh) can be used to put the events
871 in order. We have changed the tracing so it can generate ordered timestamped
872 events in the final trace, but depending on the simulator (and how much time
873 is simulated) that can lead to a huge memory utilization. It is deactivated
874 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
877 * Define a SIMGRID_VERSION macro in simgrid_config.h.
878 - We are trying hard to keep the API stable, but it may happen that
879 some things change (we're a research project after all, not a
880 nuclear plant operating system). If such things should happen, you
881 could rely on that macro to adapt.
882 - current value: 30600 for 3.06.00, aka 3.6
883 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
884 a number that can be compared with SIMGRID_VERSION.
885 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
886 assertions and verbose/debug logging events are disabled at compile time.
888 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
890 SimGrid (3.5) stable; urgency=medium
893 * New feature to any SimGrid-based simulator: Model-Checking
894 Check SIN#1 for more details.
897 * New Model SMPI (three-interval linear regression for correction factors)
898 See RR-7426, available at http://hal.inria.fr/inria-00527150
899 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
900 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
901 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
902 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
903 * See also src/smpi/README
908 - Tracing API changes: TRACE_start and TRACE_end should not be called
909 by user-code. They are automatically called by simulators created
910 with SimDAG, MSG and SMPI if the toolkit is compiled with
911 tracing_enabled=ON. Categories declaration and utilization remain the
912 same for MSG and SimDag.
913 - A function was added to the tracing API to declare categories with
915 - TRACE_category_with_color (char *category, char *color)
916 where color must be in the following format
917 "%f %f %f", red, green, blue
918 and red, green, blue are float values in the interval [0, 1]
919 - User can specify NULL as color parameter, or continue calling
921 On that case, the tracing system will define random colors
922 - The following command-line options are supported:
923 --cfg=tracing/filename:msg.trace
924 --cfg=tracing:1 (activate tracing, needed to use others)
925 --cfg=tracing/platform:1 (categorized resource use)
926 --cfg=tracing/uncategorized:1 (uncategorized resource use)
927 --cfg=tracing/msg/task:1 (task creation)
928 --cfg=tracing/msg/process:1 (process creation, migration)
929 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
930 --cfg=tracing/smpi:1 (SMPI interface tracing)
931 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
932 - examples of examples/msg/tracing updated
934 - DAXLoader and DOTLoader functions can generate tasks with categories
935 - A new function to attribute a category to SD tasks:
936 TRACE_sd_set_task_category (SD_task_t task, char *category)
937 * Tracing the MPI interface implemented by SMPI
938 - Collective operations are traced with states
939 - Point-to-Point operations are traced with states/links
940 - Tracing activated by a parameter "-trace filename" passed
941 to smpirun during execution (considering that simgrid
942 is compiled with tracing enabled)
943 - To run the simulation with gdb, the simulator
944 accepts --cfg=tracing/smpi:1 to trace SMPI
945 - tesh files to check if smpi tracing is ok
946 - See examples/smpi/NAS/DT-trace
947 * GTNetS tracing re-worked
948 - adaptation to the tracing system of GTNets to cope
949 with modifications regarding the fullduplex mode
950 - new tesh files to check if gtnets tracing is ok
953 * Asynchronous communications through the functions:
954 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
955 * New function: MSG_load_platform_script()
956 to make possible using a lua script instead of XML files to set up platforms
957 * New function: MSG_set_function
958 to associate functions to processes, used when bypassing the parser
959 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
961 Platforms: Add some more examples in examples/platforms
962 * Grid'5000: see www.grid5000.fr
963 * *_30000_hosts.xml: various huge files [mainly scalability testing]
966 * Change the XML format. This is a very important modification. SimGrid
967 3.5 introduces a new hierarchical format based on the notion of
968 Autonomous Systems. Compatibility with old format is ensured
969 through the perl script provided in the install bin directory
970 bin/simgrid_update_xml.
971 It is now possible to build platforms with specific routing
972 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
973 connect several platforms together. We will try to provide soon
974 set of realistic platforms exploiting these properties (have a
975 look at examples/platforms/ for the moment).
976 * Take the opportunity of the XML format change to be a good XML citizen:
977 rename link:ctn to link_ctn and similar changes (also dealt with by
979 * Add a new routing scheme (rule-based) using regular expressions. It
980 enables to have an extremely low memory footprint when the
981 underlying routing is simple and can be compactly described. You
982 need to have libpcre4-dev (perl regular expressions) installed if
983 you want to use this routing scheme.
984 * Revive the cluster TAG and allow to easily and efficiently (both in
985 term of memory and speed) connect clusters together. Have a look
986 at teshsuite/simdag/platforms/ to see how this can be done. With
987 this tag, you can create clusters with thousands of tasks at no
988 cost (have a look at examples/platforms/).
989 Note: clusters are implemented as ASes, so there is no need for an
990 enclosing AS tag if you have only one cluster in your platform.
991 * Add new generic functions in the public interface that allows the user
992 to call SURF 'create_resource' methods from your code (same
993 functionality as the XML bypass mechanism but with a much lighter
995 * Add a new model (enabled through command line --cfg=network/model:SMPI)
996 that uses a piecewise linear approximation to produce better
997 results when exchanging small messages.
998 * Add a new parameter to handle correctly full duplex link and account
999 for interferences between uplink and downlink communications
1000 (activate with --cfg=fullduplex:1).
1003 * Rename the SD_READY (all dependencies are satisfied and task is
1004 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1005 dependencies are satisfied) state.
1006 This prevents a confusion between the notion of "ready to schedule"
1007 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1008 simulated" (SD_RUNNABLE) used by the simulation kernel.
1009 * Change the way a task is considered as ready. Instead of removing
1010 dependencies when a task is done, a counter is decreased. This way,
1011 it is always possible to reach ancestors thanks to the
1012 SD_taks_get_parents function (even after the end of the simulation.)
1013 * Change the return type of SD_Simulate from (SD_task_t*) into
1014 xbt_dynar_t. This function was in handling a dynar internally and
1015 converted it into a NULL terminated array for historical reasons.
1016 * New function SD_dotload(char*) to load a DAG described in dot
1017 format. This loader and the corresponding examples require the
1018 installation of the graphviz library.
1019 * Fix a bug in the management of tasks of size 0 in the surf network
1020 models. This problem was only visible with SIMDAG and you should
1021 thus disregard results produced with earlier versions if you
1022 relied on this feature (some tasks were blocked because of this).
1023 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1024 even though your applications were doing only point-to-point
1025 communications and sequential computations. Now you can really use any
1026 model you want (of course, if you create real parallel tasks, which are
1027 not implemented in most models beside ptaskL07, this will abort).
1028 * Add an example that schedules a DAX on an heterogeneous platform
1029 using a Min-Min strategy.
1030 * New function SD_workstation_get_current_task() that returns the kind
1031 of task currently running on a workstation in the sequential access
1033 * Raise some warnings when unexecuted tasks remains at the end of the
1034 simulation. This is usually caused by cycles in the DAG.
1037 * New function: SIMIX_process_set_function() called by MSG_set_function
1038 * Change the underlying waiting queue in semaphores so that a process
1039 can wait on several of them simultaneously (as in waitany).
1040 * Fix the way to handle tokens in semaphores so that all access patterns
1041 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1042 * kill the dirty pimple SIMIX_message_sizes_output()
1043 Please use (proper) visualization instead
1046 * New data container: setset (set of sets of elements)
1047 * New module: mmalloc (mapped malloc, allowing to have several
1048 independent segments of malloc)
1049 * New function: xbt_dict_cursor_set_data()
1050 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1051 * New function: xbt_dynar_is_empty()
1052 * New function: xbt_fifo_get_last_item()
1053 * Fix xbt_dynar_shrink(): use the right element size.
1054 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1055 uninitialized areas during expand.
1056 * Fix semaphores: previous implementation was severely broken.
1057 * Use library init/fini functions for our initialization.
1058 - you can use logs and other feature as soon as you want in your
1059 code (even before the xbt_init / MSG_init)
1060 - xbt_exit is now a no-op and produce a warning when used.
1063 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1064 benefit from the latest improvement to the simulation kernel.
1065 * Kill measurement sockets for now. If you rely on them, sorry. This
1066 release is not for you. This feature will be reintroduced in the
1067 future, but we cannot delay the release any further.
1068 * New function: gras_msgtype_get_name().
1069 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1070 * Fix (at last) the pmm example: it should not randomly fail anymore.
1072 Build chain: bug fixes and overall polishing
1073 * Cmake is now stable enough. Hence, we killed the autotools.
1074 * Port to windows ( TM :)
1075 * Fix the 'make install' target.
1076 No need to use 'make install-simgrid' anymore
1077 * Introduce a 'make dist' target compiling a *source* archive
1078 'make package' compiles a binary archive
1079 * Compile java files only on need
1080 * Add --cd and --setenv command line options to tesh
1081 * Out of source builds are not fully supported yet, but we are close
1082 * Enable supernovae and optimization flags by default for our users
1085 * Add layer to set up environment directly from lua, without XML.
1086 * The effect of gras_stub_generator can be achieved through
1087 lua too (check examples/gras/console/ping_generator.lua)
1089 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1091 SimGrid (3.4.1) stable; urgency=low
1093 The "Polishing easter eggs is probably a good idea" release.
1094 This is a bug fixes release only.
1098 * Fix a bug preventing the tasks from begin garbage collected.
1101 * Fix a bug occuring when a host involved in a communication fails.
1102 This was not detected properly by the other peer involved in the
1103 communication. Now, it's reported as a network error.
1106 * Warn the user about loop dependencies in data flow of DAX files
1107 * Obey the control-flow dependencies of DAX files
1110 * Add option "enable_smpi" allowing to not compile SMPI.
1111 Probably useful for the (Mac) users experiencing a build error here
1112 * Improve the detection of lua5.1 and ruby1.8
1114 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1116 SimGrid (3.4) stable; urgency=low
1118 The "Easter in Cargese" release. Also known as (major changes):
1120 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1121 ~> bindings were greatly improved
1122 ~> new tracing infrastructure for better visualization introduced
1124 * the "Welcome to configury modernity" release.
1125 ~> we switched from autotools to cmake, and improved our cdash
1128 A more detailled list of changes follow (full detail in svn log).
1130 Java Bindings: Various Cleanups
1131 * (install java-gcj-compat-dev on debian-like to use them)
1132 * Remove put/get: no need to export deprecated interface in Java
1133 Use send/receive instead.
1134 * Cleanup the examples and add a README per directory
1135 * Remove example autoDestination (that's the only way to go now)
1136 * Remove example explicitDestination (was a plain copy of basic)
1137 * Make JniException a runtime exception, so that there is no need to
1138 declare the fact that you may encounter such a beast. I guess that
1139 nobody will ever want to survive such error.
1140 * Create specific errors for each MSG case of failure:
1141 host failure, transfer failure, timeout, task cancelled
1142 * Cleanup the exceptions that may get thrown by each function
1143 * Other internal cleanups in Java bindings. Performance still bad :/
1144 Ruby and Lua Bindings: create them
1145 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1146 * That's new and great, you should try them out.
1147 Same functionalities than Java bindings, only even less polished
1149 * Kill the useless "rate" argument of SD_task_get_execution_time()
1150 Everyone used to provide -1 as a value, it was not used, and the
1151 semantic of a possible use wasn't even clear.
1152 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1153 either as comm costs or compute costs to mean that there is no
1154 such thing for that specific task.
1155 * Add a SD_task_set_name() function
1156 * Fix SD_task_unschedule() on typed tasks
1157 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1158 * In DAX loader, accept useless 'level' attributes to <job> since
1159 LIGO DAGs have them (seem to be to ease graphical representation).
1161 * Add an example masterslave_mailbox.c using send/receive and not
1162 the deprecated put/get interface.
1163 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1164 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1166 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1167 (declare MSG_USE_DEPRECATED to still have the old name)
1168 * Add a MSG_task_set_data() function
1169 * About trace replay (see examples/msg/actions):
1171 - Allow to work with splitted trace files for each process
1172 Give the specific trace file as argument of each process,
1173 and call MSG_action_trace_run(NULL)
1174 You can still have one merged file for all processes.
1175 - Fix implementation of collective operations
1176 * Allow task_execute() on 0-sized tasks (closes #10063)
1178 * This is the first release of SimGrid where SMPI is not considered
1179 beta anymore (even if some corners should still be improved)
1180 * Port over the new SIMIX_network submodule (internal refactoring)
1181 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1182 * Implement more missing elements of the standard:
1184 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1186 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1187 MPI_Scan() MPI_Get_processor_name()
1188 - Added implementation of missing case for Alltoall (warning: it's
1189 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1190 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1191 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1192 SMPI_MPI_Allgatherv()
1193 * Bug fixes include:
1194 - MPI_Waitsome() was broken
1195 - Allow relative includes in smpicc
1196 - Command line cfg argument 'reference_speed' was ignored...
1197 - Some functions did not properly lead to auto-benching of user code
1198 - smpicc passes -O2 by default (just like openmpi one)
1200 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1201 * Bug fixes about timeouts during communications
1202 * add SIMIX_message_sizes_output() as a pimple to write to file the
1203 amount of messages per size. Use gnuplot to get histogram.
1204 Pimple because that's the only user-visible function of simix,
1205 defined directly in xbt.h (irk, sorry)
1207 - Add a SIMIX_sem_get_capacity() function
1208 - Fix interactions with processe resume/suspende
1209 - release_forever() was stupidly broken
1210 - Fix SIMIX_display_process_status() for processes in a semaphore
1211 - Make SIMIX_sem_block_onto() user-visible
1212 * Refactoring context stuff:
1213 - Use pseudo-OOP for better modularity
1214 - reimplement SIMIX_process_kill() without process_schedule() so
1215 that the latter can take as invariant that it is called from
1217 - Merge context_start into context_new for sake of simplicity
1219 * Add a Vivaldi network model, coded live during SUD'10 ;)
1220 * Rename configuration variables to start a hierarchy:
1221 o cpu_model -> cpu/model
1222 o network_model -> network/model
1223 o workstation_model -> workstation/model
1224 * New configuration variables:
1225 o network/bandwidth_factor: correction to bandwith
1226 o network/latency_factor: correction to latency
1227 o netwotk/weight_S: correction to the weight of competing streams
1228 * Add a long description to the models, that users can see with such
1229 argument on the command line: --cfg=cpu/model:help
1230 * --help-models display the long description of all known models
1232 * config: add the ability to set a default value after registration
1233 Does not override any previously set value (e.g. from cmd line)
1234 * dict: allow to have integer key and data.
1235 When so, you need to use the following functions
1236 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1237 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1238 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1239 In contrary to regular dicts, the key is not malloced before copy.
1240 Mixing scalar and regular elements in the same dict is not tested
1242 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1243 Tracing for Visualization:
1244 * SimGrid is now instrumented in order to generate a trace file for
1245 visualization analysis: to use it, need to compile SimGrid with the
1246 "tracing" option enabled, and instrument the program using SimGrid with
1247 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1248 (among other functions).
1249 * The instrumentation only traces the platform utilization for now
1250 * Documentation to use the tracing functions and how to analyze the
1251 traces with the Triva tool is written.
1252 * More information about: SimGrid FAQ (in the section Tracing Simulations
1255 * We moved to cmake as default build system. Autotools support will
1256 be dropped soon. Check the FAQ for more info about how to use it.
1257 * Greatly improved our cdash/ctest interactions
1258 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1259 * Added memory checking tests with valgrind; lot of memleak fixing.
1260 This may be the first release of simgrid with so few memory issues
1261 * Added code coverage tests.
1262 Our coverage is still improvable, but at least we see it on cdash.
1264 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1266 SimGrid (3.3.4) stable; urgency=low
1268 The "Desktop Grid needs love too" release (also called Xmas release).
1270 Models improvements:
1271 * Major speedup in the maxmin system solving by using lazy evaluation
1272 Instead of solving completely the maxmin system at each iteration,
1273 only invalidate (and recompute) the modified parts.
1274 This new feature is enabled in default models but you can try to
1275 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1276 * Cas01 IMproved as default CPU model
1277 This CPU model is the same Cas01 model, but it uses the
1278 maxmin-selective-update flag and a heap structure to manage
1279 actions on SURF kernel.
1280 It reduces the complexity to find the next action to finish and,
1281 consequently, it's faster than the old Cas01.
1282 This is the new default CPU model (Cas01).
1283 * Rename the old Cas01 model to Cas01_fullupdate
1284 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1285 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1286 * CpuTI (CPU Trace Integration)
1287 A new CPU model whose objective is simulate faster when using
1288 availability trace files.
1289 Instead of using a full featured, over engineered maxmin system for
1290 CPU modeling, this model does the pre-integration of traces files
1291 to calculate the amount of CPU power available, and so, executes
1292 faster than the old CPU models.
1293 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1294 * Use LV08 as default network model since it gives better accuracy
1295 for small messages and shouldn't change things for big ones.
1296 Use --cfg=network_model:CM02 to get the previous behavior.
1299 ******************************************
1300 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1301 ******************************************
1302 * The new CPU model may changes simulations!
1303 The point is that events occurring at the exact same timestamp
1304 are not scheduled in the same order with the old and new
1305 version. This may be enough to completely change the execution
1306 of simulations in some cases.
1307 * The new network model will change simulations!
1308 This new model is more realistic than the previous one, so you
1309 should consider redoing your old experiments with this model.
1310 Sorry for the inconvenience.
1313 * Introduce the supernovae compilation mode
1314 When compiled that way, the whole SimGrid (or almost) is put in a
1315 single compilation unit and compiled in one shoot.
1316 This is to help gcc which has difficulties to inline stuff from one
1318 The speedup seem to be above 15%, althrough more tests are needed on
1319 amd64 to confirm that gain.
1322 * Port of MSG's mailbox on top of SIMIX network
1323 The put/get mechanism was greatly simplified on the way.
1326 * New SIMIX network module. Provides:
1327 - Mailbox: rendez-vous mecanism to find with who you want to speak
1328 - Synchronous send/recv: easier and hopefully faster since the
1329 logic is handled in the maestro process directly now
1330 - Asynchronous send/recv: you dreamt of it? It's here now
1331 Too bad that nobody cared enough to propagate the change to MSG.
1332 * Add semaphores as SIMIX synchronization mechanism.
1335 * new function SD_daxload(char*) to load a DAX file
1336 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1337 * Introduce typed tasks. Specify its kind and cost at creation.
1338 At scheduling, just give where it should be placed, and the cost
1339 for each involved resource is automatically computed.
1340 Existing constructors so far (more to come of course):
1341 - SD_task_create_comm_e2e() for end-to-end communication
1342 - SD_task_create_comp_seq() for sequential computation
1343 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1344 * new function SD_task_dump() for debuging display
1345 * new function SD_task_dotty(task,FILE*) writing to file the info
1346 about the task in dotty format
1347 * SD_task_dependency_exists() can now cope with having one of its
1348 arguments NULL. If so, it tests whether the other argument has any
1350 * Add getters on list of preceding/following tasks:
1351 SD_task_get_parents(task) and SD_task_get_children(task)
1352 * Add getters on amount of workstations and list:
1353 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1354 * Add getter on task kind: SD_task_get_kind(task)
1355 * Update the start_time and finish_time of tasks on completion/failure
1356 * Bugfix: Remove task from state swags when destroyed
1359 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1362 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1363 dynar with one separate thread per value of the dynar.
1364 * Change the prototype of xbt_thread_create(), sorry.
1365 Added a boolean parameter indicating whether we want to join this
1366 thread (used in SG only for now)
1367 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1370 * GTNetS wrappers should now be usable again (and betterly tested too)
1371 * Fix a major regression from 3.2 where the timeout provided to
1372 MSG_task_put_with_timeout() was used as absolute time before which
1373 the comm should be done.
1374 * Start to fix the <cluster> tag.
1375 - Internal links should be good now (beside of the loopback, which
1376 use the private link instead)
1377 - paths to the external world is still rather broken
1378 - the <route:multi> tag is just broken. Actually that's brain-dead.
1379 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1380 to make it less stupid
1381 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1382 * Fix a source-level compatibility glitch from 3.2: after defining
1383 MSG_USE_DEPRECATED, you can use the old name
1384 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1385 * Allow to compile from the SVN with automake 1.11
1386 * Fix some problems when using the "start_time" tag in deployment XMLs.
1387 * Fix #8569: XBT/synchro.h has redundant declarations
1388 * Fix #8563: MSG return values and exceptions
1389 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1390 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1391 Thanks to Robson Peixoto.
1392 * Fix a vicious bug in dictionaries inducing that some elements were
1393 not freed on xbt_dict_free()
1395 Portability report of this version:
1396 * Main portability targets:
1397 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1398 on (amd64/i386/ia64)
1399 - mac leopard on i386
1400 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1401 but nothing critical.
1402 * Other platforms: windows, AIX and others were not tested for this release
1404 Timing report of this version:
1405 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1406 scenario parameters). From 8h to a few seconds in desktop grid settings.
1407 * Supernovae brings about 25% speedup on i386.
1409 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1411 SimGrid (3.3.3) stable; urgency=low
1413 The "Need for Speed" release.
1415 The timings done to validate the 3.3.2 were faulty.
1416 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1418 The problem was a conversion from a manually handled vector to
1419 xbt_dynar_t on the critical path.
1420 xbt_dynar_foreach calls functions, inducing stack management crap.
1422 We inlined these functions and xbt_dynar_foreach is now breath taking.
1423 We also inlined xbt_swag_belong on the way.
1425 Here are some approximate speedup measurements (on master/slaves
1426 simulations lasting between 10s and 20s each):
1427 3.3.1 -> 3.3.2: about same performance
1428 3.3.2 -> 3.3.3: 40% speedup
1429 3.3.1 -> 3.3.3: 40% speedup
1430 3.3.1 with inline patch -> 3.3.3: 30% speedup
1432 Our reading is that the refactoring which occurred in 3.3.2 made us
1433 suffer much more from the xbt_dynar_foreach low performance, but
1434 once we solved this, this refactoring proved to be very performance
1435 effective. From the 40% speedup, somehow, 10% are due to the
1436 inlining and 30% to the refactoring.
1438 That's a pitty that gcc cannot inline functions placed in other files
1439 alone. We have to choose between:
1440 - break the encapsulation (by putting private data structures and
1441 accessors in headers files to help gcc)
1442 - live with low performance
1443 - switch to a decent compiler such as icc (not quite possible).
1445 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1447 SimGrid (3.3.2) stable; urgency=low
1449 The "Simplicity does not preceed complexity, but follows it" release.
1451 The main contributors of this release were (lexical order):
1452 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1455 * Extract the routing logic into its own object.
1456 (was dupplicated in network.c and workstation_LV07.c;
1457 Allows to implement other ways of storing that info)
1458 => kill now useless network_card concept
1459 - Use dynar to represent routes (instead of void** + int*)
1460 - kill link_set (use surf_network_model->resource_set instead)
1461 - Add a command-line option to choose the routing schema to use
1462 - Add three new models:
1463 * Floyd (shortest path computed at initialization)
1464 * Dijikstra (shortest path recomputed all the time)
1465 * Cached Dijikstra (shortest path computed on need)
1466 All these models where contributed by Silas De Munck, and are
1467 described in his ICCS09 paper.
1469 * Simplify model declaration
1470 (less redirections, less function to write when defining a model)
1471 - Factorize stuff between models:
1474 surf_model_resource_set(model)
1475 surf_model_resource_by_name(model, name)
1476 - Unify the types of models in s_surf_model_t (using an union)
1477 - Embeed fields of common_public directly into s_surf_model_t
1478 - Rename model methods:
1479 action_free ~> action_unref
1480 action_change_state ~> action_state_set
1481 action_get_state ~> action_state_get
1482 - Change model methods into functions :
1483 (model)->common_public->action_use ~> surf_action_ref
1485 * Implement a generic resource; use it as ancestor to specific ones
1486 (allows to kill duplicated code in models)
1487 Drawback: timer command don't need no name nor properties;
1488 workstation_CLM03 don't need no properties
1489 (but I guess we can live with those few bytes wasted)
1491 * Improve the action object model
1492 - implement a constructor avoiding dupplicated code about field
1493 initialization in generic_action part.
1495 * Kill the SDP model: it has an external dependency, is deprecated
1496 in flavor of modern lmm models, and didn't compile since a while
1499 * Relocation of the context module from XBT to SIMIX.
1500 (the context were decoupled from the simix processes, duplicating a lot of code)
1501 => a lot of code was factorized
1502 - less overhead is introduced during scheduling
1503 - simpler API for the context factory
1504 - the logic for process creation,destruction and manipulation was simplified
1505 * Simplification of the s_smx_process_t data structure.
1506 => accesing the simix level data associated to a process is faster now,
1507 and the code is a lot more readable.
1510 * Implement some more MPI primitives:
1511 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1512 -implementation: Bcast: flat or 2-ary tree (default),
1513 Barrier: 4-ary tree,
1515 Allreduce: Reduce then Bcast
1516 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1517 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1518 Alltoallv: flat tree, like ompi
1520 * Add support for optimized collectives (Bcast is now binomial by default)
1521 * Port smpirun and smpicc to OS X
1524 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1525 Shout out if you used it.
1528 * Display the list of still queued messages in SG mode when existing
1532 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1533 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1534 * Add xbt_str_from_file(FILE*)
1535 * Add xbt_dict_get_key achieving a linear reverse search
1536 * Remove the context module
1538 Portability report of this version:
1539 * Main portability targets:
1540 - Linux(debian)/x86/context
1541 - Linux(debian)/x86/pthreads
1542 - Linux(debian)/amd64/context
1543 - Linux(debian)/amd64/pthreads
1544 On these, we still have the eratic breakages of gras/pmm and
1545 amok/saturate_sg reported in previous version. We still think
1546 that the tests are the cause of the fault, not the tested code.
1548 - Mac OSX Leopard/x86/context
1549 Still false negative in tesh autotesting.
1550 Smpi still fails, but this time because readlink does not accept -f
1551 Everything seems to work properly beside of that.
1554 - AIX version 5.3 (only tested contexts this time)
1555 Smpi still fails there because mktemp is not installed.
1556 Everything seems to work properly beside of that.
1558 I managed to compile it for the first time, but several breakages.
1559 Won't delay the release for this exotic platform.
1561 * Windows: it's still lagging behind. If you want to help, please
1564 Timing report of this version:
1565 This version seem to be more than 5% faster than 3.3.1 (on linux
1566 64bits with contextes). The gain is less than expected, we are
1567 investigating this for next release.
1569 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1571 SimGrid (3.3.1) stable; urgency=low
1574 * Implement a --cfg-help to show existing configuration variables
1575 * Build chain do not require doxygen in maintainer mode
1578 * fix a bug on struct sizeof computation, which prevented the
1579 exchange of arrays of structs in some conditions
1580 - added a regression test about this in datadesc_usage
1581 * Allow the exchange of 0-long dynamic vectors.
1582 - for that, use -1 as indicator of dynamic size instead of 0
1583 - This implied to change any size from unsigned long to long,
1584 reducing a bit communication abilities, but I guess that with
1585 64bits being quite common, this is more than enough.
1586 - This also induce a protocol change, thus bumping network protocol
1587 version from 0 to 1 (if we have external users, we have to get
1588 clean on that point too ;)
1589 - added two regression tests about this in datadesc_usage
1590 * Be more verbose when propagating local exceptions
1591 This helps debugging.
1592 * Display the status of simulated processes when receiving SIGINT in
1596 * Allow to control the simulation from a trace file.
1597 New functions MSG_action_register() and MSG_action_trace_run()
1598 The first one allows to associate a function execution to each
1599 kind of action while the second one parses a trace file and
1600 triggers the corresponding actions within the system.
1601 For now, only a toy example is provided in examples/msg/actions
1602 * Add an exemple of process migration in examples/msg/migration
1603 * Fix a bug in task exchange which broke MSG_task_get_sender()
1604 Add a teshsuite regression test for that.
1605 [Bug: if MSG_task_get_sender() is called after sender exit,
1607 * Fix a bug which prevented suspend/resume to work properly
1608 * Display the status of simulated processes when receiving SIGINT
1609 This fixes a regression of v3.3. due to the introduction of SIMIX
1610 * Bug fixing in failure management:
1611 - trace could not start by a failure at time 0
1612 - failure during communications were not working
1615 * Add SIMIX_process_set_name() to change the name of the current
1616 process in the log messages.
1617 * Store smx_hosts in a dict since we only retrieve them by name
1618 * Move the configuration infrastructure to surf
1621 * Move the configuration infrastructure to surf
1624 * Massive internal cleanups:
1625 - Store internal structures on processes instead of hosts (allows
1626 to have more than one process per host, in addition of being more
1628 - Cleanup the initialization/finalization process
1629 - Kill a whole bunch of unneeded synchronization:
1630 processes run in exclusive manner within the simulator
1631 - Move queues from global tables to process data fields
1633 - now accept -platform and -hostfile arguments
1634 - Pass the right rank value to processes according to the hostfile
1635 * Compile the examples by default, and use them as regression tests
1636 * Implement MPI_Wtime()
1637 * Change the reference speed to a command line option
1640 * TCP_gamma can now be specified as command line option using
1641 --cfg=TCP_gamma:10000000.0
1642 * Change the --surf-path cmd line option into --cfg=path:
1645 * Also include strbuff from xbt.h public header
1646 * xbt_ex_display(): do not free the exception after displaying
1647 This allows to do more with the given exception afterward.
1648 Users should call xbt_ex_free() themselves.
1652 Portability report of this version:
1653 * Main portability targets:
1654 - Linux(debian)/x86/context
1655 - Linux(debian)/x86/pthreads
1656 - Linux(debian)/amd64/context
1657 - Linux(debian)/amd64/pthreads
1658 These targets fail about 1/10 of times on gras/pmm, but we believe
1659 that this is because of the test, not because of simgrid.
1660 amok/saturate_sg fails even more rarely, and the test may not be
1663 - Mac OSX Leopard/x86/context
1664 The test suite still spits tons of errors because some obscure
1665 force prevents us from removing the temporary directories
1666 arguing that they still contain some metadata I've never heard of.
1667 Smpi fails because seq is not installed.
1668 Everything seems to work properly beside of that.
1671 - AIX version 5.3 (both contexts and pthread)
1672 Smpi still fails there because mktemp is not installed.
1673 XML inclusions seems rosty on AIX.
1675 * Windows: it's still lagging behind. If you want to help, please
1678 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1680 SimGrid (3.3) stable; urgency=high
1684 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1685 [Malek Cherier & Mt]
1687 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1690 Until now, GRAS were using MSG as an interface to SURF. It was
1691 quite difficult because both interface have several differences
1692 (MSG channels vs GRAS sockets were the most notable point).
1694 This also opens the gate to SMPI (which should occur soon) and speed
1695 up simulations by to 40% (even if it were not the main goal).
1697 **************************************
1698 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1699 **************************************
1700 The point is that events occuring at the exact same timestamp are
1701 not scheduled in the same order with the old and new version. This
1702 may be enough to completely change the execution of simulations in
1703 some cases. Sorry for the inconvenience.
1705 * Cleanup and upgrade the XML format to push further scalability
1706 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1708 * Improve the testing infrastructure with tesh. Now a very large part of
1709 the code is tested not only by being run but also by checking that the
1710 output match an expected output [Mt].
1712 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1713 is really less memory-demanding, which should allow you to use
1714 larger files in SimGrid [AL].
1716 * Inform valgrind about our contextes, so that it becomes usable
1717 with the default (and more effecient) version of SimGrid
1718 [contributed by Sékou Diakite, many thanks]
1721 * Introduce a listener thread in charge of receiving incoming
1722 messages from the network. It allows to overlap communication and
1723 computation but most notably, it removes some stupid deadlocks due
1724 to the fact that so far, a process could not send and receive at
1725 the same time. This made most non trivial communication schema
1727 * Convert the PIDs from long int to int to match the MSG ones (and
1728 linux ones too) [Mt]
1729 * New function: gras_agent_spawn() to launch a new process on
1730 current host. Only working in simulation for now. [Mt]
1731 * New function: gras_os_hostport() returning a constant form (ie,
1732 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1735 * Make the backtrace of exceptions more human readable [Mt]
1736 * New module: xbt/str [Mt]
1737 a ton of string utility functions (split, join, printf to a newly
1738 allocated buffer, trim, etc)
1739 * New module: xbt/hash [Mt]
1740 SHA1 hashing algorithm (more to come if needed)
1741 * New module: xbt/synchro [Mt]
1742 synchronization tools (mutex and conditions) working the same way
1743 in simulation and in real life (mainly useful for GRAS, but not
1745 * New module: xbt/queue [Mt]
1746 classical producer/consumer synchronization scheme
1747 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1748 (using the classical functions will get serialized) [Mt]
1749 * Make dictionary internal table dynamic. No need to specify its size
1750 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1752 * Make sure the log channels are organized as a tree under windows
1753 (because of ANSI C compatibility issue, any channel were child of
1754 root directly) [Mt].
1757 * Cleaned many thing in surf and fixed a few bugs [AL].
1758 * Add a nice command line configuration mechanism to compose models [AL].
1759 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1760 the previous one (KCCFLN05). It relies on something that looks like
1761 a max-min sharing mechanism but cannot be written as such. A new solver
1762 was thus designed [AL].
1763 * Add a new solver to lmm. Based on Lagrange optimization and
1764 gradient-based descent, it enables to efficiently maximise systems s.a
1766 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1768 This solver enables to propose two new network models for TCP Reno and
1769 TCP Vegas based on Low's work. These models still need to be fully
1770 tested though [Pedro Velho].
1773 * Bug fix in SD_simulate. Now the time bound given as argument is
1775 * Use the new parallel task model (ptask_L07) as default.
1776 * Use the SURF command line configuration mechanism.
1777 * 0-size tasks (for synchronization) should now work.
1779 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1781 SimGrid (3.2) stable; urgency=high
1785 We still experience issues on this platform, but we believe that at
1786 least MSG is usable.
1788 GRAS API BREAKAGE (for simplification purpose, sorry):
1789 * the gras_msgtype_by_name is not used anymore. Instead of
1790 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1791 you can write (and must)
1792 gras_msg_send(toserver, "request", &request);
1793 - If you still want to pass a gras_msgtype_t to the function (to cache
1794 the type and avoid the lookup time), use the gras_msg_send_() variant.
1795 - Impacted functions:
1796 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1797 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1798 * The callbacks are now expected to return 0 when everything went well
1799 (just like the main() function)
1801 GRAS new features and improvements:
1802 * New module mecanism where user code can use per process globals [Mt]
1803 This is similar to gras_userdata_*() functions, but for libraries. It
1804 factorize some code developped over and over in the examples and AMOK.
1805 It has still to be documented and used (only amok/peermanagement is
1807 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1808 loss when they were small enough to fit into the buffer and sent quickly
1809 enough so that they can all get received in one shoot.
1810 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1811 if not found. Use the *_or_null() variant for the old semantic.
1812 * In gras_msg_handle, do not discard messages without callback.
1813 They are probably messages to be explicitly awaited later (ie, proofs of
1814 mis-synchronization in userland since they are sent before being awaited)
1815 No big deal usually.
1816 * gras_socket_meas_send/recv: semantic changed!
1817 The numerical arguments used to be (1) the total amount of data to send
1818 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1819 messages. This was need for the fool willing to send more than MAXINT
1820 bytes on quite fat pipes.
1823 * Do really rename the hostmanagement module to peermanagement. [Mt]
1824 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1825 but this is rather new and this was documented in the module
1826 documentation (poor excuses, I admit)
1827 * Bandwidth measurement semantic changed! This follows the changes to
1828 gras_socket_meas_send/recv explained above.
1831 * A sequential mode has been added to the workstations. When a workstation
1832 is in sequential mode, it can execute only one task, and the other tasks
1833 are waiting in a FIFO. [Christophe Thiery]
1836 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1837 model for SIMDAG. [Christophe Thiery]
1838 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1839 equal to 0, which caused some bad behaviors in
1840 saturated_constraint_set_update. I now use a threshold mechanism like in
1844 * When running manually src/testall, you select specific units [Mt]
1845 testall is the result of our cunit mecanism, and should replace all
1846 the scripty thingy around since bash don't run easily on billware.
1848 * A mallocator system has been added. [Christophe Thiery]
1849 Mallocators allow you to recycle your unused objects instead of freeing them
1850 and allocating new ones.
1852 Documentation update:
1853 * FAQ reworking + New FAQs:
1854 - "Valgrind spits tons of errors!" [Mt]
1855 - "How to repport bugs" [Mt]
1856 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1857 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1858 - Communication time measurement within MSG [AL]
1859 - I experience weird communication times when I change the latency [AL]
1860 * GRAS tutorial [Mt]
1862 - an introduction to the framework and to the used communication model
1863 - an initiatic tour introducing the most proheminent features:
1865 . Lesson 0: Installing GRAS
1866 . Lesson 1: Setting up your own project
1867 o Part 2: Message passing
1868 . Lesson 2: Exchanging simple messages
1869 . Lesson 3: Passing arguments to the processes (in SG)
1870 . Lesson 4: Attaching callbacks to messages
1871 . Lesson 5: Using globals in processes
1872 . Lesson 6: Logging informations properly
1873 . Lesson 7: Using internal timers
1874 . Lesson 8: Handling errors through exceptions
1875 . Lesson 9: Exchanging simple data
1876 . Lesson 10: Remote Procedure Calling (RPC)
1877 . Lesson 11: Explicitely waiting for messages
1878 . Recapping of message passing features in GRAS
1879 - A HOWTO section containing:
1880 o HOWTO design a GRAS application
1881 More are due, of course. They will come latter. In the meanwhile, you can
1882 check the examples which are still here.
1884 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1886 SimGrid (3.1) stable; urgency=high
1890 There was a stack corruption somewhere, visible only when optimizing
1891 with these versions. [Vince]
1894 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1895 module that enabled to play with Directed Acyclic Graphs. It is built
1896 directly on top of SURF and provides an API rather close to the old
1897 SG. Some old codes using SG are currently under rewrite to check that
1898 all needful functions are provided. [Christophe Thiery]
1901 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1902 extension of the classical CLM03 model that gracefully handles
1903 failures. This is now the default model for MSG and GRAS. It doesn't
1904 handle parallel tasks yet though. [AL]
1905 * Bug fix: Weights were not correctly set in the network part.
1906 WARNING: This may have resulted in incorrect results with simulations
1907 where there are more than one flow on a given link. [AL]
1910 * After a (long ?) discussion on simgrid-devel, we have decided that the
1911 convention we had on units was stupid. That is why it has been decided
1912 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1913 WARNING : This means that all previous platform files will not work as
1914 such with this version! A warning is issued to ask users to update
1916 A conversion script can be found in the contrib module of the CVS, under
1917 the name contrib/platform_generation/surfxml_update.pl [MQ]
1920 * Bug fix: Processes were started in reverse order, wrt deployment file.
1921 WARNING: if your code relies on this bug, please fix it. [AL]
1922 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1923 being executed on two different locations. [AL]
1924 * Bug fix: Failures are now better supported thanks to Derrick's tests
1925 (there was many failure situations I hadn't thought of and that weren't
1926 correctly handled). [AL]
1927 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1931 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1932 Exception occurring on server-side are propagated back to client (!).
1934 API CHANGE: the callback changed their prototype. Change:
1935 int my_handler(gras_socket_t expeditor, void *payload_data) {
1937 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1938 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1940 * New! function: gras_msg_handleall to deal with all messages arriving
1941 within a given period.
1942 * New! function: gras_socket_server_range to get a server socket in a
1943 range of port numbers (ease to avoid port number conflicts) [MQ]
1944 * New! gras processes display their backtrace when they get a SIGUSR1
1945 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1946 Sweet to debug RL processes [MQ]
1950 - Do not force experiment sizes to be expressed in kb, or it becomes
1951 impossible to measure the latency this way (needs one byte-long tests)
1952 WARNING: this changes the amok_bw_* function semantic. [MQ]
1953 - Implements the link saturation stuff. [MQ]
1954 * Peer management module:
1955 New! module factorizing code that we wrote over and over [MQ].
1958 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1959 - Test units are placed directly into the library code, they get extracted
1960 automatically and placed into the src/testall binary.
1961 - Convert most of the XBT tests to this system.
1962 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1963 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1964 live in the same C code. This required to change a little bit the API
1965 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1966 * New module: sparse graph structure with basic algorithms (this is work
1967 in progress and the API is not considered to be frozen yet). [AL]
1968 * Display more information on backtraces: source line & function names are
1969 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1970 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1971 * Reworked a little bit some #include statements to load only required
1972 headers. Some user code that relied on SimGrid to include stdlib or
1973 stdio may need to include it by themselves. [AL]
1974 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1976 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1977 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1978 but I guess that almost nobody used those parts. [MQ]
1980 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1982 SimGrid (3.0.1) stable; urgency=low
1985 * Unfortunately, I had missed 5 misnamed functions:
1986 xbt_fifo_item_t xbt_fifo_newitem(void);
1987 void xbt_fifo_freeitem(xbt_fifo_item_t);
1988 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1989 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1990 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1991 They're now deprecated. Please use their new versions:
1992 xbt_fifo_item_t xbt_fifo_new_item(void);
1993 void xbt_fifo_free_item(xbt_fifo_item_t);
1994 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1995 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1996 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1998 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1999 * Documentation: xbt_log module unmercifully reworked [MQ]
2000 * Bugfix: there was a problem with the ending of contexts with
2001 the pthread backend. It caused some weird deadlock or behavior
2002 depending on the pthread implementation. [AL]
2003 * Bugfix: get the exceptions raised in the simulator repport where
2004 and why they come from when they are not catched in time [AL, MQ]
2007 * Bugfix: Do repport the error when two non-connected hosts try to
2008 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2011 * Add additionnal checkings on communications. Assert that two
2012 communicating hosts are connected by a set of links... [AL]
2015 * Add additionnal checkings on channel values in communication [AL]
2016 * New: MSG_task_get_source to see on which host a task was generated [HC]
2017 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2018 the number of tasks waiting to be received on channel and sent
2020 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2021 waits for the first task coming from a given host.. [AL]
2023 GRAS new functionnalities: [MQ]
2024 * Enhance the parsing macro to allow the size of multidimentional objects
2025 to be given thru annotations.
2026 * New example (and documentation): Matrix Multiplication a la RPC
2027 (as when I was young!) and fix a bunch of bugs found on the way.
2029 GRAS performance improvements: [MQ]
2031 * Reduce the amount of cbps creation/destruction by making it static to
2032 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2035 * Change libdata to a set so that we can search for stuff by ID (and thus
2036 reduce the insane amount of dict lookups)
2039 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2040 RL communication on top of SG since it's so uneffective.
2041 It may also be used for inter-thread communication in RL, one day.
2042 * Use gras_datadesc_copy() to exchange messages on top of SG
2044 - improve message exchange performance on top of SG
2045 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2046 * Don't exchange on the network the size of the used part of buffer,
2047 instead, specify the possible buffer size to read().
2049 - reduces the amount of read/write calls (one pair per exchange)
2050 - reduces the amount of exchanged data (the size)
2051 - allows to retrieve all arrived data on receiver side, if we don't need
2052 it right now (subsequent read will peek the buffer)
2053 - allows the receiver to proceed with the begining of the stream before
2054 everything is arrived
2055 - make it possible to build an iov transport (using readv/writev)
2057 - take care of the data with non-stable storage (like stacked data),
2059 * If possible, TCP send uses vector I/O (when writev() is here)
2060 - Don't use it for receive since we send data sizes and data on the
2061 same stream, so we wouldn't be able to chain large amount of chunks
2062 before having to flush the stuff to read the size.
2063 * Rework the transport plugin mecanism to simplify it and reduce the
2064 amount of pointer dereferencement when searching for the right function
2067 * I guess that now, we do almost as few system calls as possible while
2068 doing as few data copy as possible.
2070 To improve it further, we could try to send all the sizes first and then
2071 all the data (to use iov on receiving size), but it's only a partial
2072 solution: when you have 2 dimensional data, the sizes of the second
2073 dimension is data of the first dimension, so you need 3 streams.
2075 I'm not sure the potential performance gains justify the coding burden.
2077 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2079 SimGrid (3.00) stable; urgency=high
2082 * New! Give the possibility to hijack the surf parser and thus bypass
2083 MSG_create_environment and MSG_launch_application. Have a look at
2084 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2086 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2088 SimGrid (2.96) unstable; urgency=low
2093 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2094 This deprecates the xbt_error_t mecanisms.
2095 It modifies (simplifies) all XBT and GRAS API.
2096 MSG API keeps unchanged (exceptions raised by XBT are catched from
2097 within MSG and masked with existing error handling facilities)
2100 * New! Add a FATPIPE model. [AL]
2101 * New! Add a parallel task model. [AL]
2102 * New! Add automatically a loopback interface (in the default
2103 network model) if none was precised.
2106 * Bugfix: MSG_process_resume now works with the current running process.
2108 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2109 * Modification of MSG_task_get_compute_duration. Once a task has been
2110 processed, the value returned by this function is now equal to 0. [AL]
2111 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2112 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2113 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2114 of a m_task. For now, MSG_task_cancel only works with computation
2116 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2117 of the processor (in Mflop/s) regardless of the current load on the
2119 * API Change: use proper naming convention for MSG_getClock and
2120 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2122 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2123 This function changes the priority of a computation task. This priority
2124 doesn't affect the transfer rate. A priority of 2 will make a task
2125 receive two times more cpu power than the other ones. This function
2126 has been added to suit the needs of Nguyen The Loc and hasn't been that
2127 much tested yet. So if it fails, please report it and send me your code.
2129 * API Change: removed all functions and types that were marked "deprecated"
2130 since many months. Renamed MSG_global_init_args to MSG_global_init.
2132 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2134 SimGrid (2.95) unstable; urgency=low
2137 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2138 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2139 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2142 * Bugfix: Allow absolute paths for platform description files [MQ]
2143 * Bugfix: do free the variables after use. Leads to drastic performance
2145 * Implement max_duration (ie, timeouts) on resources [AL]
2148 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2150 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2151 the needed support (timeouts) [AL]
2154 * Implement measurement sockets. You can now get the bandwidth between two
2155 hosts thanks to AMOK (see below). [MQ]
2156 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2157 dynar over the network (yeah) [MQ]
2158 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2159 * simple/static token-ring example. [Alexandre Colucci and MQ]
2160 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2161 in place before [MQ]
2164 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2165 type parsing. "short" and "long" modifiers were also ignored; other
2166 modifier (such as reference level) are still ignored. [MQ]
2167 * Embeed the buffer size within the buffer itself on SG. [MQ]
2168 That way, send() are atomic and cannot get intermixed anymore (at least
2169 the ones which are less than 100k; bigger messages still have the issue)
2170 * Array size pushed by the field, not by the field type (or each
2171 and every long int will push stuff to the cbps) [MQ]
2172 * use select() to sleep since it allows to portably sleep less than one
2175 GRAS (minor cleanups)
2176 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2179 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2180 * type category 'ignored' killed as it was never used and were difficult
2182 * whether a type can cycle or not is now a flag, leaving room for more
2183 flags (as "ignored", if we feel the need one day ;)
2184 * Rename raw sockets to measurement sockets since "raw" has another
2185 meaning in networking community.
2188 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2189 and offers features not belonging to GRAS but that most applications
2190 need. One day, it may be a set of plugins loadable at runtime.
2191 * New module: bandwidth
2192 bandwidth measurement between arbitrary nodes running this module. [MQ]
2194 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2196 SimGrid (2.94) unstable; urgency=low
2198 The first beta release of SimGrid 3 !
2202 * Update the main page and the FAQ. Adding references to gforge.
2205 * Add a gras_os_getpid function.
2208 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2209 * Extend the logs so that they also print PID, hostname, date, ... if
2211 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2212 and kill the old version which were in testsuite/
2213 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2217 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2218 for users. Instead, each programming environment (GRAS, MSG,...) use it
2219 when needed to provide such a feature to users.
2220 Don't shortcut the mecanism or you will also shortcut the virtualization
2221 you need on the simulator.
2225 * Cleanups in configury with regard to compile optimization/warning flags.
2226 Also add -fno-loop-optimize to any powerpc since it's the optimization
2227 killing gcc (< 3.4.0).
2228 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2229 needed by MSG examples complications
2230 * Borrow configury beautifications from PHP
2233 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2234 warning (thanks loris for stumbling into this one).
2235 * Bugfix: stop loading private headers (gras_config.h) from the public
2239 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2240 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2241 RL would imply a DNS resolver, which is *hard* to do in a portable way
2242 (and therefore delayed).
2243 * Implement a real timer mecanism and use it in timing macros. This allows
2244 to avoid rounding errors and get a 0.000005 sec precision in timing
2245 macros. While I was at it, various cleanups:
2246 - allow to declare more than one timed section per file (fix a stupid bug)
2247 - move some private declaration to the right place
2248 - merge conditional execution and timing macros into emulation module
2249 - document the module
2250 - make sure the module cleanups its mess on gras_exit
2251 * Documentation improvements:
2252 - (new) how to compile applications using GRAS
2253 - (new) emulation support (timing macros)
2255 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2257 SimGrid (2.93) unstable; urgency=low
2259 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2262 - Use Paje properly where used. Still to be sanitized properly.
2263 - Portability fix: Add an implementation of the contexts using pthread
2267 - Add xbt_procname(): returns the name of the current process.
2268 Use it to show the current process's name in all logging.
2270 - fix detection of older flex version and the reaction, since we do
2271 depend on modern ones (we use lex_destroy)
2272 - Better separation of SG and RL in the libs: remove all simulation code
2273 from libgras. As a result, this lib is now only 200k when stripped.
2274 Some of the xbt modules may also be duplicated (two sets and such) and
2275 should be cleaned/killed before SG3.
2276 - Insist on using xlC on AIX because of weird problems involving gcc there.
2277 - Cleanup the make remote stuff. This is now done by scripts
2278 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2279 mainly for our private use, but we're working on changing them to user
2282 - Bugfix: flush the socket on close only if there is some *output*.
2283 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2285 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2288 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2290 SimGrid (2.92) unstable; urgency=low
2292 Alpha 3 on the path to SimGrid 3
2296 - New! First try of benchmarking macros.
2297 - New! First try so that gras_stub_generator generate deployment and
2298 remote compilation helpers.
2300 - Bugfix: Initialization fix in msg_test.
2304 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2306 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2308 - API Change: gras_os_sleep to take the amount of seconds as a double.
2309 Accepting an int was error prone since it was the only location where
2310 seconds were coded as such. It leaded to damn rounding errors.
2311 - Bugfix: Hard to belive that timers ever worked before this.
2313 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2315 SimGrid (2.91) unstable; urgency=low
2317 Alpha 2 on the path to SimGrid 3
2321 - Bug fix in the lmm_solver.
2323 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2324 through the function MSG_paje_output.
2325 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2326 - It is possible to bound the rate of a communication in MSG with
2327 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2328 to write it in the changelog).
2329 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2334 - Various cleanups to the autotools stuff
2335 - Begin to move Gras examples to examples/gras/
2336 - Let make distcheck work again (yeah!)
2338 - documentation overhauled using doxygen.
2339 gtk-doc-tools is dead in SimGrid now.
2340 - Automatically extract all existing logging categories, and add the list
2341 to the documentation (long standing one, to say the less)
2343 - Cleanup the known architecture table. Reorder the entries to group what
2344 should be, and use a more consistent naming scheme.
2345 (some of the test dataset are still to be regenerated)
2346 - New! Allow library to register globals on each process just as userdata
2348 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2349 lookup time (for now).
2350 Use it in msg and trp.
2351 This cleans a lot the internals and helps enforcing privacy of the
2352 headers between the gras components.
2353 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2354 - Bugfix: gras_os_time was delirious in RL.
2355 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2357 - Reenable GRAS now that it works.
2359 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2361 SimGrid (2.90) unstable; urgency=low
2363 Alpha 1 on the path to SimGrid 3
2365 * It is a long time since the last release of SimGrid. I'm sorry about
2366 that but as I had told you, I was rewriting a lot of things. I apologize
2367 to those who had been reporting bugs to me and that I had not answered.
2368 If your bug is still in the new version, please tell me. Here is a
2369 summary of the main changes.
2371 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2372 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2373 portability, speed, and a lot more but you'll figure it out later.
2374 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2375 I wanted to release the new MSG as soon as possible and I have
2376 broken GRAS, which is the reason why, for now, only MSG is fully
2377 functional. A laconic description of these projects is available
2378 in the documentation.
2380 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2381 kernel optimized for our needs (called SURF but only the developers
2382 should use it). Hence, MSG is now roughly 30 times faster and I think
2383 that by rewriting a little bit MSG, I could event speed it up a little
2384 bit more. Beside the gain in speed, it is also much easier to encode a
2385 new platform model with SURF than it was with SG. More to come...
2387 * REVOLUTION 3: I have tried to change a little as possible the API of
2388 MSG but a few things really had to disappear. The main differences
2389 with the previous version are :
2390 1) no more m_links_t and the corresponding functions. Platforms are
2391 directly read from a XML description and cannot be hard-coded
2392 anymore. The same format is used for application deployment
2393 description. The new format is described in the documentation.
2394 Have a look in tools/platform_generation. There is a tiny script
2395 that converts from the old platform format to the new one. Concerning
2396 the application deployment format, parsing the old one is tricky.
2397 I think most of you should however be able to convert your files. If
2398 it is really an issue, I can write a C code that does the conversion.
2400 2) the toolbox tbx does not exist anymore. We now have a library
2401 with much more data-structures but without the hash-tables (we have
2402 dictionaries that are much faster).
2404 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2406 *****************************************************************************
2407 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2408 * don't feel like converting the oldies. (Mt) *
2409 *****************************************************************************
2412 Version 2.90: "the long awaited one"
2413 - Finished rewriting and debugging MSG. Rewrote the documentation.
2414 - disable GRAS for now since it needs to be ported to the newest SG
2417 - Finish the port to windows (using mingw32 for cross-compile)
2420 - Main loop and datastructures of SURF. A cpu resource object is
2421 functional. Surf can thus be used to create cpu's with variable
2422 performance on which you can execute some actions.
2424 2004-11-15 Martin Quinson
2425 - Port to ARM. Simply added the alignment and size descriptions. Should
2426 work, but the ARM machines are so slow that I didn't had the opportunity
2427 to 'make check' over there yet.
2429 2004-11-15 Arnaud Legrand
2430 - Trace manager now written. It uses a heap structure and is therefore
2431 expected to be efficient. It may however be speeded up (particularly
2432 when many events occur at the same date) by using red and black
2433 trees. One day maybe...
2434 - Max-min linear system solver written. It uses a sparse matrix
2435 structure taking advantage of its expected use. Most operations are
2436 O(1) and free/calloc are called as few as possible. The computation of
2437 the minimum could however be improved by using a red and black tree
2440 2004-11-03 Arnaud Legrand
2441 - Rename every gras_* function that was in xbt/ to its xbt_
2443 - Add a heap and a doubly-linked list to xbt
2444 - Added a dichotomy to the dictionaries. make check works as well before
2445 so I assume that the patch is correct. I do not know however if things
2446 run effectively faster than before now. :)
2448 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2449 SimGrid, and the version number is bumped to 2.x
2451 2004-10-29 Martin Quinson
2452 - Introduction of the remote errors.
2453 They are the result of a RMI/RPC on the remote machine.
2454 ErrCodes being scalar values, you can't get the host on which those
2455 errors did happen. Extending the error mechanism as in Gnome is possible.
2456 No idea yet whether it is a good idea.
2458 2004-10-28 Martin Quinson
2459 - Interface revolution: the Starred Structure Eradication.
2460 I used to do typedef struct {} toto_t; and then handle *toto_t.
2461 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2462 - ???_t is a valid type (builded with typedef)
2463 - s_toto_t is a structure (access to fields with .)
2464 - s_toto is a structure needing 'struct' keyword to be used
2465 - e_toto_t is an enum
2466 - toto_t is an 'object' (struct*)
2468 typedef struct s_toto {} s_toto_t, *toto_t;
2469 typedef enum {} e_toto_t;
2470 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2471 s_toto_t) is private.
2473 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2474 so that it can set the variable to NULL. It was so for dicts and sets,
2475 it changed for dynars.
2477 - Fix a bunch of memleaks in dict_remove
2478 - Fix a bug in sg/server_socket opening: it failed all the time.
2480 2004-10-07 Martin Quinson
2481 - Speed up dynar lookup operation a bit.
2483 gras_dynar_get is dead.
2485 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2486 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2487 (which gives you the address of the stored data).
2489 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2490 copy of the data using an affectation to do the job and not a memcpy.
2492 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2493 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2495 It does not really speedup the dynar test because they are
2496 setting elements all the time (and look them seldom). But the dict does
2497 far more lookup than setting.
2499 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2501 2004-10-05 Martin Quinson
2502 - Allow to (en/dis)able the cycle detection at run time.
2504 Whether we should check for cycle or not is now a property of each
2505 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2506 datadesc_cycle_unset allow to remove this property when previously set.
2508 Note that the cycle detection is off by default since it impacts the
2509 performance. Watch the data you feed GRAS with ;)
2511 This property is hereditary. Any element embedded in a structure having it
2512 set have it set for the time of this data exchange.
2514 You should set it both on sender and receiver side. If you don't set it on
2515 sender side, it will enter an endless loop. If you forget on receiver
2516 side, the cycles won't be recreated after communication.
2518 - Header reorganization.
2519 Kill gras_private.h, each submodule must load the headers it needs.
2521 2004-10-04 Martin Quinson
2522 - Interface revolution: do not try to survive to malloc failure.
2524 Now, gras_malloc and friends call gras_abort() on failure.
2525 As a conclusion, malloc_error is not a valid error anymore, and all
2526 functions for which it was the only gras_error_t return value are
2527 changed. They now return void, or there result directly.
2528 This simplify the API a lot.
2530 2004-09-29 Martin Quinson
2531 - Re-enable raw sockets.
2532 Created by gras_socket_{client,server}_ext;
2533 Used with gras_raw_{send,recv}
2536 It should allow to kill the last bits of gras first version soon.
2538 This is not completely satisfactory yet (duplicate code with
2539 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2542 - Simplify transport plugin (internal) interface by not passing any
2543 argument to _server and _client, but embedding them in the socket
2546 2004-09-28 Martin Quinson
2547 - Finish the port to AIX.
2548 autoconf was my problem (segfault within the malloc replacement
2549 function. No idea why)
2551 2004-09-16 Martin Quinson
2552 - Fix some size_t madness on 64bit architectures.
2554 2004-09-08 Martin Quinson
2555 - Reduce the number of system headers loaded, overload some more system
2556 calls (such as malloc to cast the result of the system one, and work
2558 - Fix and reintroduce the config support
2560 2004-09-07 Martin Quinson
2561 - Source code reorganization to allow Arnaud to surf all over there.
2562 - Allow to document the logging categories.
2563 - Remove all uppercase from logging categories and useless cleanup in names.
2565 2004-08-18 Martin Quinson
2566 Version 0.6.2 (protocol not changed; API changed)
2567 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2568 gras_error_t), and NULL when not found. Functions expecting a msgtype
2569 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2570 hopefully usefull message.
2571 - Portability to prehistoric sparcs again
2573 2004-08-17 Martin Quinson
2574 Version 0.6.1 (protocol not changed; ABI not changed)
2575 - prealloc some buffers to speed things up
2577 2004-08-11 Martin Quinson
2578 Version 0.6 (protocol not changed; ABI expended)
2579 - The parsing macro can deal with the references, provided that you add
2580 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2582 2004-08-09 Martin Quinson
2583 Version 0.5 (protocol not changed; ABI changed)
2584 - Allow to off turn the cycle detection code in data exchange at
2585 compilation time. It should be at run time, but I'm short of time (and
2586 the config stuff is still broken). That way, we keep dict out of the
2587 critical path, which is good because the performance is poor:
2588 - search not dichotomial yet
2589 - dynar give no way to access their content and memcpy everytime
2590 - In composed data description (struct, ref and so on), stop foolness of
2591 keeping the subtype's ID, but store the type itself. This keeps sets out
2592 of the critical path, which is good since they rely on dynar and
2593 dictionnaries. The only loose of that is that we cannot detect the
2594 redeclaration of a structure/union with another content (but I'm not sure
2595 the code detected well this error before anyway). We still can detect
2596 the redefinition discrepancy for the other types.
2597 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2598 it breaks the code because of type-punning used all over the place).
2599 This breaks on all non-gcc architectures (for now).
2601 All those changes (plus the buffer of last time) allow me to gain 2 order
2602 of magnitude on cruel tests consisting of 800000 array of integers on two
2603 level of a hierarchical structure (200 secondes -> 4 secondes)
2606 - the selector of reference must now return the type it points to, not
2607 the ID of this type.
2609 2004-08-06 Martin Quinson
2610 Version 0.4 (protocol changed; ABI not changed)
2611 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2612 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2613 - Implement a buffer transport, and use it by default (it relies on tcp in
2614 real life and on sg in simulation).
2615 That's a bit hackish since I had a new field to the structure to store
2616 its data without interfering with the subtype ones. Inheritance
2617 is tricky in C. And that's a kind of reverse inheritance with one class
2618 derivating two classes. Or maybe a game with java interfaces. Anyway,
2619 that's damn hard in C (at least).
2620 Moreover, I got tired while trying to ensure plugin separation and
2621 genericity in SG mode. MSG wants me to do weird things, so let's go for
2622 cruel hacks (temporarily of course ;).
2623 See comment in transport_private.h:71
2624 - do not include all the _interface headers in private but in the files
2625 which really need them (to cut the compilation time when they are
2628 2004-07-26 Martin Quinson
2629 Version 0.3 (protocol not changed; ABI changed)
2630 - Major overhault of the datadesc interface to simplify it:
2631 - shorted the function names:
2632 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2633 - add a trivial way to push/pop integers into the cbps without malloc.
2634 This allows to make really generic sub_type description, which simply
2635 pop their size of the stack.
2636 - add a function gras_datadesc_ref_pop_arr() which does what users want
2637 most of the time: Declare a dynamic array (which pops its size of the
2638 stack) and declare a reference to it. Poor name, but anyway.
2639 - kill the post-send callback, add a post-receive one
2641 2004-07-23 Martin Quinson
2642 Version 0.2 (protocol changed; ABI changed)
2643 - add some testing for cpbs in the test cases, and fix some more bugs.
2644 This invalidate again the little64 data file, since I cannot regenerate
2646 - remove an awfull optimization in the logging stuff, allowing me to:
2647 - understand it again
2648 - learn gcc how to check that the argument match the provided format
2649 - fix all errors revealed by gcc after that
2650 - internal keys of dict are not \0 terminated. Deal with it properly in
2651 loggings instead of segfaulting when the user want to see the logs :-/
2653 2004-07-22 Martin Quinson
2654 - Fix some stupid bug preventing cbps (callback postit) from working
2656 2004-07-21 Martin Quinson
2657 - Some documentation cleanups
2658 - remove the useless last argument of msgtype_declare
2659 - rename the Virtu functions to fit into the 'os' namespace
2660 - move headers src/include -> src/include/gras/ and stop fooling with
2661 gras -> . symbolic link
2662 - make distcheck is now successful
2664 2004-07-19 Martin Quinson
2666 - Build shared library also
2667 - Install html doc to the right location
2668 - stop removing maintainer files in make clean
2669 - build tests only on make check
2671 2004-07-13 Martin Quinson
2673 - No major issue in previous version => change versionning schema
2674 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2675 corresponding dataset.
2677 2004-07-11 Martin Quinson
2679 - Get it working with any kind of structure (we can compute the padding
2680 bytes remotely for all the architectures I have access to)
2681 - Implement the structure parsing macro (still not quite robust/complete)
2682 - Improvement to the remote testing toysuite
2684 2004-07-10 Martin Quinson
2685 [autoconf mechanism]
2686 - get ride of a bunch of deprecated macros
2687 - actually run the test for two-compliment, not only compile it :-/
2688 - test whether the structures get packed (and bail out if yes. Damn.
2689 Alignment is a serious matter)
2690 - test whether the structures get compacted (but respecting the alignment
2691 constraints of each types)
2692 - test whether the array fields of structures can straddle alignment boundaries
2694 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2695 (took me 2 hours to find that bug, looking at the wrong place)
2696 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2697 before sending/receiving and are used to compute the offsets of fields
2698 - Given that padding size depend even on compiler options, keep track of
2699 alignment and aligned_size only for the current architecture. Not a big
2700 deal since we send structure fields one after the other (seems
2702 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2703 program, let it work on linux/gcc/little32. portability todo.
2705 2004-07-08 Martin Quinson
2706 - import and improve remote compilation support from FAST
2707 - make sure make check works on half a dozen of machines out there
2709 2004-07-07 Martin Quinson
2710 Let's say it's version 0.0.3 ;)
2711 - Implement conversions (yuhu!)
2712 - Let it work on solaris (beside conversion, of course)
2713 - Stupid me, using rand() to generate the conversion datatests in not wise.
2715 2004-07-06 Martin Quinson
2716 - Let make dist work, since I'm gonna need it to compile on remote hosts
2717 - Let Tests/datadesc_usage write the architecture on which the file was
2718 generated as first byte.
2719 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2720 architecture descriptions.
2721 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2722 on those architectures.
2723 - Optimization: send/recv array of scalar in one shoot
2725 2004-07-05 Martin Quinson
2726 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2728 - Plug a whole bunch of memleaks
2729 - each process now have to call gras_{init,exit}. One day, their log
2730 settings will be separated
2731 - Continue the code factorisation between SG, RL and common in Transport.
2733 2004-07-04 Martin Quinson
2735 - Redistribution between SG and RL.
2736 We wanna have to accept in SG, so move accepted related parts of RL in
2737 the common part. (more precisely, the dynar of all known sockets is no
2738 more a static in transport.c, but part of the process_data)
2740 [gras_stub_generator]
2741 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2743 2004-07-03 Martin Quinson
2744 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2745 them all up in one shot)
2747 - Fix the ugly case of reference to dynamic array.
2748 - New (semi-public) function gras_datadesc_size to allow the messaging
2749 layer to malloc the needed space for the buffer.
2751 - gras_socket_close now expect the socket to close (and not its address to
2752 put NULL in it after it). This is because the socket passed to handlers
2753 is one of their argument (=> not writable).
2755 - propagate the interface cleanup from last week in datadesc, ie remove a
2756 superfluous level of indirection. User pass adress of variable
2757 containing data (both when sending and receiving), and not of a variable
2758 being a pointer to the data. Let's say that I like it better ;)
2759 The price for that is constructs like "int msg=*(int*)payload" in
2760 handlers, but it's a fine price, IMHO.
2762 - Let it work in RL (yuhu)
2764 2004-06-21 Martin Quinson
2766 - porting SG plugin and SG select to new standards (works almost).
2767 - plug memleaks and fix bugs around.
2770 - cleanup the prototype of data recv and force users to specify when they
2771 want to handle references to objects. Test case working even for cycles.
2772 - plug memleaks. Valgrind is perfectly ok with this.
2774 2004-06-12 Martin Quinson
2776 - cleanup the separation between plugin and main code in plugin creation
2778 2004-06-11 Martin Quinson
2780 - Reput hook for raw sockets, needed for BW experiments
2781 - kill a few lines of dead code
2782 [Data description] Interface cleanup
2783 - gras_datadesc_by_name returns the searched type or NULL.
2784 That way, no variable is needed to use a type desc once, which makes
2786 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2787 two parameters were strings (field name, type name), leading to
2789 [Dicos] Interface cleanup
2790 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2791 This is consistant with the dynar API.
2793 2004-04-21 Martin Quinson
2795 - Porting to new standards.
2797 - interface cleanup.
2798 There is no bag anymore, no need to take extra provision to mask the
2799 pointers behind "ID".
2800 Better splitup of functions between files create/exchange/convert.
2801 This is still a bit artificial since convert and receive are so
2802 interleaved, but anyway.
2804 - add a queued message list to procdata (the ones not matching criteria
2806 - factorize some more code between SG and RL wrt procdata
2808 - use gras_exit in example to track memleaks
2809 - get rid of gs_example now that GS is properly integrated into gras
2810 - update run_test to integrate the latest tests (datadesc)
2812 - rename WARNINGn macros to WARNn since it prooved error-prone
2814 2004-04-19 Martin Quinson
2816 - register init/exit functions within gras module mechanism
2817 - send/receive function.
2818 Convertion is not implemented, but short-cutted if not needed.
2819 struct/array elements are sent one by one (instead of block-wise), but
2820 nobody really cares (yet). Get a prototype before optimizing.
2821 - tests (using a file socket) for DD send/receive on:
2822 - base types: int, float
2823 - array: fixed size, string (ie ref to dynamic string)
2824 - structure: homogeneous, heterogeneous
2825 - chained list, graph with cycle
2826 Believe it or not, valgrind is not too unhappy with the results. The
2827 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2831 - Bugfix when using a filename explicitely (instead of '-')
2833 2004-04-09 Martin Quinson
2835 - factorize more code between RL and SG in socket creation
2836 - Complete the implementation and tests of:
2838 o file (only in RL, and mainly for debugging)
2840 I lost 3 days to design a portable address resolver, and then decided
2841 that the prototype mainly have to run on my box.
2842 Addressing portability too early may be like optimizing too early :-/
2844 - use gras_init in the Tests instead of the crappy parse_log_opt
2845 (the latter function is removed)
2846 [Conditional execution]
2847 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2848 [Code reorganisation]
2849 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2850 Build examples against the RL library, since there is no way to disable
2851 its creation for now.
2853 For information, the beginning of coding on GRAS was back in june
2854 2003. I guess that every line has been rewritten at least twice since