1 SimGrid (3.12) NOT RELEASED; urgency=low
4 * Interface improvement:
5 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
6 - Sanitize the interface in MSG_task_ module:
7 - Merge two functions that were close enough but misleading:
8 set_compute_duration(t) -> set_flops_amount(t)
9 get_remaining_computation(t) -> set_flops_amount(t)
10 - set_data_size(t) -> set_bytes_amount(t)
11 get_data_size(t) -> get_bytes_amount(t)
12 - Massive cleanups in the functions related to the energy
14 * Interface improvement:
15 - Rename Host.isAvail() to Host.isOn()
16 - Rename Process.currentProcess() to Process.getCurrentProcess()
17 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
21 - SIMIX_process_throw: raises an exception in a remote process
22 * Refactoring: Separate submodules
23 - libsmx: the public interface, as libc in a real system
24 - popping: the strange dance that converts a user request into a kernel handling
25 - smx_context_*: the virtualization mechanisms that embeed the user code
26 - smx_*: the handling of each simcalls
27 * Rename smx_action_t into smx_synchro_t, making explicit that these
28 things are used to synchronize processes with their environment.
29 For example, a communication is a sort of synchronization involving
30 the communicating processes (that may block until the exchange) and
31 the platform. The same can be said from computations, etc.
35 - 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.
36 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
37 - MPI_Type_set_name, MPI_Type_get_name
38 - MPI_*_c2f and MPI_*_f2c functions
39 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
40 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
41 - Activate a lot of new tests from the mpich 3 testsuite
43 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
44 - InfiniBand network model added : Based on the works of Jerome Vienne (http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf )
45 - When smpi/display_timing is set, also display global simulation time and application times
46 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
47 * Collective communications
48 - 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
49 - 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/).
50 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
52 - Replay now uses algorithms from wanted collective selector
53 - Replay can be used with SMP-aware algorithms
54 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
55 - 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.
57 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
58 - Correctly use loopback on fat-tree clusters
59 - Asynchronous small messages shouldn't trigger deadlocks anymore
62 - "Full" network optimization flag was broken since Surf++
63 - Better handling of precision flags in maxmin
64 - Fix bug causing sometimes "Impossible" errors
65 - Properly pass cluster properties to included hosts
66 * Improvement of the Energy plugin.
67 - Always update the consumption before returning that value
68 - New property: watt_off to denote the disipation when the host is off
71 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
72 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
73 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
75 * Tracing is now always enabled (no way to turn it out)
76 * Move headers around to sort them out on installed systems:
77 - instr/instr.h -> simgrid/instr.h
78 - instr/jedule/* -> simgrid/jedule
79 - simdag/datatypes.h was removed
80 - simdag/simdag.h -> simgrid/simdag.h
81 - msg/datatypes.h was removed
82 - msg/msg.h -> simgrid/msg.h
84 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
86 SimGrid (3.11) stable; urgency=low
91 * Normalizing pointers addresses tool for better diff between logs
94 * Add cloud examples using new VMs
95 - examples/msg/cloud/two_tasks_vm.tesh
96 - examples/msg/cloud/simple_vm.tesh
97 - examples/java/cloud/cloud.tesh
98 - examples/java/cloud/migration/migration.tesh
99 * Add java surf examples:
100 - examples/java/surfPlugin/surf_plugin.tesh
101 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
102 - examples/java/surfCpuModel/surf_cpu_model.tesh
103 * Add SMPI+MSG example:
104 - examples/smpi/smpi_msg_masterslave/
109 - msg task destroy cancel test
110 - msg_host on/off test
111 * Move all tests in testsuite to teshsuite (adding tesh files)
112 * Restructure teshsuites
113 - one folder for each kind of test
114 * Restructure AddTests.cmake
116 - structure the order of tests (with sections)
119 * Add virtual machine
120 - creation of a VM on a PM
121 - migration of a VM from a PM to another PM
123 - MSG_process_join(msg_process_t process, double timeout)
124 - msg_bar_t MSG_barrier_init(unsigned int count)
125 - int MSG_barrier_wait(msg_bar_t barrier)
126 - void MSG_barrier_destroy(msg_bar_t barrier)
127 - msg_as_t MSG_environment_as_get_by_name(const char * name)
128 * New option "msg/debug_multiple_use" to help debugging when a task is used
131 - Improvements and finalization of MSG_storage, MSG_file APIs and their
133 - Increase code coverage in test suites
137 * Protect context stacks against stack overflow. The number of protected memory
138 pages allocated on the top of each stack (1 by default) can be configured
139 with the new command line option --cfg=contexts/guard_size:<value>.
140 * Simcalls are now generated by a python script that
141 - generates files included by SimGrid
142 - checks that all the functions exist, and proposes prototypes
144 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
145 host_set_data, host_get_data
147 - simcall_process_join(smx_process_t process, double timeout)
148 * Fix bug where sleeping processing could not be suspended.
151 * Translate surf models from C to C++
152 - Generic classes for all models: Model, Resource, Action
153 - A generic interface for each kind of model (CPU, Network, Storage
154 Workstation, WorkstationVM)
156 * Translate surf routings from C to C++
157 * Add callbacks using sigc++ or boost::signals2
158 - Add callback functions for resource creation/destruction
159 - Add callback functions for action state change
160 - Handle Energy as a plugin
161 * Replace swag by boost::intrusive
162 * Add new routing models for clusters. For documentation, see
163 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
164 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
165 parameters for cluster tag
166 - Fat trees, with topology="FAT_TREE" and
167 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
168 - see examples/platforms/torus_cluster.xml and
169 examples/platforms/fat_tree_cluster.xml
173 * Hostfiles support host:nb_processes construct to deploy several processes on
175 * Collective communication algorithms should not crash if used with
176 improper number of nodes and report the error.
177 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
178 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
179 MPI_Dims_create, MPI_Cart_sub are supported.
180 * New interface to use SMPI programmatically (still depends on MSG for
181 some parts, see examples/smpi/smpi_msg_masterslave) :
182 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
186 * Global variables privatization in MPI executables is now performed at runtime
187 with the option smpi/privatize_global_variables (default:no).
188 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
189 dynamic libraries loaded by the application are not privatized (static
190 linking with these libraries is advised in this case)
193 * Options defined in XML work correctly now.
196 * New cmake option, enable_lib_in_jar, to control whether native libraries are
197 copied into simgrid.jar or not (ON by default). Use this option if you want
198 to reduce the size of the installed simgrid.jar, *and* the native libraries
199 are kept installed elsewhere.
200 * Surf binding with SWIG (code generated in maintainer mode only):
201 - plugin to handle callbacks
202 - CPU model only for the moment
205 * Supernovae build mode is definitively removed. It was used to improve
206 inlining and inter-module optimizations. It is nowadays superseded by
207 link-time optimizations commonly available in compilers.
208 * Update ns-3 find lib. Bindings for ns-3 should work again now.
209 * Add boost dependency for surf++
210 * Add new macro for tests
211 - ADD_TESH(name <tesh_args>)
212 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
216 - xbt_bar_t XBT_barrier_init(unsigned int count)
217 - int XBT_barrier_wait(xbt_bar_t barrier)
218 - void XBT_barrier_destroy(xbt_bar_t barrier)
219 * Make the xbt_os_time module public
221 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
223 SimGrid (3.10) stable; urgency=low
225 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
228 * Reintegrate Java to the main archive as desynchronizing these
229 package is not acceptable anymore (Java is now considered stable)
230 * Add explicit synchronization facilities through semaphores
231 * Bug fix: Task.setDataSize() only changed the C world, not the value
232 cached in the Java world
235 * Dramatically change the way files are handled. API and internals changed, but
236 this part of MSG was not considered as production grade either.
237 * Add explicit synchronization facilities through semaphores
238 * Add a new function MSG_host_get_process_list()
239 * Preliminary DVFS support (see examples/msg/energy/ for details)
242 * SMPI is now included directly in the libsimgrid as the windows
243 linker doesn't force us on splitting it anymore.
244 * Improvements of the SMPI replay tool:
245 - Most of the collective communications are now rooted in the same process as
246 in the original application.
247 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
248 used until now). Multiple data types can now be used in a trace.
249 - The replay tool now supports traces produce either by TAU or a modified
251 - Bug Fix: the compute part of the reduce action is now taken into account.
252 - Gatherv collective is now supported
253 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
254 of smpirun outputs time independent traces for the current run. One file
255 is created per process. If too many processes are simulated, this behavior
256 can be changed to one file for all processes by using the
257 tracing/smpi/format/ti_one_file flag
258 * smpirun generates the host file if needed (with given host count and platform)
259 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
260 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
261 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
262 --cfg=smpi/coll_selector:(mpich/ompi)
263 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
265 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
266 happening inside a collective SMPI call.
267 * Fix the behavior of complex data types handling.
268 * Make MPI_Wtime another synchronization point to take computations into
270 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
271 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
272 * Add all missing Fortran bindings, SMPI should work with Fortran 90
273 (no privatization of global variables yet)
274 * Preliminary DVFS support (see examples/smpi/energy/ for details)
277 * Verification of liveness properties is now available for SMPI applications
278 (in addition to MSG applications)
279 * Bugged examples using SMPI in examples/smpi/mc/
280 * Add --cfg=model-check/visited option. Allows the verification of infinite
281 programs. Detection of loops in the execution thanks to the system state
282 comparison and reduction of the state space to explore. Can be combined with
283 DPOR for safety properties.
286 * Allow to change SimGrid configuration (see --help) within the code
287 thanks to SD_config() as it can be done in MSG.
288 * Add a new function SD_task_set_amount() upon user request.
291 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
292 * Remove rule based routing (no more PCRE dependency)
293 * Add a limiter_link option to cluster tag, to specify a maximum reachable
294 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
295 * Add a loopback_bw and loopback_lat options to cluster tag.
296 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
297 routing. Now peers are encapsulated in an AS and have their own private
298 router but this is transparent.
301 * Our own implementation of getline is renamed xbt_getline, and gets
302 used even if the OS provide a getline(). This should reduce the
303 configuration complexity by using the same code on all platforms.
304 * New type: xbt_cfg_elm_boolean.
305 * Allow to use yes/no for boolean configuration options in the command line.
306 * Allow to disable SimGrid cleanups at exit from command line option.
307 There are situations where one may want a simulation to end with an exit.
308 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
309 annoying when scripting around the simulator. Adding a
310 --cfg=clean_atexit:no allows to circumvent this issue.
313 * Lots of memory leaks were corrected in this release.
314 There is no known memory leaks anymore, in all of our 600+ tests.
315 * New command line option --version, to get SimGrid version information.
316 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
318 * Supernovae builds are deprecated, and expected to be removed in the next
321 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
323 SimGrid (3.9) stable; urgency=low
325 The Grasgory release: GRAS is really dead now.
327 * Complete overhaul of the internal host structures scheme.
330 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
331 was considered as experimental and badly maintained since then.
332 * Keeping it was thus a trap to our potential users, that could take
333 it instead of MSG or SMPI by mistake despite is pity state.
334 * GRAS seems to have very few users (if any), and no one volunteered
335 to maintain it further. It also induces a lot of XBT code (for
336 portability sake), that must be maintained too.
337 * For all these reasons, we killed GRAS. If someone wants to revive it
338 in the future, don't cry, our git history still remembers of GRAS.
341 * Major overhaul. Merge our documentation again as time proved that
342 splitting it was really not helping our users.
343 * Further improve the developer documentation to help newcomers
344 hacking on SimGrid itself. The user documentation (and in
345 particular, the beginner documentation) is still in a sorry state.
348 * Now works on Windows too!
349 * Much more extensive test suite, from MPICH
352 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
353 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
354 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
355 * Bug fix: let task be scheduled when the last dependency to be solved is
356 a control dependency.
357 * Remove SD_load_environment_script function.
358 Use the C sg_platf function if you want to declare a platform
362 * New function: MSG_process_get_number()
363 * Old function documented: MSG_config()
364 * Remove MSG_load_platform_script function
365 Use the C sg_platf function if you want to declare a platform
369 * Change the default value of the TCP_gamma constant (maximal size of TCP
370 congestion window) to a more realistic 4MiB value. If you notice changes in
371 your simulation results, you can fall back to the previous 20k tiny window
372 by adding --cfg=network/TCP_gamma:20000 on command line.
373 * (Hopefully) fix a bug wrt periodic availability and state traces
374 * Bug fix: use default values at start when first event in availability/state
375 trace is not at time 0.
378 * remove the "new_" part of function name sg_platf_new_trace_connect
379 (resulting in sg_platf_trace_connect), since it does not create
383 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
384 use case where it's really mandatory, and maintaining it was a pain in
386 * New: xbt_fifo_search(), search an item with a user-provided
387 comparison function instead of dumb pointer comparison.
390 * Fix the lua deployment:
391 Use `simgrid.init_application()` before deployment instead of
392 `simgrid.msg_register_application()` after.
395 * Transfer the tracing files into the corresponding modules.
397 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
399 SimGrid (3.8.1) stable; urgency=low
401 The "we are told that some people want to also *install* the simgrid
404 * Add missing file "tesh.1" to the archive.
406 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
408 SimGrid (3.8) stable; urgency=low
410 The Psssshiiiit release: SimGrid jumps into the Cloud.
413 * Add an experimental interface to manipulate VMs. They are mainly
414 process groups with very few intrinsic semantic, but they should
415 allow you to build the semantic you want easily.
416 * New function: MSG_host_set_property_value()
417 * New function: MSG_process_on_exit(). To clean memory in all cases.
418 * Bug fixes that made the host (and link) failures unusable.
419 * Add a way to auto-restart process when the host in which they are
420 executing comes back (ON_FAILURE="RESTART" on deployment file,
421 MSG_process_auto_restart_set).
422 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
423 please stop using the old ones, they are DEPRECATED.
425 * Deprecate functions MSG_global_init() / MSG_global_init_args()
426 Please use MSG_init() instead. (reducing the amount of entry
427 points in the library helps us).
428 * Make it impossible to link against the wrong version of the lib
429 * Deprecate MSG_clean(). No need to call it anymore.
430 * Function MSG_get_host_number() is not deprecated anymore.
433 * Split the doc into a user guide and a reference guide.
434 * Start a developper guide to help people hacking on SimGrid.
437 * Enable tracing by default. This modules rocks you should use it.
438 * Remove option custom_flags. Now use environment variables CFLAGS
440 * Use default cmake things to detect lua instead of home grown ones.
441 * New option "enable_mallocators" to disable mallocators, for debugging
442 purpose ("on" by default).
445 * Bug fixes around the resource failures: don't let the processes
446 survive the host they are running onto.
447 * Add an interface to auto-restart processes when the host in which they are
448 executing comes back.
449 * Ensures that SIMIX_clean is called automatically. It's not part of
450 the public interface anymore (bindings should be updated).
453 * Bug fix for when SD_Simulate is called with a positive value: be careful
454 when comparing doubles. Sometimes they are different for non significant
456 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
457 parallel task whose initial work is distributed among host according
458 to the Amdahl's law. Such tasks are created with a parameter alpha
459 that corresponds to the non-parallelizable part of the computation.
460 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
461 between two sets of workstations assuming a 1D block distribution (each
462 workstation owns a similar share of data) on both sides.
464 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
465 Data redistribution will be automatically scheduled once parent and child
466 are both scheduled. The filling of computation_amount and
467 communication_amount structures is now done seamlessly thanks to the chosen
469 * New function SD_workstation_dump to display various information
470 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
471 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
472 bandwidth on the route onto which the task is scheduled and the amount of
474 To divide the nominal bandwidth by 2, the rate then has to be :
475 rate = bandwidth/(2*amount)
476 * Compute tasks that have failed can now be rescheduled and executed again
477 (from their beginning)
478 * Increasing source code coverage (src/simdag is now covered at 95.8%
482 * Re-implement time-independent trace replay using SMPI (at the
483 smpi_smp_* level) instead of MSG. This should replace
484 examples/msg/actions/actions.c
485 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
486 hindexed and structs)
487 * Implement the exchange of non-contiguous data.
488 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
489 * Correct behavior of smpi/sender_gap and set its default value to 0
490 * Add option to asynchronously send small messages to allow better
491 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
492 specifies the size in bytes under which messages will be asynchronously sent.
493 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
494 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
497 * New C interface to define a platform: XML is now optional.
498 For more info, please check include/simgrid/platf.h.
499 * New interface to define random platforms from the C:
500 For more info, please check include/simgrid/platf_generator.h and
501 examples/msg/masterslave/masterslave_platfgen.c
502 * Export a sg_cmdline dynar containing all the arguments we got from
506 * Two new tracing options for adding comments to trace file so you
507 can track your experiments (see --help-tracing for details).
508 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
509 * Adding the SimGrid version that generated the trace file as a comment.
510 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
511 * Fix to avoid key clashes on Paje links
512 * Other minor fixes related to the Paje specification
515 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
516 and renamed to xbt_str_hash() and xbt_str_hash_ext().
517 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
518 * Greatly improve the robustness of mmalloc to user errors (such as
519 using an area after freeing it, or freeing it twice)
521 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
523 SimGrid (3.7.1) stable; urgency=low
526 * Restore the prototype of MSG_process_create_with_environment() to
527 the pre-3.7 situation by removing the kill_time argument.
528 * Add a MSG_process_set_kill_time() function instead.
531 * Fix weird behaviors when dealing with parallel tasks.
534 * Simgrid is now built as a dll.
535 * Simgrid-java now works on Windows.
536 * Simgrid-Java is now included into Windows package.
539 * First pre-build package for MacOSX.
542 * Fix compilation when using MSG_USE_DEPRECATED.
543 * Fix some compilation issues on Macs and Windows.
544 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
545 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
547 We discovered that the Lua console is broken, but we are missing the
548 manpower to fix it right now. The problem existed in 3.7 too, so we
549 are not blocking the release for that. Sorry if you depended on this
550 feature, any help would be really welcome.
552 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
554 SimGrid (3.7) stable; urgency=low
556 The "spring cleanups (before the next Big Project kicks in)" release.
559 * We can specify the SMPI latency/bandwidth factor with command line
560 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
561 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
562 You can also use the "config tag" from platform file by adding this line
563 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
564 (see "example/platforms/tag_config.xml" to use "config tag").
565 Note that the command line supersedes the platform file configuration.
566 * Change the correction factors used in LMM model, according to
567 the latest experiments described in INRIA RR-7821.
568 Accuracy should be improved this way.
569 * Use the partial invalidation optimization by default for the
570 network too. Should produce the exact same results, only faster.
571 * Major cleanup in surf to merge models and split some optimization
572 mechanisms from the core of the models. As a result you can now
573 specify which model to use (e.g., --cfg=network/model:LV08
574 --cfg=cpu/model:Cas01) and which optimization mode to use
575 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
576 Incompatible combinations should err at initialization. See
577 --help-models for the list of all models and optimization modes.
578 * The CLM03 workstation model was dropped for simplicity because it
579 used the deprecated CM02 network model. Use default instead.
580 * Rename the TCP_gamma configuration option to network/TCP_gamma
581 * Rename the coordinates configuration option to
582 network/coordinates, and document it
583 * Use now crosstraffic keyword instead of the terribly misleading
584 fullduplex keyword. It is activated by default now in the current
585 default model, use --cfg=network/crosstraffic:0 to turn it off.
586 * Ongoing refactoring the model parsing to make XML files optional
587 See include/simgrid/platf.h for details (still to be completed)
590 * Major overhaul of the documentation. Almost instructive now :/
591 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
592 functions and friends. This interface was considered as
593 deprecated since over 2 years, it's time to inform our users that it is.
594 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
595 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
596 use these (crappy) functions in your code.
597 These functions will be removed soon. Stop using them now.
598 * Deprecate MSG_get_host_{table,number}
599 Implement MSG_hosts_as_dynar() instead.
600 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
601 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
604 * Stabilize the parallel execution mode of user contexts
605 * Introduce configuration variables to control parallel execution:
606 - contexts/synchro: Synchronization mode to use when running
607 contexts in parallel (either futex, posix or busy_wait)
608 - contexts/parallel_threshold: Minimal number of user contexts
609 that must be part of a scheduling round to switch to parallel
610 execution mode (raw contexts only)
611 * Fix bugs that prevented to use suspend/resume along with
612 synchronization structures.
613 * Fix bugs in process termination that lead to invalid memory access
614 in very specific conditions.
617 * Introduce a parallel mode for the models (controlled by surf/nthreads
618 configuration item). In our tests, running the models in parallel
619 never lead to any speedups because they are so fast that the gain
620 of computing each model in parallel does not amortizes the
621 synchronization costs, even when ultra fast futexes are used.
622 This is released anyway because YMMV.
625 * Performance boost by using a swag internally to compute the set of
626 tasks that are finished and should constitute the return value of
630 * Enable it by default now that it is considered rather stable.
633 * Documentation of the tracing functions.
634 * Performance gains when tracing categorized/uncategorized resource
635 utilization by avoiding calls to get route when updating resource
636 variables. LMM constraints are being used instead.
637 * API changed to set task categories. Use MSG_task_set_category instead
638 of TRACE_msg_set_task_category, and SD_task_set_category instead
639 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
640 * Bugfix for graphicator, routes not correctly obtained, memory leaks
641 * Examples for link user variables added (see at examples/msg/tracing/)
642 * Deprecated function TRACE_msg_set_process_category completely removed
643 * Trace header updated according to the latest Paje file format
644 * Tracing network lazy updates, no longer obligate users to use full updates
645 * --cfg=tracing/platform:1 also registers power/bandwidth variables
646 * Experimental: let user code declare/set/push/pop application states for hosts
647 * API changed to allow the manual creation of graph configuration files
648 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
651 * Improve the API of Lua MSG bindings, using the Lua spirit.
652 * Each simulated process now lives in its own Lua world (globals are
653 automatically duplicated). It helps writing simulators. Will allow
654 to run Splay programs within SimGrid in the future.
655 * Add a Chord example in Lua, equivalent to the MSG one.
658 * Start the implementation of a solution to express temporal
659 properties, not only local assertions. This is still an
660 experimental work in progress, stay clear from it to be safe.
664 - Add new runtime parameters --help-logs and --help-log-categories
665 to display informations about supported logging parameters and
667 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
668 don't exists anymore.
669 * Mallocators: allow value NULL for the reset function.
671 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
672 create homogeneous dictionaries, where all the elements share the
673 same free function. Non homogeneous dictionaries will be
674 deprecated in the next release.
675 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
676 - Multi-level dictionaries are deprecated.
678 - new function xbt_dynar_search_or_negative() that is useful when
679 you have less than 2 million elements in your dynar and don't
680 want of the extra complexity of catching exceptions when the
681 element is not found.
683 - Make xbt_os_thread module (for thread portability) public.
684 Documentation is still to come, sorry.
686 - Cleanups and simplifications to make it maintainable again.
687 - Exotic features (such as memalign and valloc) were removed.
688 - The metadata were extended and improved so that the
689 model-checker becomes able to explore and inspect the heaps.
690 - This may induce a performance drop when enable_model-checking is
691 ON in cmake (even if it's not used in the simulation), but it is
692 necessary at this point to get MC working.
694 Turn model-checking OFF if simulation performance matters to you.
695 Not enabling it at runtime is not enough, disable it in cmake.
697 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
699 SimGrid (3.6.2) stable; urgency=low
701 The "Not coding new stuff allows to polish old things" release.
704 * New bindings to the NS3 packet level simulator (experimental)
705 * Use the raw (efficient) execution contexts instead of the sysv
706 (portable) ones when possible.
707 * libpcre is now mandatory in any cases since not using it led to
708 severe performance loss and possibly other issues
709 * Update the XML platforms:
710 - G5K: include the latest machine in Nancy
711 - GridPP and LCG: new platforms
712 * Documentation was partially updated, at least (more to come)
714 Bug fixes, cosmetics and small improvements
715 * Free terminated processes before the end of the simulation to avoid
716 exhausting the memory of users having very dynamic amount of
718 * Bug fix and cosmetics about canceling non-running tasks
719 * Bug fix about the dot loader's issues when using libcgraph
722 * Create an installer for windows with nsis (amd64 and win32)
723 - Add an hello world project to illustrate simgrid project creation.
724 - Embed libpcre into the Simgrid installer to avoid
725 its compilation burden
726 * The raw execution contexts should work on Apple now
727 * Port to Windows 64 bits
728 - Sysv contexts now have an implementation for this arch
729 - GRAS communication features now support this arch
730 * Drop support for borland compiler on windows
731 - this code was not maintained, and we kinda depend on gcc nowadays
732 * Fix portability issues on kfreebsd/gnu: build error about semaphores
733 * Fix portability issue on unstable ubuntu: linker became picky on
736 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
738 SimGrid (3.6.1) stable; urgency=low
740 The "Oops, we broke Macs too" release
743 * Fixed contexts detection so that raw ones are used when possible
744 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
745 leads to a strange error, with user code segfaulting sometimes when
746 the generated code is not perfectly aligned (which is not
747 controllable from the user side, depends on the amount of code)
750 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
752 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
754 SimGrid (3.6) stable; urgency=medium
756 The Summer Release, also known as the "OMG! They Killed Kenny!" version
759 * Bindings now constitute their own package, separated from the main one.
760 Rationale: reduce our maintenance nightmare by reducing the module coupling
761 They will soon be released on their own on gforge.
763 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
764 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
766 GRAS: It is not considered as stable anymore, but experimental. Sorry.
767 * It's not quite deprecated for now because we have no replacement,
768 but it may soon become the case.
771 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
772 * Fortran: New user-level cache variable to store the rank of the running
773 process. This improves performance by an order of magnitude.
774 * C: New coccinelle script to automatically locate and modify global and
775 local static variables.
776 * Improved SMPI network model with a sender-side gap to account for multiple
780 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
781 only say if a communication is finished, no matter whether it succeeded or
782 failed. You can call MSG_comm_get_status() to know the status of a finished
784 * New function MSG_task_dsend() to send a task and detach it. When a
785 communication is detached, you are never notified of its success or failure
786 and the memory is released automatically once it is finished. This function
787 is useful when you don't care about the end nor the success of a
789 * Change the prototypes of action replay. Sorry for inconvenience,
790 but this is really more efficient this way (and to adapt your code,
791 you just have to fix the initialization, that shouldn't be too long)
792 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
793 ever managed to do anything useful with it.
794 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
795 takes a dynar of communications. It returns immediately and gives the
796 index of a finished communication (if any).
797 * New example: a basic implementation of the Chord P2P algorithm.
800 * New model for multi-core CPUs. You can now use the core attribute to
801 precise the number of cores of a host. This is a basic model. Every
802 process running on the host receives at most the power provided in
803 the DTD (throughput<=power). Total throughput of process cannot exceed
805 * New peer tag. This peer tag creates a tiny AS comprising a host and a
806 router linked by an up-link and a down-link (possibly asymmetrical).
807 This kind of pattern allows to easily build last-mile model style platforms.
808 Aggregating such patterns in a rule-based AS is thus the technique of
809 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
810 * New model for Vivaldi routing. We transformed the Vivaldi network model
811 into a Vivaldi routing model (based on the rule-based model). This allows to
812 combine Vivaldi based latencies with last-mile platforms.
815 * Added a check for NaN of IEEE754 infinite in the double entries of
817 * Introduce a new context factory "raw", highly inspirated from the
818 ucontext factory, but using manually crafted functions in assembly to
819 do the work in an efficient manner.
820 * Allow to change the used context factory at run time, not only at
821 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
822 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
823 contexts at runtime (only with raw contexts or ucontexts).
824 * Completely rewrote this module to allow parallel execution of user
825 processes. Use --cfg=contexts/nthreads:N to execute user processes
826 with N parallel threads (the default is 1, meaning no parallelism).
827 * Allow to decide dynamically between sequential and parallel modes.
828 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
829 processes in parallel only when their number is greater than or equal to P
831 * Added a check for NaN of IEEE754 infinite in the double entries of
835 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
836 won't output the state of processes when interrupted with Ctrl-C
837 * Add a new function xbt_dynar_to_array that transforms a dynar into a
838 NULL-terminated array. This may solve backward compatibility issues
839 due to the change to return type of SD_simulate. See also:
840 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
841 * Add new macros with variable number of arguments.
842 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
843 - in xbt/asserts.h: xbt_assert
844 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
845 - in xbt/ex.h: THROWF and RETHROWF.
846 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
848 * Change xbt_die() to accept a format string with arguments, just like printf.
849 * New data structure: xbt_lib_t, like a dict but more general and with better
853 * New configuration options
854 Options triva/categorized and triva/uncategorized can be used to generate
855 graph configuration files for Triva visualization tool.
856 * Configuration option tracing/platform is renamed to tracing/categorized
857 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
858 * New TRACE_user_link_variable interface:
859 User provides the name of the link and the tracing variable to attach to it
860 * the declaration of tracing categories must be done after the environment
862 * simpler tracing interface, just one way to declare categories
863 TRACE_category or TRACE_category_with_color, it is up to you
864 * links in the trace file are again identified by their names
865 * trace contains the full platform hierarchy exactly as declared using the ASes
866 * Options tracing/msg/[task|process]:1 groups the process by hosts
867 for both cases, tasks and processes must have names that are unique during the simulation
868 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
869 * The experimental option tracing/msg/volume is deprecated
870 its functionality may be reincorporated if needed
872 The tracing generates a trace file with unordered timestamped events,
873 because of the way the core simulator (surf) works. A script available
874 at the tools directory (fix-paje-trace.sh) can be used to put the events
875 in order. We have changed the tracing so it can generate ordered timestamped
876 events in the final trace, but depending on the simulator (and how much time
877 is simulated) that can lead to a huge memory utilization. It is deactivated
878 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
881 * Define a SIMGRID_VERSION macro in simgrid_config.h.
882 - We are trying hard to keep the API stable, but it may happen that
883 some things change (we're a research project after all, not a
884 nuclear plant operating system). If such things should happen, you
885 could rely on that macro to adapt.
886 - current value: 30600 for 3.06.00, aka 3.6
887 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
888 a number that can be compared with SIMGRID_VERSION.
889 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
890 assertions and verbose/debug logging events are disabled at compile time.
892 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
894 SimGrid (3.5) stable; urgency=medium
897 * New feature to any SimGrid-based simulator: Model-Checking
898 Check SIN#1 for more details.
901 * New Model SMPI (three-interval linear regression for correction factors)
902 See RR-7426, available at http://hal.inria.fr/inria-00527150
903 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
904 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
905 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
906 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
907 * See also src/smpi/README
912 - Tracing API changes: TRACE_start and TRACE_end should not be called
913 by user-code. They are automatically called by simulators created
914 with SimDAG, MSG and SMPI if the toolkit is compiled with
915 tracing_enabled=ON. Categories declaration and utilization remain the
916 same for MSG and SimDag.
917 - A function was added to the tracing API to declare categories with
919 - TRACE_category_with_color (char *category, char *color)
920 where color must be in the following format
921 "%f %f %f", red, green, blue
922 and red, green, blue are float values in the interval [0, 1]
923 - User can specify NULL as color parameter, or continue calling
925 On that case, the tracing system will define random colors
926 - The following command-line options are supported:
927 --cfg=tracing/filename:msg.trace
928 --cfg=tracing:1 (activate tracing, needed to use others)
929 --cfg=tracing/platform:1 (categorized resource use)
930 --cfg=tracing/uncategorized:1 (uncategorized resource use)
931 --cfg=tracing/msg/task:1 (task creation)
932 --cfg=tracing/msg/process:1 (process creation, migration)
933 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
934 --cfg=tracing/smpi:1 (SMPI interface tracing)
935 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
936 - examples of examples/msg/tracing updated
938 - DAXLoader and DOTLoader functions can generate tasks with categories
939 - A new function to attribute a category to SD tasks:
940 TRACE_sd_set_task_category (SD_task_t task, char *category)
941 * Tracing the MPI interface implemented by SMPI
942 - Collective operations are traced with states
943 - Point-to-Point operations are traced with states/links
944 - Tracing activated by a parameter "-trace filename" passed
945 to smpirun during execution (considering that simgrid
946 is compiled with tracing enabled)
947 - To run the simulation with gdb, the simulator
948 accepts --cfg=tracing/smpi:1 to trace SMPI
949 - tesh files to check if smpi tracing is ok
950 - See examples/smpi/NAS/DT-trace
951 * GTNetS tracing re-worked
952 - adaptation to the tracing system of GTNets to cope
953 with modifications regarding the fullduplex mode
954 - new tesh files to check if gtnets tracing is ok
957 * Asynchronous communications through the functions:
958 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
959 * New function: MSG_load_platform_script()
960 to make possible using a lua script instead of XML files to set up platforms
961 * New function: MSG_set_function
962 to associate functions to processes, used when bypassing the parser
963 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
965 Platforms: Add some more examples in examples/platforms
966 * Grid'5000: see www.grid5000.fr
967 * *_30000_hosts.xml: various huge files [mainly scalability testing]
970 * Change the XML format. This is a very important modification. SimGrid
971 3.5 introduces a new hierarchical format based on the notion of
972 Autonomous Systems. Compatibility with old format is ensured
973 through the perl script provided in the install bin directory
974 bin/simgrid_update_xml.
975 It is now possible to build platforms with specific routing
976 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
977 connect several platforms together. We will try to provide soon
978 set of realistic platforms exploiting these properties (have a
979 look at examples/platforms/ for the moment).
980 * Take the opportunity of the XML format change to be a good XML citizen:
981 rename link:ctn to link_ctn and similar changes (also dealt with by
983 * Add a new routing scheme (rule-based) using regular expressions. It
984 enables to have an extremely low memory footprint when the
985 underlying routing is simple and can be compactly described. You
986 need to have libpcre4-dev (perl regular expressions) installed if
987 you want to use this routing scheme.
988 * Revive the cluster TAG and allow to easily and efficiently (both in
989 term of memory and speed) connect clusters together. Have a look
990 at teshsuite/simdag/platforms/ to see how this can be done. With
991 this tag, you can create clusters with thousands of tasks at no
992 cost (have a look at examples/platforms/).
993 Note: clusters are implemented as ASes, so there is no need for an
994 enclosing AS tag if you have only one cluster in your platform.
995 * Add new generic functions in the public interface that allows the user
996 to call SURF 'create_resource' methods from your code (same
997 functionality as the XML bypass mechanism but with a much lighter
999 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1000 that uses a piecewise linear approximation to produce better
1001 results when exchanging small messages.
1002 * Add a new parameter to handle correctly full duplex link and account
1003 for interferences between uplink and downlink communications
1004 (activate with --cfg=fullduplex:1).
1007 * Rename the SD_READY (all dependencies are satisfied and task is
1008 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1009 dependencies are satisfied) state.
1010 This prevents a confusion between the notion of "ready to schedule"
1011 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1012 simulated" (SD_RUNNABLE) used by the simulation kernel.
1013 * Change the way a task is considered as ready. Instead of removing
1014 dependencies when a task is done, a counter is decreased. This way,
1015 it is always possible to reach ancestors thanks to the
1016 SD_taks_get_parents function (even after the end of the simulation.)
1017 * Change the return type of SD_Simulate from (SD_task_t*) into
1018 xbt_dynar_t. This function was in handling a dynar internally and
1019 converted it into a NULL terminated array for historical reasons.
1020 * New function SD_dotload(char*) to load a DAG described in dot
1021 format. This loader and the corresponding examples require the
1022 installation of the graphviz library.
1023 * Fix a bug in the management of tasks of size 0 in the surf network
1024 models. This problem was only visible with SIMDAG and you should
1025 thus disregard results produced with earlier versions if you
1026 relied on this feature (some tasks were blocked because of this).
1027 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1028 even though your applications were doing only point-to-point
1029 communications and sequential computations. Now you can really use any
1030 model you want (of course, if you create real parallel tasks, which are
1031 not implemented in most models beside ptaskL07, this will abort).
1032 * Add an example that schedules a DAX on an heterogeneous platform
1033 using a Min-Min strategy.
1034 * New function SD_workstation_get_current_task() that returns the kind
1035 of task currently running on a workstation in the sequential access
1037 * Raise some warnings when unexecuted tasks remains at the end of the
1038 simulation. This is usually caused by cycles in the DAG.
1041 * New function: SIMIX_process_set_function() called by MSG_set_function
1042 * Change the underlying waiting queue in semaphores so that a process
1043 can wait on several of them simultaneously (as in waitany).
1044 * Fix the way to handle tokens in semaphores so that all access patterns
1045 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1046 * kill the dirty pimple SIMIX_message_sizes_output()
1047 Please use (proper) visualization instead
1050 * New data container: setset (set of sets of elements)
1051 * New module: mmalloc (mapped malloc, allowing to have several
1052 independent segments of malloc)
1053 * New function: xbt_dict_cursor_set_data()
1054 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1055 * New function: xbt_dynar_is_empty()
1056 * New function: xbt_fifo_get_last_item()
1057 * Fix xbt_dynar_shrink(): use the right element size.
1058 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1059 uninitialized areas during expand.
1060 * Fix semaphores: previous implementation was severely broken.
1061 * Use library init/fini functions for our initialization.
1062 - you can use logs and other feature as soon as you want in your
1063 code (even before the xbt_init / MSG_init)
1064 - xbt_exit is now a no-op and produce a warning when used.
1067 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1068 benefit from the latest improvement to the simulation kernel.
1069 * Kill measurement sockets for now. If you rely on them, sorry. This
1070 release is not for you. This feature will be reintroduced in the
1071 future, but we cannot delay the release any further.
1072 * New function: gras_msgtype_get_name().
1073 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1074 * Fix (at last) the pmm example: it should not randomly fail anymore.
1076 Build chain: bug fixes and overall polishing
1077 * Cmake is now stable enough. Hence, we killed the autotools.
1078 * Port to windows ( TM :)
1079 * Fix the 'make install' target.
1080 No need to use 'make install-simgrid' anymore
1081 * Introduce a 'make dist' target compiling a *source* archive
1082 'make package' compiles a binary archive
1083 * Compile java files only on need
1084 * Add --cd and --setenv command line options to tesh
1085 * Out of source builds are not fully supported yet, but we are close
1086 * Enable supernovae and optimization flags by default for our users
1089 * Add layer to set up environment directly from lua, without XML.
1090 * The effect of gras_stub_generator can be achieved through
1091 lua too (check examples/gras/console/ping_generator.lua)
1093 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1095 SimGrid (3.4.1) stable; urgency=low
1097 The "Polishing easter eggs is probably a good idea" release.
1098 This is a bug fixes release only.
1102 * Fix a bug preventing the tasks from begin garbage collected.
1105 * Fix a bug occuring when a host involved in a communication fails.
1106 This was not detected properly by the other peer involved in the
1107 communication. Now, it's reported as a network error.
1110 * Warn the user about loop dependencies in data flow of DAX files
1111 * Obey the control-flow dependencies of DAX files
1114 * Add option "enable_smpi" allowing to not compile SMPI.
1115 Probably useful for the (Mac) users experiencing a build error here
1116 * Improve the detection of lua5.1 and ruby1.8
1118 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1120 SimGrid (3.4) stable; urgency=low
1122 The "Easter in Cargese" release. Also known as (major changes):
1124 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1125 ~> bindings were greatly improved
1126 ~> new tracing infrastructure for better visualization introduced
1128 * the "Welcome to configury modernity" release.
1129 ~> we switched from autotools to cmake, and improved our cdash
1132 A more detailled list of changes follow (full detail in svn log).
1134 Java Bindings: Various Cleanups
1135 * (install java-gcj-compat-dev on debian-like to use them)
1136 * Remove put/get: no need to export deprecated interface in Java
1137 Use send/receive instead.
1138 * Cleanup the examples and add a README per directory
1139 * Remove example autoDestination (that's the only way to go now)
1140 * Remove example explicitDestination (was a plain copy of basic)
1141 * Make JniException a runtime exception, so that there is no need to
1142 declare the fact that you may encounter such a beast. I guess that
1143 nobody will ever want to survive such error.
1144 * Create specific errors for each MSG case of failure:
1145 host failure, transfer failure, timeout, task cancelled
1146 * Cleanup the exceptions that may get thrown by each function
1147 * Other internal cleanups in Java bindings. Performance still bad :/
1148 Ruby and Lua Bindings: create them
1149 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1150 * That's new and great, you should try them out.
1151 Same functionalities than Java bindings, only even less polished
1153 * Kill the useless "rate" argument of SD_task_get_execution_time()
1154 Everyone used to provide -1 as a value, it was not used, and the
1155 semantic of a possible use wasn't even clear.
1156 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1157 either as comm costs or compute costs to mean that there is no
1158 such thing for that specific task.
1159 * Add a SD_task_set_name() function
1160 * Fix SD_task_unschedule() on typed tasks
1161 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1162 * In DAX loader, accept useless 'level' attributes to <job> since
1163 LIGO DAGs have them (seem to be to ease graphical representation).
1165 * Add an example masterslave_mailbox.c using send/receive and not
1166 the deprecated put/get interface.
1167 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1168 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1170 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1171 (declare MSG_USE_DEPRECATED to still have the old name)
1172 * Add a MSG_task_set_data() function
1173 * About trace replay (see examples/msg/actions):
1175 - Allow to work with splitted trace files for each process
1176 Give the specific trace file as argument of each process,
1177 and call MSG_action_trace_run(NULL)
1178 You can still have one merged file for all processes.
1179 - Fix implementation of collective operations
1180 * Allow task_execute() on 0-sized tasks (closes #10063)
1182 * This is the first release of SimGrid where SMPI is not considered
1183 beta anymore (even if some corners should still be improved)
1184 * Port over the new SIMIX_network submodule (internal refactoring)
1185 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1186 * Implement more missing elements of the standard:
1188 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1190 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1191 MPI_Scan() MPI_Get_processor_name()
1192 - Added implementation of missing case for Alltoall (warning: it's
1193 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1194 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1195 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1196 SMPI_MPI_Allgatherv()
1197 * Bug fixes include:
1198 - MPI_Waitsome() was broken
1199 - Allow relative includes in smpicc
1200 - Command line cfg argument 'reference_speed' was ignored...
1201 - Some functions did not properly lead to auto-benching of user code
1202 - smpicc passes -O2 by default (just like openmpi one)
1204 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1205 * Bug fixes about timeouts during communications
1206 * add SIMIX_message_sizes_output() as a pimple to write to file the
1207 amount of messages per size. Use gnuplot to get histogram.
1208 Pimple because that's the only user-visible function of simix,
1209 defined directly in xbt.h (irk, sorry)
1211 - Add a SIMIX_sem_get_capacity() function
1212 - Fix interactions with processe resume/suspende
1213 - release_forever() was stupidly broken
1214 - Fix SIMIX_display_process_status() for processes in a semaphore
1215 - Make SIMIX_sem_block_onto() user-visible
1216 * Refactoring context stuff:
1217 - Use pseudo-OOP for better modularity
1218 - reimplement SIMIX_process_kill() without process_schedule() so
1219 that the latter can take as invariant that it is called from
1221 - Merge context_start into context_new for sake of simplicity
1223 * Add a Vivaldi network model, coded live during SUD'10 ;)
1224 * Rename configuration variables to start a hierarchy:
1225 o cpu_model -> cpu/model
1226 o network_model -> network/model
1227 o workstation_model -> workstation/model
1228 * New configuration variables:
1229 o network/bandwidth_factor: correction to bandwith
1230 o network/latency_factor: correction to latency
1231 o netwotk/weight_S: correction to the weight of competing streams
1232 * Add a long description to the models, that users can see with such
1233 argument on the command line: --cfg=cpu/model:help
1234 * --help-models display the long description of all known models
1236 * config: add the ability to set a default value after registration
1237 Does not override any previously set value (e.g. from cmd line)
1238 * dict: allow to have integer key and data.
1239 When so, you need to use the following functions
1240 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1241 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1242 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1243 In contrary to regular dicts, the key is not malloced before copy.
1244 Mixing scalar and regular elements in the same dict is not tested
1246 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1247 Tracing for Visualization:
1248 * SimGrid is now instrumented in order to generate a trace file for
1249 visualization analysis: to use it, need to compile SimGrid with the
1250 "tracing" option enabled, and instrument the program using SimGrid with
1251 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1252 (among other functions).
1253 * The instrumentation only traces the platform utilization for now
1254 * Documentation to use the tracing functions and how to analyze the
1255 traces with the Triva tool is written.
1256 * More information about: SimGrid FAQ (in the section Tracing Simulations
1259 * We moved to cmake as default build system. Autotools support will
1260 be dropped soon. Check the FAQ for more info about how to use it.
1261 * Greatly improved our cdash/ctest interactions
1262 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1263 * Added memory checking tests with valgrind; lot of memleak fixing.
1264 This may be the first release of simgrid with so few memory issues
1265 * Added code coverage tests.
1266 Our coverage is still improvable, but at least we see it on cdash.
1268 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1270 SimGrid (3.3.4) stable; urgency=low
1272 The "Desktop Grid needs love too" release (also called Xmas release).
1274 Models improvements:
1275 * Major speedup in the maxmin system solving by using lazy evaluation
1276 Instead of solving completely the maxmin system at each iteration,
1277 only invalidate (and recompute) the modified parts.
1278 This new feature is enabled in default models but you can try to
1279 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1280 * Cas01 IMproved as default CPU model
1281 This CPU model is the same Cas01 model, but it uses the
1282 maxmin-selective-update flag and a heap structure to manage
1283 actions on SURF kernel.
1284 It reduces the complexity to find the next action to finish and,
1285 consequently, it's faster than the old Cas01.
1286 This is the new default CPU model (Cas01).
1287 * Rename the old Cas01 model to Cas01_fullupdate
1288 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1289 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1290 * CpuTI (CPU Trace Integration)
1291 A new CPU model whose objective is simulate faster when using
1292 availability trace files.
1293 Instead of using a full featured, over engineered maxmin system for
1294 CPU modeling, this model does the pre-integration of traces files
1295 to calculate the amount of CPU power available, and so, executes
1296 faster than the old CPU models.
1297 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1298 * Use LV08 as default network model since it gives better accuracy
1299 for small messages and shouldn't change things for big ones.
1300 Use --cfg=network_model:CM02 to get the previous behavior.
1303 ******************************************
1304 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1305 ******************************************
1306 * The new CPU model may changes simulations!
1307 The point is that events occurring at the exact same timestamp
1308 are not scheduled in the same order with the old and new
1309 version. This may be enough to completely change the execution
1310 of simulations in some cases.
1311 * The new network model will change simulations!
1312 This new model is more realistic than the previous one, so you
1313 should consider redoing your old experiments with this model.
1314 Sorry for the inconvenience.
1317 * Introduce the supernovae compilation mode
1318 When compiled that way, the whole SimGrid (or almost) is put in a
1319 single compilation unit and compiled in one shoot.
1320 This is to help gcc which has difficulties to inline stuff from one
1322 The speedup seem to be above 15%, althrough more tests are needed on
1323 amd64 to confirm that gain.
1326 * Port of MSG's mailbox on top of SIMIX network
1327 The put/get mechanism was greatly simplified on the way.
1330 * New SIMIX network module. Provides:
1331 - Mailbox: rendez-vous mecanism to find with who you want to speak
1332 - Synchronous send/recv: easier and hopefully faster since the
1333 logic is handled in the maestro process directly now
1334 - Asynchronous send/recv: you dreamt of it? It's here now
1335 Too bad that nobody cared enough to propagate the change to MSG.
1336 * Add semaphores as SIMIX synchronization mechanism.
1339 * new function SD_daxload(char*) to load a DAX file
1340 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1341 * Introduce typed tasks. Specify its kind and cost at creation.
1342 At scheduling, just give where it should be placed, and the cost
1343 for each involved resource is automatically computed.
1344 Existing constructors so far (more to come of course):
1345 - SD_task_create_comm_e2e() for end-to-end communication
1346 - SD_task_create_comp_seq() for sequential computation
1347 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1348 * new function SD_task_dump() for debuging display
1349 * new function SD_task_dotty(task,FILE*) writing to file the info
1350 about the task in dotty format
1351 * SD_task_dependency_exists() can now cope with having one of its
1352 arguments NULL. If so, it tests whether the other argument has any
1354 * Add getters on list of preceding/following tasks:
1355 SD_task_get_parents(task) and SD_task_get_children(task)
1356 * Add getters on amount of workstations and list:
1357 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1358 * Add getter on task kind: SD_task_get_kind(task)
1359 * Update the start_time and finish_time of tasks on completion/failure
1360 * Bugfix: Remove task from state swags when destroyed
1363 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1366 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1367 dynar with one separate thread per value of the dynar.
1368 * Change the prototype of xbt_thread_create(), sorry.
1369 Added a boolean parameter indicating whether we want to join this
1370 thread (used in SG only for now)
1371 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1374 * GTNetS wrappers should now be usable again (and betterly tested too)
1375 * Fix a major regression from 3.2 where the timeout provided to
1376 MSG_task_put_with_timeout() was used as absolute time before which
1377 the comm should be done.
1378 * Start to fix the <cluster> tag.
1379 - Internal links should be good now (beside of the loopback, which
1380 use the private link instead)
1381 - paths to the external world is still rather broken
1382 - the <route:multi> tag is just broken. Actually that's brain-dead.
1383 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1384 to make it less stupid
1385 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1386 * Fix a source-level compatibility glitch from 3.2: after defining
1387 MSG_USE_DEPRECATED, you can use the old name
1388 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1389 * Allow to compile from the SVN with automake 1.11
1390 * Fix some problems when using the "start_time" tag in deployment XMLs.
1391 * Fix #8569: XBT/synchro.h has redundant declarations
1392 * Fix #8563: MSG return values and exceptions
1393 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1394 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1395 Thanks to Robson Peixoto.
1396 * Fix a vicious bug in dictionaries inducing that some elements were
1397 not freed on xbt_dict_free()
1399 Portability report of this version:
1400 * Main portability targets:
1401 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1402 on (amd64/i386/ia64)
1403 - mac leopard on i386
1404 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1405 but nothing critical.
1406 * Other platforms: windows, AIX and others were not tested for this release
1408 Timing report of this version:
1409 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1410 scenario parameters). From 8h to a few seconds in desktop grid settings.
1411 * Supernovae brings about 25% speedup on i386.
1413 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1415 SimGrid (3.3.3) stable; urgency=low
1417 The "Need for Speed" release.
1419 The timings done to validate the 3.3.2 were faulty.
1420 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1422 The problem was a conversion from a manually handled vector to
1423 xbt_dynar_t on the critical path.
1424 xbt_dynar_foreach calls functions, inducing stack management crap.
1426 We inlined these functions and xbt_dynar_foreach is now breath taking.
1427 We also inlined xbt_swag_belong on the way.
1429 Here are some approximate speedup measurements (on master/slaves
1430 simulations lasting between 10s and 20s each):
1431 3.3.1 -> 3.3.2: about same performance
1432 3.3.2 -> 3.3.3: 40% speedup
1433 3.3.1 -> 3.3.3: 40% speedup
1434 3.3.1 with inline patch -> 3.3.3: 30% speedup
1436 Our reading is that the refactoring which occurred in 3.3.2 made us
1437 suffer much more from the xbt_dynar_foreach low performance, but
1438 once we solved this, this refactoring proved to be very performance
1439 effective. From the 40% speedup, somehow, 10% are due to the
1440 inlining and 30% to the refactoring.
1442 That's a pitty that gcc cannot inline functions placed in other files
1443 alone. We have to choose between:
1444 - break the encapsulation (by putting private data structures and
1445 accessors in headers files to help gcc)
1446 - live with low performance
1447 - switch to a decent compiler such as icc (not quite possible).
1449 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1451 SimGrid (3.3.2) stable; urgency=low
1453 The "Simplicity does not preceed complexity, but follows it" release.
1455 The main contributors of this release were (lexical order):
1456 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1459 * Extract the routing logic into its own object.
1460 (was dupplicated in network.c and workstation_LV07.c;
1461 Allows to implement other ways of storing that info)
1462 => kill now useless network_card concept
1463 - Use dynar to represent routes (instead of void** + int*)
1464 - kill link_set (use surf_network_model->resource_set instead)
1465 - Add a command-line option to choose the routing schema to use
1466 - Add three new models:
1467 * Floyd (shortest path computed at initialization)
1468 * Dijikstra (shortest path recomputed all the time)
1469 * Cached Dijikstra (shortest path computed on need)
1470 All these models where contributed by Silas De Munck, and are
1471 described in his ICCS09 paper.
1473 * Simplify model declaration
1474 (less redirections, less function to write when defining a model)
1475 - Factorize stuff between models:
1478 surf_model_resource_set(model)
1479 surf_model_resource_by_name(model, name)
1480 - Unify the types of models in s_surf_model_t (using an union)
1481 - Embeed fields of common_public directly into s_surf_model_t
1482 - Rename model methods:
1483 action_free ~> action_unref
1484 action_change_state ~> action_state_set
1485 action_get_state ~> action_state_get
1486 - Change model methods into functions :
1487 (model)->common_public->action_use ~> surf_action_ref
1489 * Implement a generic resource; use it as ancestor to specific ones
1490 (allows to kill duplicated code in models)
1491 Drawback: timer command don't need no name nor properties;
1492 workstation_CLM03 don't need no properties
1493 (but I guess we can live with those few bytes wasted)
1495 * Improve the action object model
1496 - implement a constructor avoiding dupplicated code about field
1497 initialization in generic_action part.
1499 * Kill the SDP model: it has an external dependency, is deprecated
1500 in flavor of modern lmm models, and didn't compile since a while
1503 * Relocation of the context module from XBT to SIMIX.
1504 (the context were decoupled from the simix processes, duplicating a lot of code)
1505 => a lot of code was factorized
1506 - less overhead is introduced during scheduling
1507 - simpler API for the context factory
1508 - the logic for process creation,destruction and manipulation was simplified
1509 * Simplification of the s_smx_process_t data structure.
1510 => accesing the simix level data associated to a process is faster now,
1511 and the code is a lot more readable.
1514 * Implement some more MPI primitives:
1515 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1516 -implementation: Bcast: flat or 2-ary tree (default),
1517 Barrier: 4-ary tree,
1519 Allreduce: Reduce then Bcast
1520 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1521 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1522 Alltoallv: flat tree, like ompi
1524 * Add support for optimized collectives (Bcast is now binomial by default)
1525 * Port smpirun and smpicc to OS X
1528 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1529 Shout out if you used it.
1532 * Display the list of still queued messages in SG mode when existing
1536 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1537 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1538 * Add xbt_str_from_file(FILE*)
1539 * Add xbt_dict_get_key achieving a linear reverse search
1540 * Remove the context module
1542 Portability report of this version:
1543 * Main portability targets:
1544 - Linux(debian)/x86/context
1545 - Linux(debian)/x86/pthreads
1546 - Linux(debian)/amd64/context
1547 - Linux(debian)/amd64/pthreads
1548 On these, we still have the eratic breakages of gras/pmm and
1549 amok/saturate_sg reported in previous version. We still think
1550 that the tests are the cause of the fault, not the tested code.
1552 - Mac OSX Leopard/x86/context
1553 Still false negative in tesh autotesting.
1554 Smpi still fails, but this time because readlink does not accept -f
1555 Everything seems to work properly beside of that.
1558 - AIX version 5.3 (only tested contexts this time)
1559 Smpi still fails there because mktemp is not installed.
1560 Everything seems to work properly beside of that.
1562 I managed to compile it for the first time, but several breakages.
1563 Won't delay the release for this exotic platform.
1565 * Windows: it's still lagging behind. If you want to help, please
1568 Timing report of this version:
1569 This version seem to be more than 5% faster than 3.3.1 (on linux
1570 64bits with contextes). The gain is less than expected, we are
1571 investigating this for next release.
1573 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1575 SimGrid (3.3.1) stable; urgency=low
1578 * Implement a --cfg-help to show existing configuration variables
1579 * Build chain do not require doxygen in maintainer mode
1582 * fix a bug on struct sizeof computation, which prevented the
1583 exchange of arrays of structs in some conditions
1584 - added a regression test about this in datadesc_usage
1585 * Allow the exchange of 0-long dynamic vectors.
1586 - for that, use -1 as indicator of dynamic size instead of 0
1587 - This implied to change any size from unsigned long to long,
1588 reducing a bit communication abilities, but I guess that with
1589 64bits being quite common, this is more than enough.
1590 - This also induce a protocol change, thus bumping network protocol
1591 version from 0 to 1 (if we have external users, we have to get
1592 clean on that point too ;)
1593 - added two regression tests about this in datadesc_usage
1594 * Be more verbose when propagating local exceptions
1595 This helps debugging.
1596 * Display the status of simulated processes when receiving SIGINT in
1600 * Allow to control the simulation from a trace file.
1601 New functions MSG_action_register() and MSG_action_trace_run()
1602 The first one allows to associate a function execution to each
1603 kind of action while the second one parses a trace file and
1604 triggers the corresponding actions within the system.
1605 For now, only a toy example is provided in examples/msg/actions
1606 * Add an exemple of process migration in examples/msg/migration
1607 * Fix a bug in task exchange which broke MSG_task_get_sender()
1608 Add a teshsuite regression test for that.
1609 [Bug: if MSG_task_get_sender() is called after sender exit,
1611 * Fix a bug which prevented suspend/resume to work properly
1612 * Display the status of simulated processes when receiving SIGINT
1613 This fixes a regression of v3.3. due to the introduction of SIMIX
1614 * Bug fixing in failure management:
1615 - trace could not start by a failure at time 0
1616 - failure during communications were not working
1619 * Add SIMIX_process_set_name() to change the name of the current
1620 process in the log messages.
1621 * Store smx_hosts in a dict since we only retrieve them by name
1622 * Move the configuration infrastructure to surf
1625 * Move the configuration infrastructure to surf
1628 * Massive internal cleanups:
1629 - Store internal structures on processes instead of hosts (allows
1630 to have more than one process per host, in addition of being more
1632 - Cleanup the initialization/finalization process
1633 - Kill a whole bunch of unneeded synchronization:
1634 processes run in exclusive manner within the simulator
1635 - Move queues from global tables to process data fields
1637 - now accept -platform and -hostfile arguments
1638 - Pass the right rank value to processes according to the hostfile
1639 * Compile the examples by default, and use them as regression tests
1640 * Implement MPI_Wtime()
1641 * Change the reference speed to a command line option
1644 * TCP_gamma can now be specified as command line option using
1645 --cfg=TCP_gamma:10000000.0
1646 * Change the --surf-path cmd line option into --cfg=path:
1649 * Also include strbuff from xbt.h public header
1650 * xbt_ex_display(): do not free the exception after displaying
1651 This allows to do more with the given exception afterward.
1652 Users should call xbt_ex_free() themselves.
1656 Portability report of this version:
1657 * Main portability targets:
1658 - Linux(debian)/x86/context
1659 - Linux(debian)/x86/pthreads
1660 - Linux(debian)/amd64/context
1661 - Linux(debian)/amd64/pthreads
1662 These targets fail about 1/10 of times on gras/pmm, but we believe
1663 that this is because of the test, not because of simgrid.
1664 amok/saturate_sg fails even more rarely, and the test may not be
1667 - Mac OSX Leopard/x86/context
1668 The test suite still spits tons of errors because some obscure
1669 force prevents us from removing the temporary directories
1670 arguing that they still contain some metadata I've never heard of.
1671 Smpi fails because seq is not installed.
1672 Everything seems to work properly beside of that.
1675 - AIX version 5.3 (both contexts and pthread)
1676 Smpi still fails there because mktemp is not installed.
1677 XML inclusions seems rosty on AIX.
1679 * Windows: it's still lagging behind. If you want to help, please
1682 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1684 SimGrid (3.3) stable; urgency=high
1688 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1689 [Malek Cherier & Mt]
1691 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1694 Until now, GRAS were using MSG as an interface to SURF. It was
1695 quite difficult because both interface have several differences
1696 (MSG channels vs GRAS sockets were the most notable point).
1698 This also opens the gate to SMPI (which should occur soon) and speed
1699 up simulations by to 40% (even if it were not the main goal).
1701 **************************************
1702 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1703 **************************************
1704 The point is that events occuring at the exact same timestamp are
1705 not scheduled in the same order with the old and new version. This
1706 may be enough to completely change the execution of simulations in
1707 some cases. Sorry for the inconvenience.
1709 * Cleanup and upgrade the XML format to push further scalability
1710 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1712 * Improve the testing infrastructure with tesh. Now a very large part of
1713 the code is tested not only by being run but also by checking that the
1714 output match an expected output [Mt].
1716 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1717 is really less memory-demanding, which should allow you to use
1718 larger files in SimGrid [AL].
1720 * Inform valgrind about our contextes, so that it becomes usable
1721 with the default (and more effecient) version of SimGrid
1722 [contributed by Sékou Diakite, many thanks]
1725 * Introduce a listener thread in charge of receiving incoming
1726 messages from the network. It allows to overlap communication and
1727 computation but most notably, it removes some stupid deadlocks due
1728 to the fact that so far, a process could not send and receive at
1729 the same time. This made most non trivial communication schema
1731 * Convert the PIDs from long int to int to match the MSG ones (and
1732 linux ones too) [Mt]
1733 * New function: gras_agent_spawn() to launch a new process on
1734 current host. Only working in simulation for now. [Mt]
1735 * New function: gras_os_hostport() returning a constant form (ie,
1736 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1739 * Make the backtrace of exceptions more human readable [Mt]
1740 * New module: xbt/str [Mt]
1741 a ton of string utility functions (split, join, printf to a newly
1742 allocated buffer, trim, etc)
1743 * New module: xbt/hash [Mt]
1744 SHA1 hashing algorithm (more to come if needed)
1745 * New module: xbt/synchro [Mt]
1746 synchronization tools (mutex and conditions) working the same way
1747 in simulation and in real life (mainly useful for GRAS, but not
1749 * New module: xbt/queue [Mt]
1750 classical producer/consumer synchronization scheme
1751 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1752 (using the classical functions will get serialized) [Mt]
1753 * Make dictionary internal table dynamic. No need to specify its size
1754 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1756 * Make sure the log channels are organized as a tree under windows
1757 (because of ANSI C compatibility issue, any channel were child of
1758 root directly) [Mt].
1761 * Cleaned many thing in surf and fixed a few bugs [AL].
1762 * Add a nice command line configuration mechanism to compose models [AL].
1763 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1764 the previous one (KCCFLN05). It relies on something that looks like
1765 a max-min sharing mechanism but cannot be written as such. A new solver
1766 was thus designed [AL].
1767 * Add a new solver to lmm. Based on Lagrange optimization and
1768 gradient-based descent, it enables to efficiently maximise systems s.a
1770 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1772 This solver enables to propose two new network models for TCP Reno and
1773 TCP Vegas based on Low's work. These models still need to be fully
1774 tested though [Pedro Velho].
1777 * Bug fix in SD_simulate. Now the time bound given as argument is
1779 * Use the new parallel task model (ptask_L07) as default.
1780 * Use the SURF command line configuration mechanism.
1781 * 0-size tasks (for synchronization) should now work.
1783 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1785 SimGrid (3.2) stable; urgency=high
1789 We still experience issues on this platform, but we believe that at
1790 least MSG is usable.
1792 GRAS API BREAKAGE (for simplification purpose, sorry):
1793 * the gras_msgtype_by_name is not used anymore. Instead of
1794 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1795 you can write (and must)
1796 gras_msg_send(toserver, "request", &request);
1797 - If you still want to pass a gras_msgtype_t to the function (to cache
1798 the type and avoid the lookup time), use the gras_msg_send_() variant.
1799 - Impacted functions:
1800 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1801 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1802 * The callbacks are now expected to return 0 when everything went well
1803 (just like the main() function)
1805 GRAS new features and improvements:
1806 * New module mecanism where user code can use per process globals [Mt]
1807 This is similar to gras_userdata_*() functions, but for libraries. It
1808 factorize some code developped over and over in the examples and AMOK.
1809 It has still to be documented and used (only amok/peermanagement is
1811 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1812 loss when they were small enough to fit into the buffer and sent quickly
1813 enough so that they can all get received in one shoot.
1814 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1815 if not found. Use the *_or_null() variant for the old semantic.
1816 * In gras_msg_handle, do not discard messages without callback.
1817 They are probably messages to be explicitly awaited later (ie, proofs of
1818 mis-synchronization in userland since they are sent before being awaited)
1819 No big deal usually.
1820 * gras_socket_meas_send/recv: semantic changed!
1821 The numerical arguments used to be (1) the total amount of data to send
1822 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1823 messages. This was need for the fool willing to send more than MAXINT
1824 bytes on quite fat pipes.
1827 * Do really rename the hostmanagement module to peermanagement. [Mt]
1828 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1829 but this is rather new and this was documented in the module
1830 documentation (poor excuses, I admit)
1831 * Bandwidth measurement semantic changed! This follows the changes to
1832 gras_socket_meas_send/recv explained above.
1835 * A sequential mode has been added to the workstations. When a workstation
1836 is in sequential mode, it can execute only one task, and the other tasks
1837 are waiting in a FIFO. [Christophe Thiery]
1840 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1841 model for SIMDAG. [Christophe Thiery]
1842 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1843 equal to 0, which caused some bad behaviors in
1844 saturated_constraint_set_update. I now use a threshold mechanism like in
1848 * When running manually src/testall, you select specific units [Mt]
1849 testall is the result of our cunit mecanism, and should replace all
1850 the scripty thingy around since bash don't run easily on billware.
1852 * A mallocator system has been added. [Christophe Thiery]
1853 Mallocators allow you to recycle your unused objects instead of freeing them
1854 and allocating new ones.
1856 Documentation update:
1857 * FAQ reworking + New FAQs:
1858 - "Valgrind spits tons of errors!" [Mt]
1859 - "How to repport bugs" [Mt]
1860 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1861 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1862 - Communication time measurement within MSG [AL]
1863 - I experience weird communication times when I change the latency [AL]
1864 * GRAS tutorial [Mt]
1866 - an introduction to the framework and to the used communication model
1867 - an initiatic tour introducing the most proheminent features:
1869 . Lesson 0: Installing GRAS
1870 . Lesson 1: Setting up your own project
1871 o Part 2: Message passing
1872 . Lesson 2: Exchanging simple messages
1873 . Lesson 3: Passing arguments to the processes (in SG)
1874 . Lesson 4: Attaching callbacks to messages
1875 . Lesson 5: Using globals in processes
1876 . Lesson 6: Logging informations properly
1877 . Lesson 7: Using internal timers
1878 . Lesson 8: Handling errors through exceptions
1879 . Lesson 9: Exchanging simple data
1880 . Lesson 10: Remote Procedure Calling (RPC)
1881 . Lesson 11: Explicitely waiting for messages
1882 . Recapping of message passing features in GRAS
1883 - A HOWTO section containing:
1884 o HOWTO design a GRAS application
1885 More are due, of course. They will come latter. In the meanwhile, you can
1886 check the examples which are still here.
1888 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1890 SimGrid (3.1) stable; urgency=high
1894 There was a stack corruption somewhere, visible only when optimizing
1895 with these versions. [Vince]
1898 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1899 module that enabled to play with Directed Acyclic Graphs. It is built
1900 directly on top of SURF and provides an API rather close to the old
1901 SG. Some old codes using SG are currently under rewrite to check that
1902 all needful functions are provided. [Christophe Thiery]
1905 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1906 extension of the classical CLM03 model that gracefully handles
1907 failures. This is now the default model for MSG and GRAS. It doesn't
1908 handle parallel tasks yet though. [AL]
1909 * Bug fix: Weights were not correctly set in the network part.
1910 WARNING: This may have resulted in incorrect results with simulations
1911 where there are more than one flow on a given link. [AL]
1914 * After a (long ?) discussion on simgrid-devel, we have decided that the
1915 convention we had on units was stupid. That is why it has been decided
1916 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1917 WARNING : This means that all previous platform files will not work as
1918 such with this version! A warning is issued to ask users to update
1920 A conversion script can be found in the contrib module of the CVS, under
1921 the name contrib/platform_generation/surfxml_update.pl [MQ]
1924 * Bug fix: Processes were started in reverse order, wrt deployment file.
1925 WARNING: if your code relies on this bug, please fix it. [AL]
1926 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1927 being executed on two different locations. [AL]
1928 * Bug fix: Failures are now better supported thanks to Derrick's tests
1929 (there was many failure situations I hadn't thought of and that weren't
1930 correctly handled). [AL]
1931 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1935 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1936 Exception occurring on server-side are propagated back to client (!).
1938 API CHANGE: the callback changed their prototype. Change:
1939 int my_handler(gras_socket_t expeditor, void *payload_data) {
1941 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1942 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1944 * New! function: gras_msg_handleall to deal with all messages arriving
1945 within a given period.
1946 * New! function: gras_socket_server_range to get a server socket in a
1947 range of port numbers (ease to avoid port number conflicts) [MQ]
1948 * New! gras processes display their backtrace when they get a SIGUSR1
1949 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1950 Sweet to debug RL processes [MQ]
1954 - Do not force experiment sizes to be expressed in kb, or it becomes
1955 impossible to measure the latency this way (needs one byte-long tests)
1956 WARNING: this changes the amok_bw_* function semantic. [MQ]
1957 - Implements the link saturation stuff. [MQ]
1958 * Peer management module:
1959 New! module factorizing code that we wrote over and over [MQ].
1962 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1963 - Test units are placed directly into the library code, they get extracted
1964 automatically and placed into the src/testall binary.
1965 - Convert most of the XBT tests to this system.
1966 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1967 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1968 live in the same C code. This required to change a little bit the API
1969 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1970 * New module: sparse graph structure with basic algorithms (this is work
1971 in progress and the API is not considered to be frozen yet). [AL]
1972 * Display more information on backtraces: source line & function names are
1973 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1974 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1975 * Reworked a little bit some #include statements to load only required
1976 headers. Some user code that relied on SimGrid to include stdlib or
1977 stdio may need to include it by themselves. [AL]
1978 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1980 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1981 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1982 but I guess that almost nobody used those parts. [MQ]
1984 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1986 SimGrid (3.0.1) stable; urgency=low
1989 * Unfortunately, I had missed 5 misnamed functions:
1990 xbt_fifo_item_t xbt_fifo_newitem(void);
1991 void xbt_fifo_freeitem(xbt_fifo_item_t);
1992 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1993 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1994 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1995 They're now deprecated. Please use their new versions:
1996 xbt_fifo_item_t xbt_fifo_new_item(void);
1997 void xbt_fifo_free_item(xbt_fifo_item_t);
1998 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1999 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2000 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2002 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2003 * Documentation: xbt_log module unmercifully reworked [MQ]
2004 * Bugfix: there was a problem with the ending of contexts with
2005 the pthread backend. It caused some weird deadlock or behavior
2006 depending on the pthread implementation. [AL]
2007 * Bugfix: get the exceptions raised in the simulator repport where
2008 and why they come from when they are not catched in time [AL, MQ]
2011 * Bugfix: Do repport the error when two non-connected hosts try to
2012 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2015 * Add additionnal checkings on communications. Assert that two
2016 communicating hosts are connected by a set of links... [AL]
2019 * Add additionnal checkings on channel values in communication [AL]
2020 * New: MSG_task_get_source to see on which host a task was generated [HC]
2021 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2022 the number of tasks waiting to be received on channel and sent
2024 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2025 waits for the first task coming from a given host.. [AL]
2027 GRAS new functionnalities: [MQ]
2028 * Enhance the parsing macro to allow the size of multidimentional objects
2029 to be given thru annotations.
2030 * New example (and documentation): Matrix Multiplication a la RPC
2031 (as when I was young!) and fix a bunch of bugs found on the way.
2033 GRAS performance improvements: [MQ]
2035 * Reduce the amount of cbps creation/destruction by making it static to
2036 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2039 * Change libdata to a set so that we can search for stuff by ID (and thus
2040 reduce the insane amount of dict lookups)
2043 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2044 RL communication on top of SG since it's so uneffective.
2045 It may also be used for inter-thread communication in RL, one day.
2046 * Use gras_datadesc_copy() to exchange messages on top of SG
2048 - improve message exchange performance on top of SG
2049 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2050 * Don't exchange on the network the size of the used part of buffer,
2051 instead, specify the possible buffer size to read().
2053 - reduces the amount of read/write calls (one pair per exchange)
2054 - reduces the amount of exchanged data (the size)
2055 - allows to retrieve all arrived data on receiver side, if we don't need
2056 it right now (subsequent read will peek the buffer)
2057 - allows the receiver to proceed with the begining of the stream before
2058 everything is arrived
2059 - make it possible to build an iov transport (using readv/writev)
2061 - take care of the data with non-stable storage (like stacked data),
2063 * If possible, TCP send uses vector I/O (when writev() is here)
2064 - Don't use it for receive since we send data sizes and data on the
2065 same stream, so we wouldn't be able to chain large amount of chunks
2066 before having to flush the stuff to read the size.
2067 * Rework the transport plugin mecanism to simplify it and reduce the
2068 amount of pointer dereferencement when searching for the right function
2071 * I guess that now, we do almost as few system calls as possible while
2072 doing as few data copy as possible.
2074 To improve it further, we could try to send all the sizes first and then
2075 all the data (to use iov on receiving size), but it's only a partial
2076 solution: when you have 2 dimensional data, the sizes of the second
2077 dimension is data of the first dimension, so you need 3 streams.
2079 I'm not sure the potential performance gains justify the coding burden.
2081 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2083 SimGrid (3.00) stable; urgency=high
2086 * New! Give the possibility to hijack the surf parser and thus bypass
2087 MSG_create_environment and MSG_launch_application. Have a look at
2088 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2090 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2092 SimGrid (2.96) unstable; urgency=low
2097 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2098 This deprecates the xbt_error_t mecanisms.
2099 It modifies (simplifies) all XBT and GRAS API.
2100 MSG API keeps unchanged (exceptions raised by XBT are catched from
2101 within MSG and masked with existing error handling facilities)
2104 * New! Add a FATPIPE model. [AL]
2105 * New! Add a parallel task model. [AL]
2106 * New! Add automatically a loopback interface (in the default
2107 network model) if none was precised.
2110 * Bugfix: MSG_process_resume now works with the current running process.
2112 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2113 * Modification of MSG_task_get_compute_duration. Once a task has been
2114 processed, the value returned by this function is now equal to 0. [AL]
2115 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2116 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2117 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2118 of a m_task. For now, MSG_task_cancel only works with computation
2120 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2121 of the processor (in Mflop/s) regardless of the current load on the
2123 * API Change: use proper naming convention for MSG_getClock and
2124 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2126 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2127 This function changes the priority of a computation task. This priority
2128 doesn't affect the transfer rate. A priority of 2 will make a task
2129 receive two times more cpu power than the other ones. This function
2130 has been added to suit the needs of Nguyen The Loc and hasn't been that
2131 much tested yet. So if it fails, please report it and send me your code.
2133 * API Change: removed all functions and types that were marked "deprecated"
2134 since many months. Renamed MSG_global_init_args to MSG_global_init.
2136 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2138 SimGrid (2.95) unstable; urgency=low
2141 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2142 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2143 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2146 * Bugfix: Allow absolute paths for platform description files [MQ]
2147 * Bugfix: do free the variables after use. Leads to drastic performance
2149 * Implement max_duration (ie, timeouts) on resources [AL]
2152 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2154 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2155 the needed support (timeouts) [AL]
2158 * Implement measurement sockets. You can now get the bandwidth between two
2159 hosts thanks to AMOK (see below). [MQ]
2160 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2161 dynar over the network (yeah) [MQ]
2162 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2163 * simple/static token-ring example. [Alexandre Colucci and MQ]
2164 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2165 in place before [MQ]
2168 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2169 type parsing. "short" and "long" modifiers were also ignored; other
2170 modifier (such as reference level) are still ignored. [MQ]
2171 * Embeed the buffer size within the buffer itself on SG. [MQ]
2172 That way, send() are atomic and cannot get intermixed anymore (at least
2173 the ones which are less than 100k; bigger messages still have the issue)
2174 * Array size pushed by the field, not by the field type (or each
2175 and every long int will push stuff to the cbps) [MQ]
2176 * use select() to sleep since it allows to portably sleep less than one
2179 GRAS (minor cleanups)
2180 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2183 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2184 * type category 'ignored' killed as it was never used and were difficult
2186 * whether a type can cycle or not is now a flag, leaving room for more
2187 flags (as "ignored", if we feel the need one day ;)
2188 * Rename raw sockets to measurement sockets since "raw" has another
2189 meaning in networking community.
2192 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2193 and offers features not belonging to GRAS but that most applications
2194 need. One day, it may be a set of plugins loadable at runtime.
2195 * New module: bandwidth
2196 bandwidth measurement between arbitrary nodes running this module. [MQ]
2198 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2200 SimGrid (2.94) unstable; urgency=low
2202 The first beta release of SimGrid 3 !
2206 * Update the main page and the FAQ. Adding references to gforge.
2209 * Add a gras_os_getpid function.
2212 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2213 * Extend the logs so that they also print PID, hostname, date, ... if
2215 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2216 and kill the old version which were in testsuite/
2217 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2221 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2222 for users. Instead, each programming environment (GRAS, MSG,...) use it
2223 when needed to provide such a feature to users.
2224 Don't shortcut the mecanism or you will also shortcut the virtualization
2225 you need on the simulator.
2229 * Cleanups in configury with regard to compile optimization/warning flags.
2230 Also add -fno-loop-optimize to any powerpc since it's the optimization
2231 killing gcc (< 3.4.0).
2232 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2233 needed by MSG examples complications
2234 * Borrow configury beautifications from PHP
2237 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2238 warning (thanks loris for stumbling into this one).
2239 * Bugfix: stop loading private headers (gras_config.h) from the public
2243 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2244 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2245 RL would imply a DNS resolver, which is *hard* to do in a portable way
2246 (and therefore delayed).
2247 * Implement a real timer mecanism and use it in timing macros. This allows
2248 to avoid rounding errors and get a 0.000005 sec precision in timing
2249 macros. While I was at it, various cleanups:
2250 - allow to declare more than one timed section per file (fix a stupid bug)
2251 - move some private declaration to the right place
2252 - merge conditional execution and timing macros into emulation module
2253 - document the module
2254 - make sure the module cleanups its mess on gras_exit
2255 * Documentation improvements:
2256 - (new) how to compile applications using GRAS
2257 - (new) emulation support (timing macros)
2259 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2261 SimGrid (2.93) unstable; urgency=low
2263 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2266 - Use Paje properly where used. Still to be sanitized properly.
2267 - Portability fix: Add an implementation of the contexts using pthread
2271 - Add xbt_procname(): returns the name of the current process.
2272 Use it to show the current process's name in all logging.
2274 - fix detection of older flex version and the reaction, since we do
2275 depend on modern ones (we use lex_destroy)
2276 - Better separation of SG and RL in the libs: remove all simulation code
2277 from libgras. As a result, this lib is now only 200k when stripped.
2278 Some of the xbt modules may also be duplicated (two sets and such) and
2279 should be cleaned/killed before SG3.
2280 - Insist on using xlC on AIX because of weird problems involving gcc there.
2281 - Cleanup the make remote stuff. This is now done by scripts
2282 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2283 mainly for our private use, but we're working on changing them to user
2286 - Bugfix: flush the socket on close only if there is some *output*.
2287 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2289 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2292 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2294 SimGrid (2.92) unstable; urgency=low
2296 Alpha 3 on the path to SimGrid 3
2300 - New! First try of benchmarking macros.
2301 - New! First try so that gras_stub_generator generate deployment and
2302 remote compilation helpers.
2304 - Bugfix: Initialization fix in msg_test.
2308 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2310 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2312 - API Change: gras_os_sleep to take the amount of seconds as a double.
2313 Accepting an int was error prone since it was the only location where
2314 seconds were coded as such. It leaded to damn rounding errors.
2315 - Bugfix: Hard to belive that timers ever worked before this.
2317 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2319 SimGrid (2.91) unstable; urgency=low
2321 Alpha 2 on the path to SimGrid 3
2325 - Bug fix in the lmm_solver.
2327 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2328 through the function MSG_paje_output.
2329 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2330 - It is possible to bound the rate of a communication in MSG with
2331 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2332 to write it in the changelog).
2333 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2338 - Various cleanups to the autotools stuff
2339 - Begin to move Gras examples to examples/gras/
2340 - Let make distcheck work again (yeah!)
2342 - documentation overhauled using doxygen.
2343 gtk-doc-tools is dead in SimGrid now.
2344 - Automatically extract all existing logging categories, and add the list
2345 to the documentation (long standing one, to say the less)
2347 - Cleanup the known architecture table. Reorder the entries to group what
2348 should be, and use a more consistent naming scheme.
2349 (some of the test dataset are still to be regenerated)
2350 - New! Allow library to register globals on each process just as userdata
2352 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2353 lookup time (for now).
2354 Use it in msg and trp.
2355 This cleans a lot the internals and helps enforcing privacy of the
2356 headers between the gras components.
2357 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2358 - Bugfix: gras_os_time was delirious in RL.
2359 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2361 - Reenable GRAS now that it works.
2363 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2365 SimGrid (2.90) unstable; urgency=low
2367 Alpha 1 on the path to SimGrid 3
2369 * It is a long time since the last release of SimGrid. I'm sorry about
2370 that but as I had told you, I was rewriting a lot of things. I apologize
2371 to those who had been reporting bugs to me and that I had not answered.
2372 If your bug is still in the new version, please tell me. Here is a
2373 summary of the main changes.
2375 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2376 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2377 portability, speed, and a lot more but you'll figure it out later.
2378 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2379 I wanted to release the new MSG as soon as possible and I have
2380 broken GRAS, which is the reason why, for now, only MSG is fully
2381 functional. A laconic description of these projects is available
2382 in the documentation.
2384 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2385 kernel optimized for our needs (called SURF but only the developers
2386 should use it). Hence, MSG is now roughly 30 times faster and I think
2387 that by rewriting a little bit MSG, I could event speed it up a little
2388 bit more. Beside the gain in speed, it is also much easier to encode a
2389 new platform model with SURF than it was with SG. More to come...
2391 * REVOLUTION 3: I have tried to change a little as possible the API of
2392 MSG but a few things really had to disappear. The main differences
2393 with the previous version are :
2394 1) no more m_links_t and the corresponding functions. Platforms are
2395 directly read from a XML description and cannot be hard-coded
2396 anymore. The same format is used for application deployment
2397 description. The new format is described in the documentation.
2398 Have a look in tools/platform_generation. There is a tiny script
2399 that converts from the old platform format to the new one. Concerning
2400 the application deployment format, parsing the old one is tricky.
2401 I think most of you should however be able to convert your files. If
2402 it is really an issue, I can write a C code that does the conversion.
2404 2) the toolbox tbx does not exist anymore. We now have a library
2405 with much more data-structures but without the hash-tables (we have
2406 dictionaries that are much faster).
2408 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2410 *****************************************************************************
2411 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2412 * don't feel like converting the oldies. (Mt) *
2413 *****************************************************************************
2416 Version 2.90: "the long awaited one"
2417 - Finished rewriting and debugging MSG. Rewrote the documentation.
2418 - disable GRAS for now since it needs to be ported to the newest SG
2421 - Finish the port to windows (using mingw32 for cross-compile)
2424 - Main loop and datastructures of SURF. A cpu resource object is
2425 functional. Surf can thus be used to create cpu's with variable
2426 performance on which you can execute some actions.
2428 2004-11-15 Martin Quinson
2429 - Port to ARM. Simply added the alignment and size descriptions. Should
2430 work, but the ARM machines are so slow that I didn't had the opportunity
2431 to 'make check' over there yet.
2433 2004-11-15 Arnaud Legrand
2434 - Trace manager now written. It uses a heap structure and is therefore
2435 expected to be efficient. It may however be speeded up (particularly
2436 when many events occur at the same date) by using red and black
2437 trees. One day maybe...
2438 - Max-min linear system solver written. It uses a sparse matrix
2439 structure taking advantage of its expected use. Most operations are
2440 O(1) and free/calloc are called as few as possible. The computation of
2441 the minimum could however be improved by using a red and black tree
2444 2004-11-03 Arnaud Legrand
2445 - Rename every gras_* function that was in xbt/ to its xbt_
2447 - Add a heap and a doubly-linked list to xbt
2448 - Added a dichotomy to the dictionaries. make check works as well before
2449 so I assume that the patch is correct. I do not know however if things
2450 run effectively faster than before now. :)
2452 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2453 SimGrid, and the version number is bumped to 2.x
2455 2004-10-29 Martin Quinson
2456 - Introduction of the remote errors.
2457 They are the result of a RMI/RPC on the remote machine.
2458 ErrCodes being scalar values, you can't get the host on which those
2459 errors did happen. Extending the error mechanism as in Gnome is possible.
2460 No idea yet whether it is a good idea.
2462 2004-10-28 Martin Quinson
2463 - Interface revolution: the Starred Structure Eradication.
2464 I used to do typedef struct {} toto_t; and then handle *toto_t.
2465 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2466 - ???_t is a valid type (builded with typedef)
2467 - s_toto_t is a structure (access to fields with .)
2468 - s_toto is a structure needing 'struct' keyword to be used
2469 - e_toto_t is an enum
2470 - toto_t is an 'object' (struct*)
2472 typedef struct s_toto {} s_toto_t, *toto_t;
2473 typedef enum {} e_toto_t;
2474 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2475 s_toto_t) is private.
2477 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2478 so that it can set the variable to NULL. It was so for dicts and sets,
2479 it changed for dynars.
2481 - Fix a bunch of memleaks in dict_remove
2482 - Fix a bug in sg/server_socket opening: it failed all the time.
2484 2004-10-07 Martin Quinson
2485 - Speed up dynar lookup operation a bit.
2487 gras_dynar_get is dead.
2489 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2490 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2491 (which gives you the address of the stored data).
2493 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2494 copy of the data using an affectation to do the job and not a memcpy.
2496 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2497 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2499 It does not really speedup the dynar test because they are
2500 setting elements all the time (and look them seldom). But the dict does
2501 far more lookup than setting.
2503 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2505 2004-10-05 Martin Quinson
2506 - Allow to (en/dis)able the cycle detection at run time.
2508 Whether we should check for cycle or not is now a property of each
2509 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2510 datadesc_cycle_unset allow to remove this property when previously set.
2512 Note that the cycle detection is off by default since it impacts the
2513 performance. Watch the data you feed GRAS with ;)
2515 This property is hereditary. Any element embedded in a structure having it
2516 set have it set for the time of this data exchange.
2518 You should set it both on sender and receiver side. If you don't set it on
2519 sender side, it will enter an endless loop. If you forget on receiver
2520 side, the cycles won't be recreated after communication.
2522 - Header reorganization.
2523 Kill gras_private.h, each submodule must load the headers it needs.
2525 2004-10-04 Martin Quinson
2526 - Interface revolution: do not try to survive to malloc failure.
2528 Now, gras_malloc and friends call gras_abort() on failure.
2529 As a conclusion, malloc_error is not a valid error anymore, and all
2530 functions for which it was the only gras_error_t return value are
2531 changed. They now return void, or there result directly.
2532 This simplify the API a lot.
2534 2004-09-29 Martin Quinson
2535 - Re-enable raw sockets.
2536 Created by gras_socket_{client,server}_ext;
2537 Used with gras_raw_{send,recv}
2540 It should allow to kill the last bits of gras first version soon.
2542 This is not completely satisfactory yet (duplicate code with
2543 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2546 - Simplify transport plugin (internal) interface by not passing any
2547 argument to _server and _client, but embedding them in the socket
2550 2004-09-28 Martin Quinson
2551 - Finish the port to AIX.
2552 autoconf was my problem (segfault within the malloc replacement
2553 function. No idea why)
2555 2004-09-16 Martin Quinson
2556 - Fix some size_t madness on 64bit architectures.
2558 2004-09-08 Martin Quinson
2559 - Reduce the number of system headers loaded, overload some more system
2560 calls (such as malloc to cast the result of the system one, and work
2562 - Fix and reintroduce the config support
2564 2004-09-07 Martin Quinson
2565 - Source code reorganization to allow Arnaud to surf all over there.
2566 - Allow to document the logging categories.
2567 - Remove all uppercase from logging categories and useless cleanup in names.
2569 2004-08-18 Martin Quinson
2570 Version 0.6.2 (protocol not changed; API changed)
2571 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2572 gras_error_t), and NULL when not found. Functions expecting a msgtype
2573 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2574 hopefully usefull message.
2575 - Portability to prehistoric sparcs again
2577 2004-08-17 Martin Quinson
2578 Version 0.6.1 (protocol not changed; ABI not changed)
2579 - prealloc some buffers to speed things up
2581 2004-08-11 Martin Quinson
2582 Version 0.6 (protocol not changed; ABI expended)
2583 - The parsing macro can deal with the references, provided that you add
2584 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2586 2004-08-09 Martin Quinson
2587 Version 0.5 (protocol not changed; ABI changed)
2588 - Allow to off turn the cycle detection code in data exchange at
2589 compilation time. It should be at run time, but I'm short of time (and
2590 the config stuff is still broken). That way, we keep dict out of the
2591 critical path, which is good because the performance is poor:
2592 - search not dichotomial yet
2593 - dynar give no way to access their content and memcpy everytime
2594 - In composed data description (struct, ref and so on), stop foolness of
2595 keeping the subtype's ID, but store the type itself. This keeps sets out
2596 of the critical path, which is good since they rely on dynar and
2597 dictionnaries. The only loose of that is that we cannot detect the
2598 redeclaration of a structure/union with another content (but I'm not sure
2599 the code detected well this error before anyway). We still can detect
2600 the redefinition discrepancy for the other types.
2601 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2602 it breaks the code because of type-punning used all over the place).
2603 This breaks on all non-gcc architectures (for now).
2605 All those changes (plus the buffer of last time) allow me to gain 2 order
2606 of magnitude on cruel tests consisting of 800000 array of integers on two
2607 level of a hierarchical structure (200 secondes -> 4 secondes)
2610 - the selector of reference must now return the type it points to, not
2611 the ID of this type.
2613 2004-08-06 Martin Quinson
2614 Version 0.4 (protocol changed; ABI not changed)
2615 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2616 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2617 - Implement a buffer transport, and use it by default (it relies on tcp in
2618 real life and on sg in simulation).
2619 That's a bit hackish since I had a new field to the structure to store
2620 its data without interfering with the subtype ones. Inheritance
2621 is tricky in C. And that's a kind of reverse inheritance with one class
2622 derivating two classes. Or maybe a game with java interfaces. Anyway,
2623 that's damn hard in C (at least).
2624 Moreover, I got tired while trying to ensure plugin separation and
2625 genericity in SG mode. MSG wants me to do weird things, so let's go for
2626 cruel hacks (temporarily of course ;).
2627 See comment in transport_private.h:71
2628 - do not include all the _interface headers in private but in the files
2629 which really need them (to cut the compilation time when they are
2632 2004-07-26 Martin Quinson
2633 Version 0.3 (protocol not changed; ABI changed)
2634 - Major overhault of the datadesc interface to simplify it:
2635 - shorted the function names:
2636 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2637 - add a trivial way to push/pop integers into the cbps without malloc.
2638 This allows to make really generic sub_type description, which simply
2639 pop their size of the stack.
2640 - add a function gras_datadesc_ref_pop_arr() which does what users want
2641 most of the time: Declare a dynamic array (which pops its size of the
2642 stack) and declare a reference to it. Poor name, but anyway.
2643 - kill the post-send callback, add a post-receive one
2645 2004-07-23 Martin Quinson
2646 Version 0.2 (protocol changed; ABI changed)
2647 - add some testing for cpbs in the test cases, and fix some more bugs.
2648 This invalidate again the little64 data file, since I cannot regenerate
2650 - remove an awfull optimization in the logging stuff, allowing me to:
2651 - understand it again
2652 - learn gcc how to check that the argument match the provided format
2653 - fix all errors revealed by gcc after that
2654 - internal keys of dict are not \0 terminated. Deal with it properly in
2655 loggings instead of segfaulting when the user want to see the logs :-/
2657 2004-07-22 Martin Quinson
2658 - Fix some stupid bug preventing cbps (callback postit) from working
2660 2004-07-21 Martin Quinson
2661 - Some documentation cleanups
2662 - remove the useless last argument of msgtype_declare
2663 - rename the Virtu functions to fit into the 'os' namespace
2664 - move headers src/include -> src/include/gras/ and stop fooling with
2665 gras -> . symbolic link
2666 - make distcheck is now successful
2668 2004-07-19 Martin Quinson
2670 - Build shared library also
2671 - Install html doc to the right location
2672 - stop removing maintainer files in make clean
2673 - build tests only on make check
2675 2004-07-13 Martin Quinson
2677 - No major issue in previous version => change versionning schema
2678 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2679 corresponding dataset.
2681 2004-07-11 Martin Quinson
2683 - Get it working with any kind of structure (we can compute the padding
2684 bytes remotely for all the architectures I have access to)
2685 - Implement the structure parsing macro (still not quite robust/complete)
2686 - Improvement to the remote testing toysuite
2688 2004-07-10 Martin Quinson
2689 [autoconf mechanism]
2690 - get ride of a bunch of deprecated macros
2691 - actually run the test for two-compliment, not only compile it :-/
2692 - test whether the structures get packed (and bail out if yes. Damn.
2693 Alignment is a serious matter)
2694 - test whether the structures get compacted (but respecting the alignment
2695 constraints of each types)
2696 - test whether the array fields of structures can straddle alignment boundaries
2698 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2699 (took me 2 hours to find that bug, looking at the wrong place)
2700 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2701 before sending/receiving and are used to compute the offsets of fields
2702 - Given that padding size depend even on compiler options, keep track of
2703 alignment and aligned_size only for the current architecture. Not a big
2704 deal since we send structure fields one after the other (seems
2706 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2707 program, let it work on linux/gcc/little32. portability todo.
2709 2004-07-08 Martin Quinson
2710 - import and improve remote compilation support from FAST
2711 - make sure make check works on half a dozen of machines out there
2713 2004-07-07 Martin Quinson
2714 Let's say it's version 0.0.3 ;)
2715 - Implement conversions (yuhu!)
2716 - Let it work on solaris (beside conversion, of course)
2717 - Stupid me, using rand() to generate the conversion datatests in not wise.
2719 2004-07-06 Martin Quinson
2720 - Let make dist work, since I'm gonna need it to compile on remote hosts
2721 - Let Tests/datadesc_usage write the architecture on which the file was
2722 generated as first byte.
2723 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2724 architecture descriptions.
2725 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2726 on those architectures.
2727 - Optimization: send/recv array of scalar in one shoot
2729 2004-07-05 Martin Quinson
2730 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2732 - Plug a whole bunch of memleaks
2733 - each process now have to call gras_{init,exit}. One day, their log
2734 settings will be separated
2735 - Continue the code factorisation between SG, RL and common in Transport.
2737 2004-07-04 Martin Quinson
2739 - Redistribution between SG and RL.
2740 We wanna have to accept in SG, so move accepted related parts of RL in
2741 the common part. (more precisely, the dynar of all known sockets is no
2742 more a static in transport.c, but part of the process_data)
2744 [gras_stub_generator]
2745 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2747 2004-07-03 Martin Quinson
2748 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2749 them all up in one shot)
2751 - Fix the ugly case of reference to dynamic array.
2752 - New (semi-public) function gras_datadesc_size to allow the messaging
2753 layer to malloc the needed space for the buffer.
2755 - gras_socket_close now expect the socket to close (and not its address to
2756 put NULL in it after it). This is because the socket passed to handlers
2757 is one of their argument (=> not writable).
2759 - propagate the interface cleanup from last week in datadesc, ie remove a
2760 superfluous level of indirection. User pass adress of variable
2761 containing data (both when sending and receiving), and not of a variable
2762 being a pointer to the data. Let's say that I like it better ;)
2763 The price for that is constructs like "int msg=*(int*)payload" in
2764 handlers, but it's a fine price, IMHO.
2766 - Let it work in RL (yuhu)
2768 2004-06-21 Martin Quinson
2770 - porting SG plugin and SG select to new standards (works almost).
2771 - plug memleaks and fix bugs around.
2774 - cleanup the prototype of data recv and force users to specify when they
2775 want to handle references to objects. Test case working even for cycles.
2776 - plug memleaks. Valgrind is perfectly ok with this.
2778 2004-06-12 Martin Quinson
2780 - cleanup the separation between plugin and main code in plugin creation
2782 2004-06-11 Martin Quinson
2784 - Reput hook for raw sockets, needed for BW experiments
2785 - kill a few lines of dead code
2786 [Data description] Interface cleanup
2787 - gras_datadesc_by_name returns the searched type or NULL.
2788 That way, no variable is needed to use a type desc once, which makes
2790 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2791 two parameters were strings (field name, type name), leading to
2793 [Dicos] Interface cleanup
2794 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2795 This is consistant with the dynar API.
2797 2004-04-21 Martin Quinson
2799 - Porting to new standards.
2801 - interface cleanup.
2802 There is no bag anymore, no need to take extra provision to mask the
2803 pointers behind "ID".
2804 Better splitup of functions between files create/exchange/convert.
2805 This is still a bit artificial since convert and receive are so
2806 interleaved, but anyway.
2808 - add a queued message list to procdata (the ones not matching criteria
2810 - factorize some more code between SG and RL wrt procdata
2812 - use gras_exit in example to track memleaks
2813 - get rid of gs_example now that GS is properly integrated into gras
2814 - update run_test to integrate the latest tests (datadesc)
2816 - rename WARNINGn macros to WARNn since it prooved error-prone
2818 2004-04-19 Martin Quinson
2820 - register init/exit functions within gras module mechanism
2821 - send/receive function.
2822 Convertion is not implemented, but short-cutted if not needed.
2823 struct/array elements are sent one by one (instead of block-wise), but
2824 nobody really cares (yet). Get a prototype before optimizing.
2825 - tests (using a file socket) for DD send/receive on:
2826 - base types: int, float
2827 - array: fixed size, string (ie ref to dynamic string)
2828 - structure: homogeneous, heterogeneous
2829 - chained list, graph with cycle
2830 Believe it or not, valgrind is not too unhappy with the results. The
2831 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2835 - Bugfix when using a filename explicitely (instead of '-')
2837 2004-04-09 Martin Quinson
2839 - factorize more code between RL and SG in socket creation
2840 - Complete the implementation and tests of:
2842 o file (only in RL, and mainly for debugging)
2844 I lost 3 days to design a portable address resolver, and then decided
2845 that the prototype mainly have to run on my box.
2846 Addressing portability too early may be like optimizing too early :-/
2848 - use gras_init in the Tests instead of the crappy parse_log_opt
2849 (the latter function is removed)
2850 [Conditional execution]
2851 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2852 [Code reorganisation]
2853 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2854 Build examples against the RL library, since there is no way to disable
2855 its creation for now.
2857 For information, the beginning of coding on GRAS was back in june
2858 2003. I guess that every line has been rewritten at least twice since