1 SimGrid (3.12) NOT RELEASED; urgency=low
4 * Interface improvement:
5 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
6 - Sanitize the interface in MSG_task_ module:
7 - Merge two functions that were close enough but misleading:
8 set_compute_duration(t) -> set_flops_amount(t)
9 get_remaining_computation(t) -> set_flops_amount(t)
10 - set_data_size(t) -> set_bytes_amount(t)
11 get_data_size(t) -> get_bytes_amount(t)
12 - Massive cleanups in the functions related to the energy
13 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
14 - New: MSG_host_get_pstate()
16 - msg/energy/onoff: switching hosts on and off
18 * Interface improvement:
19 - Rename Host.isAvail() to Host.isOn()
20 - Rename Process.currentProcess() to Process.getCurrentProcess()
21 - Rename Task.setDataSize() to Task.setBytesAmount()
22 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
24 - #18874: Actually allows the GC to reclaim tasks
28 - SIMIX_process_throw: raises an exception in a remote process
29 * Refactoring: Separate submodules
30 - libsmx: the public interface, as libc in a real system
31 - popping: the strange dance that converts a user request into a kernel handling
32 - smx_context_*: the virtualization mechanisms that embeed the user code
33 - smx_*: the handling of each simcalls
35 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
36 * Rename smx_action_t into smx_synchro_t, making explicit that these
37 things are used to synchronize processes with their environment.
38 For example, a communication is a sort of synchronization involving
39 the communicating processes (that may block until the exchange) and
40 the platform. The same can be said from computations, etc.
42 - #18888: segfault when a process ends before its kill_time
46 - 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.
47 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
48 - MPI_Type_set_name, MPI_Type_get_name
49 - MPI_*_c2f and MPI_*_f2c functions
50 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
51 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
52 - Activate a lot of new tests from the mpich 3 testsuite
54 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
55 - InfiniBand network model added : Based on the works of Jerome Vienne (http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf )
56 - When smpi/display_timing is set, also display global simulation time and application times
57 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
58 * Collective communications
59 - 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
60 - 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/).
61 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
63 - Replay now uses algorithms from wanted collective selector
64 - Replay can be used with SMP-aware algorithms
65 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
66 - 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.
68 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
69 - Correctly use loopback on fat-tree clusters
70 - Asynchronous small messages shouldn't trigger deadlocks anymore
71 * Energy/DVFS cleanup and improvement
72 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
73 - new: smpi_get_host_pstate()
76 - "Full" network optimization flag was broken since Surf++
77 - Better handling of precision flags in maxmin
78 - Fix bug causing sometimes "Impossible" errors
79 - Properly pass cluster properties to included hosts
80 * Improvement of the Energy plugin.
81 - Always update the consumption before returning that value
82 - New property: watt_off to denote the disipation when the host is off
83 - New functions getWattMinAt and getWattMaxAt to retrieve the
84 dissipation of pstates that we are not currently at.
87 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
88 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
89 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
91 * Tracing is now always enabled (no way to turn it out)
92 * Move headers around to sort them out on installed systems:
93 - instr/instr.h -> simgrid/instr.h
94 - instr/jedule/* -> simgrid/jedule
95 - simdag/datatypes.h was removed
96 - simdag/simdag.h -> simgrid/simdag.h
97 - msg/datatypes.h was removed
98 - msg/msg.h -> simgrid/msg.h
100 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
102 SimGrid (3.11) stable; urgency=low
107 * Normalizing pointers addresses tool for better diff between logs
110 * Add cloud examples using new VMs
111 - examples/msg/cloud/two_tasks_vm.tesh
112 - examples/msg/cloud/simple_vm.tesh
113 - examples/java/cloud/cloud.tesh
114 - examples/java/cloud/migration/migration.tesh
115 * Add java surf examples:
116 - examples/java/surfPlugin/surf_plugin.tesh
117 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
118 - examples/java/surfCpuModel/surf_cpu_model.tesh
119 * Add SMPI+MSG example:
120 - examples/smpi/smpi_msg_masterslave/
125 - msg task destroy cancel test
126 - msg_host on/off test
127 * Move all tests in testsuite to teshsuite (adding tesh files)
128 * Restructure teshsuites
129 - one folder for each kind of test
130 * Restructure AddTests.cmake
132 - structure the order of tests (with sections)
135 * Add virtual machine
136 - creation of a VM on a PM
137 - migration of a VM from a PM to another PM
139 - MSG_process_join(msg_process_t process, double timeout)
140 - msg_bar_t MSG_barrier_init(unsigned int count)
141 - int MSG_barrier_wait(msg_bar_t barrier)
142 - void MSG_barrier_destroy(msg_bar_t barrier)
143 - msg_as_t MSG_environment_as_get_by_name(const char * name)
144 * New option "msg/debug_multiple_use" to help debugging when a task is used
147 - Improvements and finalization of MSG_storage, MSG_file APIs and their
149 - Increase code coverage in test suites
153 * Protect context stacks against stack overflow. The number of protected memory
154 pages allocated on the top of each stack (1 by default) can be configured
155 with the new command line option --cfg=contexts/guard_size:<value>.
156 * Simcalls are now generated by a python script that
157 - generates files included by SimGrid
158 - checks that all the functions exist, and proposes prototypes
160 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
161 host_set_data, host_get_data
163 - simcall_process_join(smx_process_t process, double timeout)
164 * Fix bug where sleeping processing could not be suspended.
167 * Translate surf models from C to C++
168 - Generic classes for all models: Model, Resource, Action
169 - A generic interface for each kind of model (CPU, Network, Storage
170 Workstation, WorkstationVM)
172 * Translate surf routings from C to C++
173 * Add callbacks using sigc++ or boost::signals2
174 - Add callback functions for resource creation/destruction
175 - Add callback functions for action state change
176 - Handle Energy as a plugin
177 * Replace swag by boost::intrusive
178 * Add new routing models for clusters. For documentation, see
179 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
180 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
181 parameters for cluster tag
182 - Fat trees, with topology="FAT_TREE" and
183 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
184 - see examples/platforms/torus_cluster.xml and
185 examples/platforms/fat_tree_cluster.xml
189 * Hostfiles support host:nb_processes construct to deploy several processes on
191 * Collective communication algorithms should not crash if used with
192 improper number of nodes and report the error.
193 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
194 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
195 MPI_Dims_create, MPI_Cart_sub are supported.
196 * New interface to use SMPI programmatically (still depends on MSG for
197 some parts, see examples/smpi/smpi_msg_masterslave) :
198 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
202 * Global variables privatization in MPI executables is now performed at runtime
203 with the option smpi/privatize_global_variables (default:no).
204 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
205 dynamic libraries loaded by the application are not privatized (static
206 linking with these libraries is advised in this case)
209 * Options defined in XML work correctly now.
212 * New cmake option, enable_lib_in_jar, to control whether native libraries are
213 copied into simgrid.jar or not (ON by default). Use this option if you want
214 to reduce the size of the installed simgrid.jar, *and* the native libraries
215 are kept installed elsewhere.
216 * Surf binding with SWIG (code generated in maintainer mode only):
217 - plugin to handle callbacks
218 - CPU model only for the moment
221 * Supernovae build mode is definitively removed. It was used to improve
222 inlining and inter-module optimizations. It is nowadays superseded by
223 link-time optimizations commonly available in compilers.
224 * Update ns-3 find lib. Bindings for ns-3 should work again now.
225 * Add boost dependency for surf++
226 * Add new macro for tests
227 - ADD_TESH(name <tesh_args>)
228 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
232 - xbt_bar_t XBT_barrier_init(unsigned int count)
233 - int XBT_barrier_wait(xbt_bar_t barrier)
234 - void XBT_barrier_destroy(xbt_bar_t barrier)
235 * Make the xbt_os_time module public
237 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
239 SimGrid (3.10) stable; urgency=low
241 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
244 * Reintegrate Java to the main archive as desynchronizing these
245 package is not acceptable anymore (Java is now considered stable)
246 * Add explicit synchronization facilities through semaphores
247 * Bug fix: Task.setDataSize() only changed the C world, not the value
248 cached in the Java world
251 * Dramatically change the way files are handled. API and internals changed, but
252 this part of MSG was not considered as production grade either.
253 * Add explicit synchronization facilities through semaphores
254 * Add a new function MSG_host_get_process_list()
255 * Preliminary DVFS support (see examples/msg/energy/ for details)
258 * SMPI is now included directly in the libsimgrid as the windows
259 linker doesn't force us on splitting it anymore.
260 * Improvements of the SMPI replay tool:
261 - Most of the collective communications are now rooted in the same process as
262 in the original application.
263 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
264 used until now). Multiple data types can now be used in a trace.
265 - The replay tool now supports traces produce either by TAU or a modified
267 - Bug Fix: the compute part of the reduce action is now taken into account.
268 - Gatherv collective is now supported
269 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
270 of smpirun outputs time independent traces for the current run. One file
271 is created per process. If too many processes are simulated, this behavior
272 can be changed to one file for all processes by using the
273 tracing/smpi/format/ti_one_file flag
274 * smpirun generates the host file if needed (with given host count and platform)
275 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
276 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
277 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
278 --cfg=smpi/coll_selector:(mpich/ompi)
279 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
281 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
282 happening inside a collective SMPI call.
283 * Fix the behavior of complex data types handling.
284 * Make MPI_Wtime another synchronization point to take computations into
286 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
287 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
288 * Add all missing Fortran bindings, SMPI should work with Fortran 90
289 (no privatization of global variables yet)
290 * Preliminary DVFS support (see examples/smpi/energy/ for details)
293 * Verification of liveness properties is now available for SMPI applications
294 (in addition to MSG applications)
295 * Bugged examples using SMPI in examples/smpi/mc/
296 * Add --cfg=model-check/visited option. Allows the verification of infinite
297 programs. Detection of loops in the execution thanks to the system state
298 comparison and reduction of the state space to explore. Can be combined with
299 DPOR for safety properties.
302 * Allow to change SimGrid configuration (see --help) within the code
303 thanks to SD_config() as it can be done in MSG.
304 * Add a new function SD_task_set_amount() upon user request.
307 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
308 * Remove rule based routing (no more PCRE dependency)
309 * Add a limiter_link option to cluster tag, to specify a maximum reachable
310 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
311 * Add a loopback_bw and loopback_lat options to cluster tag.
312 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
313 routing. Now peers are encapsulated in an AS and have their own private
314 router but this is transparent.
317 * Our own implementation of getline is renamed xbt_getline, and gets
318 used even if the OS provide a getline(). This should reduce the
319 configuration complexity by using the same code on all platforms.
320 * New type: xbt_cfg_elm_boolean.
321 * Allow to use yes/no for boolean configuration options in the command line.
322 * Allow to disable SimGrid cleanups at exit from command line option.
323 There are situations where one may want a simulation to end with an exit.
324 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
325 annoying when scripting around the simulator. Adding a
326 --cfg=clean_atexit:no allows to circumvent this issue.
329 * Lots of memory leaks were corrected in this release.
330 There is no known memory leaks anymore, in all of our 600+ tests.
331 * New command line option --version, to get SimGrid version information.
332 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
334 * Supernovae builds are deprecated, and expected to be removed in the next
337 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
339 SimGrid (3.9) stable; urgency=low
341 The Grasgory release: GRAS is really dead now.
343 * Complete overhaul of the internal host structures scheme.
346 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
347 was considered as experimental and badly maintained since then.
348 * Keeping it was thus a trap to our potential users, that could take
349 it instead of MSG or SMPI by mistake despite is pity state.
350 * GRAS seems to have very few users (if any), and no one volunteered
351 to maintain it further. It also induces a lot of XBT code (for
352 portability sake), that must be maintained too.
353 * For all these reasons, we killed GRAS. If someone wants to revive it
354 in the future, don't cry, our git history still remembers of GRAS.
357 * Major overhaul. Merge our documentation again as time proved that
358 splitting it was really not helping our users.
359 * Further improve the developer documentation to help newcomers
360 hacking on SimGrid itself. The user documentation (and in
361 particular, the beginner documentation) is still in a sorry state.
364 * Now works on Windows too!
365 * Much more extensive test suite, from MPICH
368 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
369 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
370 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
371 * Bug fix: let task be scheduled when the last dependency to be solved is
372 a control dependency.
373 * Remove SD_load_environment_script function.
374 Use the C sg_platf function if you want to declare a platform
378 * New function: MSG_process_get_number()
379 * Old function documented: MSG_config()
380 * Remove MSG_load_platform_script function
381 Use the C sg_platf function if you want to declare a platform
385 * Change the default value of the TCP_gamma constant (maximal size of TCP
386 congestion window) to a more realistic 4MiB value. If you notice changes in
387 your simulation results, you can fall back to the previous 20k tiny window
388 by adding --cfg=network/TCP_gamma:20000 on command line.
389 * (Hopefully) fix a bug wrt periodic availability and state traces
390 * Bug fix: use default values at start when first event in availability/state
391 trace is not at time 0.
394 * remove the "new_" part of function name sg_platf_new_trace_connect
395 (resulting in sg_platf_trace_connect), since it does not create
399 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
400 use case where it's really mandatory, and maintaining it was a pain in
402 * New: xbt_fifo_search(), search an item with a user-provided
403 comparison function instead of dumb pointer comparison.
406 * Fix the lua deployment:
407 Use `simgrid.init_application()` before deployment instead of
408 `simgrid.msg_register_application()` after.
411 * Transfer the tracing files into the corresponding modules.
413 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
415 SimGrid (3.8.1) stable; urgency=low
417 The "we are told that some people want to also *install* the simgrid
420 * Add missing file "tesh.1" to the archive.
422 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
424 SimGrid (3.8) stable; urgency=low
426 The Psssshiiiit release: SimGrid jumps into the Cloud.
429 * Add an experimental interface to manipulate VMs. They are mainly
430 process groups with very few intrinsic semantic, but they should
431 allow you to build the semantic you want easily.
432 * New function: MSG_host_set_property_value()
433 * New function: MSG_process_on_exit(). To clean memory in all cases.
434 * Bug fixes that made the host (and link) failures unusable.
435 * Add a way to auto-restart process when the host in which they are
436 executing comes back (ON_FAILURE="RESTART" on deployment file,
437 MSG_process_auto_restart_set).
438 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
439 please stop using the old ones, they are DEPRECATED.
441 * Deprecate functions MSG_global_init() / MSG_global_init_args()
442 Please use MSG_init() instead. (reducing the amount of entry
443 points in the library helps us).
444 * Make it impossible to link against the wrong version of the lib
445 * Deprecate MSG_clean(). No need to call it anymore.
446 * Function MSG_get_host_number() is not deprecated anymore.
449 * Split the doc into a user guide and a reference guide.
450 * Start a developper guide to help people hacking on SimGrid.
453 * Enable tracing by default. This modules rocks you should use it.
454 * Remove option custom_flags. Now use environment variables CFLAGS
456 * Use default cmake things to detect lua instead of home grown ones.
457 * New option "enable_mallocators" to disable mallocators, for debugging
458 purpose ("on" by default).
461 * Bug fixes around the resource failures: don't let the processes
462 survive the host they are running onto.
463 * Add an interface to auto-restart processes when the host in which they are
464 executing comes back.
465 * Ensures that SIMIX_clean is called automatically. It's not part of
466 the public interface anymore (bindings should be updated).
469 * Bug fix for when SD_Simulate is called with a positive value: be careful
470 when comparing doubles. Sometimes they are different for non significant
472 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
473 parallel task whose initial work is distributed among host according
474 to the Amdahl's law. Such tasks are created with a parameter alpha
475 that corresponds to the non-parallelizable part of the computation.
476 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
477 between two sets of workstations assuming a 1D block distribution (each
478 workstation owns a similar share of data) on both sides.
480 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
481 Data redistribution will be automatically scheduled once parent and child
482 are both scheduled. The filling of computation_amount and
483 communication_amount structures is now done seamlessly thanks to the chosen
485 * New function SD_workstation_dump to display various information
486 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
487 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
488 bandwidth on the route onto which the task is scheduled and the amount of
490 To divide the nominal bandwidth by 2, the rate then has to be :
491 rate = bandwidth/(2*amount)
492 * Compute tasks that have failed can now be rescheduled and executed again
493 (from their beginning)
494 * Increasing source code coverage (src/simdag is now covered at 95.8%
498 * Re-implement time-independent trace replay using SMPI (at the
499 smpi_smp_* level) instead of MSG. This should replace
500 examples/msg/actions/actions.c
501 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
502 hindexed and structs)
503 * Implement the exchange of non-contiguous data.
504 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
505 * Correct behavior of smpi/sender_gap and set its default value to 0
506 * Add option to asynchronously send small messages to allow better
507 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
508 specifies the size in bytes under which messages will be asynchronously sent.
509 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
510 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
513 * New C interface to define a platform: XML is now optional.
514 For more info, please check include/simgrid/platf.h.
515 * New interface to define random platforms from the C:
516 For more info, please check include/simgrid/platf_generator.h and
517 examples/msg/masterslave/masterslave_platfgen.c
518 * Export a sg_cmdline dynar containing all the arguments we got from
522 * Two new tracing options for adding comments to trace file so you
523 can track your experiments (see --help-tracing for details).
524 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
525 * Adding the SimGrid version that generated the trace file as a comment.
526 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
527 * Fix to avoid key clashes on Paje links
528 * Other minor fixes related to the Paje specification
531 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
532 and renamed to xbt_str_hash() and xbt_str_hash_ext().
533 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
534 * Greatly improve the robustness of mmalloc to user errors (such as
535 using an area after freeing it, or freeing it twice)
537 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
539 SimGrid (3.7.1) stable; urgency=low
542 * Restore the prototype of MSG_process_create_with_environment() to
543 the pre-3.7 situation by removing the kill_time argument.
544 * Add a MSG_process_set_kill_time() function instead.
547 * Fix weird behaviors when dealing with parallel tasks.
550 * Simgrid is now built as a dll.
551 * Simgrid-java now works on Windows.
552 * Simgrid-Java is now included into Windows package.
555 * First pre-build package for MacOSX.
558 * Fix compilation when using MSG_USE_DEPRECATED.
559 * Fix some compilation issues on Macs and Windows.
560 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
561 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
563 We discovered that the Lua console is broken, but we are missing the
564 manpower to fix it right now. The problem existed in 3.7 too, so we
565 are not blocking the release for that. Sorry if you depended on this
566 feature, any help would be really welcome.
568 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
570 SimGrid (3.7) stable; urgency=low
572 The "spring cleanups (before the next Big Project kicks in)" release.
575 * We can specify the SMPI latency/bandwidth factor with command line
576 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
577 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
578 You can also use the "config tag" from platform file by adding this line
579 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
580 (see "example/platforms/tag_config.xml" to use "config tag").
581 Note that the command line supersedes the platform file configuration.
582 * Change the correction factors used in LMM model, according to
583 the latest experiments described in INRIA RR-7821.
584 Accuracy should be improved this way.
585 * Use the partial invalidation optimization by default for the
586 network too. Should produce the exact same results, only faster.
587 * Major cleanup in surf to merge models and split some optimization
588 mechanisms from the core of the models. As a result you can now
589 specify which model to use (e.g., --cfg=network/model:LV08
590 --cfg=cpu/model:Cas01) and which optimization mode to use
591 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
592 Incompatible combinations should err at initialization. See
593 --help-models for the list of all models and optimization modes.
594 * The CLM03 workstation model was dropped for simplicity because it
595 used the deprecated CM02 network model. Use default instead.
596 * Rename the TCP_gamma configuration option to network/TCP_gamma
597 * Rename the coordinates configuration option to
598 network/coordinates, and document it
599 * Use now crosstraffic keyword instead of the terribly misleading
600 fullduplex keyword. It is activated by default now in the current
601 default model, use --cfg=network/crosstraffic:0 to turn it off.
602 * Ongoing refactoring the model parsing to make XML files optional
603 See include/simgrid/platf.h for details (still to be completed)
606 * Major overhaul of the documentation. Almost instructive now :/
607 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
608 functions and friends. This interface was considered as
609 deprecated since over 2 years, it's time to inform our users that it is.
610 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
611 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
612 use these (crappy) functions in your code.
613 These functions will be removed soon. Stop using them now.
614 * Deprecate MSG_get_host_{table,number}
615 Implement MSG_hosts_as_dynar() instead.
616 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
617 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
620 * Stabilize the parallel execution mode of user contexts
621 * Introduce configuration variables to control parallel execution:
622 - contexts/synchro: Synchronization mode to use when running
623 contexts in parallel (either futex, posix or busy_wait)
624 - contexts/parallel_threshold: Minimal number of user contexts
625 that must be part of a scheduling round to switch to parallel
626 execution mode (raw contexts only)
627 * Fix bugs that prevented to use suspend/resume along with
628 synchronization structures.
629 * Fix bugs in process termination that lead to invalid memory access
630 in very specific conditions.
633 * Introduce a parallel mode for the models (controlled by surf/nthreads
634 configuration item). In our tests, running the models in parallel
635 never lead to any speedups because they are so fast that the gain
636 of computing each model in parallel does not amortizes the
637 synchronization costs, even when ultra fast futexes are used.
638 This is released anyway because YMMV.
641 * Performance boost by using a swag internally to compute the set of
642 tasks that are finished and should constitute the return value of
646 * Enable it by default now that it is considered rather stable.
649 * Documentation of the tracing functions.
650 * Performance gains when tracing categorized/uncategorized resource
651 utilization by avoiding calls to get route when updating resource
652 variables. LMM constraints are being used instead.
653 * API changed to set task categories. Use MSG_task_set_category instead
654 of TRACE_msg_set_task_category, and SD_task_set_category instead
655 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
656 * Bugfix for graphicator, routes not correctly obtained, memory leaks
657 * Examples for link user variables added (see at examples/msg/tracing/)
658 * Deprecated function TRACE_msg_set_process_category completely removed
659 * Trace header updated according to the latest Paje file format
660 * Tracing network lazy updates, no longer obligate users to use full updates
661 * --cfg=tracing/platform:1 also registers power/bandwidth variables
662 * Experimental: let user code declare/set/push/pop application states for hosts
663 * API changed to allow the manual creation of graph configuration files
664 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
667 * Improve the API of Lua MSG bindings, using the Lua spirit.
668 * Each simulated process now lives in its own Lua world (globals are
669 automatically duplicated). It helps writing simulators. Will allow
670 to run Splay programs within SimGrid in the future.
671 * Add a Chord example in Lua, equivalent to the MSG one.
674 * Start the implementation of a solution to express temporal
675 properties, not only local assertions. This is still an
676 experimental work in progress, stay clear from it to be safe.
680 - Add new runtime parameters --help-logs and --help-log-categories
681 to display informations about supported logging parameters and
683 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
684 don't exists anymore.
685 * Mallocators: allow value NULL for the reset function.
687 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
688 create homogeneous dictionaries, where all the elements share the
689 same free function. Non homogeneous dictionaries will be
690 deprecated in the next release.
691 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
692 - Multi-level dictionaries are deprecated.
694 - new function xbt_dynar_search_or_negative() that is useful when
695 you have less than 2 million elements in your dynar and don't
696 want of the extra complexity of catching exceptions when the
697 element is not found.
699 - Make xbt_os_thread module (for thread portability) public.
700 Documentation is still to come, sorry.
702 - Cleanups and simplifications to make it maintainable again.
703 - Exotic features (such as memalign and valloc) were removed.
704 - The metadata were extended and improved so that the
705 model-checker becomes able to explore and inspect the heaps.
706 - This may induce a performance drop when enable_model-checking is
707 ON in cmake (even if it's not used in the simulation), but it is
708 necessary at this point to get MC working.
710 Turn model-checking OFF if simulation performance matters to you.
711 Not enabling it at runtime is not enough, disable it in cmake.
713 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
715 SimGrid (3.6.2) stable; urgency=low
717 The "Not coding new stuff allows to polish old things" release.
720 * New bindings to the NS3 packet level simulator (experimental)
721 * Use the raw (efficient) execution contexts instead of the sysv
722 (portable) ones when possible.
723 * libpcre is now mandatory in any cases since not using it led to
724 severe performance loss and possibly other issues
725 * Update the XML platforms:
726 - G5K: include the latest machine in Nancy
727 - GridPP and LCG: new platforms
728 * Documentation was partially updated, at least (more to come)
730 Bug fixes, cosmetics and small improvements
731 * Free terminated processes before the end of the simulation to avoid
732 exhausting the memory of users having very dynamic amount of
734 * Bug fix and cosmetics about canceling non-running tasks
735 * Bug fix about the dot loader's issues when using libcgraph
738 * Create an installer for windows with nsis (amd64 and win32)
739 - Add an hello world project to illustrate simgrid project creation.
740 - Embed libpcre into the Simgrid installer to avoid
741 its compilation burden
742 * The raw execution contexts should work on Apple now
743 * Port to Windows 64 bits
744 - Sysv contexts now have an implementation for this arch
745 - GRAS communication features now support this arch
746 * Drop support for borland compiler on windows
747 - this code was not maintained, and we kinda depend on gcc nowadays
748 * Fix portability issues on kfreebsd/gnu: build error about semaphores
749 * Fix portability issue on unstable ubuntu: linker became picky on
752 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
754 SimGrid (3.6.1) stable; urgency=low
756 The "Oops, we broke Macs too" release
759 * Fixed contexts detection so that raw ones are used when possible
760 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
761 leads to a strange error, with user code segfaulting sometimes when
762 the generated code is not perfectly aligned (which is not
763 controllable from the user side, depends on the amount of code)
766 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
768 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
770 SimGrid (3.6) stable; urgency=medium
772 The Summer Release, also known as the "OMG! They Killed Kenny!" version
775 * Bindings now constitute their own package, separated from the main one.
776 Rationale: reduce our maintenance nightmare by reducing the module coupling
777 They will soon be released on their own on gforge.
779 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
780 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
782 GRAS: It is not considered as stable anymore, but experimental. Sorry.
783 * It's not quite deprecated for now because we have no replacement,
784 but it may soon become the case.
787 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
788 * Fortran: New user-level cache variable to store the rank of the running
789 process. This improves performance by an order of magnitude.
790 * C: New coccinelle script to automatically locate and modify global and
791 local static variables.
792 * Improved SMPI network model with a sender-side gap to account for multiple
796 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
797 only say if a communication is finished, no matter whether it succeeded or
798 failed. You can call MSG_comm_get_status() to know the status of a finished
800 * New function MSG_task_dsend() to send a task and detach it. When a
801 communication is detached, you are never notified of its success or failure
802 and the memory is released automatically once it is finished. This function
803 is useful when you don't care about the end nor the success of a
805 * Change the prototypes of action replay. Sorry for inconvenience,
806 but this is really more efficient this way (and to adapt your code,
807 you just have to fix the initialization, that shouldn't be too long)
808 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
809 ever managed to do anything useful with it.
810 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
811 takes a dynar of communications. It returns immediately and gives the
812 index of a finished communication (if any).
813 * New example: a basic implementation of the Chord P2P algorithm.
816 * New model for multi-core CPUs. You can now use the core attribute to
817 precise the number of cores of a host. This is a basic model. Every
818 process running on the host receives at most the power provided in
819 the DTD (throughput<=power). Total throughput of process cannot exceed
821 * New peer tag. This peer tag creates a tiny AS comprising a host and a
822 router linked by an up-link and a down-link (possibly asymmetrical).
823 This kind of pattern allows to easily build last-mile model style platforms.
824 Aggregating such patterns in a rule-based AS is thus the technique of
825 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
826 * New model for Vivaldi routing. We transformed the Vivaldi network model
827 into a Vivaldi routing model (based on the rule-based model). This allows to
828 combine Vivaldi based latencies with last-mile platforms.
831 * Added a check for NaN of IEEE754 infinite in the double entries of
833 * Introduce a new context factory "raw", highly inspirated from the
834 ucontext factory, but using manually crafted functions in assembly to
835 do the work in an efficient manner.
836 * Allow to change the used context factory at run time, not only at
837 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
838 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
839 contexts at runtime (only with raw contexts or ucontexts).
840 * Completely rewrote this module to allow parallel execution of user
841 processes. Use --cfg=contexts/nthreads:N to execute user processes
842 with N parallel threads (the default is 1, meaning no parallelism).
843 * Allow to decide dynamically between sequential and parallel modes.
844 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
845 processes in parallel only when their number is greater than or equal to P
847 * Added a check for NaN of IEEE754 infinite in the double entries of
851 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
852 won't output the state of processes when interrupted with Ctrl-C
853 * Add a new function xbt_dynar_to_array that transforms a dynar into a
854 NULL-terminated array. This may solve backward compatibility issues
855 due to the change to return type of SD_simulate. See also:
856 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
857 * Add new macros with variable number of arguments.
858 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
859 - in xbt/asserts.h: xbt_assert
860 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
861 - in xbt/ex.h: THROWF and RETHROWF.
862 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
864 * Change xbt_die() to accept a format string with arguments, just like printf.
865 * New data structure: xbt_lib_t, like a dict but more general and with better
869 * New configuration options
870 Options triva/categorized and triva/uncategorized can be used to generate
871 graph configuration files for Triva visualization tool.
872 * Configuration option tracing/platform is renamed to tracing/categorized
873 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
874 * New TRACE_user_link_variable interface:
875 User provides the name of the link and the tracing variable to attach to it
876 * the declaration of tracing categories must be done after the environment
878 * simpler tracing interface, just one way to declare categories
879 TRACE_category or TRACE_category_with_color, it is up to you
880 * links in the trace file are again identified by their names
881 * trace contains the full platform hierarchy exactly as declared using the ASes
882 * Options tracing/msg/[task|process]:1 groups the process by hosts
883 for both cases, tasks and processes must have names that are unique during the simulation
884 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
885 * The experimental option tracing/msg/volume is deprecated
886 its functionality may be reincorporated if needed
888 The tracing generates a trace file with unordered timestamped events,
889 because of the way the core simulator (surf) works. A script available
890 at the tools directory (fix-paje-trace.sh) can be used to put the events
891 in order. We have changed the tracing so it can generate ordered timestamped
892 events in the final trace, but depending on the simulator (and how much time
893 is simulated) that can lead to a huge memory utilization. It is deactivated
894 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
897 * Define a SIMGRID_VERSION macro in simgrid_config.h.
898 - We are trying hard to keep the API stable, but it may happen that
899 some things change (we're a research project after all, not a
900 nuclear plant operating system). If such things should happen, you
901 could rely on that macro to adapt.
902 - current value: 30600 for 3.06.00, aka 3.6
903 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
904 a number that can be compared with SIMGRID_VERSION.
905 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
906 assertions and verbose/debug logging events are disabled at compile time.
908 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
910 SimGrid (3.5) stable; urgency=medium
913 * New feature to any SimGrid-based simulator: Model-Checking
914 Check SIN#1 for more details.
917 * New Model SMPI (three-interval linear regression for correction factors)
918 See RR-7426, available at http://hal.inria.fr/inria-00527150
919 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
920 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
921 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
922 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
923 * See also src/smpi/README
928 - Tracing API changes: TRACE_start and TRACE_end should not be called
929 by user-code. They are automatically called by simulators created
930 with SimDAG, MSG and SMPI if the toolkit is compiled with
931 tracing_enabled=ON. Categories declaration and utilization remain the
932 same for MSG and SimDag.
933 - A function was added to the tracing API to declare categories with
935 - TRACE_category_with_color (char *category, char *color)
936 where color must be in the following format
937 "%f %f %f", red, green, blue
938 and red, green, blue are float values in the interval [0, 1]
939 - User can specify NULL as color parameter, or continue calling
941 On that case, the tracing system will define random colors
942 - The following command-line options are supported:
943 --cfg=tracing/filename:msg.trace
944 --cfg=tracing:1 (activate tracing, needed to use others)
945 --cfg=tracing/platform:1 (categorized resource use)
946 --cfg=tracing/uncategorized:1 (uncategorized resource use)
947 --cfg=tracing/msg/task:1 (task creation)
948 --cfg=tracing/msg/process:1 (process creation, migration)
949 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
950 --cfg=tracing/smpi:1 (SMPI interface tracing)
951 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
952 - examples of examples/msg/tracing updated
954 - DAXLoader and DOTLoader functions can generate tasks with categories
955 - A new function to attribute a category to SD tasks:
956 TRACE_sd_set_task_category (SD_task_t task, char *category)
957 * Tracing the MPI interface implemented by SMPI
958 - Collective operations are traced with states
959 - Point-to-Point operations are traced with states/links
960 - Tracing activated by a parameter "-trace filename" passed
961 to smpirun during execution (considering that simgrid
962 is compiled with tracing enabled)
963 - To run the simulation with gdb, the simulator
964 accepts --cfg=tracing/smpi:1 to trace SMPI
965 - tesh files to check if smpi tracing is ok
966 - See examples/smpi/NAS/DT-trace
967 * GTNetS tracing re-worked
968 - adaptation to the tracing system of GTNets to cope
969 with modifications regarding the fullduplex mode
970 - new tesh files to check if gtnets tracing is ok
973 * Asynchronous communications through the functions:
974 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
975 * New function: MSG_load_platform_script()
976 to make possible using a lua script instead of XML files to set up platforms
977 * New function: MSG_set_function
978 to associate functions to processes, used when bypassing the parser
979 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
981 Platforms: Add some more examples in examples/platforms
982 * Grid'5000: see www.grid5000.fr
983 * *_30000_hosts.xml: various huge files [mainly scalability testing]
986 * Change the XML format. This is a very important modification. SimGrid
987 3.5 introduces a new hierarchical format based on the notion of
988 Autonomous Systems. Compatibility with old format is ensured
989 through the perl script provided in the install bin directory
990 bin/simgrid_update_xml.
991 It is now possible to build platforms with specific routing
992 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
993 connect several platforms together. We will try to provide soon
994 set of realistic platforms exploiting these properties (have a
995 look at examples/platforms/ for the moment).
996 * Take the opportunity of the XML format change to be a good XML citizen:
997 rename link:ctn to link_ctn and similar changes (also dealt with by
999 * Add a new routing scheme (rule-based) using regular expressions. It
1000 enables to have an extremely low memory footprint when the
1001 underlying routing is simple and can be compactly described. You
1002 need to have libpcre4-dev (perl regular expressions) installed if
1003 you want to use this routing scheme.
1004 * Revive the cluster TAG and allow to easily and efficiently (both in
1005 term of memory and speed) connect clusters together. Have a look
1006 at teshsuite/simdag/platforms/ to see how this can be done. With
1007 this tag, you can create clusters with thousands of tasks at no
1008 cost (have a look at examples/platforms/).
1009 Note: clusters are implemented as ASes, so there is no need for an
1010 enclosing AS tag if you have only one cluster in your platform.
1011 * Add new generic functions in the public interface that allows the user
1012 to call SURF 'create_resource' methods from your code (same
1013 functionality as the XML bypass mechanism but with a much lighter
1015 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1016 that uses a piecewise linear approximation to produce better
1017 results when exchanging small messages.
1018 * Add a new parameter to handle correctly full duplex link and account
1019 for interferences between uplink and downlink communications
1020 (activate with --cfg=fullduplex:1).
1023 * Rename the SD_READY (all dependencies are satisfied and task is
1024 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1025 dependencies are satisfied) state.
1026 This prevents a confusion between the notion of "ready to schedule"
1027 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1028 simulated" (SD_RUNNABLE) used by the simulation kernel.
1029 * Change the way a task is considered as ready. Instead of removing
1030 dependencies when a task is done, a counter is decreased. This way,
1031 it is always possible to reach ancestors thanks to the
1032 SD_taks_get_parents function (even after the end of the simulation.)
1033 * Change the return type of SD_Simulate from (SD_task_t*) into
1034 xbt_dynar_t. This function was in handling a dynar internally and
1035 converted it into a NULL terminated array for historical reasons.
1036 * New function SD_dotload(char*) to load a DAG described in dot
1037 format. This loader and the corresponding examples require the
1038 installation of the graphviz library.
1039 * Fix a bug in the management of tasks of size 0 in the surf network
1040 models. This problem was only visible with SIMDAG and you should
1041 thus disregard results produced with earlier versions if you
1042 relied on this feature (some tasks were blocked because of this).
1043 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1044 even though your applications were doing only point-to-point
1045 communications and sequential computations. Now you can really use any
1046 model you want (of course, if you create real parallel tasks, which are
1047 not implemented in most models beside ptaskL07, this will abort).
1048 * Add an example that schedules a DAX on an heterogeneous platform
1049 using a Min-Min strategy.
1050 * New function SD_workstation_get_current_task() that returns the kind
1051 of task currently running on a workstation in the sequential access
1053 * Raise some warnings when unexecuted tasks remains at the end of the
1054 simulation. This is usually caused by cycles in the DAG.
1057 * New function: SIMIX_process_set_function() called by MSG_set_function
1058 * Change the underlying waiting queue in semaphores so that a process
1059 can wait on several of them simultaneously (as in waitany).
1060 * Fix the way to handle tokens in semaphores so that all access patterns
1061 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1062 * kill the dirty pimple SIMIX_message_sizes_output()
1063 Please use (proper) visualization instead
1066 * New data container: setset (set of sets of elements)
1067 * New module: mmalloc (mapped malloc, allowing to have several
1068 independent segments of malloc)
1069 * New function: xbt_dict_cursor_set_data()
1070 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1071 * New function: xbt_dynar_is_empty()
1072 * New function: xbt_fifo_get_last_item()
1073 * Fix xbt_dynar_shrink(): use the right element size.
1074 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1075 uninitialized areas during expand.
1076 * Fix semaphores: previous implementation was severely broken.
1077 * Use library init/fini functions for our initialization.
1078 - you can use logs and other feature as soon as you want in your
1079 code (even before the xbt_init / MSG_init)
1080 - xbt_exit is now a no-op and produce a warning when used.
1083 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1084 benefit from the latest improvement to the simulation kernel.
1085 * Kill measurement sockets for now. If you rely on them, sorry. This
1086 release is not for you. This feature will be reintroduced in the
1087 future, but we cannot delay the release any further.
1088 * New function: gras_msgtype_get_name().
1089 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1090 * Fix (at last) the pmm example: it should not randomly fail anymore.
1092 Build chain: bug fixes and overall polishing
1093 * Cmake is now stable enough. Hence, we killed the autotools.
1094 * Port to windows ( TM :)
1095 * Fix the 'make install' target.
1096 No need to use 'make install-simgrid' anymore
1097 * Introduce a 'make dist' target compiling a *source* archive
1098 'make package' compiles a binary archive
1099 * Compile java files only on need
1100 * Add --cd and --setenv command line options to tesh
1101 * Out of source builds are not fully supported yet, but we are close
1102 * Enable supernovae and optimization flags by default for our users
1105 * Add layer to set up environment directly from lua, without XML.
1106 * The effect of gras_stub_generator can be achieved through
1107 lua too (check examples/gras/console/ping_generator.lua)
1109 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1111 SimGrid (3.4.1) stable; urgency=low
1113 The "Polishing easter eggs is probably a good idea" release.
1114 This is a bug fixes release only.
1118 * Fix a bug preventing the tasks from begin garbage collected.
1121 * Fix a bug occuring when a host involved in a communication fails.
1122 This was not detected properly by the other peer involved in the
1123 communication. Now, it's reported as a network error.
1126 * Warn the user about loop dependencies in data flow of DAX files
1127 * Obey the control-flow dependencies of DAX files
1130 * Add option "enable_smpi" allowing to not compile SMPI.
1131 Probably useful for the (Mac) users experiencing a build error here
1132 * Improve the detection of lua5.1 and ruby1.8
1134 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1136 SimGrid (3.4) stable; urgency=low
1138 The "Easter in Cargese" release. Also known as (major changes):
1140 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1141 ~> bindings were greatly improved
1142 ~> new tracing infrastructure for better visualization introduced
1144 * the "Welcome to configury modernity" release.
1145 ~> we switched from autotools to cmake, and improved our cdash
1148 A more detailled list of changes follow (full detail in svn log).
1150 Java Bindings: Various Cleanups
1151 * (install java-gcj-compat-dev on debian-like to use them)
1152 * Remove put/get: no need to export deprecated interface in Java
1153 Use send/receive instead.
1154 * Cleanup the examples and add a README per directory
1155 * Remove example autoDestination (that's the only way to go now)
1156 * Remove example explicitDestination (was a plain copy of basic)
1157 * Make JniException a runtime exception, so that there is no need to
1158 declare the fact that you may encounter such a beast. I guess that
1159 nobody will ever want to survive such error.
1160 * Create specific errors for each MSG case of failure:
1161 host failure, transfer failure, timeout, task cancelled
1162 * Cleanup the exceptions that may get thrown by each function
1163 * Other internal cleanups in Java bindings. Performance still bad :/
1164 Ruby and Lua Bindings: create them
1165 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1166 * That's new and great, you should try them out.
1167 Same functionalities than Java bindings, only even less polished
1169 * Kill the useless "rate" argument of SD_task_get_execution_time()
1170 Everyone used to provide -1 as a value, it was not used, and the
1171 semantic of a possible use wasn't even clear.
1172 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1173 either as comm costs or compute costs to mean that there is no
1174 such thing for that specific task.
1175 * Add a SD_task_set_name() function
1176 * Fix SD_task_unschedule() on typed tasks
1177 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1178 * In DAX loader, accept useless 'level' attributes to <job> since
1179 LIGO DAGs have them (seem to be to ease graphical representation).
1181 * Add an example masterslave_mailbox.c using send/receive and not
1182 the deprecated put/get interface.
1183 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1184 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1186 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1187 (declare MSG_USE_DEPRECATED to still have the old name)
1188 * Add a MSG_task_set_data() function
1189 * About trace replay (see examples/msg/actions):
1191 - Allow to work with splitted trace files for each process
1192 Give the specific trace file as argument of each process,
1193 and call MSG_action_trace_run(NULL)
1194 You can still have one merged file for all processes.
1195 - Fix implementation of collective operations
1196 * Allow task_execute() on 0-sized tasks (closes #10063)
1198 * This is the first release of SimGrid where SMPI is not considered
1199 beta anymore (even if some corners should still be improved)
1200 * Port over the new SIMIX_network submodule (internal refactoring)
1201 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1202 * Implement more missing elements of the standard:
1204 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1206 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1207 MPI_Scan() MPI_Get_processor_name()
1208 - Added implementation of missing case for Alltoall (warning: it's
1209 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1210 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1211 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1212 SMPI_MPI_Allgatherv()
1213 * Bug fixes include:
1214 - MPI_Waitsome() was broken
1215 - Allow relative includes in smpicc
1216 - Command line cfg argument 'reference_speed' was ignored...
1217 - Some functions did not properly lead to auto-benching of user code
1218 - smpicc passes -O2 by default (just like openmpi one)
1220 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1221 * Bug fixes about timeouts during communications
1222 * add SIMIX_message_sizes_output() as a pimple to write to file the
1223 amount of messages per size. Use gnuplot to get histogram.
1224 Pimple because that's the only user-visible function of simix,
1225 defined directly in xbt.h (irk, sorry)
1227 - Add a SIMIX_sem_get_capacity() function
1228 - Fix interactions with processe resume/suspende
1229 - release_forever() was stupidly broken
1230 - Fix SIMIX_display_process_status() for processes in a semaphore
1231 - Make SIMIX_sem_block_onto() user-visible
1232 * Refactoring context stuff:
1233 - Use pseudo-OOP for better modularity
1234 - reimplement SIMIX_process_kill() without process_schedule() so
1235 that the latter can take as invariant that it is called from
1237 - Merge context_start into context_new for sake of simplicity
1239 * Add a Vivaldi network model, coded live during SUD'10 ;)
1240 * Rename configuration variables to start a hierarchy:
1241 o cpu_model -> cpu/model
1242 o network_model -> network/model
1243 o workstation_model -> workstation/model
1244 * New configuration variables:
1245 o network/bandwidth_factor: correction to bandwith
1246 o network/latency_factor: correction to latency
1247 o netwotk/weight_S: correction to the weight of competing streams
1248 * Add a long description to the models, that users can see with such
1249 argument on the command line: --cfg=cpu/model:help
1250 * --help-models display the long description of all known models
1252 * config: add the ability to set a default value after registration
1253 Does not override any previously set value (e.g. from cmd line)
1254 * dict: allow to have integer key and data.
1255 When so, you need to use the following functions
1256 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1257 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1258 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1259 In contrary to regular dicts, the key is not malloced before copy.
1260 Mixing scalar and regular elements in the same dict is not tested
1262 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1263 Tracing for Visualization:
1264 * SimGrid is now instrumented in order to generate a trace file for
1265 visualization analysis: to use it, need to compile SimGrid with the
1266 "tracing" option enabled, and instrument the program using SimGrid with
1267 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1268 (among other functions).
1269 * The instrumentation only traces the platform utilization for now
1270 * Documentation to use the tracing functions and how to analyze the
1271 traces with the Triva tool is written.
1272 * More information about: SimGrid FAQ (in the section Tracing Simulations
1275 * We moved to cmake as default build system. Autotools support will
1276 be dropped soon. Check the FAQ for more info about how to use it.
1277 * Greatly improved our cdash/ctest interactions
1278 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1279 * Added memory checking tests with valgrind; lot of memleak fixing.
1280 This may be the first release of simgrid with so few memory issues
1281 * Added code coverage tests.
1282 Our coverage is still improvable, but at least we see it on cdash.
1284 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1286 SimGrid (3.3.4) stable; urgency=low
1288 The "Desktop Grid needs love too" release (also called Xmas release).
1290 Models improvements:
1291 * Major speedup in the maxmin system solving by using lazy evaluation
1292 Instead of solving completely the maxmin system at each iteration,
1293 only invalidate (and recompute) the modified parts.
1294 This new feature is enabled in default models but you can try to
1295 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1296 * Cas01 IMproved as default CPU model
1297 This CPU model is the same Cas01 model, but it uses the
1298 maxmin-selective-update flag and a heap structure to manage
1299 actions on SURF kernel.
1300 It reduces the complexity to find the next action to finish and,
1301 consequently, it's faster than the old Cas01.
1302 This is the new default CPU model (Cas01).
1303 * Rename the old Cas01 model to Cas01_fullupdate
1304 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1305 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1306 * CpuTI (CPU Trace Integration)
1307 A new CPU model whose objective is simulate faster when using
1308 availability trace files.
1309 Instead of using a full featured, over engineered maxmin system for
1310 CPU modeling, this model does the pre-integration of traces files
1311 to calculate the amount of CPU power available, and so, executes
1312 faster than the old CPU models.
1313 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1314 * Use LV08 as default network model since it gives better accuracy
1315 for small messages and shouldn't change things for big ones.
1316 Use --cfg=network_model:CM02 to get the previous behavior.
1319 ******************************************
1320 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1321 ******************************************
1322 * The new CPU model may changes simulations!
1323 The point is that events occurring at the exact same timestamp
1324 are not scheduled in the same order with the old and new
1325 version. This may be enough to completely change the execution
1326 of simulations in some cases.
1327 * The new network model will change simulations!
1328 This new model is more realistic than the previous one, so you
1329 should consider redoing your old experiments with this model.
1330 Sorry for the inconvenience.
1333 * Introduce the supernovae compilation mode
1334 When compiled that way, the whole SimGrid (or almost) is put in a
1335 single compilation unit and compiled in one shoot.
1336 This is to help gcc which has difficulties to inline stuff from one
1338 The speedup seem to be above 15%, althrough more tests are needed on
1339 amd64 to confirm that gain.
1342 * Port of MSG's mailbox on top of SIMIX network
1343 The put/get mechanism was greatly simplified on the way.
1346 * New SIMIX network module. Provides:
1347 - Mailbox: rendez-vous mecanism to find with who you want to speak
1348 - Synchronous send/recv: easier and hopefully faster since the
1349 logic is handled in the maestro process directly now
1350 - Asynchronous send/recv: you dreamt of it? It's here now
1351 Too bad that nobody cared enough to propagate the change to MSG.
1352 * Add semaphores as SIMIX synchronization mechanism.
1355 * new function SD_daxload(char*) to load a DAX file
1356 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1357 * Introduce typed tasks. Specify its kind and cost at creation.
1358 At scheduling, just give where it should be placed, and the cost
1359 for each involved resource is automatically computed.
1360 Existing constructors so far (more to come of course):
1361 - SD_task_create_comm_e2e() for end-to-end communication
1362 - SD_task_create_comp_seq() for sequential computation
1363 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1364 * new function SD_task_dump() for debuging display
1365 * new function SD_task_dotty(task,FILE*) writing to file the info
1366 about the task in dotty format
1367 * SD_task_dependency_exists() can now cope with having one of its
1368 arguments NULL. If so, it tests whether the other argument has any
1370 * Add getters on list of preceding/following tasks:
1371 SD_task_get_parents(task) and SD_task_get_children(task)
1372 * Add getters on amount of workstations and list:
1373 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1374 * Add getter on task kind: SD_task_get_kind(task)
1375 * Update the start_time and finish_time of tasks on completion/failure
1376 * Bugfix: Remove task from state swags when destroyed
1379 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1382 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1383 dynar with one separate thread per value of the dynar.
1384 * Change the prototype of xbt_thread_create(), sorry.
1385 Added a boolean parameter indicating whether we want to join this
1386 thread (used in SG only for now)
1387 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1390 * GTNetS wrappers should now be usable again (and betterly tested too)
1391 * Fix a major regression from 3.2 where the timeout provided to
1392 MSG_task_put_with_timeout() was used as absolute time before which
1393 the comm should be done.
1394 * Start to fix the <cluster> tag.
1395 - Internal links should be good now (beside of the loopback, which
1396 use the private link instead)
1397 - paths to the external world is still rather broken
1398 - the <route:multi> tag is just broken. Actually that's brain-dead.
1399 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1400 to make it less stupid
1401 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1402 * Fix a source-level compatibility glitch from 3.2: after defining
1403 MSG_USE_DEPRECATED, you can use the old name
1404 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1405 * Allow to compile from the SVN with automake 1.11
1406 * Fix some problems when using the "start_time" tag in deployment XMLs.
1407 * Fix #8569: XBT/synchro.h has redundant declarations
1408 * Fix #8563: MSG return values and exceptions
1409 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1410 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1411 Thanks to Robson Peixoto.
1412 * Fix a vicious bug in dictionaries inducing that some elements were
1413 not freed on xbt_dict_free()
1415 Portability report of this version:
1416 * Main portability targets:
1417 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1418 on (amd64/i386/ia64)
1419 - mac leopard on i386
1420 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1421 but nothing critical.
1422 * Other platforms: windows, AIX and others were not tested for this release
1424 Timing report of this version:
1425 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1426 scenario parameters). From 8h to a few seconds in desktop grid settings.
1427 * Supernovae brings about 25% speedup on i386.
1429 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1431 SimGrid (3.3.3) stable; urgency=low
1433 The "Need for Speed" release.
1435 The timings done to validate the 3.3.2 were faulty.
1436 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1438 The problem was a conversion from a manually handled vector to
1439 xbt_dynar_t on the critical path.
1440 xbt_dynar_foreach calls functions, inducing stack management crap.
1442 We inlined these functions and xbt_dynar_foreach is now breath taking.
1443 We also inlined xbt_swag_belong on the way.
1445 Here are some approximate speedup measurements (on master/slaves
1446 simulations lasting between 10s and 20s each):
1447 3.3.1 -> 3.3.2: about same performance
1448 3.3.2 -> 3.3.3: 40% speedup
1449 3.3.1 -> 3.3.3: 40% speedup
1450 3.3.1 with inline patch -> 3.3.3: 30% speedup
1452 Our reading is that the refactoring which occurred in 3.3.2 made us
1453 suffer much more from the xbt_dynar_foreach low performance, but
1454 once we solved this, this refactoring proved to be very performance
1455 effective. From the 40% speedup, somehow, 10% are due to the
1456 inlining and 30% to the refactoring.
1458 That's a pitty that gcc cannot inline functions placed in other files
1459 alone. We have to choose between:
1460 - break the encapsulation (by putting private data structures and
1461 accessors in headers files to help gcc)
1462 - live with low performance
1463 - switch to a decent compiler such as icc (not quite possible).
1465 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1467 SimGrid (3.3.2) stable; urgency=low
1469 The "Simplicity does not preceed complexity, but follows it" release.
1471 The main contributors of this release were (lexical order):
1472 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1475 * Extract the routing logic into its own object.
1476 (was dupplicated in network.c and workstation_LV07.c;
1477 Allows to implement other ways of storing that info)
1478 => kill now useless network_card concept
1479 - Use dynar to represent routes (instead of void** + int*)
1480 - kill link_set (use surf_network_model->resource_set instead)
1481 - Add a command-line option to choose the routing schema to use
1482 - Add three new models:
1483 * Floyd (shortest path computed at initialization)
1484 * Dijikstra (shortest path recomputed all the time)
1485 * Cached Dijikstra (shortest path computed on need)
1486 All these models where contributed by Silas De Munck, and are
1487 described in his ICCS09 paper.
1489 * Simplify model declaration
1490 (less redirections, less function to write when defining a model)
1491 - Factorize stuff between models:
1494 surf_model_resource_set(model)
1495 surf_model_resource_by_name(model, name)
1496 - Unify the types of models in s_surf_model_t (using an union)
1497 - Embeed fields of common_public directly into s_surf_model_t
1498 - Rename model methods:
1499 action_free ~> action_unref
1500 action_change_state ~> action_state_set
1501 action_get_state ~> action_state_get
1502 - Change model methods into functions :
1503 (model)->common_public->action_use ~> surf_action_ref
1505 * Implement a generic resource; use it as ancestor to specific ones
1506 (allows to kill duplicated code in models)
1507 Drawback: timer command don't need no name nor properties;
1508 workstation_CLM03 don't need no properties
1509 (but I guess we can live with those few bytes wasted)
1511 * Improve the action object model
1512 - implement a constructor avoiding dupplicated code about field
1513 initialization in generic_action part.
1515 * Kill the SDP model: it has an external dependency, is deprecated
1516 in flavor of modern lmm models, and didn't compile since a while
1519 * Relocation of the context module from XBT to SIMIX.
1520 (the context were decoupled from the simix processes, duplicating a lot of code)
1521 => a lot of code was factorized
1522 - less overhead is introduced during scheduling
1523 - simpler API for the context factory
1524 - the logic for process creation,destruction and manipulation was simplified
1525 * Simplification of the s_smx_process_t data structure.
1526 => accesing the simix level data associated to a process is faster now,
1527 and the code is a lot more readable.
1530 * Implement some more MPI primitives:
1531 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1532 -implementation: Bcast: flat or 2-ary tree (default),
1533 Barrier: 4-ary tree,
1535 Allreduce: Reduce then Bcast
1536 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1537 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1538 Alltoallv: flat tree, like ompi
1540 * Add support for optimized collectives (Bcast is now binomial by default)
1541 * Port smpirun and smpicc to OS X
1544 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1545 Shout out if you used it.
1548 * Display the list of still queued messages in SG mode when existing
1552 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1553 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1554 * Add xbt_str_from_file(FILE*)
1555 * Add xbt_dict_get_key achieving a linear reverse search
1556 * Remove the context module
1558 Portability report of this version:
1559 * Main portability targets:
1560 - Linux(debian)/x86/context
1561 - Linux(debian)/x86/pthreads
1562 - Linux(debian)/amd64/context
1563 - Linux(debian)/amd64/pthreads
1564 On these, we still have the eratic breakages of gras/pmm and
1565 amok/saturate_sg reported in previous version. We still think
1566 that the tests are the cause of the fault, not the tested code.
1568 - Mac OSX Leopard/x86/context
1569 Still false negative in tesh autotesting.
1570 Smpi still fails, but this time because readlink does not accept -f
1571 Everything seems to work properly beside of that.
1574 - AIX version 5.3 (only tested contexts this time)
1575 Smpi still fails there because mktemp is not installed.
1576 Everything seems to work properly beside of that.
1578 I managed to compile it for the first time, but several breakages.
1579 Won't delay the release for this exotic platform.
1581 * Windows: it's still lagging behind. If you want to help, please
1584 Timing report of this version:
1585 This version seem to be more than 5% faster than 3.3.1 (on linux
1586 64bits with contextes). The gain is less than expected, we are
1587 investigating this for next release.
1589 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1591 SimGrid (3.3.1) stable; urgency=low
1594 * Implement a --cfg-help to show existing configuration variables
1595 * Build chain do not require doxygen in maintainer mode
1598 * fix a bug on struct sizeof computation, which prevented the
1599 exchange of arrays of structs in some conditions
1600 - added a regression test about this in datadesc_usage
1601 * Allow the exchange of 0-long dynamic vectors.
1602 - for that, use -1 as indicator of dynamic size instead of 0
1603 - This implied to change any size from unsigned long to long,
1604 reducing a bit communication abilities, but I guess that with
1605 64bits being quite common, this is more than enough.
1606 - This also induce a protocol change, thus bumping network protocol
1607 version from 0 to 1 (if we have external users, we have to get
1608 clean on that point too ;)
1609 - added two regression tests about this in datadesc_usage
1610 * Be more verbose when propagating local exceptions
1611 This helps debugging.
1612 * Display the status of simulated processes when receiving SIGINT in
1616 * Allow to control the simulation from a trace file.
1617 New functions MSG_action_register() and MSG_action_trace_run()
1618 The first one allows to associate a function execution to each
1619 kind of action while the second one parses a trace file and
1620 triggers the corresponding actions within the system.
1621 For now, only a toy example is provided in examples/msg/actions
1622 * Add an exemple of process migration in examples/msg/migration
1623 * Fix a bug in task exchange which broke MSG_task_get_sender()
1624 Add a teshsuite regression test for that.
1625 [Bug: if MSG_task_get_sender() is called after sender exit,
1627 * Fix a bug which prevented suspend/resume to work properly
1628 * Display the status of simulated processes when receiving SIGINT
1629 This fixes a regression of v3.3. due to the introduction of SIMIX
1630 * Bug fixing in failure management:
1631 - trace could not start by a failure at time 0
1632 - failure during communications were not working
1635 * Add SIMIX_process_set_name() to change the name of the current
1636 process in the log messages.
1637 * Store smx_hosts in a dict since we only retrieve them by name
1638 * Move the configuration infrastructure to surf
1641 * Move the configuration infrastructure to surf
1644 * Massive internal cleanups:
1645 - Store internal structures on processes instead of hosts (allows
1646 to have more than one process per host, in addition of being more
1648 - Cleanup the initialization/finalization process
1649 - Kill a whole bunch of unneeded synchronization:
1650 processes run in exclusive manner within the simulator
1651 - Move queues from global tables to process data fields
1653 - now accept -platform and -hostfile arguments
1654 - Pass the right rank value to processes according to the hostfile
1655 * Compile the examples by default, and use them as regression tests
1656 * Implement MPI_Wtime()
1657 * Change the reference speed to a command line option
1660 * TCP_gamma can now be specified as command line option using
1661 --cfg=TCP_gamma:10000000.0
1662 * Change the --surf-path cmd line option into --cfg=path:
1665 * Also include strbuff from xbt.h public header
1666 * xbt_ex_display(): do not free the exception after displaying
1667 This allows to do more with the given exception afterward.
1668 Users should call xbt_ex_free() themselves.
1672 Portability report of this version:
1673 * Main portability targets:
1674 - Linux(debian)/x86/context
1675 - Linux(debian)/x86/pthreads
1676 - Linux(debian)/amd64/context
1677 - Linux(debian)/amd64/pthreads
1678 These targets fail about 1/10 of times on gras/pmm, but we believe
1679 that this is because of the test, not because of simgrid.
1680 amok/saturate_sg fails even more rarely, and the test may not be
1683 - Mac OSX Leopard/x86/context
1684 The test suite still spits tons of errors because some obscure
1685 force prevents us from removing the temporary directories
1686 arguing that they still contain some metadata I've never heard of.
1687 Smpi fails because seq is not installed.
1688 Everything seems to work properly beside of that.
1691 - AIX version 5.3 (both contexts and pthread)
1692 Smpi still fails there because mktemp is not installed.
1693 XML inclusions seems rosty on AIX.
1695 * Windows: it's still lagging behind. If you want to help, please
1698 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1700 SimGrid (3.3) stable; urgency=high
1704 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1705 [Malek Cherier & Mt]
1707 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1710 Until now, GRAS were using MSG as an interface to SURF. It was
1711 quite difficult because both interface have several differences
1712 (MSG channels vs GRAS sockets were the most notable point).
1714 This also opens the gate to SMPI (which should occur soon) and speed
1715 up simulations by to 40% (even if it were not the main goal).
1717 **************************************
1718 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1719 **************************************
1720 The point is that events occuring at the exact same timestamp are
1721 not scheduled in the same order with the old and new version. This
1722 may be enough to completely change the execution of simulations in
1723 some cases. Sorry for the inconvenience.
1725 * Cleanup and upgrade the XML format to push further scalability
1726 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1728 * Improve the testing infrastructure with tesh. Now a very large part of
1729 the code is tested not only by being run but also by checking that the
1730 output match an expected output [Mt].
1732 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1733 is really less memory-demanding, which should allow you to use
1734 larger files in SimGrid [AL].
1736 * Inform valgrind about our contextes, so that it becomes usable
1737 with the default (and more effecient) version of SimGrid
1738 [contributed by Sékou Diakite, many thanks]
1741 * Introduce a listener thread in charge of receiving incoming
1742 messages from the network. It allows to overlap communication and
1743 computation but most notably, it removes some stupid deadlocks due
1744 to the fact that so far, a process could not send and receive at
1745 the same time. This made most non trivial communication schema
1747 * Convert the PIDs from long int to int to match the MSG ones (and
1748 linux ones too) [Mt]
1749 * New function: gras_agent_spawn() to launch a new process on
1750 current host. Only working in simulation for now. [Mt]
1751 * New function: gras_os_hostport() returning a constant form (ie,
1752 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1755 * Make the backtrace of exceptions more human readable [Mt]
1756 * New module: xbt/str [Mt]
1757 a ton of string utility functions (split, join, printf to a newly
1758 allocated buffer, trim, etc)
1759 * New module: xbt/hash [Mt]
1760 SHA1 hashing algorithm (more to come if needed)
1761 * New module: xbt/synchro [Mt]
1762 synchronization tools (mutex and conditions) working the same way
1763 in simulation and in real life (mainly useful for GRAS, but not
1765 * New module: xbt/queue [Mt]
1766 classical producer/consumer synchronization scheme
1767 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1768 (using the classical functions will get serialized) [Mt]
1769 * Make dictionary internal table dynamic. No need to specify its size
1770 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1772 * Make sure the log channels are organized as a tree under windows
1773 (because of ANSI C compatibility issue, any channel were child of
1774 root directly) [Mt].
1777 * Cleaned many thing in surf and fixed a few bugs [AL].
1778 * Add a nice command line configuration mechanism to compose models [AL].
1779 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1780 the previous one (KCCFLN05). It relies on something that looks like
1781 a max-min sharing mechanism but cannot be written as such. A new solver
1782 was thus designed [AL].
1783 * Add a new solver to lmm. Based on Lagrange optimization and
1784 gradient-based descent, it enables to efficiently maximise systems s.a
1786 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1788 This solver enables to propose two new network models for TCP Reno and
1789 TCP Vegas based on Low's work. These models still need to be fully
1790 tested though [Pedro Velho].
1793 * Bug fix in SD_simulate. Now the time bound given as argument is
1795 * Use the new parallel task model (ptask_L07) as default.
1796 * Use the SURF command line configuration mechanism.
1797 * 0-size tasks (for synchronization) should now work.
1799 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1801 SimGrid (3.2) stable; urgency=high
1805 We still experience issues on this platform, but we believe that at
1806 least MSG is usable.
1808 GRAS API BREAKAGE (for simplification purpose, sorry):
1809 * the gras_msgtype_by_name is not used anymore. Instead of
1810 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1811 you can write (and must)
1812 gras_msg_send(toserver, "request", &request);
1813 - If you still want to pass a gras_msgtype_t to the function (to cache
1814 the type and avoid the lookup time), use the gras_msg_send_() variant.
1815 - Impacted functions:
1816 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1817 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1818 * The callbacks are now expected to return 0 when everything went well
1819 (just like the main() function)
1821 GRAS new features and improvements:
1822 * New module mecanism where user code can use per process globals [Mt]
1823 This is similar to gras_userdata_*() functions, but for libraries. It
1824 factorize some code developped over and over in the examples and AMOK.
1825 It has still to be documented and used (only amok/peermanagement is
1827 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1828 loss when they were small enough to fit into the buffer and sent quickly
1829 enough so that they can all get received in one shoot.
1830 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1831 if not found. Use the *_or_null() variant for the old semantic.
1832 * In gras_msg_handle, do not discard messages without callback.
1833 They are probably messages to be explicitly awaited later (ie, proofs of
1834 mis-synchronization in userland since they are sent before being awaited)
1835 No big deal usually.
1836 * gras_socket_meas_send/recv: semantic changed!
1837 The numerical arguments used to be (1) the total amount of data to send
1838 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1839 messages. This was need for the fool willing to send more than MAXINT
1840 bytes on quite fat pipes.
1843 * Do really rename the hostmanagement module to peermanagement. [Mt]
1844 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1845 but this is rather new and this was documented in the module
1846 documentation (poor excuses, I admit)
1847 * Bandwidth measurement semantic changed! This follows the changes to
1848 gras_socket_meas_send/recv explained above.
1851 * A sequential mode has been added to the workstations. When a workstation
1852 is in sequential mode, it can execute only one task, and the other tasks
1853 are waiting in a FIFO. [Christophe Thiery]
1856 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1857 model for SIMDAG. [Christophe Thiery]
1858 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1859 equal to 0, which caused some bad behaviors in
1860 saturated_constraint_set_update. I now use a threshold mechanism like in
1864 * When running manually src/testall, you select specific units [Mt]
1865 testall is the result of our cunit mecanism, and should replace all
1866 the scripty thingy around since bash don't run easily on billware.
1868 * A mallocator system has been added. [Christophe Thiery]
1869 Mallocators allow you to recycle your unused objects instead of freeing them
1870 and allocating new ones.
1872 Documentation update:
1873 * FAQ reworking + New FAQs:
1874 - "Valgrind spits tons of errors!" [Mt]
1875 - "How to repport bugs" [Mt]
1876 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1877 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1878 - Communication time measurement within MSG [AL]
1879 - I experience weird communication times when I change the latency [AL]
1880 * GRAS tutorial [Mt]
1882 - an introduction to the framework and to the used communication model
1883 - an initiatic tour introducing the most proheminent features:
1885 . Lesson 0: Installing GRAS
1886 . Lesson 1: Setting up your own project
1887 o Part 2: Message passing
1888 . Lesson 2: Exchanging simple messages
1889 . Lesson 3: Passing arguments to the processes (in SG)
1890 . Lesson 4: Attaching callbacks to messages
1891 . Lesson 5: Using globals in processes
1892 . Lesson 6: Logging informations properly
1893 . Lesson 7: Using internal timers
1894 . Lesson 8: Handling errors through exceptions
1895 . Lesson 9: Exchanging simple data
1896 . Lesson 10: Remote Procedure Calling (RPC)
1897 . Lesson 11: Explicitely waiting for messages
1898 . Recapping of message passing features in GRAS
1899 - A HOWTO section containing:
1900 o HOWTO design a GRAS application
1901 More are due, of course. They will come latter. In the meanwhile, you can
1902 check the examples which are still here.
1904 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1906 SimGrid (3.1) stable; urgency=high
1910 There was a stack corruption somewhere, visible only when optimizing
1911 with these versions. [Vince]
1914 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1915 module that enabled to play with Directed Acyclic Graphs. It is built
1916 directly on top of SURF and provides an API rather close to the old
1917 SG. Some old codes using SG are currently under rewrite to check that
1918 all needful functions are provided. [Christophe Thiery]
1921 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1922 extension of the classical CLM03 model that gracefully handles
1923 failures. This is now the default model for MSG and GRAS. It doesn't
1924 handle parallel tasks yet though. [AL]
1925 * Bug fix: Weights were not correctly set in the network part.
1926 WARNING: This may have resulted in incorrect results with simulations
1927 where there are more than one flow on a given link. [AL]
1930 * After a (long ?) discussion on simgrid-devel, we have decided that the
1931 convention we had on units was stupid. That is why it has been decided
1932 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1933 WARNING : This means that all previous platform files will not work as
1934 such with this version! A warning is issued to ask users to update
1936 A conversion script can be found in the contrib module of the CVS, under
1937 the name contrib/platform_generation/surfxml_update.pl [MQ]
1940 * Bug fix: Processes were started in reverse order, wrt deployment file.
1941 WARNING: if your code relies on this bug, please fix it. [AL]
1942 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1943 being executed on two different locations. [AL]
1944 * Bug fix: Failures are now better supported thanks to Derrick's tests
1945 (there was many failure situations I hadn't thought of and that weren't
1946 correctly handled). [AL]
1947 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1951 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1952 Exception occurring on server-side are propagated back to client (!).
1954 API CHANGE: the callback changed their prototype. Change:
1955 int my_handler(gras_socket_t expeditor, void *payload_data) {
1957 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1958 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1960 * New! function: gras_msg_handleall to deal with all messages arriving
1961 within a given period.
1962 * New! function: gras_socket_server_range to get a server socket in a
1963 range of port numbers (ease to avoid port number conflicts) [MQ]
1964 * New! gras processes display their backtrace when they get a SIGUSR1
1965 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1966 Sweet to debug RL processes [MQ]
1970 - Do not force experiment sizes to be expressed in kb, or it becomes
1971 impossible to measure the latency this way (needs one byte-long tests)
1972 WARNING: this changes the amok_bw_* function semantic. [MQ]
1973 - Implements the link saturation stuff. [MQ]
1974 * Peer management module:
1975 New! module factorizing code that we wrote over and over [MQ].
1978 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1979 - Test units are placed directly into the library code, they get extracted
1980 automatically and placed into the src/testall binary.
1981 - Convert most of the XBT tests to this system.
1982 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1983 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1984 live in the same C code. This required to change a little bit the API
1985 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1986 * New module: sparse graph structure with basic algorithms (this is work
1987 in progress and the API is not considered to be frozen yet). [AL]
1988 * Display more information on backtraces: source line & function names are
1989 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1990 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1991 * Reworked a little bit some #include statements to load only required
1992 headers. Some user code that relied on SimGrid to include stdlib or
1993 stdio may need to include it by themselves. [AL]
1994 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1996 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1997 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1998 but I guess that almost nobody used those parts. [MQ]
2000 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2002 SimGrid (3.0.1) stable; urgency=low
2005 * Unfortunately, I had missed 5 misnamed functions:
2006 xbt_fifo_item_t xbt_fifo_newitem(void);
2007 void xbt_fifo_freeitem(xbt_fifo_item_t);
2008 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2009 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2010 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2011 They're now deprecated. Please use their new versions:
2012 xbt_fifo_item_t xbt_fifo_new_item(void);
2013 void xbt_fifo_free_item(xbt_fifo_item_t);
2014 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2015 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2016 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2018 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2019 * Documentation: xbt_log module unmercifully reworked [MQ]
2020 * Bugfix: there was a problem with the ending of contexts with
2021 the pthread backend. It caused some weird deadlock or behavior
2022 depending on the pthread implementation. [AL]
2023 * Bugfix: get the exceptions raised in the simulator repport where
2024 and why they come from when they are not catched in time [AL, MQ]
2027 * Bugfix: Do repport the error when two non-connected hosts try to
2028 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2031 * Add additionnal checkings on communications. Assert that two
2032 communicating hosts are connected by a set of links... [AL]
2035 * Add additionnal checkings on channel values in communication [AL]
2036 * New: MSG_task_get_source to see on which host a task was generated [HC]
2037 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2038 the number of tasks waiting to be received on channel and sent
2040 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2041 waits for the first task coming from a given host.. [AL]
2043 GRAS new functionnalities: [MQ]
2044 * Enhance the parsing macro to allow the size of multidimentional objects
2045 to be given thru annotations.
2046 * New example (and documentation): Matrix Multiplication a la RPC
2047 (as when I was young!) and fix a bunch of bugs found on the way.
2049 GRAS performance improvements: [MQ]
2051 * Reduce the amount of cbps creation/destruction by making it static to
2052 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2055 * Change libdata to a set so that we can search for stuff by ID (and thus
2056 reduce the insane amount of dict lookups)
2059 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2060 RL communication on top of SG since it's so uneffective.
2061 It may also be used for inter-thread communication in RL, one day.
2062 * Use gras_datadesc_copy() to exchange messages on top of SG
2064 - improve message exchange performance on top of SG
2065 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2066 * Don't exchange on the network the size of the used part of buffer,
2067 instead, specify the possible buffer size to read().
2069 - reduces the amount of read/write calls (one pair per exchange)
2070 - reduces the amount of exchanged data (the size)
2071 - allows to retrieve all arrived data on receiver side, if we don't need
2072 it right now (subsequent read will peek the buffer)
2073 - allows the receiver to proceed with the begining of the stream before
2074 everything is arrived
2075 - make it possible to build an iov transport (using readv/writev)
2077 - take care of the data with non-stable storage (like stacked data),
2079 * If possible, TCP send uses vector I/O (when writev() is here)
2080 - Don't use it for receive since we send data sizes and data on the
2081 same stream, so we wouldn't be able to chain large amount of chunks
2082 before having to flush the stuff to read the size.
2083 * Rework the transport plugin mecanism to simplify it and reduce the
2084 amount of pointer dereferencement when searching for the right function
2087 * I guess that now, we do almost as few system calls as possible while
2088 doing as few data copy as possible.
2090 To improve it further, we could try to send all the sizes first and then
2091 all the data (to use iov on receiving size), but it's only a partial
2092 solution: when you have 2 dimensional data, the sizes of the second
2093 dimension is data of the first dimension, so you need 3 streams.
2095 I'm not sure the potential performance gains justify the coding burden.
2097 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2099 SimGrid (3.00) stable; urgency=high
2102 * New! Give the possibility to hijack the surf parser and thus bypass
2103 MSG_create_environment and MSG_launch_application. Have a look at
2104 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2106 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2108 SimGrid (2.96) unstable; urgency=low
2113 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2114 This deprecates the xbt_error_t mecanisms.
2115 It modifies (simplifies) all XBT and GRAS API.
2116 MSG API keeps unchanged (exceptions raised by XBT are catched from
2117 within MSG and masked with existing error handling facilities)
2120 * New! Add a FATPIPE model. [AL]
2121 * New! Add a parallel task model. [AL]
2122 * New! Add automatically a loopback interface (in the default
2123 network model) if none was precised.
2126 * Bugfix: MSG_process_resume now works with the current running process.
2128 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2129 * Modification of MSG_task_get_compute_duration. Once a task has been
2130 processed, the value returned by this function is now equal to 0. [AL]
2131 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2132 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2133 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2134 of a m_task. For now, MSG_task_cancel only works with computation
2136 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2137 of the processor (in Mflop/s) regardless of the current load on the
2139 * API Change: use proper naming convention for MSG_getClock and
2140 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2142 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2143 This function changes the priority of a computation task. This priority
2144 doesn't affect the transfer rate. A priority of 2 will make a task
2145 receive two times more cpu power than the other ones. This function
2146 has been added to suit the needs of Nguyen The Loc and hasn't been that
2147 much tested yet. So if it fails, please report it and send me your code.
2149 * API Change: removed all functions and types that were marked "deprecated"
2150 since many months. Renamed MSG_global_init_args to MSG_global_init.
2152 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2154 SimGrid (2.95) unstable; urgency=low
2157 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2158 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2159 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2162 * Bugfix: Allow absolute paths for platform description files [MQ]
2163 * Bugfix: do free the variables after use. Leads to drastic performance
2165 * Implement max_duration (ie, timeouts) on resources [AL]
2168 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2170 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2171 the needed support (timeouts) [AL]
2174 * Implement measurement sockets. You can now get the bandwidth between two
2175 hosts thanks to AMOK (see below). [MQ]
2176 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2177 dynar over the network (yeah) [MQ]
2178 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2179 * simple/static token-ring example. [Alexandre Colucci and MQ]
2180 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2181 in place before [MQ]
2184 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2185 type parsing. "short" and "long" modifiers were also ignored; other
2186 modifier (such as reference level) are still ignored. [MQ]
2187 * Embeed the buffer size within the buffer itself on SG. [MQ]
2188 That way, send() are atomic and cannot get intermixed anymore (at least
2189 the ones which are less than 100k; bigger messages still have the issue)
2190 * Array size pushed by the field, not by the field type (or each
2191 and every long int will push stuff to the cbps) [MQ]
2192 * use select() to sleep since it allows to portably sleep less than one
2195 GRAS (minor cleanups)
2196 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2199 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2200 * type category 'ignored' killed as it was never used and were difficult
2202 * whether a type can cycle or not is now a flag, leaving room for more
2203 flags (as "ignored", if we feel the need one day ;)
2204 * Rename raw sockets to measurement sockets since "raw" has another
2205 meaning in networking community.
2208 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2209 and offers features not belonging to GRAS but that most applications
2210 need. One day, it may be a set of plugins loadable at runtime.
2211 * New module: bandwidth
2212 bandwidth measurement between arbitrary nodes running this module. [MQ]
2214 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2216 SimGrid (2.94) unstable; urgency=low
2218 The first beta release of SimGrid 3 !
2222 * Update the main page and the FAQ. Adding references to gforge.
2225 * Add a gras_os_getpid function.
2228 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2229 * Extend the logs so that they also print PID, hostname, date, ... if
2231 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2232 and kill the old version which were in testsuite/
2233 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2237 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2238 for users. Instead, each programming environment (GRAS, MSG,...) use it
2239 when needed to provide such a feature to users.
2240 Don't shortcut the mecanism or you will also shortcut the virtualization
2241 you need on the simulator.
2245 * Cleanups in configury with regard to compile optimization/warning flags.
2246 Also add -fno-loop-optimize to any powerpc since it's the optimization
2247 killing gcc (< 3.4.0).
2248 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2249 needed by MSG examples complications
2250 * Borrow configury beautifications from PHP
2253 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2254 warning (thanks loris for stumbling into this one).
2255 * Bugfix: stop loading private headers (gras_config.h) from the public
2259 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2260 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2261 RL would imply a DNS resolver, which is *hard* to do in a portable way
2262 (and therefore delayed).
2263 * Implement a real timer mecanism and use it in timing macros. This allows
2264 to avoid rounding errors and get a 0.000005 sec precision in timing
2265 macros. While I was at it, various cleanups:
2266 - allow to declare more than one timed section per file (fix a stupid bug)
2267 - move some private declaration to the right place
2268 - merge conditional execution and timing macros into emulation module
2269 - document the module
2270 - make sure the module cleanups its mess on gras_exit
2271 * Documentation improvements:
2272 - (new) how to compile applications using GRAS
2273 - (new) emulation support (timing macros)
2275 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2277 SimGrid (2.93) unstable; urgency=low
2279 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2282 - Use Paje properly where used. Still to be sanitized properly.
2283 - Portability fix: Add an implementation of the contexts using pthread
2287 - Add xbt_procname(): returns the name of the current process.
2288 Use it to show the current process's name in all logging.
2290 - fix detection of older flex version and the reaction, since we do
2291 depend on modern ones (we use lex_destroy)
2292 - Better separation of SG and RL in the libs: remove all simulation code
2293 from libgras. As a result, this lib is now only 200k when stripped.
2294 Some of the xbt modules may also be duplicated (two sets and such) and
2295 should be cleaned/killed before SG3.
2296 - Insist on using xlC on AIX because of weird problems involving gcc there.
2297 - Cleanup the make remote stuff. This is now done by scripts
2298 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2299 mainly for our private use, but we're working on changing them to user
2302 - Bugfix: flush the socket on close only if there is some *output*.
2303 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2305 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2308 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2310 SimGrid (2.92) unstable; urgency=low
2312 Alpha 3 on the path to SimGrid 3
2316 - New! First try of benchmarking macros.
2317 - New! First try so that gras_stub_generator generate deployment and
2318 remote compilation helpers.
2320 - Bugfix: Initialization fix in msg_test.
2324 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2326 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2328 - API Change: gras_os_sleep to take the amount of seconds as a double.
2329 Accepting an int was error prone since it was the only location where
2330 seconds were coded as such. It leaded to damn rounding errors.
2331 - Bugfix: Hard to belive that timers ever worked before this.
2333 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2335 SimGrid (2.91) unstable; urgency=low
2337 Alpha 2 on the path to SimGrid 3
2341 - Bug fix in the lmm_solver.
2343 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2344 through the function MSG_paje_output.
2345 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2346 - It is possible to bound the rate of a communication in MSG with
2347 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2348 to write it in the changelog).
2349 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2354 - Various cleanups to the autotools stuff
2355 - Begin to move Gras examples to examples/gras/
2356 - Let make distcheck work again (yeah!)
2358 - documentation overhauled using doxygen.
2359 gtk-doc-tools is dead in SimGrid now.
2360 - Automatically extract all existing logging categories, and add the list
2361 to the documentation (long standing one, to say the less)
2363 - Cleanup the known architecture table. Reorder the entries to group what
2364 should be, and use a more consistent naming scheme.
2365 (some of the test dataset are still to be regenerated)
2366 - New! Allow library to register globals on each process just as userdata
2368 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2369 lookup time (for now).
2370 Use it in msg and trp.
2371 This cleans a lot the internals and helps enforcing privacy of the
2372 headers between the gras components.
2373 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2374 - Bugfix: gras_os_time was delirious in RL.
2375 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2377 - Reenable GRAS now that it works.
2379 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2381 SimGrid (2.90) unstable; urgency=low
2383 Alpha 1 on the path to SimGrid 3
2385 * It is a long time since the last release of SimGrid. I'm sorry about
2386 that but as I had told you, I was rewriting a lot of things. I apologize
2387 to those who had been reporting bugs to me and that I had not answered.
2388 If your bug is still in the new version, please tell me. Here is a
2389 summary of the main changes.
2391 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2392 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2393 portability, speed, and a lot more but you'll figure it out later.
2394 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2395 I wanted to release the new MSG as soon as possible and I have
2396 broken GRAS, which is the reason why, for now, only MSG is fully
2397 functional. A laconic description of these projects is available
2398 in the documentation.
2400 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2401 kernel optimized for our needs (called SURF but only the developers
2402 should use it). Hence, MSG is now roughly 30 times faster and I think
2403 that by rewriting a little bit MSG, I could event speed it up a little
2404 bit more. Beside the gain in speed, it is also much easier to encode a
2405 new platform model with SURF than it was with SG. More to come...
2407 * REVOLUTION 3: I have tried to change a little as possible the API of
2408 MSG but a few things really had to disappear. The main differences
2409 with the previous version are :
2410 1) no more m_links_t and the corresponding functions. Platforms are
2411 directly read from a XML description and cannot be hard-coded
2412 anymore. The same format is used for application deployment
2413 description. The new format is described in the documentation.
2414 Have a look in tools/platform_generation. There is a tiny script
2415 that converts from the old platform format to the new one. Concerning
2416 the application deployment format, parsing the old one is tricky.
2417 I think most of you should however be able to convert your files. If
2418 it is really an issue, I can write a C code that does the conversion.
2420 2) the toolbox tbx does not exist anymore. We now have a library
2421 with much more data-structures but without the hash-tables (we have
2422 dictionaries that are much faster).
2424 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2426 *****************************************************************************
2427 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2428 * don't feel like converting the oldies. (Mt) *
2429 *****************************************************************************
2432 Version 2.90: "the long awaited one"
2433 - Finished rewriting and debugging MSG. Rewrote the documentation.
2434 - disable GRAS for now since it needs to be ported to the newest SG
2437 - Finish the port to windows (using mingw32 for cross-compile)
2440 - Main loop and datastructures of SURF. A cpu resource object is
2441 functional. Surf can thus be used to create cpu's with variable
2442 performance on which you can execute some actions.
2444 2004-11-15 Martin Quinson
2445 - Port to ARM. Simply added the alignment and size descriptions. Should
2446 work, but the ARM machines are so slow that I didn't had the opportunity
2447 to 'make check' over there yet.
2449 2004-11-15 Arnaud Legrand
2450 - Trace manager now written. It uses a heap structure and is therefore
2451 expected to be efficient. It may however be speeded up (particularly
2452 when many events occur at the same date) by using red and black
2453 trees. One day maybe...
2454 - Max-min linear system solver written. It uses a sparse matrix
2455 structure taking advantage of its expected use. Most operations are
2456 O(1) and free/calloc are called as few as possible. The computation of
2457 the minimum could however be improved by using a red and black tree
2460 2004-11-03 Arnaud Legrand
2461 - Rename every gras_* function that was in xbt/ to its xbt_
2463 - Add a heap and a doubly-linked list to xbt
2464 - Added a dichotomy to the dictionaries. make check works as well before
2465 so I assume that the patch is correct. I do not know however if things
2466 run effectively faster than before now. :)
2468 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2469 SimGrid, and the version number is bumped to 2.x
2471 2004-10-29 Martin Quinson
2472 - Introduction of the remote errors.
2473 They are the result of a RMI/RPC on the remote machine.
2474 ErrCodes being scalar values, you can't get the host on which those
2475 errors did happen. Extending the error mechanism as in Gnome is possible.
2476 No idea yet whether it is a good idea.
2478 2004-10-28 Martin Quinson
2479 - Interface revolution: the Starred Structure Eradication.
2480 I used to do typedef struct {} toto_t; and then handle *toto_t.
2481 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2482 - ???_t is a valid type (builded with typedef)
2483 - s_toto_t is a structure (access to fields with .)
2484 - s_toto is a structure needing 'struct' keyword to be used
2485 - e_toto_t is an enum
2486 - toto_t is an 'object' (struct*)
2488 typedef struct s_toto {} s_toto_t, *toto_t;
2489 typedef enum {} e_toto_t;
2490 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2491 s_toto_t) is private.
2493 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2494 so that it can set the variable to NULL. It was so for dicts and sets,
2495 it changed for dynars.
2497 - Fix a bunch of memleaks in dict_remove
2498 - Fix a bug in sg/server_socket opening: it failed all the time.
2500 2004-10-07 Martin Quinson
2501 - Speed up dynar lookup operation a bit.
2503 gras_dynar_get is dead.
2505 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2506 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2507 (which gives you the address of the stored data).
2509 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2510 copy of the data using an affectation to do the job and not a memcpy.
2512 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2513 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2515 It does not really speedup the dynar test because they are
2516 setting elements all the time (and look them seldom). But the dict does
2517 far more lookup than setting.
2519 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2521 2004-10-05 Martin Quinson
2522 - Allow to (en/dis)able the cycle detection at run time.
2524 Whether we should check for cycle or not is now a property of each
2525 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2526 datadesc_cycle_unset allow to remove this property when previously set.
2528 Note that the cycle detection is off by default since it impacts the
2529 performance. Watch the data you feed GRAS with ;)
2531 This property is hereditary. Any element embedded in a structure having it
2532 set have it set for the time of this data exchange.
2534 You should set it both on sender and receiver side. If you don't set it on
2535 sender side, it will enter an endless loop. If you forget on receiver
2536 side, the cycles won't be recreated after communication.
2538 - Header reorganization.
2539 Kill gras_private.h, each submodule must load the headers it needs.
2541 2004-10-04 Martin Quinson
2542 - Interface revolution: do not try to survive to malloc failure.
2544 Now, gras_malloc and friends call gras_abort() on failure.
2545 As a conclusion, malloc_error is not a valid error anymore, and all
2546 functions for which it was the only gras_error_t return value are
2547 changed. They now return void, or there result directly.
2548 This simplify the API a lot.
2550 2004-09-29 Martin Quinson
2551 - Re-enable raw sockets.
2552 Created by gras_socket_{client,server}_ext;
2553 Used with gras_raw_{send,recv}
2556 It should allow to kill the last bits of gras first version soon.
2558 This is not completely satisfactory yet (duplicate code with
2559 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2562 - Simplify transport plugin (internal) interface by not passing any
2563 argument to _server and _client, but embedding them in the socket
2566 2004-09-28 Martin Quinson
2567 - Finish the port to AIX.
2568 autoconf was my problem (segfault within the malloc replacement
2569 function. No idea why)
2571 2004-09-16 Martin Quinson
2572 - Fix some size_t madness on 64bit architectures.
2574 2004-09-08 Martin Quinson
2575 - Reduce the number of system headers loaded, overload some more system
2576 calls (such as malloc to cast the result of the system one, and work
2578 - Fix and reintroduce the config support
2580 2004-09-07 Martin Quinson
2581 - Source code reorganization to allow Arnaud to surf all over there.
2582 - Allow to document the logging categories.
2583 - Remove all uppercase from logging categories and useless cleanup in names.
2585 2004-08-18 Martin Quinson
2586 Version 0.6.2 (protocol not changed; API changed)
2587 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2588 gras_error_t), and NULL when not found. Functions expecting a msgtype
2589 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2590 hopefully usefull message.
2591 - Portability to prehistoric sparcs again
2593 2004-08-17 Martin Quinson
2594 Version 0.6.1 (protocol not changed; ABI not changed)
2595 - prealloc some buffers to speed things up
2597 2004-08-11 Martin Quinson
2598 Version 0.6 (protocol not changed; ABI expended)
2599 - The parsing macro can deal with the references, provided that you add
2600 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2602 2004-08-09 Martin Quinson
2603 Version 0.5 (protocol not changed; ABI changed)
2604 - Allow to off turn the cycle detection code in data exchange at
2605 compilation time. It should be at run time, but I'm short of time (and
2606 the config stuff is still broken). That way, we keep dict out of the
2607 critical path, which is good because the performance is poor:
2608 - search not dichotomial yet
2609 - dynar give no way to access their content and memcpy everytime
2610 - In composed data description (struct, ref and so on), stop foolness of
2611 keeping the subtype's ID, but store the type itself. This keeps sets out
2612 of the critical path, which is good since they rely on dynar and
2613 dictionnaries. The only loose of that is that we cannot detect the
2614 redeclaration of a structure/union with another content (but I'm not sure
2615 the code detected well this error before anyway). We still can detect
2616 the redefinition discrepancy for the other types.
2617 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2618 it breaks the code because of type-punning used all over the place).
2619 This breaks on all non-gcc architectures (for now).
2621 All those changes (plus the buffer of last time) allow me to gain 2 order
2622 of magnitude on cruel tests consisting of 800000 array of integers on two
2623 level of a hierarchical structure (200 secondes -> 4 secondes)
2626 - the selector of reference must now return the type it points to, not
2627 the ID of this type.
2629 2004-08-06 Martin Quinson
2630 Version 0.4 (protocol changed; ABI not changed)
2631 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2632 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2633 - Implement a buffer transport, and use it by default (it relies on tcp in
2634 real life and on sg in simulation).
2635 That's a bit hackish since I had a new field to the structure to store
2636 its data without interfering with the subtype ones. Inheritance
2637 is tricky in C. And that's a kind of reverse inheritance with one class
2638 derivating two classes. Or maybe a game with java interfaces. Anyway,
2639 that's damn hard in C (at least).
2640 Moreover, I got tired while trying to ensure plugin separation and
2641 genericity in SG mode. MSG wants me to do weird things, so let's go for
2642 cruel hacks (temporarily of course ;).
2643 See comment in transport_private.h:71
2644 - do not include all the _interface headers in private but in the files
2645 which really need them (to cut the compilation time when they are
2648 2004-07-26 Martin Quinson
2649 Version 0.3 (protocol not changed; ABI changed)
2650 - Major overhault of the datadesc interface to simplify it:
2651 - shorted the function names:
2652 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2653 - add a trivial way to push/pop integers into the cbps without malloc.
2654 This allows to make really generic sub_type description, which simply
2655 pop their size of the stack.
2656 - add a function gras_datadesc_ref_pop_arr() which does what users want
2657 most of the time: Declare a dynamic array (which pops its size of the
2658 stack) and declare a reference to it. Poor name, but anyway.
2659 - kill the post-send callback, add a post-receive one
2661 2004-07-23 Martin Quinson
2662 Version 0.2 (protocol changed; ABI changed)
2663 - add some testing for cpbs in the test cases, and fix some more bugs.
2664 This invalidate again the little64 data file, since I cannot regenerate
2666 - remove an awfull optimization in the logging stuff, allowing me to:
2667 - understand it again
2668 - learn gcc how to check that the argument match the provided format
2669 - fix all errors revealed by gcc after that
2670 - internal keys of dict are not \0 terminated. Deal with it properly in
2671 loggings instead of segfaulting when the user want to see the logs :-/
2673 2004-07-22 Martin Quinson
2674 - Fix some stupid bug preventing cbps (callback postit) from working
2676 2004-07-21 Martin Quinson
2677 - Some documentation cleanups
2678 - remove the useless last argument of msgtype_declare
2679 - rename the Virtu functions to fit into the 'os' namespace
2680 - move headers src/include -> src/include/gras/ and stop fooling with
2681 gras -> . symbolic link
2682 - make distcheck is now successful
2684 2004-07-19 Martin Quinson
2686 - Build shared library also
2687 - Install html doc to the right location
2688 - stop removing maintainer files in make clean
2689 - build tests only on make check
2691 2004-07-13 Martin Quinson
2693 - No major issue in previous version => change versionning schema
2694 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2695 corresponding dataset.
2697 2004-07-11 Martin Quinson
2699 - Get it working with any kind of structure (we can compute the padding
2700 bytes remotely for all the architectures I have access to)
2701 - Implement the structure parsing macro (still not quite robust/complete)
2702 - Improvement to the remote testing toysuite
2704 2004-07-10 Martin Quinson
2705 [autoconf mechanism]
2706 - get ride of a bunch of deprecated macros
2707 - actually run the test for two-compliment, not only compile it :-/
2708 - test whether the structures get packed (and bail out if yes. Damn.
2709 Alignment is a serious matter)
2710 - test whether the structures get compacted (but respecting the alignment
2711 constraints of each types)
2712 - test whether the array fields of structures can straddle alignment boundaries
2714 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2715 (took me 2 hours to find that bug, looking at the wrong place)
2716 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2717 before sending/receiving and are used to compute the offsets of fields
2718 - Given that padding size depend even on compiler options, keep track of
2719 alignment and aligned_size only for the current architecture. Not a big
2720 deal since we send structure fields one after the other (seems
2722 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2723 program, let it work on linux/gcc/little32. portability todo.
2725 2004-07-08 Martin Quinson
2726 - import and improve remote compilation support from FAST
2727 - make sure make check works on half a dozen of machines out there
2729 2004-07-07 Martin Quinson
2730 Let's say it's version 0.0.3 ;)
2731 - Implement conversions (yuhu!)
2732 - Let it work on solaris (beside conversion, of course)
2733 - Stupid me, using rand() to generate the conversion datatests in not wise.
2735 2004-07-06 Martin Quinson
2736 - Let make dist work, since I'm gonna need it to compile on remote hosts
2737 - Let Tests/datadesc_usage write the architecture on which the file was
2738 generated as first byte.
2739 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2740 architecture descriptions.
2741 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2742 on those architectures.
2743 - Optimization: send/recv array of scalar in one shoot
2745 2004-07-05 Martin Quinson
2746 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2748 - Plug a whole bunch of memleaks
2749 - each process now have to call gras_{init,exit}. One day, their log
2750 settings will be separated
2751 - Continue the code factorisation between SG, RL and common in Transport.
2753 2004-07-04 Martin Quinson
2755 - Redistribution between SG and RL.
2756 We wanna have to accept in SG, so move accepted related parts of RL in
2757 the common part. (more precisely, the dynar of all known sockets is no
2758 more a static in transport.c, but part of the process_data)
2760 [gras_stub_generator]
2761 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2763 2004-07-03 Martin Quinson
2764 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2765 them all up in one shot)
2767 - Fix the ugly case of reference to dynamic array.
2768 - New (semi-public) function gras_datadesc_size to allow the messaging
2769 layer to malloc the needed space for the buffer.
2771 - gras_socket_close now expect the socket to close (and not its address to
2772 put NULL in it after it). This is because the socket passed to handlers
2773 is one of their argument (=> not writable).
2775 - propagate the interface cleanup from last week in datadesc, ie remove a
2776 superfluous level of indirection. User pass adress of variable
2777 containing data (both when sending and receiving), and not of a variable
2778 being a pointer to the data. Let's say that I like it better ;)
2779 The price for that is constructs like "int msg=*(int*)payload" in
2780 handlers, but it's a fine price, IMHO.
2782 - Let it work in RL (yuhu)
2784 2004-06-21 Martin Quinson
2786 - porting SG plugin and SG select to new standards (works almost).
2787 - plug memleaks and fix bugs around.
2790 - cleanup the prototype of data recv and force users to specify when they
2791 want to handle references to objects. Test case working even for cycles.
2792 - plug memleaks. Valgrind is perfectly ok with this.
2794 2004-06-12 Martin Quinson
2796 - cleanup the separation between plugin and main code in plugin creation
2798 2004-06-11 Martin Quinson
2800 - Reput hook for raw sockets, needed for BW experiments
2801 - kill a few lines of dead code
2802 [Data description] Interface cleanup
2803 - gras_datadesc_by_name returns the searched type or NULL.
2804 That way, no variable is needed to use a type desc once, which makes
2806 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2807 two parameters were strings (field name, type name), leading to
2809 [Dicos] Interface cleanup
2810 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2811 This is consistant with the dynar API.
2813 2004-04-21 Martin Quinson
2815 - Porting to new standards.
2817 - interface cleanup.
2818 There is no bag anymore, no need to take extra provision to mask the
2819 pointers behind "ID".
2820 Better splitup of functions between files create/exchange/convert.
2821 This is still a bit artificial since convert and receive are so
2822 interleaved, but anyway.
2824 - add a queued message list to procdata (the ones not matching criteria
2826 - factorize some more code between SG and RL wrt procdata
2828 - use gras_exit in example to track memleaks
2829 - get rid of gs_example now that GS is properly integrated into gras
2830 - update run_test to integrate the latest tests (datadesc)
2832 - rename WARNINGn macros to WARNn since it prooved error-prone
2834 2004-04-19 Martin Quinson
2836 - register init/exit functions within gras module mechanism
2837 - send/receive function.
2838 Convertion is not implemented, but short-cutted if not needed.
2839 struct/array elements are sent one by one (instead of block-wise), but
2840 nobody really cares (yet). Get a prototype before optimizing.
2841 - tests (using a file socket) for DD send/receive on:
2842 - base types: int, float
2843 - array: fixed size, string (ie ref to dynamic string)
2844 - structure: homogeneous, heterogeneous
2845 - chained list, graph with cycle
2846 Believe it or not, valgrind is not too unhappy with the results. The
2847 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2851 - Bugfix when using a filename explicitely (instead of '-')
2853 2004-04-09 Martin Quinson
2855 - factorize more code between RL and SG in socket creation
2856 - Complete the implementation and tests of:
2858 o file (only in RL, and mainly for debugging)
2860 I lost 3 days to design a portable address resolver, and then decided
2861 that the prototype mainly have to run on my box.
2862 Addressing portability too early may be like optimizing too early :-/
2864 - use gras_init in the Tests instead of the crappy parse_log_opt
2865 (the latter function is removed)
2866 [Conditional execution]
2867 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2868 [Code reorganisation]
2869 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2870 Build examples against the RL library, since there is no way to disable
2871 its creation for now.
2873 For information, the beginning of coding on GRAS was back in june
2874 2003. I guess that every line has been rewritten at least twice since