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 - Massive cleanups in the functions related to the energy
8 * Interface improvement:
9 - Rename Host.isAvail() to Host.isOn()
10 - Rename Process.currentProcess() to Process.getCurrentProcess()
14 - SIMIX_process_throw: raises an exception in a remote process
15 * Refactoring: Separate submodules
16 - libsmx: the public interface, as libc in a real system
17 - popping: the strange dance that converts a user request into a kernel handling
18 - smx_context_*: the virtualization mechanisms that embeed the user code
19 - smx_*: the handling of each simcalls
20 * Rename smx_action_t into smx_synchro_t, making explicit that these
21 things are used to synchronize processes with their environment.
22 For example, a communication is a sort of synchronization involving
23 the communicating processes (that may block until the exchange) and
24 the platform. The same can be said from computations, etc.
28 - 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.
29 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
30 - MPI_Type_set_name, MPI_Type_get_name
31 - MPI_*_c2f and MPI_*_f2c functions
32 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
33 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
34 - Activate a lot of new tests from the mpich 3 testsuite
36 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
37 - InfiniBand network model added : Based on the works of Jerome Vienne (http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf )
38 - When smpi/display_timing is set, also display global simulation time and application times
39 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
40 * Collective communications
41 - 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
42 - 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/).
43 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
45 - Replay now uses algorithms from wanted collective selector
46 - Replay can be used with SMP-aware algorithms
47 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
48 - 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.
50 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
51 - Correctly use loopback on fat-tree clusters
52 - Asynchronous small messages shouldn't trigger deadlocks anymore
55 - "Full" network optimization flag was broken since Surf++
56 - Better handling of precision flags in maxmin
57 - Fix bug causing sometimes "Impossible" errors
58 - Properly pass cluster properties to included hosts
59 * Improvement of the Energy plugin.
60 - Always update the consumption before returning that value
61 - New property: watt_off to denote the disipation when the host is off
64 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
65 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
66 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
69 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
71 SimGrid (3.11) stable; urgency=low
76 * Normalizing pointers addresses tool for better diff between logs
79 * Add cloud examples using new VMs
80 - examples/msg/cloud/two_tasks_vm.tesh
81 - examples/msg/cloud/simple_vm.tesh
82 - examples/java/cloud/cloud.tesh
83 - examples/java/cloud/migration/migration.tesh
84 * Add java surf examples:
85 - examples/java/surfPlugin/surf_plugin.tesh
86 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
87 - examples/java/surfCpuModel/surf_cpu_model.tesh
88 * Add SMPI+MSG example:
89 - examples/smpi/smpi_msg_masterslave/
94 - msg task destroy cancel test
95 - msg_host on/off test
96 * Move all tests in testsuite to teshsuite (adding tesh files)
97 * Restructure teshsuites
98 - one folder for each kind of test
99 * Restructure AddTests.cmake
101 - structure the order of tests (with sections)
104 * Add virtual machine
105 - creation of a VM on a PM
106 - migration of a VM from a PM to another PM
108 - MSG_process_join(msg_process_t process, double timeout)
109 - msg_bar_t MSG_barrier_init(unsigned int count)
110 - int MSG_barrier_wait(msg_bar_t barrier)
111 - void MSG_barrier_destroy(msg_bar_t barrier)
112 - msg_as_t MSG_environment_as_get_by_name(const char * name)
113 * New option "msg/debug_multiple_use" to help debugging when a task is used
116 - Improvements and finalization of MSG_storage, MSG_file APIs and their
118 - Increase code coverage in test suites
122 * Protect context stacks against stack overflow. The number of protected memory
123 pages allocated on the top of each stack (1 by default) can be configured
124 with the new command line option --cfg=contexts/guard_size:<value>.
125 * Simcalls are now generated by a python script that
126 - generates files included by SimGrid
127 - checks that all the functions exist, and proposes prototypes
129 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
130 host_set_data, host_get_data
132 - simcall_process_join(smx_process_t process, double timeout)
133 * Fix bug where sleeping processing could not be suspended.
136 * Translate surf models from C to C++
137 - Generic classes for all models: Model, Resource, Action
138 - A generic interface for each kind of model (CPU, Network, Storage
139 Workstation, WorkstationVM)
141 * Translate surf routings from C to C++
142 * Add callbacks using sigc++ or boost::signals2
143 - Add callback functions for resource creation/destruction
144 - Add callback functions for action state change
145 - Handle Energy as a plugin
146 * Replace swag by boost::intrusive
147 * Add new routing models for clusters. For documentation, see
148 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
149 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
150 parameters for cluster tag
151 - Fat trees, with topology="FAT_TREE" and
152 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
153 - see examples/platforms/torus_cluster.xml and
154 examples/platforms/fat_tree_cluster.xml
158 * Hostfiles support host:nb_processes construct to deploy several processes on
160 * Collective communication algorithms should not crash if used with
161 improper number of nodes and report the error.
162 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
163 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
164 MPI_Dims_create, MPI_Cart_sub are supported.
165 * New interface to use SMPI programmatically (still depends on MSG for
166 some parts, see examples/smpi/smpi_msg_masterslave) :
167 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
171 * Global variables privatization in MPI executables is now performed at runtime
172 with the option smpi/privatize_global_variables (default:no).
173 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
174 dynamic libraries loaded by the application are not privatized (static
175 linking with these libraries is advised in this case)
178 * Options defined in XML work correctly now.
181 * New cmake option, enable_lib_in_jar, to control whether native libraries are
182 copied into simgrid.jar or not (ON by default). Use this option if you want
183 to reduce the size of the installed simgrid.jar, *and* the native libraries
184 are kept installed elsewhere.
185 * Surf binding with SWIG (code generated in maintainer mode only):
186 - plugin to handle callbacks
187 - CPU model only for the moment
190 * Supernovae build mode is definitively removed. It was used to improve
191 inlining and inter-module optimizations. It is nowadays superseded by
192 link-time optimizations commonly available in compilers.
193 * Update ns-3 find lib. Bindings for ns-3 should work again now.
194 * Add boost dependency for surf++
195 * Add new macro for tests
196 - ADD_TESH(name <tesh_args>)
197 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
201 - xbt_bar_t XBT_barrier_init(unsigned int count)
202 - int XBT_barrier_wait(xbt_bar_t barrier)
203 - void XBT_barrier_destroy(xbt_bar_t barrier)
204 * Make the xbt_os_time module public
206 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
208 SimGrid (3.10) stable; urgency=low
210 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
213 * Reintegrate Java to the main archive as desynchronizing these
214 package is not acceptable anymore (Java is now considered stable)
215 * Add explicit synchronization facilities through semaphores
216 * Bug fix: Task.setDataSize() only changed the C world, not the value
217 cached in the Java world
220 * Dramatically change the way files are handled. API and internals changed, but
221 this part of MSG was not considered as production grade either.
222 * Add explicit synchronization facilities through semaphores
223 * Add a new function MSG_host_get_process_list()
224 * Preliminary DVFS support (see examples/msg/energy/ for details)
227 * SMPI is now included directly in the libsimgrid as the windows
228 linker doesn't force us on splitting it anymore.
229 * Improvements of the SMPI replay tool:
230 - Most of the collective communications are now rooted in the same process as
231 in the original application.
232 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
233 used until now). Multiple data types can now be used in a trace.
234 - The replay tool now supports traces produce either by TAU or a modified
236 - Bug Fix: the compute part of the reduce action is now taken into account.
237 - Gatherv collective is now supported
238 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
239 of smpirun outputs time independent traces for the current run. One file
240 is created per process. If too many processes are simulated, this behavior
241 can be changed to one file for all processes by using the
242 tracing/smpi/format/ti_one_file flag
243 * smpirun generates the host file if needed (with given host count and platform)
244 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
245 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
246 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
247 --cfg=smpi/coll_selector:(mpich/ompi)
248 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
250 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
251 happening inside a collective SMPI call.
252 * Fix the behavior of complex data types handling.
253 * Make MPI_Wtime another synchronization point to take computations into
255 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
256 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
257 * Add all missing Fortran bindings, SMPI should work with Fortran 90
258 (no privatization of global variables yet)
259 * Preliminary DVFS support (see examples/smpi/energy/ for details)
262 * Verification of liveness properties is now available for SMPI applications
263 (in addition to MSG applications)
264 * Bugged examples using SMPI in examples/smpi/mc/
265 * Add --cfg=model-check/visited option. Allows the verification of infinite
266 programs. Detection of loops in the execution thanks to the system state
267 comparison and reduction of the state space to explore. Can be combined with
268 DPOR for safety properties.
271 * Allow to change SimGrid configuration (see --help) within the code
272 thanks to SD_config() as it can be done in MSG.
273 * Add a new function SD_task_set_amount() upon user request.
276 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
277 * Remove rule based routing (no more PCRE dependency)
278 * Add a limiter_link option to cluster tag, to specify a maximum reachable
279 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
280 * Add a loopback_bw and loopback_lat options to cluster tag.
281 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
282 routing. Now peers are encapsulated in an AS and have their own private
283 router but this is transparent.
286 * Our own implementation of getline is renamed xbt_getline, and gets
287 used even if the OS provide a getline(). This should reduce the
288 configuration complexity by using the same code on all platforms.
289 * New type: xbt_cfg_elm_boolean.
290 * Allow to use yes/no for boolean configuration options in the command line.
291 * Allow to disable SimGrid cleanups at exit from command line option.
292 There are situations where one may want a simulation to end with an exit.
293 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
294 annoying when scripting around the simulator. Adding a
295 --cfg=clean_atexit:no allows to circumvent this issue.
298 * Lots of memory leaks were corrected in this release.
299 There is no known memory leaks anymore, in all of our 600+ tests.
300 * New command line option --version, to get SimGrid version information.
301 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
303 * Supernovae builds are deprecated, and expected to be removed in the next
306 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
308 SimGrid (3.9) stable; urgency=low
310 The Grasgory release: GRAS is really dead now.
312 * Complete overhaul of the internal host structures scheme.
315 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
316 was considered as experimental and badly maintained since then.
317 * Keeping it was thus a trap to our potential users, that could take
318 it instead of MSG or SMPI by mistake despite is pity state.
319 * GRAS seems to have very few users (if any), and no one volunteered
320 to maintain it further. It also induces a lot of XBT code (for
321 portability sake), that must be maintained too.
322 * For all these reasons, we killed GRAS. If someone wants to revive it
323 in the future, don't cry, our git history still remembers of GRAS.
326 * Major overhaul. Merge our documentation again as time proved that
327 splitting it was really not helping our users.
328 * Further improve the developer documentation to help newcomers
329 hacking on SimGrid itself. The user documentation (and in
330 particular, the beginner documentation) is still in a sorry state.
333 * Now works on Windows too!
334 * Much more extensive test suite, from MPICH
337 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
338 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
339 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
340 * Bug fix: let task be scheduled when the last dependency to be solved is
341 a control dependency.
342 * Remove SD_load_environment_script function.
343 Use the C sg_platf function if you want to declare a platform
347 * New function: MSG_process_get_number()
348 * Old function documented: MSG_config()
349 * Remove MSG_load_platform_script function
350 Use the C sg_platf function if you want to declare a platform
354 * Change the default value of the TCP_gamma constant (maximal size of TCP
355 congestion window) to a more realistic 4MiB value. If you notice changes in
356 your simulation results, you can fall back to the previous 20k tiny window
357 by adding --cfg=network/TCP_gamma:20000 on command line.
358 * (Hopefully) fix a bug wrt periodic availability and state traces
359 * Bug fix: use default values at start when first event in availability/state
360 trace is not at time 0.
363 * remove the "new_" part of function name sg_platf_new_trace_connect
364 (resulting in sg_platf_trace_connect), since it does not create
368 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
369 use case where it's really mandatory, and maintaining it was a pain in
371 * New: xbt_fifo_search(), search an item with a user-provided
372 comparison function instead of dumb pointer comparison.
375 * Fix the lua deployment:
376 Use `simgrid.init_application()` before deployment instead of
377 `simgrid.msg_register_application()` after.
380 * Transfer the tracing files into the corresponding modules.
382 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
384 SimGrid (3.8.1) stable; urgency=low
386 The "we are told that some people want to also *install* the simgrid
389 * Add missing file "tesh.1" to the archive.
391 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
393 SimGrid (3.8) stable; urgency=low
395 The Psssshiiiit release: SimGrid jumps into the Cloud.
398 * Add an experimental interface to manipulate VMs. They are mainly
399 process groups with very few intrinsic semantic, but they should
400 allow you to build the semantic you want easily.
401 * New function: MSG_host_set_property_value()
402 * New function: MSG_process_on_exit(). To clean memory in all cases.
403 * Bug fixes that made the host (and link) failures unusable.
404 * Add a way to auto-restart process when the host in which they are
405 executing comes back (ON_FAILURE="RESTART" on deployment file,
406 MSG_process_auto_restart_set).
407 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
408 please stop using the old ones, they are DEPRECATED.
410 * Deprecate functions MSG_global_init() / MSG_global_init_args()
411 Please use MSG_init() instead. (reducing the amount of entry
412 points in the library helps us).
413 * Make it impossible to link against the wrong version of the lib
414 * Deprecate MSG_clean(). No need to call it anymore.
415 * Function MSG_get_host_number() is not deprecated anymore.
418 * Split the doc into a user guide and a reference guide.
419 * Start a developper guide to help people hacking on SimGrid.
422 * Enable tracing by default. This modules rocks you should use it.
423 * Remove option custom_flags. Now use environment variables CFLAGS
425 * Use default cmake things to detect lua instead of home grown ones.
426 * New option "enable_mallocators" to disable mallocators, for debugging
427 purpose ("on" by default).
430 * Bug fixes around the resource failures: don't let the processes
431 survive the host they are running onto.
432 * Add an interface to auto-restart processes when the host in which they are
433 executing comes back.
434 * Ensures that SIMIX_clean is called automatically. It's not part of
435 the public interface anymore (bindings should be updated).
438 * Bug fix for when SD_Simulate is called with a positive value: be careful
439 when comparing doubles. Sometimes they are different for non significant
441 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
442 parallel task whose initial work is distributed among host according
443 to the Amdahl's law. Such tasks are created with a parameter alpha
444 that corresponds to the non-parallelizable part of the computation.
445 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
446 between two sets of workstations assuming a 1D block distribution (each
447 workstation owns a similar share of data) on both sides.
449 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
450 Data redistribution will be automatically scheduled once parent and child
451 are both scheduled. The filling of computation_amount and
452 communication_amount structures is now done seamlessly thanks to the chosen
454 * New function SD_workstation_dump to display various information
455 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
456 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
457 bandwidth on the route onto which the task is scheduled and the amount of
459 To divide the nominal bandwidth by 2, the rate then has to be :
460 rate = bandwidth/(2*amount)
461 * Compute tasks that have failed can now be rescheduled and executed again
462 (from their beginning)
463 * Increasing source code coverage (src/simdag is now covered at 95.8%
467 * Re-implement time-independent trace replay using SMPI (at the
468 smpi_smp_* level) instead of MSG. This should replace
469 examples/msg/actions/actions.c
470 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
471 hindexed and structs)
472 * Implement the exchange of non-contiguous data.
473 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
474 * Correct behavior of smpi/sender_gap and set its default value to 0
475 * Add option to asynchronously send small messages to allow better
476 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
477 specifies the size in bytes under which messages will be asynchronously sent.
478 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
479 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
482 * New C interface to define a platform: XML is now optional.
483 For more info, please check include/simgrid/platf.h.
484 * New interface to define random platforms from the C:
485 For more info, please check include/simgrid/platf_generator.h and
486 examples/msg/masterslave/masterslave_platfgen.c
487 * Export a sg_cmdline dynar containing all the arguments we got from
491 * Two new tracing options for adding comments to trace file so you
492 can track your experiments (see --help-tracing for details).
493 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
494 * Adding the SimGrid version that generated the trace file as a comment.
495 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
496 * Fix to avoid key clashes on Paje links
497 * Other minor fixes related to the Paje specification
500 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
501 and renamed to xbt_str_hash() and xbt_str_hash_ext().
502 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
503 * Greatly improve the robustness of mmalloc to user errors (such as
504 using an area after freeing it, or freeing it twice)
506 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
508 SimGrid (3.7.1) stable; urgency=low
511 * Restore the prototype of MSG_process_create_with_environment() to
512 the pre-3.7 situation by removing the kill_time argument.
513 * Add a MSG_process_set_kill_time() function instead.
516 * Fix weird behaviors when dealing with parallel tasks.
519 * Simgrid is now built as a dll.
520 * Simgrid-java now works on Windows.
521 * Simgrid-Java is now included into Windows package.
524 * First pre-build package for MacOSX.
527 * Fix compilation when using MSG_USE_DEPRECATED.
528 * Fix some compilation issues on Macs and Windows.
529 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
530 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
532 We discovered that the Lua console is broken, but we are missing the
533 manpower to fix it right now. The problem existed in 3.7 too, so we
534 are not blocking the release for that. Sorry if you depended on this
535 feature, any help would be really welcome.
537 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
539 SimGrid (3.7) stable; urgency=low
541 The "spring cleanups (before the next Big Project kicks in)" release.
544 * We can specify the SMPI latency/bandwidth factor with command line
545 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
546 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
547 You can also use the "config tag" from platform file by adding this line
548 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
549 (see "example/platforms/tag_config.xml" to use "config tag").
550 Note that the command line supersedes the platform file configuration.
551 * Change the correction factors used in LMM model, according to
552 the latest experiments described in INRIA RR-7821.
553 Accuracy should be improved this way.
554 * Use the partial invalidation optimization by default for the
555 network too. Should produce the exact same results, only faster.
556 * Major cleanup in surf to merge models and split some optimization
557 mechanisms from the core of the models. As a result you can now
558 specify which model to use (e.g., --cfg=network/model:LV08
559 --cfg=cpu/model:Cas01) and which optimization mode to use
560 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
561 Incompatible combinations should err at initialization. See
562 --help-models for the list of all models and optimization modes.
563 * The CLM03 workstation model was dropped for simplicity because it
564 used the deprecated CM02 network model. Use default instead.
565 * Rename the TCP_gamma configuration option to network/TCP_gamma
566 * Rename the coordinates configuration option to
567 network/coordinates, and document it
568 * Use now crosstraffic keyword instead of the terribly misleading
569 fullduplex keyword. It is activated by default now in the current
570 default model, use --cfg=network/crosstraffic:0 to turn it off.
571 * Ongoing refactoring the model parsing to make XML files optional
572 See include/simgrid/platf.h for details (still to be completed)
575 * Major overhaul of the documentation. Almost instructive now :/
576 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
577 functions and friends. This interface was considered as
578 deprecated since over 2 years, it's time to inform our users that it is.
579 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
580 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
581 use these (crappy) functions in your code.
582 These functions will be removed soon. Stop using them now.
583 * Deprecate MSG_get_host_{table,number}
584 Implement MSG_hosts_as_dynar() instead.
585 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
586 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
589 * Stabilize the parallel execution mode of user contexts
590 * Introduce configuration variables to control parallel execution:
591 - contexts/synchro: Synchronization mode to use when running
592 contexts in parallel (either futex, posix or busy_wait)
593 - contexts/parallel_threshold: Minimal number of user contexts
594 that must be part of a scheduling round to switch to parallel
595 execution mode (raw contexts only)
596 * Fix bugs that prevented to use suspend/resume along with
597 synchronization structures.
598 * Fix bugs in process termination that lead to invalid memory access
599 in very specific conditions.
602 * Introduce a parallel mode for the models (controlled by surf/nthreads
603 configuration item). In our tests, running the models in parallel
604 never lead to any speedups because they are so fast that the gain
605 of computing each model in parallel does not amortizes the
606 synchronization costs, even when ultra fast futexes are used.
607 This is released anyway because YMMV.
610 * Performance boost by using a swag internally to compute the set of
611 tasks that are finished and should constitute the return value of
615 * Enable it by default now that it is considered rather stable.
618 * Documentation of the tracing functions.
619 * Performance gains when tracing categorized/uncategorized resource
620 utilization by avoiding calls to get route when updating resource
621 variables. LMM constraints are being used instead.
622 * API changed to set task categories. Use MSG_task_set_category instead
623 of TRACE_msg_set_task_category, and SD_task_set_category instead
624 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
625 * Bugfix for graphicator, routes not correctly obtained, memory leaks
626 * Examples for link user variables added (see at examples/msg/tracing/)
627 * Deprecated function TRACE_msg_set_process_category completely removed
628 * Trace header updated according to the latest Paje file format
629 * Tracing network lazy updates, no longer obligate users to use full updates
630 * --cfg=tracing/platform:1 also registers power/bandwidth variables
631 * Experimental: let user code declare/set/push/pop application states for hosts
632 * API changed to allow the manual creation of graph configuration files
633 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
636 * Improve the API of Lua MSG bindings, using the Lua spirit.
637 * Each simulated process now lives in its own Lua world (globals are
638 automatically duplicated). It helps writing simulators. Will allow
639 to run Splay programs within SimGrid in the future.
640 * Add a Chord example in Lua, equivalent to the MSG one.
643 * Start the implementation of a solution to express temporal
644 properties, not only local assertions. This is still an
645 experimental work in progress, stay clear from it to be safe.
649 - Add new runtime parameters --help-logs and --help-log-categories
650 to display informations about supported logging parameters and
652 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
653 don't exists anymore.
654 * Mallocators: allow value NULL for the reset function.
656 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
657 create homogeneous dictionaries, where all the elements share the
658 same free function. Non homogeneous dictionaries will be
659 deprecated in the next release.
660 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
661 - Multi-level dictionaries are deprecated.
663 - new function xbt_dynar_search_or_negative() that is useful when
664 you have less than 2 million elements in your dynar and don't
665 want of the extra complexity of catching exceptions when the
666 element is not found.
668 - Make xbt_os_thread module (for thread portability) public.
669 Documentation is still to come, sorry.
671 - Cleanups and simplifications to make it maintainable again.
672 - Exotic features (such as memalign and valloc) were removed.
673 - The metadata were extended and improved so that the
674 model-checker becomes able to explore and inspect the heaps.
675 - This may induce a performance drop when enable_model-checking is
676 ON in cmake (even if it's not used in the simulation), but it is
677 necessary at this point to get MC working.
679 Turn model-checking OFF if simulation performance matters to you.
680 Not enabling it at runtime is not enough, disable it in cmake.
682 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
684 SimGrid (3.6.2) stable; urgency=low
686 The "Not coding new stuff allows to polish old things" release.
689 * New bindings to the NS3 packet level simulator (experimental)
690 * Use the raw (efficient) execution contexts instead of the sysv
691 (portable) ones when possible.
692 * libpcre is now mandatory in any cases since not using it led to
693 severe performance loss and possibly other issues
694 * Update the XML platforms:
695 - G5K: include the latest machine in Nancy
696 - GridPP and LCG: new platforms
697 * Documentation was partially updated, at least (more to come)
699 Bug fixes, cosmetics and small improvements
700 * Free terminated processes before the end of the simulation to avoid
701 exhausting the memory of users having very dynamic amount of
703 * Bug fix and cosmetics about canceling non-running tasks
704 * Bug fix about the dot loader's issues when using libcgraph
707 * Create an installer for windows with nsis (amd64 and win32)
708 - Add an hello world project to illustrate simgrid project creation.
709 - Embed libpcre into the Simgrid installer to avoid
710 its compilation burden
711 * The raw execution contexts should work on Apple now
712 * Port to Windows 64 bits
713 - Sysv contexts now have an implementation for this arch
714 - GRAS communication features now support this arch
715 * Drop support for borland compiler on windows
716 - this code was not maintained, and we kinda depend on gcc nowadays
717 * Fix portability issues on kfreebsd/gnu: build error about semaphores
718 * Fix portability issue on unstable ubuntu: linker became picky on
721 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
723 SimGrid (3.6.1) stable; urgency=low
725 The "Oops, we broke Macs too" release
728 * Fixed contexts detection so that raw ones are used when possible
729 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
730 leads to a strange error, with user code segfaulting sometimes when
731 the generated code is not perfectly aligned (which is not
732 controllable from the user side, depends on the amount of code)
735 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
737 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
739 SimGrid (3.6) stable; urgency=medium
741 The Summer Release, also known as the "OMG! They Killed Kenny!" version
744 * Bindings now constitute their own package, separated from the main one.
745 Rationale: reduce our maintenance nightmare by reducing the module coupling
746 They will soon be released on their own on gforge.
748 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
749 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
751 GRAS: It is not considered as stable anymore, but experimental. Sorry.
752 * It's not quite deprecated for now because we have no replacement,
753 but it may soon become the case.
756 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
757 * Fortran: New user-level cache variable to store the rank of the running
758 process. This improves performance by an order of magnitude.
759 * C: New coccinelle script to automatically locate and modify global and
760 local static variables.
761 * Improved SMPI network model with a sender-side gap to account for multiple
765 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
766 only say if a communication is finished, no matter whether it succeeded or
767 failed. You can call MSG_comm_get_status() to know the status of a finished
769 * New function MSG_task_dsend() to send a task and detach it. When a
770 communication is detached, you are never notified of its success or failure
771 and the memory is released automatically once it is finished. This function
772 is useful when you don't care about the end nor the success of a
774 * Change the prototypes of action replay. Sorry for inconvenience,
775 but this is really more efficient this way (and to adapt your code,
776 you just have to fix the initialization, that shouldn't be too long)
777 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
778 ever managed to do anything useful with it.
779 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
780 takes a dynar of communications. It returns immediately and gives the
781 index of a finished communication (if any).
782 * New example: a basic implementation of the Chord P2P algorithm.
785 * New model for multi-core CPUs. You can now use the core attribute to
786 precise the number of cores of a host. This is a basic model. Every
787 process running on the host receives at most the power provided in
788 the DTD (throughput<=power). Total throughput of process cannot exceed
790 * New peer tag. This peer tag creates a tiny AS comprising a host and a
791 router linked by an up-link and a down-link (possibly asymmetrical).
792 This kind of pattern allows to easily build last-mile model style platforms.
793 Aggregating such patterns in a rule-based AS is thus the technique of
794 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
795 * New model for Vivaldi routing. We transformed the Vivaldi network model
796 into a Vivaldi routing model (based on the rule-based model). This allows to
797 combine Vivaldi based latencies with last-mile platforms.
800 * Added a check for NaN of IEEE754 infinite in the double entries of
802 * Introduce a new context factory "raw", highly inspirated from the
803 ucontext factory, but using manually crafted functions in assembly to
804 do the work in an efficient manner.
805 * Allow to change the used context factory at run time, not only at
806 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
807 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
808 contexts at runtime (only with raw contexts or ucontexts).
809 * Completely rewrote this module to allow parallel execution of user
810 processes. Use --cfg=contexts/nthreads:N to execute user processes
811 with N parallel threads (the default is 1, meaning no parallelism).
812 * Allow to decide dynamically between sequential and parallel modes.
813 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
814 processes in parallel only when their number is greater than or equal to P
816 * Added a check for NaN of IEEE754 infinite in the double entries of
820 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
821 won't output the state of processes when interrupted with Ctrl-C
822 * Add a new function xbt_dynar_to_array that transforms a dynar into a
823 NULL-terminated array. This may solve backward compatibility issues
824 due to the change to return type of SD_simulate. See also:
825 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
826 * Add new macros with variable number of arguments.
827 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
828 - in xbt/asserts.h: xbt_assert
829 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
830 - in xbt/ex.h: THROWF and RETHROWF.
831 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
833 * Change xbt_die() to accept a format string with arguments, just like printf.
834 * New data structure: xbt_lib_t, like a dict but more general and with better
838 * New configuration options
839 Options triva/categorized and triva/uncategorized can be used to generate
840 graph configuration files for Triva visualization tool.
841 * Configuration option tracing/platform is renamed to tracing/categorized
842 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
843 * New TRACE_user_link_variable interface:
844 User provides the name of the link and the tracing variable to attach to it
845 * the declaration of tracing categories must be done after the environment
847 * simpler tracing interface, just one way to declare categories
848 TRACE_category or TRACE_category_with_color, it is up to you
849 * links in the trace file are again identified by their names
850 * trace contains the full platform hierarchy exactly as declared using the ASes
851 * Options tracing/msg/[task|process]:1 groups the process by hosts
852 for both cases, tasks and processes must have names that are unique during the simulation
853 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
854 * The experimental option tracing/msg/volume is deprecated
855 its functionality may be reincorporated if needed
857 The tracing generates a trace file with unordered timestamped events,
858 because of the way the core simulator (surf) works. A script available
859 at the tools directory (fix-paje-trace.sh) can be used to put the events
860 in order. We have changed the tracing so it can generate ordered timestamped
861 events in the final trace, but depending on the simulator (and how much time
862 is simulated) that can lead to a huge memory utilization. It is deactivated
863 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
866 * Define a SIMGRID_VERSION macro in simgrid_config.h.
867 - We are trying hard to keep the API stable, but it may happen that
868 some things change (we're a research project after all, not a
869 nuclear plant operating system). If such things should happen, you
870 could rely on that macro to adapt.
871 - current value: 30600 for 3.06.00, aka 3.6
872 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
873 a number that can be compared with SIMGRID_VERSION.
874 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
875 assertions and verbose/debug logging events are disabled at compile time.
877 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
879 SimGrid (3.5) stable; urgency=medium
882 * New feature to any SimGrid-based simulator: Model-Checking
883 Check SIN#1 for more details.
886 * New Model SMPI (three-interval linear regression for correction factors)
887 See RR-7426, available at http://hal.inria.fr/inria-00527150
888 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
889 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
890 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
891 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
892 * See also src/smpi/README
897 - Tracing API changes: TRACE_start and TRACE_end should not be called
898 by user-code. They are automatically called by simulators created
899 with SimDAG, MSG and SMPI if the toolkit is compiled with
900 tracing_enabled=ON. Categories declaration and utilization remain the
901 same for MSG and SimDag.
902 - A function was added to the tracing API to declare categories with
904 - TRACE_category_with_color (char *category, char *color)
905 where color must be in the following format
906 "%f %f %f", red, green, blue
907 and red, green, blue are float values in the interval [0, 1]
908 - User can specify NULL as color parameter, or continue calling
910 On that case, the tracing system will define random colors
911 - The following command-line options are supported:
912 --cfg=tracing/filename:msg.trace
913 --cfg=tracing:1 (activate tracing, needed to use others)
914 --cfg=tracing/platform:1 (categorized resource use)
915 --cfg=tracing/uncategorized:1 (uncategorized resource use)
916 --cfg=tracing/msg/task:1 (task creation)
917 --cfg=tracing/msg/process:1 (process creation, migration)
918 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
919 --cfg=tracing/smpi:1 (SMPI interface tracing)
920 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
921 - examples of examples/msg/tracing updated
923 - DAXLoader and DOTLoader functions can generate tasks with categories
924 - A new function to attribute a category to SD tasks:
925 TRACE_sd_set_task_category (SD_task_t task, char *category)
926 * Tracing the MPI interface implemented by SMPI
927 - Collective operations are traced with states
928 - Point-to-Point operations are traced with states/links
929 - Tracing activated by a parameter "-trace filename" passed
930 to smpirun during execution (considering that simgrid
931 is compiled with tracing enabled)
932 - To run the simulation with gdb, the simulator
933 accepts --cfg=tracing/smpi:1 to trace SMPI
934 - tesh files to check if smpi tracing is ok
935 - See examples/smpi/NAS/DT-trace
936 * GTNetS tracing re-worked
937 - adaptation to the tracing system of GTNets to cope
938 with modifications regarding the fullduplex mode
939 - new tesh files to check if gtnets tracing is ok
942 * Asynchronous communications through the functions:
943 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
944 * New function: MSG_load_platform_script()
945 to make possible using a lua script instead of XML files to set up platforms
946 * New function: MSG_set_function
947 to associate functions to processes, used when bypassing the parser
948 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
950 Platforms: Add some more examples in examples/platforms
951 * Grid'5000: see www.grid5000.fr
952 * *_30000_hosts.xml: various huge files [mainly scalability testing]
955 * Change the XML format. This is a very important modification. SimGrid
956 3.5 introduces a new hierarchical format based on the notion of
957 Autonomous Systems. Compatibility with old format is ensured
958 through the perl script provided in the install bin directory
959 bin/simgrid_update_xml.
960 It is now possible to build platforms with specific routing
961 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
962 connect several platforms together. We will try to provide soon
963 set of realistic platforms exploiting these properties (have a
964 look at examples/platforms/ for the moment).
965 * Take the opportunity of the XML format change to be a good XML citizen:
966 rename link:ctn to link_ctn and similar changes (also dealt with by
968 * Add a new routing scheme (rule-based) using regular expressions. It
969 enables to have an extremely low memory footprint when the
970 underlying routing is simple and can be compactly described. You
971 need to have libpcre4-dev (perl regular expressions) installed if
972 you want to use this routing scheme.
973 * Revive the cluster TAG and allow to easily and efficiently (both in
974 term of memory and speed) connect clusters together. Have a look
975 at teshsuite/simdag/platforms/ to see how this can be done. With
976 this tag, you can create clusters with thousands of tasks at no
977 cost (have a look at examples/platforms/).
978 Note: clusters are implemented as ASes, so there is no need for an
979 enclosing AS tag if you have only one cluster in your platform.
980 * Add new generic functions in the public interface that allows the user
981 to call SURF 'create_resource' methods from your code (same
982 functionality as the XML bypass mechanism but with a much lighter
984 * Add a new model (enabled through command line --cfg=network/model:SMPI)
985 that uses a piecewise linear approximation to produce better
986 results when exchanging small messages.
987 * Add a new parameter to handle correctly full duplex link and account
988 for interferences between uplink and downlink communications
989 (activate with --cfg=fullduplex:1).
992 * Rename the SD_READY (all dependencies are satisfied and task is
993 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
994 dependencies are satisfied) state.
995 This prevents a confusion between the notion of "ready to schedule"
996 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
997 simulated" (SD_RUNNABLE) used by the simulation kernel.
998 * Change the way a task is considered as ready. Instead of removing
999 dependencies when a task is done, a counter is decreased. This way,
1000 it is always possible to reach ancestors thanks to the
1001 SD_taks_get_parents function (even after the end of the simulation.)
1002 * Change the return type of SD_Simulate from (SD_task_t*) into
1003 xbt_dynar_t. This function was in handling a dynar internally and
1004 converted it into a NULL terminated array for historical reasons.
1005 * New function SD_dotload(char*) to load a DAG described in dot
1006 format. This loader and the corresponding examples require the
1007 installation of the graphviz library.
1008 * Fix a bug in the management of tasks of size 0 in the surf network
1009 models. This problem was only visible with SIMDAG and you should
1010 thus disregard results produced with earlier versions if you
1011 relied on this feature (some tasks were blocked because of this).
1012 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1013 even though your applications were doing only point-to-point
1014 communications and sequential computations. Now you can really use any
1015 model you want (of course, if you create real parallel tasks, which are
1016 not implemented in most models beside ptaskL07, this will abort).
1017 * Add an example that schedules a DAX on an heterogeneous platform
1018 using a Min-Min strategy.
1019 * New function SD_workstation_get_current_task() that returns the kind
1020 of task currently running on a workstation in the sequential access
1022 * Raise some warnings when unexecuted tasks remains at the end of the
1023 simulation. This is usually caused by cycles in the DAG.
1026 * New function: SIMIX_process_set_function() called by MSG_set_function
1027 * Change the underlying waiting queue in semaphores so that a process
1028 can wait on several of them simultaneously (as in waitany).
1029 * Fix the way to handle tokens in semaphores so that all access patterns
1030 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1031 * kill the dirty pimple SIMIX_message_sizes_output()
1032 Please use (proper) visualization instead
1035 * New data container: setset (set of sets of elements)
1036 * New module: mmalloc (mapped malloc, allowing to have several
1037 independent segments of malloc)
1038 * New function: xbt_dict_cursor_set_data()
1039 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1040 * New function: xbt_dynar_is_empty()
1041 * New function: xbt_fifo_get_last_item()
1042 * Fix xbt_dynar_shrink(): use the right element size.
1043 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1044 uninitialized areas during expand.
1045 * Fix semaphores: previous implementation was severely broken.
1046 * Use library init/fini functions for our initialization.
1047 - you can use logs and other feature as soon as you want in your
1048 code (even before the xbt_init / MSG_init)
1049 - xbt_exit is now a no-op and produce a warning when used.
1052 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1053 benefit from the latest improvement to the simulation kernel.
1054 * Kill measurement sockets for now. If you rely on them, sorry. This
1055 release is not for you. This feature will be reintroduced in the
1056 future, but we cannot delay the release any further.
1057 * New function: gras_msgtype_get_name().
1058 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1059 * Fix (at last) the pmm example: it should not randomly fail anymore.
1061 Build chain: bug fixes and overall polishing
1062 * Cmake is now stable enough. Hence, we killed the autotools.
1063 * Port to windows ( TM :)
1064 * Fix the 'make install' target.
1065 No need to use 'make install-simgrid' anymore
1066 * Introduce a 'make dist' target compiling a *source* archive
1067 'make package' compiles a binary archive
1068 * Compile java files only on need
1069 * Add --cd and --setenv command line options to tesh
1070 * Out of source builds are not fully supported yet, but we are close
1071 * Enable supernovae and optimization flags by default for our users
1074 * Add layer to set up environment directly from lua, without XML.
1075 * The effect of gras_stub_generator can be achieved through
1076 lua too (check examples/gras/console/ping_generator.lua)
1078 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1080 SimGrid (3.4.1) stable; urgency=low
1082 The "Polishing easter eggs is probably a good idea" release.
1083 This is a bug fixes release only.
1087 * Fix a bug preventing the tasks from begin garbage collected.
1090 * Fix a bug occuring when a host involved in a communication fails.
1091 This was not detected properly by the other peer involved in the
1092 communication. Now, it's reported as a network error.
1095 * Warn the user about loop dependencies in data flow of DAX files
1096 * Obey the control-flow dependencies of DAX files
1099 * Add option "enable_smpi" allowing to not compile SMPI.
1100 Probably useful for the (Mac) users experiencing a build error here
1101 * Improve the detection of lua5.1 and ruby1.8
1103 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1105 SimGrid (3.4) stable; urgency=low
1107 The "Easter in Cargese" release. Also known as (major changes):
1109 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1110 ~> bindings were greatly improved
1111 ~> new tracing infrastructure for better visualization introduced
1113 * the "Welcome to configury modernity" release.
1114 ~> we switched from autotools to cmake, and improved our cdash
1117 A more detailled list of changes follow (full detail in svn log).
1119 Java Bindings: Various Cleanups
1120 * (install java-gcj-compat-dev on debian-like to use them)
1121 * Remove put/get: no need to export deprecated interface in Java
1122 Use send/receive instead.
1123 * Cleanup the examples and add a README per directory
1124 * Remove example autoDestination (that's the only way to go now)
1125 * Remove example explicitDestination (was a plain copy of basic)
1126 * Make JniException a runtime exception, so that there is no need to
1127 declare the fact that you may encounter such a beast. I guess that
1128 nobody will ever want to survive such error.
1129 * Create specific errors for each MSG case of failure:
1130 host failure, transfer failure, timeout, task cancelled
1131 * Cleanup the exceptions that may get thrown by each function
1132 * Other internal cleanups in Java bindings. Performance still bad :/
1133 Ruby and Lua Bindings: create them
1134 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1135 * That's new and great, you should try them out.
1136 Same functionalities than Java bindings, only even less polished
1138 * Kill the useless "rate" argument of SD_task_get_execution_time()
1139 Everyone used to provide -1 as a value, it was not used, and the
1140 semantic of a possible use wasn't even clear.
1141 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1142 either as comm costs or compute costs to mean that there is no
1143 such thing for that specific task.
1144 * Add a SD_task_set_name() function
1145 * Fix SD_task_unschedule() on typed tasks
1146 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1147 * In DAX loader, accept useless 'level' attributes to <job> since
1148 LIGO DAGs have them (seem to be to ease graphical representation).
1150 * Add an example masterslave_mailbox.c using send/receive and not
1151 the deprecated put/get interface.
1152 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1153 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1155 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1156 (declare MSG_USE_DEPRECATED to still have the old name)
1157 * Add a MSG_task_set_data() function
1158 * About trace replay (see examples/msg/actions):
1160 - Allow to work with splitted trace files for each process
1161 Give the specific trace file as argument of each process,
1162 and call MSG_action_trace_run(NULL)
1163 You can still have one merged file for all processes.
1164 - Fix implementation of collective operations
1165 * Allow task_execute() on 0-sized tasks (closes #10063)
1167 * This is the first release of SimGrid where SMPI is not considered
1168 beta anymore (even if some corners should still be improved)
1169 * Port over the new SIMIX_network submodule (internal refactoring)
1170 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1171 * Implement more missing elements of the standard:
1173 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1175 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1176 MPI_Scan() MPI_Get_processor_name()
1177 - Added implementation of missing case for Alltoall (warning: it's
1178 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1179 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1180 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1181 SMPI_MPI_Allgatherv()
1182 * Bug fixes include:
1183 - MPI_Waitsome() was broken
1184 - Allow relative includes in smpicc
1185 - Command line cfg argument 'reference_speed' was ignored...
1186 - Some functions did not properly lead to auto-benching of user code
1187 - smpicc passes -O2 by default (just like openmpi one)
1189 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1190 * Bug fixes about timeouts during communications
1191 * add SIMIX_message_sizes_output() as a pimple to write to file the
1192 amount of messages per size. Use gnuplot to get histogram.
1193 Pimple because that's the only user-visible function of simix,
1194 defined directly in xbt.h (irk, sorry)
1196 - Add a SIMIX_sem_get_capacity() function
1197 - Fix interactions with processe resume/suspende
1198 - release_forever() was stupidly broken
1199 - Fix SIMIX_display_process_status() for processes in a semaphore
1200 - Make SIMIX_sem_block_onto() user-visible
1201 * Refactoring context stuff:
1202 - Use pseudo-OOP for better modularity
1203 - reimplement SIMIX_process_kill() without process_schedule() so
1204 that the latter can take as invariant that it is called from
1206 - Merge context_start into context_new for sake of simplicity
1208 * Add a Vivaldi network model, coded live during SUD'10 ;)
1209 * Rename configuration variables to start a hierarchy:
1210 o cpu_model -> cpu/model
1211 o network_model -> network/model
1212 o workstation_model -> workstation/model
1213 * New configuration variables:
1214 o network/bandwidth_factor: correction to bandwith
1215 o network/latency_factor: correction to latency
1216 o netwotk/weight_S: correction to the weight of competing streams
1217 * Add a long description to the models, that users can see with such
1218 argument on the command line: --cfg=cpu/model:help
1219 * --help-models display the long description of all known models
1221 * config: add the ability to set a default value after registration
1222 Does not override any previously set value (e.g. from cmd line)
1223 * dict: allow to have integer key and data.
1224 When so, you need to use the following functions
1225 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1226 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1227 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1228 In contrary to regular dicts, the key is not malloced before copy.
1229 Mixing scalar and regular elements in the same dict is not tested
1231 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1232 Tracing for Visualization:
1233 * SimGrid is now instrumented in order to generate a trace file for
1234 visualization analysis: to use it, need to compile SimGrid with the
1235 "tracing" option enabled, and instrument the program using SimGrid with
1236 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1237 (among other functions).
1238 * The instrumentation only traces the platform utilization for now
1239 * Documentation to use the tracing functions and how to analyze the
1240 traces with the Triva tool is written.
1241 * More information about: SimGrid FAQ (in the section Tracing Simulations
1244 * We moved to cmake as default build system. Autotools support will
1245 be dropped soon. Check the FAQ for more info about how to use it.
1246 * Greatly improved our cdash/ctest interactions
1247 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1248 * Added memory checking tests with valgrind; lot of memleak fixing.
1249 This may be the first release of simgrid with so few memory issues
1250 * Added code coverage tests.
1251 Our coverage is still improvable, but at least we see it on cdash.
1253 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1255 SimGrid (3.3.4) stable; urgency=low
1257 The "Desktop Grid needs love too" release (also called Xmas release).
1259 Models improvements:
1260 * Major speedup in the maxmin system solving by using lazy evaluation
1261 Instead of solving completely the maxmin system at each iteration,
1262 only invalidate (and recompute) the modified parts.
1263 This new feature is enabled in default models but you can try to
1264 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1265 * Cas01 IMproved as default CPU model
1266 This CPU model is the same Cas01 model, but it uses the
1267 maxmin-selective-update flag and a heap structure to manage
1268 actions on SURF kernel.
1269 It reduces the complexity to find the next action to finish and,
1270 consequently, it's faster than the old Cas01.
1271 This is the new default CPU model (Cas01).
1272 * Rename the old Cas01 model to Cas01_fullupdate
1273 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1274 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1275 * CpuTI (CPU Trace Integration)
1276 A new CPU model whose objective is simulate faster when using
1277 availability trace files.
1278 Instead of using a full featured, over engineered maxmin system for
1279 CPU modeling, this model does the pre-integration of traces files
1280 to calculate the amount of CPU power available, and so, executes
1281 faster than the old CPU models.
1282 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1283 * Use LV08 as default network model since it gives better accuracy
1284 for small messages and shouldn't change things for big ones.
1285 Use --cfg=network_model:CM02 to get the previous behavior.
1288 ******************************************
1289 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1290 ******************************************
1291 * The new CPU model may changes simulations!
1292 The point is that events occurring at the exact same timestamp
1293 are not scheduled in the same order with the old and new
1294 version. This may be enough to completely change the execution
1295 of simulations in some cases.
1296 * The new network model will change simulations!
1297 This new model is more realistic than the previous one, so you
1298 should consider redoing your old experiments with this model.
1299 Sorry for the inconvenience.
1302 * Introduce the supernovae compilation mode
1303 When compiled that way, the whole SimGrid (or almost) is put in a
1304 single compilation unit and compiled in one shoot.
1305 This is to help gcc which has difficulties to inline stuff from one
1307 The speedup seem to be above 15%, althrough more tests are needed on
1308 amd64 to confirm that gain.
1311 * Port of MSG's mailbox on top of SIMIX network
1312 The put/get mechanism was greatly simplified on the way.
1315 * New SIMIX network module. Provides:
1316 - Mailbox: rendez-vous mecanism to find with who you want to speak
1317 - Synchronous send/recv: easier and hopefully faster since the
1318 logic is handled in the maestro process directly now
1319 - Asynchronous send/recv: you dreamt of it? It's here now
1320 Too bad that nobody cared enough to propagate the change to MSG.
1321 * Add semaphores as SIMIX synchronization mechanism.
1324 * new function SD_daxload(char*) to load a DAX file
1325 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1326 * Introduce typed tasks. Specify its kind and cost at creation.
1327 At scheduling, just give where it should be placed, and the cost
1328 for each involved resource is automatically computed.
1329 Existing constructors so far (more to come of course):
1330 - SD_task_create_comm_e2e() for end-to-end communication
1331 - SD_task_create_comp_seq() for sequential computation
1332 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1333 * new function SD_task_dump() for debuging display
1334 * new function SD_task_dotty(task,FILE*) writing to file the info
1335 about the task in dotty format
1336 * SD_task_dependency_exists() can now cope with having one of its
1337 arguments NULL. If so, it tests whether the other argument has any
1339 * Add getters on list of preceding/following tasks:
1340 SD_task_get_parents(task) and SD_task_get_children(task)
1341 * Add getters on amount of workstations and list:
1342 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1343 * Add getter on task kind: SD_task_get_kind(task)
1344 * Update the start_time and finish_time of tasks on completion/failure
1345 * Bugfix: Remove task from state swags when destroyed
1348 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1351 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1352 dynar with one separate thread per value of the dynar.
1353 * Change the prototype of xbt_thread_create(), sorry.
1354 Added a boolean parameter indicating whether we want to join this
1355 thread (used in SG only for now)
1356 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1359 * GTNetS wrappers should now be usable again (and betterly tested too)
1360 * Fix a major regression from 3.2 where the timeout provided to
1361 MSG_task_put_with_timeout() was used as absolute time before which
1362 the comm should be done.
1363 * Start to fix the <cluster> tag.
1364 - Internal links should be good now (beside of the loopback, which
1365 use the private link instead)
1366 - paths to the external world is still rather broken
1367 - the <route:multi> tag is just broken. Actually that's brain-dead.
1368 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1369 to make it less stupid
1370 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1371 * Fix a source-level compatibility glitch from 3.2: after defining
1372 MSG_USE_DEPRECATED, you can use the old name
1373 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1374 * Allow to compile from the SVN with automake 1.11
1375 * Fix some problems when using the "start_time" tag in deployment XMLs.
1376 * Fix #8569: XBT/synchro.h has redundant declarations
1377 * Fix #8563: MSG return values and exceptions
1378 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1379 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1380 Thanks to Robson Peixoto.
1381 * Fix a vicious bug in dictionaries inducing that some elements were
1382 not freed on xbt_dict_free()
1384 Portability report of this version:
1385 * Main portability targets:
1386 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1387 on (amd64/i386/ia64)
1388 - mac leopard on i386
1389 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1390 but nothing critical.
1391 * Other platforms: windows, AIX and others were not tested for this release
1393 Timing report of this version:
1394 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1395 scenario parameters). From 8h to a few seconds in desktop grid settings.
1396 * Supernovae brings about 25% speedup on i386.
1398 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1400 SimGrid (3.3.3) stable; urgency=low
1402 The "Need for Speed" release.
1404 The timings done to validate the 3.3.2 were faulty.
1405 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1407 The problem was a conversion from a manually handled vector to
1408 xbt_dynar_t on the critical path.
1409 xbt_dynar_foreach calls functions, inducing stack management crap.
1411 We inlined these functions and xbt_dynar_foreach is now breath taking.
1412 We also inlined xbt_swag_belong on the way.
1414 Here are some approximate speedup measurements (on master/slaves
1415 simulations lasting between 10s and 20s each):
1416 3.3.1 -> 3.3.2: about same performance
1417 3.3.2 -> 3.3.3: 40% speedup
1418 3.3.1 -> 3.3.3: 40% speedup
1419 3.3.1 with inline patch -> 3.3.3: 30% speedup
1421 Our reading is that the refactoring which occurred in 3.3.2 made us
1422 suffer much more from the xbt_dynar_foreach low performance, but
1423 once we solved this, this refactoring proved to be very performance
1424 effective. From the 40% speedup, somehow, 10% are due to the
1425 inlining and 30% to the refactoring.
1427 That's a pitty that gcc cannot inline functions placed in other files
1428 alone. We have to choose between:
1429 - break the encapsulation (by putting private data structures and
1430 accessors in headers files to help gcc)
1431 - live with low performance
1432 - switch to a decent compiler such as icc (not quite possible).
1434 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1436 SimGrid (3.3.2) stable; urgency=low
1438 The "Simplicity does not preceed complexity, but follows it" release.
1440 The main contributors of this release were (lexical order):
1441 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1444 * Extract the routing logic into its own object.
1445 (was dupplicated in network.c and workstation_LV07.c;
1446 Allows to implement other ways of storing that info)
1447 => kill now useless network_card concept
1448 - Use dynar to represent routes (instead of void** + int*)
1449 - kill link_set (use surf_network_model->resource_set instead)
1450 - Add a command-line option to choose the routing schema to use
1451 - Add three new models:
1452 * Floyd (shortest path computed at initialization)
1453 * Dijikstra (shortest path recomputed all the time)
1454 * Cached Dijikstra (shortest path computed on need)
1455 All these models where contributed by Silas De Munck, and are
1456 described in his ICCS09 paper.
1458 * Simplify model declaration
1459 (less redirections, less function to write when defining a model)
1460 - Factorize stuff between models:
1463 surf_model_resource_set(model)
1464 surf_model_resource_by_name(model, name)
1465 - Unify the types of models in s_surf_model_t (using an union)
1466 - Embeed fields of common_public directly into s_surf_model_t
1467 - Rename model methods:
1468 action_free ~> action_unref
1469 action_change_state ~> action_state_set
1470 action_get_state ~> action_state_get
1471 - Change model methods into functions :
1472 (model)->common_public->action_use ~> surf_action_ref
1474 * Implement a generic resource; use it as ancestor to specific ones
1475 (allows to kill duplicated code in models)
1476 Drawback: timer command don't need no name nor properties;
1477 workstation_CLM03 don't need no properties
1478 (but I guess we can live with those few bytes wasted)
1480 * Improve the action object model
1481 - implement a constructor avoiding dupplicated code about field
1482 initialization in generic_action part.
1484 * Kill the SDP model: it has an external dependency, is deprecated
1485 in flavor of modern lmm models, and didn't compile since a while
1488 * Relocation of the context module from XBT to SIMIX.
1489 (the context were decoupled from the simix processes, duplicating a lot of code)
1490 => a lot of code was factorized
1491 - less overhead is introduced during scheduling
1492 - simpler API for the context factory
1493 - the logic for process creation,destruction and manipulation was simplified
1494 * Simplification of the s_smx_process_t data structure.
1495 => accesing the simix level data associated to a process is faster now,
1496 and the code is a lot more readable.
1499 * Implement some more MPI primitives:
1500 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1501 -implementation: Bcast: flat or 2-ary tree (default),
1502 Barrier: 4-ary tree,
1504 Allreduce: Reduce then Bcast
1505 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1506 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1507 Alltoallv: flat tree, like ompi
1509 * Add support for optimized collectives (Bcast is now binomial by default)
1510 * Port smpirun and smpicc to OS X
1513 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1514 Shout out if you used it.
1517 * Display the list of still queued messages in SG mode when existing
1521 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1522 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1523 * Add xbt_str_from_file(FILE*)
1524 * Add xbt_dict_get_key achieving a linear reverse search
1525 * Remove the context module
1527 Portability report of this version:
1528 * Main portability targets:
1529 - Linux(debian)/x86/context
1530 - Linux(debian)/x86/pthreads
1531 - Linux(debian)/amd64/context
1532 - Linux(debian)/amd64/pthreads
1533 On these, we still have the eratic breakages of gras/pmm and
1534 amok/saturate_sg reported in previous version. We still think
1535 that the tests are the cause of the fault, not the tested code.
1537 - Mac OSX Leopard/x86/context
1538 Still false negative in tesh autotesting.
1539 Smpi still fails, but this time because readlink does not accept -f
1540 Everything seems to work properly beside of that.
1543 - AIX version 5.3 (only tested contexts this time)
1544 Smpi still fails there because mktemp is not installed.
1545 Everything seems to work properly beside of that.
1547 I managed to compile it for the first time, but several breakages.
1548 Won't delay the release for this exotic platform.
1550 * Windows: it's still lagging behind. If you want to help, please
1553 Timing report of this version:
1554 This version seem to be more than 5% faster than 3.3.1 (on linux
1555 64bits with contextes). The gain is less than expected, we are
1556 investigating this for next release.
1558 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1560 SimGrid (3.3.1) stable; urgency=low
1563 * Implement a --cfg-help to show existing configuration variables
1564 * Build chain do not require doxygen in maintainer mode
1567 * fix a bug on struct sizeof computation, which prevented the
1568 exchange of arrays of structs in some conditions
1569 - added a regression test about this in datadesc_usage
1570 * Allow the exchange of 0-long dynamic vectors.
1571 - for that, use -1 as indicator of dynamic size instead of 0
1572 - This implied to change any size from unsigned long to long,
1573 reducing a bit communication abilities, but I guess that with
1574 64bits being quite common, this is more than enough.
1575 - This also induce a protocol change, thus bumping network protocol
1576 version from 0 to 1 (if we have external users, we have to get
1577 clean on that point too ;)
1578 - added two regression tests about this in datadesc_usage
1579 * Be more verbose when propagating local exceptions
1580 This helps debugging.
1581 * Display the status of simulated processes when receiving SIGINT in
1585 * Allow to control the simulation from a trace file.
1586 New functions MSG_action_register() and MSG_action_trace_run()
1587 The first one allows to associate a function execution to each
1588 kind of action while the second one parses a trace file and
1589 triggers the corresponding actions within the system.
1590 For now, only a toy example is provided in examples/msg/actions
1591 * Add an exemple of process migration in examples/msg/migration
1592 * Fix a bug in task exchange which broke MSG_task_get_sender()
1593 Add a teshsuite regression test for that.
1594 [Bug: if MSG_task_get_sender() is called after sender exit,
1596 * Fix a bug which prevented suspend/resume to work properly
1597 * Display the status of simulated processes when receiving SIGINT
1598 This fixes a regression of v3.3. due to the introduction of SIMIX
1599 * Bug fixing in failure management:
1600 - trace could not start by a failure at time 0
1601 - failure during communications were not working
1604 * Add SIMIX_process_set_name() to change the name of the current
1605 process in the log messages.
1606 * Store smx_hosts in a dict since we only retrieve them by name
1607 * Move the configuration infrastructure to surf
1610 * Move the configuration infrastructure to surf
1613 * Massive internal cleanups:
1614 - Store internal structures on processes instead of hosts (allows
1615 to have more than one process per host, in addition of being more
1617 - Cleanup the initialization/finalization process
1618 - Kill a whole bunch of unneeded synchronization:
1619 processes run in exclusive manner within the simulator
1620 - Move queues from global tables to process data fields
1622 - now accept -platform and -hostfile arguments
1623 - Pass the right rank value to processes according to the hostfile
1624 * Compile the examples by default, and use them as regression tests
1625 * Implement MPI_Wtime()
1626 * Change the reference speed to a command line option
1629 * TCP_gamma can now be specified as command line option using
1630 --cfg=TCP_gamma:10000000.0
1631 * Change the --surf-path cmd line option into --cfg=path:
1634 * Also include strbuff from xbt.h public header
1635 * xbt_ex_display(): do not free the exception after displaying
1636 This allows to do more with the given exception afterward.
1637 Users should call xbt_ex_free() themselves.
1641 Portability report of this version:
1642 * Main portability targets:
1643 - Linux(debian)/x86/context
1644 - Linux(debian)/x86/pthreads
1645 - Linux(debian)/amd64/context
1646 - Linux(debian)/amd64/pthreads
1647 These targets fail about 1/10 of times on gras/pmm, but we believe
1648 that this is because of the test, not because of simgrid.
1649 amok/saturate_sg fails even more rarely, and the test may not be
1652 - Mac OSX Leopard/x86/context
1653 The test suite still spits tons of errors because some obscure
1654 force prevents us from removing the temporary directories
1655 arguing that they still contain some metadata I've never heard of.
1656 Smpi fails because seq is not installed.
1657 Everything seems to work properly beside of that.
1660 - AIX version 5.3 (both contexts and pthread)
1661 Smpi still fails there because mktemp is not installed.
1662 XML inclusions seems rosty on AIX.
1664 * Windows: it's still lagging behind. If you want to help, please
1667 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1669 SimGrid (3.3) stable; urgency=high
1673 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1674 [Malek Cherier & Mt]
1676 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1679 Until now, GRAS were using MSG as an interface to SURF. It was
1680 quite difficult because both interface have several differences
1681 (MSG channels vs GRAS sockets were the most notable point).
1683 This also opens the gate to SMPI (which should occur soon) and speed
1684 up simulations by to 40% (even if it were not the main goal).
1686 **************************************
1687 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1688 **************************************
1689 The point is that events occuring at the exact same timestamp are
1690 not scheduled in the same order with the old and new version. This
1691 may be enough to completely change the execution of simulations in
1692 some cases. Sorry for the inconvenience.
1694 * Cleanup and upgrade the XML format to push further scalability
1695 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1697 * Improve the testing infrastructure with tesh. Now a very large part of
1698 the code is tested not only by being run but also by checking that the
1699 output match an expected output [Mt].
1701 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1702 is really less memory-demanding, which should allow you to use
1703 larger files in SimGrid [AL].
1705 * Inform valgrind about our contextes, so that it becomes usable
1706 with the default (and more effecient) version of SimGrid
1707 [contributed by Sékou Diakite, many thanks]
1710 * Introduce a listener thread in charge of receiving incoming
1711 messages from the network. It allows to overlap communication and
1712 computation but most notably, it removes some stupid deadlocks due
1713 to the fact that so far, a process could not send and receive at
1714 the same time. This made most non trivial communication schema
1716 * Convert the PIDs from long int to int to match the MSG ones (and
1717 linux ones too) [Mt]
1718 * New function: gras_agent_spawn() to launch a new process on
1719 current host. Only working in simulation for now. [Mt]
1720 * New function: gras_os_hostport() returning a constant form (ie,
1721 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1724 * Make the backtrace of exceptions more human readable [Mt]
1725 * New module: xbt/str [Mt]
1726 a ton of string utility functions (split, join, printf to a newly
1727 allocated buffer, trim, etc)
1728 * New module: xbt/hash [Mt]
1729 SHA1 hashing algorithm (more to come if needed)
1730 * New module: xbt/synchro [Mt]
1731 synchronization tools (mutex and conditions) working the same way
1732 in simulation and in real life (mainly useful for GRAS, but not
1734 * New module: xbt/queue [Mt]
1735 classical producer/consumer synchronization scheme
1736 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1737 (using the classical functions will get serialized) [Mt]
1738 * Make dictionary internal table dynamic. No need to specify its size
1739 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1741 * Make sure the log channels are organized as a tree under windows
1742 (because of ANSI C compatibility issue, any channel were child of
1743 root directly) [Mt].
1746 * Cleaned many thing in surf and fixed a few bugs [AL].
1747 * Add a nice command line configuration mechanism to compose models [AL].
1748 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1749 the previous one (KCCFLN05). It relies on something that looks like
1750 a max-min sharing mechanism but cannot be written as such. A new solver
1751 was thus designed [AL].
1752 * Add a new solver to lmm. Based on Lagrange optimization and
1753 gradient-based descent, it enables to efficiently maximise systems s.a
1755 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1757 This solver enables to propose two new network models for TCP Reno and
1758 TCP Vegas based on Low's work. These models still need to be fully
1759 tested though [Pedro Velho].
1762 * Bug fix in SD_simulate. Now the time bound given as argument is
1764 * Use the new parallel task model (ptask_L07) as default.
1765 * Use the SURF command line configuration mechanism.
1766 * 0-size tasks (for synchronization) should now work.
1768 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1770 SimGrid (3.2) stable; urgency=high
1774 We still experience issues on this platform, but we believe that at
1775 least MSG is usable.
1777 GRAS API BREAKAGE (for simplification purpose, sorry):
1778 * the gras_msgtype_by_name is not used anymore. Instead of
1779 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1780 you can write (and must)
1781 gras_msg_send(toserver, "request", &request);
1782 - If you still want to pass a gras_msgtype_t to the function (to cache
1783 the type and avoid the lookup time), use the gras_msg_send_() variant.
1784 - Impacted functions:
1785 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1786 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1787 * The callbacks are now expected to return 0 when everything went well
1788 (just like the main() function)
1790 GRAS new features and improvements:
1791 * New module mecanism where user code can use per process globals [Mt]
1792 This is similar to gras_userdata_*() functions, but for libraries. It
1793 factorize some code developped over and over in the examples and AMOK.
1794 It has still to be documented and used (only amok/peermanagement is
1796 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1797 loss when they were small enough to fit into the buffer and sent quickly
1798 enough so that they can all get received in one shoot.
1799 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1800 if not found. Use the *_or_null() variant for the old semantic.
1801 * In gras_msg_handle, do not discard messages without callback.
1802 They are probably messages to be explicitly awaited later (ie, proofs of
1803 mis-synchronization in userland since they are sent before being awaited)
1804 No big deal usually.
1805 * gras_socket_meas_send/recv: semantic changed!
1806 The numerical arguments used to be (1) the total amount of data to send
1807 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1808 messages. This was need for the fool willing to send more than MAXINT
1809 bytes on quite fat pipes.
1812 * Do really rename the hostmanagement module to peermanagement. [Mt]
1813 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1814 but this is rather new and this was documented in the module
1815 documentation (poor excuses, I admit)
1816 * Bandwidth measurement semantic changed! This follows the changes to
1817 gras_socket_meas_send/recv explained above.
1820 * A sequential mode has been added to the workstations. When a workstation
1821 is in sequential mode, it can execute only one task, and the other tasks
1822 are waiting in a FIFO. [Christophe Thiery]
1825 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1826 model for SIMDAG. [Christophe Thiery]
1827 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1828 equal to 0, which caused some bad behaviors in
1829 saturated_constraint_set_update. I now use a threshold mechanism like in
1833 * When running manually src/testall, you select specific units [Mt]
1834 testall is the result of our cunit mecanism, and should replace all
1835 the scripty thingy around since bash don't run easily on billware.
1837 * A mallocator system has been added. [Christophe Thiery]
1838 Mallocators allow you to recycle your unused objects instead of freeing them
1839 and allocating new ones.
1841 Documentation update:
1842 * FAQ reworking + New FAQs:
1843 - "Valgrind spits tons of errors!" [Mt]
1844 - "How to repport bugs" [Mt]
1845 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1846 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1847 - Communication time measurement within MSG [AL]
1848 - I experience weird communication times when I change the latency [AL]
1849 * GRAS tutorial [Mt]
1851 - an introduction to the framework and to the used communication model
1852 - an initiatic tour introducing the most proheminent features:
1854 . Lesson 0: Installing GRAS
1855 . Lesson 1: Setting up your own project
1856 o Part 2: Message passing
1857 . Lesson 2: Exchanging simple messages
1858 . Lesson 3: Passing arguments to the processes (in SG)
1859 . Lesson 4: Attaching callbacks to messages
1860 . Lesson 5: Using globals in processes
1861 . Lesson 6: Logging informations properly
1862 . Lesson 7: Using internal timers
1863 . Lesson 8: Handling errors through exceptions
1864 . Lesson 9: Exchanging simple data
1865 . Lesson 10: Remote Procedure Calling (RPC)
1866 . Lesson 11: Explicitely waiting for messages
1867 . Recapping of message passing features in GRAS
1868 - A HOWTO section containing:
1869 o HOWTO design a GRAS application
1870 More are due, of course. They will come latter. In the meanwhile, you can
1871 check the examples which are still here.
1873 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1875 SimGrid (3.1) stable; urgency=high
1879 There was a stack corruption somewhere, visible only when optimizing
1880 with these versions. [Vince]
1883 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1884 module that enabled to play with Directed Acyclic Graphs. It is built
1885 directly on top of SURF and provides an API rather close to the old
1886 SG. Some old codes using SG are currently under rewrite to check that
1887 all needful functions are provided. [Christophe Thiery]
1890 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1891 extension of the classical CLM03 model that gracefully handles
1892 failures. This is now the default model for MSG and GRAS. It doesn't
1893 handle parallel tasks yet though. [AL]
1894 * Bug fix: Weights were not correctly set in the network part.
1895 WARNING: This may have resulted in incorrect results with simulations
1896 where there are more than one flow on a given link. [AL]
1899 * After a (long ?) discussion on simgrid-devel, we have decided that the
1900 convention we had on units was stupid. That is why it has been decided
1901 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1902 WARNING : This means that all previous platform files will not work as
1903 such with this version! A warning is issued to ask users to update
1905 A conversion script can be found in the contrib module of the CVS, under
1906 the name contrib/platform_generation/surfxml_update.pl [MQ]
1909 * Bug fix: Processes were started in reverse order, wrt deployment file.
1910 WARNING: if your code relies on this bug, please fix it. [AL]
1911 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1912 being executed on two different locations. [AL]
1913 * Bug fix: Failures are now better supported thanks to Derrick's tests
1914 (there was many failure situations I hadn't thought of and that weren't
1915 correctly handled). [AL]
1916 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1920 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1921 Exception occurring on server-side are propagated back to client (!).
1923 API CHANGE: the callback changed their prototype. Change:
1924 int my_handler(gras_socket_t expeditor, void *payload_data) {
1926 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1927 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1929 * New! function: gras_msg_handleall to deal with all messages arriving
1930 within a given period.
1931 * New! function: gras_socket_server_range to get a server socket in a
1932 range of port numbers (ease to avoid port number conflicts) [MQ]
1933 * New! gras processes display their backtrace when they get a SIGUSR1
1934 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1935 Sweet to debug RL processes [MQ]
1939 - Do not force experiment sizes to be expressed in kb, or it becomes
1940 impossible to measure the latency this way (needs one byte-long tests)
1941 WARNING: this changes the amok_bw_* function semantic. [MQ]
1942 - Implements the link saturation stuff. [MQ]
1943 * Peer management module:
1944 New! module factorizing code that we wrote over and over [MQ].
1947 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1948 - Test units are placed directly into the library code, they get extracted
1949 automatically and placed into the src/testall binary.
1950 - Convert most of the XBT tests to this system.
1951 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1952 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1953 live in the same C code. This required to change a little bit the API
1954 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1955 * New module: sparse graph structure with basic algorithms (this is work
1956 in progress and the API is not considered to be frozen yet). [AL]
1957 * Display more information on backtraces: source line & function names are
1958 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1959 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1960 * Reworked a little bit some #include statements to load only required
1961 headers. Some user code that relied on SimGrid to include stdlib or
1962 stdio may need to include it by themselves. [AL]
1963 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1965 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1966 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1967 but I guess that almost nobody used those parts. [MQ]
1969 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1971 SimGrid (3.0.1) stable; urgency=low
1974 * Unfortunately, I had missed 5 misnamed functions:
1975 xbt_fifo_item_t xbt_fifo_newitem(void);
1976 void xbt_fifo_freeitem(xbt_fifo_item_t);
1977 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1978 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1979 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1980 They're now deprecated. Please use their new versions:
1981 xbt_fifo_item_t xbt_fifo_new_item(void);
1982 void xbt_fifo_free_item(xbt_fifo_item_t);
1983 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1984 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1985 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1987 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1988 * Documentation: xbt_log module unmercifully reworked [MQ]
1989 * Bugfix: there was a problem with the ending of contexts with
1990 the pthread backend. It caused some weird deadlock or behavior
1991 depending on the pthread implementation. [AL]
1992 * Bugfix: get the exceptions raised in the simulator repport where
1993 and why they come from when they are not catched in time [AL, MQ]
1996 * Bugfix: Do repport the error when two non-connected hosts try to
1997 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2000 * Add additionnal checkings on communications. Assert that two
2001 communicating hosts are connected by a set of links... [AL]
2004 * Add additionnal checkings on channel values in communication [AL]
2005 * New: MSG_task_get_source to see on which host a task was generated [HC]
2006 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2007 the number of tasks waiting to be received on channel and sent
2009 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2010 waits for the first task coming from a given host.. [AL]
2012 GRAS new functionnalities: [MQ]
2013 * Enhance the parsing macro to allow the size of multidimentional objects
2014 to be given thru annotations.
2015 * New example (and documentation): Matrix Multiplication a la RPC
2016 (as when I was young!) and fix a bunch of bugs found on the way.
2018 GRAS performance improvements: [MQ]
2020 * Reduce the amount of cbps creation/destruction by making it static to
2021 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2024 * Change libdata to a set so that we can search for stuff by ID (and thus
2025 reduce the insane amount of dict lookups)
2028 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2029 RL communication on top of SG since it's so uneffective.
2030 It may also be used for inter-thread communication in RL, one day.
2031 * Use gras_datadesc_copy() to exchange messages on top of SG
2033 - improve message exchange performance on top of SG
2034 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2035 * Don't exchange on the network the size of the used part of buffer,
2036 instead, specify the possible buffer size to read().
2038 - reduces the amount of read/write calls (one pair per exchange)
2039 - reduces the amount of exchanged data (the size)
2040 - allows to retrieve all arrived data on receiver side, if we don't need
2041 it right now (subsequent read will peek the buffer)
2042 - allows the receiver to proceed with the begining of the stream before
2043 everything is arrived
2044 - make it possible to build an iov transport (using readv/writev)
2046 - take care of the data with non-stable storage (like stacked data),
2048 * If possible, TCP send uses vector I/O (when writev() is here)
2049 - Don't use it for receive since we send data sizes and data on the
2050 same stream, so we wouldn't be able to chain large amount of chunks
2051 before having to flush the stuff to read the size.
2052 * Rework the transport plugin mecanism to simplify it and reduce the
2053 amount of pointer dereferencement when searching for the right function
2056 * I guess that now, we do almost as few system calls as possible while
2057 doing as few data copy as possible.
2059 To improve it further, we could try to send all the sizes first and then
2060 all the data (to use iov on receiving size), but it's only a partial
2061 solution: when you have 2 dimensional data, the sizes of the second
2062 dimension is data of the first dimension, so you need 3 streams.
2064 I'm not sure the potential performance gains justify the coding burden.
2066 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2068 SimGrid (3.00) stable; urgency=high
2071 * New! Give the possibility to hijack the surf parser and thus bypass
2072 MSG_create_environment and MSG_launch_application. Have a look at
2073 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2075 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2077 SimGrid (2.96) unstable; urgency=low
2082 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2083 This deprecates the xbt_error_t mecanisms.
2084 It modifies (simplifies) all XBT and GRAS API.
2085 MSG API keeps unchanged (exceptions raised by XBT are catched from
2086 within MSG and masked with existing error handling facilities)
2089 * New! Add a FATPIPE model. [AL]
2090 * New! Add a parallel task model. [AL]
2091 * New! Add automatically a loopback interface (in the default
2092 network model) if none was precised.
2095 * Bugfix: MSG_process_resume now works with the current running process.
2097 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2098 * Modification of MSG_task_get_compute_duration. Once a task has been
2099 processed, the value returned by this function is now equal to 0. [AL]
2100 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2101 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2102 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2103 of a m_task. For now, MSG_task_cancel only works with computation
2105 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2106 of the processor (in Mflop/s) regardless of the current load on the
2108 * API Change: use proper naming convention for MSG_getClock and
2109 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2111 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2112 This function changes the priority of a computation task. This priority
2113 doesn't affect the transfer rate. A priority of 2 will make a task
2114 receive two times more cpu power than the other ones. This function
2115 has been added to suit the needs of Nguyen The Loc and hasn't been that
2116 much tested yet. So if it fails, please report it and send me your code.
2118 * API Change: removed all functions and types that were marked "deprecated"
2119 since many months. Renamed MSG_global_init_args to MSG_global_init.
2121 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2123 SimGrid (2.95) unstable; urgency=low
2126 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2127 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2128 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2131 * Bugfix: Allow absolute paths for platform description files [MQ]
2132 * Bugfix: do free the variables after use. Leads to drastic performance
2134 * Implement max_duration (ie, timeouts) on resources [AL]
2137 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2139 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2140 the needed support (timeouts) [AL]
2143 * Implement measurement sockets. You can now get the bandwidth between two
2144 hosts thanks to AMOK (see below). [MQ]
2145 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2146 dynar over the network (yeah) [MQ]
2147 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2148 * simple/static token-ring example. [Alexandre Colucci and MQ]
2149 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2150 in place before [MQ]
2153 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2154 type parsing. "short" and "long" modifiers were also ignored; other
2155 modifier (such as reference level) are still ignored. [MQ]
2156 * Embeed the buffer size within the buffer itself on SG. [MQ]
2157 That way, send() are atomic and cannot get intermixed anymore (at least
2158 the ones which are less than 100k; bigger messages still have the issue)
2159 * Array size pushed by the field, not by the field type (or each
2160 and every long int will push stuff to the cbps) [MQ]
2161 * use select() to sleep since it allows to portably sleep less than one
2164 GRAS (minor cleanups)
2165 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2168 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2169 * type category 'ignored' killed as it was never used and were difficult
2171 * whether a type can cycle or not is now a flag, leaving room for more
2172 flags (as "ignored", if we feel the need one day ;)
2173 * Rename raw sockets to measurement sockets since "raw" has another
2174 meaning in networking community.
2177 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2178 and offers features not belonging to GRAS but that most applications
2179 need. One day, it may be a set of plugins loadable at runtime.
2180 * New module: bandwidth
2181 bandwidth measurement between arbitrary nodes running this module. [MQ]
2183 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2185 SimGrid (2.94) unstable; urgency=low
2187 The first beta release of SimGrid 3 !
2191 * Update the main page and the FAQ. Adding references to gforge.
2194 * Add a gras_os_getpid function.
2197 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2198 * Extend the logs so that they also print PID, hostname, date, ... if
2200 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2201 and kill the old version which were in testsuite/
2202 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2206 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2207 for users. Instead, each programming environment (GRAS, MSG,...) use it
2208 when needed to provide such a feature to users.
2209 Don't shortcut the mecanism or you will also shortcut the virtualization
2210 you need on the simulator.
2214 * Cleanups in configury with regard to compile optimization/warning flags.
2215 Also add -fno-loop-optimize to any powerpc since it's the optimization
2216 killing gcc (< 3.4.0).
2217 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2218 needed by MSG examples complications
2219 * Borrow configury beautifications from PHP
2222 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2223 warning (thanks loris for stumbling into this one).
2224 * Bugfix: stop loading private headers (gras_config.h) from the public
2228 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2229 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2230 RL would imply a DNS resolver, which is *hard* to do in a portable way
2231 (and therefore delayed).
2232 * Implement a real timer mecanism and use it in timing macros. This allows
2233 to avoid rounding errors and get a 0.000005 sec precision in timing
2234 macros. While I was at it, various cleanups:
2235 - allow to declare more than one timed section per file (fix a stupid bug)
2236 - move some private declaration to the right place
2237 - merge conditional execution and timing macros into emulation module
2238 - document the module
2239 - make sure the module cleanups its mess on gras_exit
2240 * Documentation improvements:
2241 - (new) how to compile applications using GRAS
2242 - (new) emulation support (timing macros)
2244 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2246 SimGrid (2.93) unstable; urgency=low
2248 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2251 - Use Paje properly where used. Still to be sanitized properly.
2252 - Portability fix: Add an implementation of the contexts using pthread
2256 - Add xbt_procname(): returns the name of the current process.
2257 Use it to show the current process's name in all logging.
2259 - fix detection of older flex version and the reaction, since we do
2260 depend on modern ones (we use lex_destroy)
2261 - Better separation of SG and RL in the libs: remove all simulation code
2262 from libgras. As a result, this lib is now only 200k when stripped.
2263 Some of the xbt modules may also be duplicated (two sets and such) and
2264 should be cleaned/killed before SG3.
2265 - Insist on using xlC on AIX because of weird problems involving gcc there.
2266 - Cleanup the make remote stuff. This is now done by scripts
2267 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2268 mainly for our private use, but we're working on changing them to user
2271 - Bugfix: flush the socket on close only if there is some *output*.
2272 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2274 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2277 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2279 SimGrid (2.92) unstable; urgency=low
2281 Alpha 3 on the path to SimGrid 3
2285 - New! First try of benchmarking macros.
2286 - New! First try so that gras_stub_generator generate deployment and
2287 remote compilation helpers.
2289 - Bugfix: Initialization fix in msg_test.
2293 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2295 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2297 - API Change: gras_os_sleep to take the amount of seconds as a double.
2298 Accepting an int was error prone since it was the only location where
2299 seconds were coded as such. It leaded to damn rounding errors.
2300 - Bugfix: Hard to belive that timers ever worked before this.
2302 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2304 SimGrid (2.91) unstable; urgency=low
2306 Alpha 2 on the path to SimGrid 3
2310 - Bug fix in the lmm_solver.
2312 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2313 through the function MSG_paje_output.
2314 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2315 - It is possible to bound the rate of a communication in MSG with
2316 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2317 to write it in the changelog).
2318 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2323 - Various cleanups to the autotools stuff
2324 - Begin to move Gras examples to examples/gras/
2325 - Let make distcheck work again (yeah!)
2327 - documentation overhauled using doxygen.
2328 gtk-doc-tools is dead in SimGrid now.
2329 - Automatically extract all existing logging categories, and add the list
2330 to the documentation (long standing one, to say the less)
2332 - Cleanup the known architecture table. Reorder the entries to group what
2333 should be, and use a more consistent naming scheme.
2334 (some of the test dataset are still to be regenerated)
2335 - New! Allow library to register globals on each process just as userdata
2337 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2338 lookup time (for now).
2339 Use it in msg and trp.
2340 This cleans a lot the internals and helps enforcing privacy of the
2341 headers between the gras components.
2342 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2343 - Bugfix: gras_os_time was delirious in RL.
2344 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2346 - Reenable GRAS now that it works.
2348 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2350 SimGrid (2.90) unstable; urgency=low
2352 Alpha 1 on the path to SimGrid 3
2354 * It is a long time since the last release of SimGrid. I'm sorry about
2355 that but as I had told you, I was rewriting a lot of things. I apologize
2356 to those who had been reporting bugs to me and that I had not answered.
2357 If your bug is still in the new version, please tell me. Here is a
2358 summary of the main changes.
2360 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2361 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2362 portability, speed, and a lot more but you'll figure it out later.
2363 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2364 I wanted to release the new MSG as soon as possible and I have
2365 broken GRAS, which is the reason why, for now, only MSG is fully
2366 functional. A laconic description of these projects is available
2367 in the documentation.
2369 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2370 kernel optimized for our needs (called SURF but only the developers
2371 should use it). Hence, MSG is now roughly 30 times faster and I think
2372 that by rewriting a little bit MSG, I could event speed it up a little
2373 bit more. Beside the gain in speed, it is also much easier to encode a
2374 new platform model with SURF than it was with SG. More to come...
2376 * REVOLUTION 3: I have tried to change a little as possible the API of
2377 MSG but a few things really had to disappear. The main differences
2378 with the previous version are :
2379 1) no more m_links_t and the corresponding functions. Platforms are
2380 directly read from a XML description and cannot be hard-coded
2381 anymore. The same format is used for application deployment
2382 description. The new format is described in the documentation.
2383 Have a look in tools/platform_generation. There is a tiny script
2384 that converts from the old platform format to the new one. Concerning
2385 the application deployment format, parsing the old one is tricky.
2386 I think most of you should however be able to convert your files. If
2387 it is really an issue, I can write a C code that does the conversion.
2389 2) the toolbox tbx does not exist anymore. We now have a library
2390 with much more data-structures but without the hash-tables (we have
2391 dictionaries that are much faster).
2393 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2395 *****************************************************************************
2396 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2397 * don't feel like converting the oldies. (Mt) *
2398 *****************************************************************************
2401 Version 2.90: "the long awaited one"
2402 - Finished rewriting and debugging MSG. Rewrote the documentation.
2403 - disable GRAS for now since it needs to be ported to the newest SG
2406 - Finish the port to windows (using mingw32 for cross-compile)
2409 - Main loop and datastructures of SURF. A cpu resource object is
2410 functional. Surf can thus be used to create cpu's with variable
2411 performance on which you can execute some actions.
2413 2004-11-15 Martin Quinson
2414 - Port to ARM. Simply added the alignment and size descriptions. Should
2415 work, but the ARM machines are so slow that I didn't had the opportunity
2416 to 'make check' over there yet.
2418 2004-11-15 Arnaud Legrand
2419 - Trace manager now written. It uses a heap structure and is therefore
2420 expected to be efficient. It may however be speeded up (particularly
2421 when many events occur at the same date) by using red and black
2422 trees. One day maybe...
2423 - Max-min linear system solver written. It uses a sparse matrix
2424 structure taking advantage of its expected use. Most operations are
2425 O(1) and free/calloc are called as few as possible. The computation of
2426 the minimum could however be improved by using a red and black tree
2429 2004-11-03 Arnaud Legrand
2430 - Rename every gras_* function that was in xbt/ to its xbt_
2432 - Add a heap and a doubly-linked list to xbt
2433 - Added a dichotomy to the dictionaries. make check works as well before
2434 so I assume that the patch is correct. I do not know however if things
2435 run effectively faster than before now. :)
2437 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2438 SimGrid, and the version number is bumped to 2.x
2440 2004-10-29 Martin Quinson
2441 - Introduction of the remote errors.
2442 They are the result of a RMI/RPC on the remote machine.
2443 ErrCodes being scalar values, you can't get the host on which those
2444 errors did happen. Extending the error mechanism as in Gnome is possible.
2445 No idea yet whether it is a good idea.
2447 2004-10-28 Martin Quinson
2448 - Interface revolution: the Starred Structure Eradication.
2449 I used to do typedef struct {} toto_t; and then handle *toto_t.
2450 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2451 - ???_t is a valid type (builded with typedef)
2452 - s_toto_t is a structure (access to fields with .)
2453 - s_toto is a structure needing 'struct' keyword to be used
2454 - e_toto_t is an enum
2455 - toto_t is an 'object' (struct*)
2457 typedef struct s_toto {} s_toto_t, *toto_t;
2458 typedef enum {} e_toto_t;
2459 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2460 s_toto_t) is private.
2462 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2463 so that it can set the variable to NULL. It was so for dicts and sets,
2464 it changed for dynars.
2466 - Fix a bunch of memleaks in dict_remove
2467 - Fix a bug in sg/server_socket opening: it failed all the time.
2469 2004-10-07 Martin Quinson
2470 - Speed up dynar lookup operation a bit.
2472 gras_dynar_get is dead.
2474 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2475 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2476 (which gives you the address of the stored data).
2478 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2479 copy of the data using an affectation to do the job and not a memcpy.
2481 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2482 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2484 It does not really speedup the dynar test because they are
2485 setting elements all the time (and look them seldom). But the dict does
2486 far more lookup than setting.
2488 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2490 2004-10-05 Martin Quinson
2491 - Allow to (en/dis)able the cycle detection at run time.
2493 Whether we should check for cycle or not is now a property of each
2494 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2495 datadesc_cycle_unset allow to remove this property when previously set.
2497 Note that the cycle detection is off by default since it impacts the
2498 performance. Watch the data you feed GRAS with ;)
2500 This property is hereditary. Any element embedded in a structure having it
2501 set have it set for the time of this data exchange.
2503 You should set it both on sender and receiver side. If you don't set it on
2504 sender side, it will enter an endless loop. If you forget on receiver
2505 side, the cycles won't be recreated after communication.
2507 - Header reorganization.
2508 Kill gras_private.h, each submodule must load the headers it needs.
2510 2004-10-04 Martin Quinson
2511 - Interface revolution: do not try to survive to malloc failure.
2513 Now, gras_malloc and friends call gras_abort() on failure.
2514 As a conclusion, malloc_error is not a valid error anymore, and all
2515 functions for which it was the only gras_error_t return value are
2516 changed. They now return void, or there result directly.
2517 This simplify the API a lot.
2519 2004-09-29 Martin Quinson
2520 - Re-enable raw sockets.
2521 Created by gras_socket_{client,server}_ext;
2522 Used with gras_raw_{send,recv}
2525 It should allow to kill the last bits of gras first version soon.
2527 This is not completely satisfactory yet (duplicate code with
2528 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2531 - Simplify transport plugin (internal) interface by not passing any
2532 argument to _server and _client, but embedding them in the socket
2535 2004-09-28 Martin Quinson
2536 - Finish the port to AIX.
2537 autoconf was my problem (segfault within the malloc replacement
2538 function. No idea why)
2540 2004-09-16 Martin Quinson
2541 - Fix some size_t madness on 64bit architectures.
2543 2004-09-08 Martin Quinson
2544 - Reduce the number of system headers loaded, overload some more system
2545 calls (such as malloc to cast the result of the system one, and work
2547 - Fix and reintroduce the config support
2549 2004-09-07 Martin Quinson
2550 - Source code reorganization to allow Arnaud to surf all over there.
2551 - Allow to document the logging categories.
2552 - Remove all uppercase from logging categories and useless cleanup in names.
2554 2004-08-18 Martin Quinson
2555 Version 0.6.2 (protocol not changed; API changed)
2556 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2557 gras_error_t), and NULL when not found. Functions expecting a msgtype
2558 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2559 hopefully usefull message.
2560 - Portability to prehistoric sparcs again
2562 2004-08-17 Martin Quinson
2563 Version 0.6.1 (protocol not changed; ABI not changed)
2564 - prealloc some buffers to speed things up
2566 2004-08-11 Martin Quinson
2567 Version 0.6 (protocol not changed; ABI expended)
2568 - The parsing macro can deal with the references, provided that you add
2569 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2571 2004-08-09 Martin Quinson
2572 Version 0.5 (protocol not changed; ABI changed)
2573 - Allow to off turn the cycle detection code in data exchange at
2574 compilation time. It should be at run time, but I'm short of time (and
2575 the config stuff is still broken). That way, we keep dict out of the
2576 critical path, which is good because the performance is poor:
2577 - search not dichotomial yet
2578 - dynar give no way to access their content and memcpy everytime
2579 - In composed data description (struct, ref and so on), stop foolness of
2580 keeping the subtype's ID, but store the type itself. This keeps sets out
2581 of the critical path, which is good since they rely on dynar and
2582 dictionnaries. The only loose of that is that we cannot detect the
2583 redeclaration of a structure/union with another content (but I'm not sure
2584 the code detected well this error before anyway). We still can detect
2585 the redefinition discrepancy for the other types.
2586 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2587 it breaks the code because of type-punning used all over the place).
2588 This breaks on all non-gcc architectures (for now).
2590 All those changes (plus the buffer of last time) allow me to gain 2 order
2591 of magnitude on cruel tests consisting of 800000 array of integers on two
2592 level of a hierarchical structure (200 secondes -> 4 secondes)
2595 - the selector of reference must now return the type it points to, not
2596 the ID of this type.
2598 2004-08-06 Martin Quinson
2599 Version 0.4 (protocol changed; ABI not changed)
2600 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2601 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2602 - Implement a buffer transport, and use it by default (it relies on tcp in
2603 real life and on sg in simulation).
2604 That's a bit hackish since I had a new field to the structure to store
2605 its data without interfering with the subtype ones. Inheritance
2606 is tricky in C. And that's a kind of reverse inheritance with one class
2607 derivating two classes. Or maybe a game with java interfaces. Anyway,
2608 that's damn hard in C (at least).
2609 Moreover, I got tired while trying to ensure plugin separation and
2610 genericity in SG mode. MSG wants me to do weird things, so let's go for
2611 cruel hacks (temporarily of course ;).
2612 See comment in transport_private.h:71
2613 - do not include all the _interface headers in private but in the files
2614 which really need them (to cut the compilation time when they are
2617 2004-07-26 Martin Quinson
2618 Version 0.3 (protocol not changed; ABI changed)
2619 - Major overhault of the datadesc interface to simplify it:
2620 - shorted the function names:
2621 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2622 - add a trivial way to push/pop integers into the cbps without malloc.
2623 This allows to make really generic sub_type description, which simply
2624 pop their size of the stack.
2625 - add a function gras_datadesc_ref_pop_arr() which does what users want
2626 most of the time: Declare a dynamic array (which pops its size of the
2627 stack) and declare a reference to it. Poor name, but anyway.
2628 - kill the post-send callback, add a post-receive one
2630 2004-07-23 Martin Quinson
2631 Version 0.2 (protocol changed; ABI changed)
2632 - add some testing for cpbs in the test cases, and fix some more bugs.
2633 This invalidate again the little64 data file, since I cannot regenerate
2635 - remove an awfull optimization in the logging stuff, allowing me to:
2636 - understand it again
2637 - learn gcc how to check that the argument match the provided format
2638 - fix all errors revealed by gcc after that
2639 - internal keys of dict are not \0 terminated. Deal with it properly in
2640 loggings instead of segfaulting when the user want to see the logs :-/
2642 2004-07-22 Martin Quinson
2643 - Fix some stupid bug preventing cbps (callback postit) from working
2645 2004-07-21 Martin Quinson
2646 - Some documentation cleanups
2647 - remove the useless last argument of msgtype_declare
2648 - rename the Virtu functions to fit into the 'os' namespace
2649 - move headers src/include -> src/include/gras/ and stop fooling with
2650 gras -> . symbolic link
2651 - make distcheck is now successful
2653 2004-07-19 Martin Quinson
2655 - Build shared library also
2656 - Install html doc to the right location
2657 - stop removing maintainer files in make clean
2658 - build tests only on make check
2660 2004-07-13 Martin Quinson
2662 - No major issue in previous version => change versionning schema
2663 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2664 corresponding dataset.
2666 2004-07-11 Martin Quinson
2668 - Get it working with any kind of structure (we can compute the padding
2669 bytes remotely for all the architectures I have access to)
2670 - Implement the structure parsing macro (still not quite robust/complete)
2671 - Improvement to the remote testing toysuite
2673 2004-07-10 Martin Quinson
2674 [autoconf mechanism]
2675 - get ride of a bunch of deprecated macros
2676 - actually run the test for two-compliment, not only compile it :-/
2677 - test whether the structures get packed (and bail out if yes. Damn.
2678 Alignment is a serious matter)
2679 - test whether the structures get compacted (but respecting the alignment
2680 constraints of each types)
2681 - test whether the array fields of structures can straddle alignment boundaries
2683 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2684 (took me 2 hours to find that bug, looking at the wrong place)
2685 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2686 before sending/receiving and are used to compute the offsets of fields
2687 - Given that padding size depend even on compiler options, keep track of
2688 alignment and aligned_size only for the current architecture. Not a big
2689 deal since we send structure fields one after the other (seems
2691 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2692 program, let it work on linux/gcc/little32. portability todo.
2694 2004-07-08 Martin Quinson
2695 - import and improve remote compilation support from FAST
2696 - make sure make check works on half a dozen of machines out there
2698 2004-07-07 Martin Quinson
2699 Let's say it's version 0.0.3 ;)
2700 - Implement conversions (yuhu!)
2701 - Let it work on solaris (beside conversion, of course)
2702 - Stupid me, using rand() to generate the conversion datatests in not wise.
2704 2004-07-06 Martin Quinson
2705 - Let make dist work, since I'm gonna need it to compile on remote hosts
2706 - Let Tests/datadesc_usage write the architecture on which the file was
2707 generated as first byte.
2708 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2709 architecture descriptions.
2710 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2711 on those architectures.
2712 - Optimization: send/recv array of scalar in one shoot
2714 2004-07-05 Martin Quinson
2715 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2717 - Plug a whole bunch of memleaks
2718 - each process now have to call gras_{init,exit}. One day, their log
2719 settings will be separated
2720 - Continue the code factorisation between SG, RL and common in Transport.
2722 2004-07-04 Martin Quinson
2724 - Redistribution between SG and RL.
2725 We wanna have to accept in SG, so move accepted related parts of RL in
2726 the common part. (more precisely, the dynar of all known sockets is no
2727 more a static in transport.c, but part of the process_data)
2729 [gras_stub_generator]
2730 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2732 2004-07-03 Martin Quinson
2733 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2734 them all up in one shot)
2736 - Fix the ugly case of reference to dynamic array.
2737 - New (semi-public) function gras_datadesc_size to allow the messaging
2738 layer to malloc the needed space for the buffer.
2740 - gras_socket_close now expect the socket to close (and not its address to
2741 put NULL in it after it). This is because the socket passed to handlers
2742 is one of their argument (=> not writable).
2744 - propagate the interface cleanup from last week in datadesc, ie remove a
2745 superfluous level of indirection. User pass adress of variable
2746 containing data (both when sending and receiving), and not of a variable
2747 being a pointer to the data. Let's say that I like it better ;)
2748 The price for that is constructs like "int msg=*(int*)payload" in
2749 handlers, but it's a fine price, IMHO.
2751 - Let it work in RL (yuhu)
2753 2004-06-21 Martin Quinson
2755 - porting SG plugin and SG select to new standards (works almost).
2756 - plug memleaks and fix bugs around.
2759 - cleanup the prototype of data recv and force users to specify when they
2760 want to handle references to objects. Test case working even for cycles.
2761 - plug memleaks. Valgrind is perfectly ok with this.
2763 2004-06-12 Martin Quinson
2765 - cleanup the separation between plugin and main code in plugin creation
2767 2004-06-11 Martin Quinson
2769 - Reput hook for raw sockets, needed for BW experiments
2770 - kill a few lines of dead code
2771 [Data description] Interface cleanup
2772 - gras_datadesc_by_name returns the searched type or NULL.
2773 That way, no variable is needed to use a type desc once, which makes
2775 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2776 two parameters were strings (field name, type name), leading to
2778 [Dicos] Interface cleanup
2779 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2780 This is consistant with the dynar API.
2782 2004-04-21 Martin Quinson
2784 - Porting to new standards.
2786 - interface cleanup.
2787 There is no bag anymore, no need to take extra provision to mask the
2788 pointers behind "ID".
2789 Better splitup of functions between files create/exchange/convert.
2790 This is still a bit artificial since convert and receive are so
2791 interleaved, but anyway.
2793 - add a queued message list to procdata (the ones not matching criteria
2795 - factorize some more code between SG and RL wrt procdata
2797 - use gras_exit in example to track memleaks
2798 - get rid of gs_example now that GS is properly integrated into gras
2799 - update run_test to integrate the latest tests (datadesc)
2801 - rename WARNINGn macros to WARNn since it prooved error-prone
2803 2004-04-19 Martin Quinson
2805 - register init/exit functions within gras module mechanism
2806 - send/receive function.
2807 Convertion is not implemented, but short-cutted if not needed.
2808 struct/array elements are sent one by one (instead of block-wise), but
2809 nobody really cares (yet). Get a prototype before optimizing.
2810 - tests (using a file socket) for DD send/receive on:
2811 - base types: int, float
2812 - array: fixed size, string (ie ref to dynamic string)
2813 - structure: homogeneous, heterogeneous
2814 - chained list, graph with cycle
2815 Believe it or not, valgrind is not too unhappy with the results. The
2816 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2820 - Bugfix when using a filename explicitely (instead of '-')
2822 2004-04-09 Martin Quinson
2824 - factorize more code between RL and SG in socket creation
2825 - Complete the implementation and tests of:
2827 o file (only in RL, and mainly for debugging)
2829 I lost 3 days to design a portable address resolver, and then decided
2830 that the prototype mainly have to run on my box.
2831 Addressing portability too early may be like optimizing too early :-/
2833 - use gras_init in the Tests instead of the crappy parse_log_opt
2834 (the latter function is removed)
2835 [Conditional execution]
2836 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2837 [Code reorganisation]
2838 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2839 Build examples against the RL library, since there is no way to disable
2840 its creation for now.
2842 For information, the beginning of coding on GRAS was back in june
2843 2003. I guess that every line has been rewritten at least twice since