1 SimGrid (3.12) NOT RELEASED; urgency=low
4 * Require g++ v4.7 at least to not speak prehistorical C++.
5 * Require Boost 1.48 (for signal2 component).
6 * Tracing is now always enabled (no way to turn it out)
7 * Remove GTNetS. It was not working anyway.
8 * Move headers around to sort them out on installed systems:
9 - instr/instr.h -> simgrid/instr.h
10 - instr/jedule/* -> simgrid/jedule
11 - simdag/datatypes.h was removed
12 - simdag/simdag.h -> simgrid/simdag.h
13 - msg/datatypes.h was removed
14 - msg/msg.h -> simgrid/msg.h
17 * Interface improvement:
18 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
19 - Sanitize the interface in MSG_task_ module:
20 - Merge two functions that were close enough but misleading:
21 set_compute_duration(t) -> set_flops_amount(t)
22 get_remaining_computation(t) -> get_flops_amount(t)
23 - set_data_size(t) -> set_bytes_amount(t)
24 get_data_size(t) -> get_bytes_amount(t)
25 - Massive cleanups in the functions related to the energy
26 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
27 - New: MSG_host_get_pstate()
29 - msg/energy/onoff: switching hosts on and off
31 * Interface improvement:
32 - Rename Host.isAvail() to Host.isOn()
33 - Rename Process.currentProcess() to Process.getCurrentProcess()
34 - Rename Task.setDataSize() to Task.setBytesAmount()
35 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
37 - #18874: Actually allows the GC to reclaim tasks
41 - SIMIX_process_throw: raises an exception in a remote process
42 * Refactoring: Separate sub-modules
43 - libsmx: the public interface, as libc in a real system
44 - popping: the strange dance that converts a user request into a kernel handling
45 - smx_context_*: the virtualization mechanisms that embed the user code
46 - smx_*: the handling of each simcalls
48 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
49 * Rename smx_action_t into smx_synchro_t, making explicit that these
50 things are used to synchronize processes with their environment.
51 For example, a communication is a sort of synchronization involving
52 the communicating processes (that may block until the exchange) and
53 the platform. The same can be said from computations, etc.
55 - #18888: segfault when a process ends before its kill_time
59 - 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.
60 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
61 - MPI_Type_set_name, MPI_Type_get_name
62 - MPI_*_c2f and MPI_*_f2c functions
63 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
64 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
65 - Activate a lot of new tests from the mpich 3 testsuite
67 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
68 - InfiniBand network model added : Based on the works of Jerome Vienne
69 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
70 - When smpi/display_timing is set, also display global simulation time and application times
71 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
72 * Collective communications
73 - 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
74 - 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/).
75 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
77 - Replay now uses algorithms from wanted collective selector
78 - Replay can be used with SMP-aware algorithms
79 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
80 - 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.
82 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
83 - Correctly use loopback on fat-tree clusters
84 - Asynchronous small messages shouldn't trigger deadlocks anymore
85 * Energy/DVFS cleanup and improvement
86 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
87 - new: smpi_get_host_pstate()
90 - "Full" network optimization flag was broken since Surf++
91 - Better handling of precision flags in maxmin
92 - Fix bug causing sometimes "Impossible" errors
93 - Properly pass cluster properties to included hosts
94 * Improvement of the Energy plugin.
95 - Always update the consumption before returning that value
96 - New property: watt_off to denote the dissipation when the host is off
97 - New functions getWattMinAt and getWattMaxAt to retrieve the
98 dissipation of pstates that we are not currently at.
99 * Java: class NetworkLink renamed to Link
100 * New function: simcall_process_get_kill_time()
101 * Massive rename s/workstation/host/
102 - That's intrusive, but that's good for the project consistency. Sorry.
103 - Change config option "workstation/model" into "host/model"
106 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
107 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
108 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
109 - xbt_dirname and xbt_basename for non-POSIX systems
111 * The model checker now runs as a separate process.
112 * The model checker runs must now be launched with the new simgrid-mc program.
113 * Record/Replay: the MC can display a textual representation of a path in the
114 execution graph. It can then be replayed outside of the model checker.
116 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
118 SimGrid (3.11) stable; urgency=low
123 * Normalizing pointers addresses tool for better diff between logs
126 * Add cloud examples using new VMs
127 - examples/msg/cloud/two_tasks_vm.tesh
128 - examples/msg/cloud/simple_vm.tesh
129 - examples/java/cloud/cloud.tesh
130 - examples/java/cloud/migration/migration.tesh
131 * Add java surf examples:
132 - examples/java/surfPlugin/surf_plugin.tesh
133 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
134 - examples/java/surfCpuModel/surf_cpu_model.tesh
135 * Add SMPI+MSG example:
136 - examples/smpi/smpi_msg_masterslave/
141 - msg task destroy cancel test
142 - msg_host on/off test
143 * Move all tests in testsuite to teshsuite (adding tesh files)
144 * Restructure teshsuites
145 - one folder for each kind of test
146 * Restructure AddTests.cmake
148 - structure the order of tests (with sections)
151 * Add virtual machine
152 - creation of a VM on a PM
153 - migration of a VM from a PM to another PM
155 - MSG_process_join(msg_process_t process, double timeout)
156 - msg_bar_t MSG_barrier_init(unsigned int count)
157 - int MSG_barrier_wait(msg_bar_t barrier)
158 - void MSG_barrier_destroy(msg_bar_t barrier)
159 - msg_as_t MSG_environment_as_get_by_name(const char * name)
160 * New option "msg/debug_multiple_use" to help debugging when a task is used
163 - Improvements and finalization of MSG_storage, MSG_file APIs and their
165 - Increase code coverage in test suites
169 * Protect context stacks against stack overflow. The number of protected memory
170 pages allocated on the top of each stack (1 by default) can be configured
171 with the new command line option --cfg=contexts/guard_size:<value>.
172 * Simcalls are now generated by a python script that
173 - generates files included by SimGrid
174 - checks that all the functions exist, and proposes prototypes
176 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
177 host_set_data, host_get_data
179 - simcall_process_join(smx_process_t process, double timeout)
180 * Fix bug where sleeping processing could not be suspended.
183 * Translate surf models from C to C++
184 - Generic classes for all models: Model, Resource, Action
185 - A generic interface for each kind of model (CPU, Network, Storage
186 Workstation, WorkstationVM)
188 * Translate surf routings from C to C++
189 * Add callbacks using sigc++ or boost::signals2
190 - Add callback functions for resource creation/destruction
191 - Add callback functions for action state change
192 - Handle Energy as a plugin
193 * Replace swag by boost::intrusive
194 * Add new routing models for clusters. For documentation, see
195 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
196 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
197 parameters for cluster tag
198 - Fat trees, with topology="FAT_TREE" and
199 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
200 - see examples/platforms/torus_cluster.xml and
201 examples/platforms/fat_tree_cluster.xml
205 * Hostfiles support host:nb_processes construct to deploy several processes on
207 * Collective communication algorithms should not crash if used with
208 improper number of nodes and report the error.
209 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
210 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
211 MPI_Dims_create, MPI_Cart_sub are supported.
212 * New interface to use SMPI programmatically (still depends on MSG for
213 some parts, see examples/smpi/smpi_msg_masterslave) :
214 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
218 * Global variables privatization in MPI executables is now performed at runtime
219 with the option smpi/privatize_global_variables (default:no).
220 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
221 dynamic libraries loaded by the application are not privatized (static
222 linking with these libraries is advised in this case)
225 * Options defined in XML work correctly now.
228 * New cmake option, enable_lib_in_jar, to control whether native libraries are
229 copied into simgrid.jar or not (ON by default). Use this option if you want
230 to reduce the size of the installed simgrid.jar, *and* the native libraries
231 are kept installed elsewhere.
232 * Surf binding with SWIG (code generated in maintainer mode only):
233 - plugin to handle callbacks
234 - CPU model only for the moment
237 * Supernovae build mode is definitively removed. It was used to improve
238 inlining and inter-module optimizations. It is nowadays superseded by
239 link-time optimizations commonly available in compilers.
240 * Update ns-3 find lib. Bindings for ns-3 should work again now.
241 * Add boost dependency for surf++
242 * Add new macro for tests
243 - ADD_TESH(name <tesh_args>)
244 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
248 - xbt_bar_t XBT_barrier_init(unsigned int count)
249 - int XBT_barrier_wait(xbt_bar_t barrier)
250 - void XBT_barrier_destroy(xbt_bar_t barrier)
251 * Make the xbt_os_time module public
253 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
255 SimGrid (3.10) stable; urgency=low
257 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
260 * Reintegrate Java to the main archive as desynchronizing these
261 package is not acceptable anymore (Java is now considered stable)
262 * Add explicit synchronization facilities through semaphores
263 * Bug fix: Task.setDataSize() only changed the C world, not the value
264 cached in the Java world
267 * Dramatically change the way files are handled. API and internals changed, but
268 this part of MSG was not considered as production grade either.
269 * Add explicit synchronization facilities through semaphores
270 * Add a new function MSG_host_get_process_list()
271 * Preliminary DVFS support (see examples/msg/energy/ for details)
274 * SMPI is now included directly in the libsimgrid as the windows
275 linker doesn't force us on splitting it anymore.
276 * Improvements of the SMPI replay tool:
277 - Most of the collective communications are now rooted in the same process as
278 in the original application.
279 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
280 used until now). Multiple data types can now be used in a trace.
281 - The replay tool now supports traces produce either by TAU or a modified
283 - Bug Fix: the compute part of the reduce action is now taken into account.
284 - Gatherv collective is now supported
285 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
286 of smpirun outputs time independent traces for the current run. One file
287 is created per process. If too many processes are simulated, this behavior
288 can be changed to one file for all processes by using the
289 tracing/smpi/format/ti_one_file flag
290 * smpirun generates the host file if needed (with given host count and platform)
291 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
292 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
293 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
294 --cfg=smpi/coll_selector:(mpich/ompi)
295 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
297 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
298 happening inside a collective SMPI call.
299 * Fix the behavior of complex data types handling.
300 * Make MPI_Wtime another synchronization point to take computations into
302 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
303 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
304 * Add all missing Fortran bindings, SMPI should work with Fortran 90
305 (no privatization of global variables yet)
306 * Preliminary DVFS support (see examples/smpi/energy/ for details)
309 * Verification of liveness properties is now available for SMPI applications
310 (in addition to MSG applications)
311 * Bugged examples using SMPI in examples/smpi/mc/
312 * Add --cfg=model-check/visited option. Allows the verification of infinite
313 programs. Detection of loops in the execution thanks to the system state
314 comparison and reduction of the state space to explore. Can be combined with
315 DPOR for safety properties.
318 * Allow to change SimGrid configuration (see --help) within the code
319 thanks to SD_config() as it can be done in MSG.
320 * Add a new function SD_task_set_amount() upon user request.
323 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
324 * Remove rule based routing (no more PCRE dependency)
325 * Add a limiter_link option to cluster tag, to specify a maximum reachable
326 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
327 * Add a loopback_bw and loopback_lat options to cluster tag.
328 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
329 routing. Now peers are encapsulated in an AS and have their own private
330 router but this is transparent.
333 * Our own implementation of getline is renamed xbt_getline, and gets
334 used even if the OS provide a getline(). This should reduce the
335 configuration complexity by using the same code on all platforms.
336 * New type: xbt_cfg_elm_boolean.
337 * Allow to use yes/no for boolean configuration options in the command line.
338 * Allow to disable SimGrid cleanups at exit from command line option.
339 There are situations where one may want a simulation to end with an exit.
340 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
341 annoying when scripting around the simulator. Adding a
342 --cfg=clean_atexit:no allows to circumvent this issue.
345 * Lots of memory leaks were corrected in this release.
346 There is no known memory leaks anymore, in all of our 600+ tests.
347 * New command line option --version, to get SimGrid version information.
348 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
350 * Supernovae builds are deprecated, and expected to be removed in the next
353 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
355 SimGrid (3.9) stable; urgency=low
357 The Grasgory release: GRAS is really dead now.
359 * Complete overhaul of the internal host structures scheme.
362 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
363 was considered as experimental and badly maintained since then.
364 * Keeping it was thus a trap to our potential users, that could take
365 it instead of MSG or SMPI by mistake despite is pity state.
366 * GRAS seems to have very few users (if any), and no one volunteered
367 to maintain it further. It also induces a lot of XBT code (for
368 portability sake), that must be maintained too.
369 * For all these reasons, we killed GRAS. If someone wants to revive it
370 in the future, don't cry, our git history still remembers of GRAS.
373 * Major overhaul. Merge our documentation again as time proved that
374 splitting it was really not helping our users.
375 * Further improve the developer documentation to help newcomers
376 hacking on SimGrid itself. The user documentation (and in
377 particular, the beginner documentation) is still in a sorry state.
380 * Now works on Windows too!
381 * Much more extensive test suite, from MPICH
384 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
385 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
386 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
387 * Bug fix: let task be scheduled when the last dependency to be solved is
388 a control dependency.
389 * Remove SD_load_environment_script function.
390 Use the C sg_platf function if you want to declare a platform
394 * New function: MSG_process_get_number()
395 * Old function documented: MSG_config()
396 * Remove MSG_load_platform_script function
397 Use the C sg_platf function if you want to declare a platform
401 * Change the default value of the TCP_gamma constant (maximal size of TCP
402 congestion window) to a more realistic 4MiB value. If you notice changes in
403 your simulation results, you can fall back to the previous 20k tiny window
404 by adding --cfg=network/TCP_gamma:20000 on command line.
405 * (Hopefully) fix a bug wrt periodic availability and state traces
406 * Bug fix: use default values at start when first event in availability/state
407 trace is not at time 0.
410 * remove the "new_" part of function name sg_platf_new_trace_connect
411 (resulting in sg_platf_trace_connect), since it does not create
415 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
416 use case where it's really mandatory, and maintaining it was a pain in
418 * New: xbt_fifo_search(), search an item with a user-provided
419 comparison function instead of dumb pointer comparison.
422 * Fix the lua deployment:
423 Use `simgrid.init_application()` before deployment instead of
424 `simgrid.msg_register_application()` after.
427 * Transfer the tracing files into the corresponding modules.
429 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
431 SimGrid (3.8.1) stable; urgency=low
433 The "we are told that some people want to also *install* the simgrid
436 * Add missing file "tesh.1" to the archive.
438 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
440 SimGrid (3.8) stable; urgency=low
442 The Psssshiiiit release: SimGrid jumps into the Cloud.
445 * Add an experimental interface to manipulate VMs. They are mainly
446 process groups with very few intrinsic semantic, but they should
447 allow you to build the semantic you want easily.
448 * New function: MSG_host_set_property_value()
449 * New function: MSG_process_on_exit(). To clean memory in all cases.
450 * Bug fixes that made the host (and link) failures unusable.
451 * Add a way to auto-restart process when the host in which they are
452 executing comes back (ON_FAILURE="RESTART" on deployment file,
453 MSG_process_auto_restart_set).
454 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
455 please stop using the old ones, they are DEPRECATED.
457 * Deprecate functions MSG_global_init() / MSG_global_init_args()
458 Please use MSG_init() instead. (reducing the amount of entry
459 points in the library helps us).
460 * Make it impossible to link against the wrong version of the lib
461 * Deprecate MSG_clean(). No need to call it anymore.
462 * Function MSG_get_host_number() is not deprecated anymore.
465 * Split the doc into a user guide and a reference guide.
466 * Start a developper guide to help people hacking on SimGrid.
469 * Enable tracing by default. This modules rocks you should use it.
470 * Remove option custom_flags. Now use environment variables CFLAGS
472 * Use default cmake things to detect lua instead of home grown ones.
473 * New option "enable_mallocators" to disable mallocators, for debugging
474 purpose ("on" by default).
477 * Bug fixes around the resource failures: don't let the processes
478 survive the host they are running onto.
479 * Add an interface to auto-restart processes when the host in which they are
480 executing comes back.
481 * Ensures that SIMIX_clean is called automatically. It's not part of
482 the public interface anymore (bindings should be updated).
485 * Bug fix for when SD_Simulate is called with a positive value: be careful
486 when comparing doubles. Sometimes they are different for non significant
488 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
489 parallel task whose initial work is distributed among host according
490 to the Amdahl's law. Such tasks are created with a parameter alpha
491 that corresponds to the non-parallelizable part of the computation.
492 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
493 between two sets of workstations assuming a 1D block distribution (each
494 workstation owns a similar share of data) on both sides.
496 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
497 Data redistribution will be automatically scheduled once parent and child
498 are both scheduled. The filling of computation_amount and
499 communication_amount structures is now done seamlessly thanks to the chosen
501 * New function SD_workstation_dump to display various information
502 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
503 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
504 bandwidth on the route onto which the task is scheduled and the amount of
506 To divide the nominal bandwidth by 2, the rate then has to be :
507 rate = bandwidth/(2*amount)
508 * Compute tasks that have failed can now be rescheduled and executed again
509 (from their beginning)
510 * Increasing source code coverage (src/simdag is now covered at 95.8%
514 * Re-implement time-independent trace replay using SMPI (at the
515 smpi_smp_* level) instead of MSG. This should replace
516 examples/msg/actions/actions.c
517 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
518 hindexed and structs)
519 * Implement the exchange of non-contiguous data.
520 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
521 * Correct behavior of smpi/sender_gap and set its default value to 0
522 * Add option to asynchronously send small messages to allow better
523 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
524 specifies the size in bytes under which messages will be asynchronously sent.
525 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
526 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
529 * New C interface to define a platform: XML is now optional.
530 For more info, please check include/simgrid/platf.h.
531 * New interface to define random platforms from the C:
532 For more info, please check include/simgrid/platf_generator.h and
533 examples/msg/masterslave/masterslave_platfgen.c
534 * Export a sg_cmdline dynar containing all the arguments we got from
538 * Two new tracing options for adding comments to trace file so you
539 can track your experiments (see --help-tracing for details).
540 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
541 * Adding the SimGrid version that generated the trace file as a comment.
542 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
543 * Fix to avoid key clashes on Paje links
544 * Other minor fixes related to the Paje specification
547 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
548 and renamed to xbt_str_hash() and xbt_str_hash_ext().
549 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
550 * Greatly improve the robustness of mmalloc to user errors (such as
551 using an area after freeing it, or freeing it twice)
553 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
555 SimGrid (3.7.1) stable; urgency=low
558 * Restore the prototype of MSG_process_create_with_environment() to
559 the pre-3.7 situation by removing the kill_time argument.
560 * Add a MSG_process_set_kill_time() function instead.
563 * Fix weird behaviors when dealing with parallel tasks.
566 * Simgrid is now built as a dll.
567 * Simgrid-java now works on Windows.
568 * Simgrid-Java is now included into Windows package.
571 * First pre-build package for MacOSX.
574 * Fix compilation when using MSG_USE_DEPRECATED.
575 * Fix some compilation issues on Macs and Windows.
576 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
577 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
579 We discovered that the Lua console is broken, but we are missing the
580 manpower to fix it right now. The problem existed in 3.7 too, so we
581 are not blocking the release for that. Sorry if you depended on this
582 feature, any help would be really welcome.
584 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
586 SimGrid (3.7) stable; urgency=low
588 The "spring cleanups (before the next Big Project kicks in)" release.
591 * We can specify the SMPI latency/bandwidth factor with command line
592 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
593 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
594 You can also use the "config tag" from platform file by adding this line
595 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
596 (see "example/platforms/tag_config.xml" to use "config tag").
597 Note that the command line supersedes the platform file configuration.
598 * Change the correction factors used in LMM model, according to
599 the latest experiments described in INRIA RR-7821.
600 Accuracy should be improved this way.
601 * Use the partial invalidation optimization by default for the
602 network too. Should produce the exact same results, only faster.
603 * Major cleanup in surf to merge models and split some optimization
604 mechanisms from the core of the models. As a result you can now
605 specify which model to use (e.g., --cfg=network/model:LV08
606 --cfg=cpu/model:Cas01) and which optimization mode to use
607 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
608 Incompatible combinations should err at initialization. See
609 --help-models for the list of all models and optimization modes.
610 * The CLM03 workstation model was dropped for simplicity because it
611 used the deprecated CM02 network model. Use default instead.
612 * Rename the TCP_gamma configuration option to network/TCP_gamma
613 * Rename the coordinates configuration option to
614 network/coordinates, and document it
615 * Use now crosstraffic keyword instead of the terribly misleading
616 fullduplex keyword. It is activated by default now in the current
617 default model, use --cfg=network/crosstraffic:0 to turn it off.
618 * Ongoing refactoring the model parsing to make XML files optional
619 See include/simgrid/platf.h for details (still to be completed)
622 * Major overhaul of the documentation. Almost instructive now :/
623 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
624 functions and friends. This interface was considered as
625 deprecated since over 2 years, it's time to inform our users that it is.
626 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
627 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
628 use these (crappy) functions in your code.
629 These functions will be removed soon. Stop using them now.
630 * Deprecate MSG_get_host_{table,number}
631 Implement MSG_hosts_as_dynar() instead.
632 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
633 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
636 * Stabilize the parallel execution mode of user contexts
637 * Introduce configuration variables to control parallel execution:
638 - contexts/synchro: Synchronization mode to use when running
639 contexts in parallel (either futex, posix or busy_wait)
640 - contexts/parallel_threshold: Minimal number of user contexts
641 that must be part of a scheduling round to switch to parallel
642 execution mode (raw contexts only)
643 * Fix bugs that prevented to use suspend/resume along with
644 synchronization structures.
645 * Fix bugs in process termination that lead to invalid memory access
646 in very specific conditions.
649 * Introduce a parallel mode for the models (controlled by surf/nthreads
650 configuration item). In our tests, running the models in parallel
651 never lead to any speedups because they are so fast that the gain
652 of computing each model in parallel does not amortizes the
653 synchronization costs, even when ultra fast futexes are used.
654 This is released anyway because YMMV.
657 * Performance boost by using a swag internally to compute the set of
658 tasks that are finished and should constitute the return value of
662 * Enable it by default now that it is considered rather stable.
665 * Documentation of the tracing functions.
666 * Performance gains when tracing categorized/uncategorized resource
667 utilization by avoiding calls to get route when updating resource
668 variables. LMM constraints are being used instead.
669 * API changed to set task categories. Use MSG_task_set_category instead
670 of TRACE_msg_set_task_category, and SD_task_set_category instead
671 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
672 * Bugfix for graphicator, routes not correctly obtained, memory leaks
673 * Examples for link user variables added (see at examples/msg/tracing/)
674 * Deprecated function TRACE_msg_set_process_category completely removed
675 * Trace header updated according to the latest Paje file format
676 * Tracing network lazy updates, no longer obligate users to use full updates
677 * --cfg=tracing/platform:1 also registers power/bandwidth variables
678 * Experimental: let user code declare/set/push/pop application states for hosts
679 * API changed to allow the manual creation of graph configuration files
680 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
683 * Improve the API of Lua MSG bindings, using the Lua spirit.
684 * Each simulated process now lives in its own Lua world (globals are
685 automatically duplicated). It helps writing simulators. Will allow
686 to run Splay programs within SimGrid in the future.
687 * Add a Chord example in Lua, equivalent to the MSG one.
690 * Start the implementation of a solution to express temporal
691 properties, not only local assertions. This is still an
692 experimental work in progress, stay clear from it to be safe.
696 - Add new runtime parameters --help-logs and --help-log-categories
697 to display informations about supported logging parameters and
699 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
700 don't exists anymore.
701 * Mallocators: allow value NULL for the reset function.
703 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
704 create homogeneous dictionaries, where all the elements share the
705 same free function. Non homogeneous dictionaries will be
706 deprecated in the next release.
707 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
708 - Multi-level dictionaries are deprecated.
710 - new function xbt_dynar_search_or_negative() that is useful when
711 you have less than 2 million elements in your dynar and don't
712 want of the extra complexity of catching exceptions when the
713 element is not found.
715 - Make xbt_os_thread module (for thread portability) public.
716 Documentation is still to come, sorry.
718 - Cleanups and simplifications to make it maintainable again.
719 - Exotic features (such as memalign and valloc) were removed.
720 - The metadata were extended and improved so that the
721 model-checker becomes able to explore and inspect the heaps.
722 - This may induce a performance drop when enable_model-checking is
723 ON in cmake (even if it's not used in the simulation), but it is
724 necessary at this point to get MC working.
726 Turn model-checking OFF if simulation performance matters to you.
727 Not enabling it at runtime is not enough, disable it in cmake.
729 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
731 SimGrid (3.6.2) stable; urgency=low
733 The "Not coding new stuff allows to polish old things" release.
736 * New bindings to the NS3 packet level simulator (experimental)
737 * Use the raw (efficient) execution contexts instead of the sysv
738 (portable) ones when possible.
739 * libpcre is now mandatory in any cases since not using it led to
740 severe performance loss and possibly other issues
741 * Update the XML platforms:
742 - G5K: include the latest machine in Nancy
743 - GridPP and LCG: new platforms
744 * Documentation was partially updated, at least (more to come)
746 Bug fixes, cosmetics and small improvements
747 * Free terminated processes before the end of the simulation to avoid
748 exhausting the memory of users having very dynamic amount of
750 * Bug fix and cosmetics about canceling non-running tasks
751 * Bug fix about the dot loader's issues when using libcgraph
754 * Create an installer for windows with nsis (amd64 and win32)
755 - Add an hello world project to illustrate simgrid project creation.
756 - Embed libpcre into the Simgrid installer to avoid
757 its compilation burden
758 * The raw execution contexts should work on Apple now
759 * Port to Windows 64 bits
760 - Sysv contexts now have an implementation for this arch
761 - GRAS communication features now support this arch
762 * Drop support for borland compiler on windows
763 - this code was not maintained, and we kinda depend on gcc nowadays
764 * Fix portability issues on kfreebsd/gnu: build error about semaphores
765 * Fix portability issue on unstable ubuntu: linker became picky on
768 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
770 SimGrid (3.6.1) stable; urgency=low
772 The "Oops, we broke Macs too" release
775 * Fixed contexts detection so that raw ones are used when possible
776 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
777 leads to a strange error, with user code segfaulting sometimes when
778 the generated code is not perfectly aligned (which is not
779 controllable from the user side, depends on the amount of code)
782 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
784 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
786 SimGrid (3.6) stable; urgency=medium
788 The Summer Release, also known as the "OMG! They Killed Kenny!" version
791 * Bindings now constitute their own package, separated from the main one.
792 Rationale: reduce our maintenance nightmare by reducing the module coupling
793 They will soon be released on their own on gforge.
795 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
796 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
798 GRAS: It is not considered as stable anymore, but experimental. Sorry.
799 * It's not quite deprecated for now because we have no replacement,
800 but it may soon become the case.
803 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
804 * Fortran: New user-level cache variable to store the rank of the running
805 process. This improves performance by an order of magnitude.
806 * C: New coccinelle script to automatically locate and modify global and
807 local static variables.
808 * Improved SMPI network model with a sender-side gap to account for multiple
812 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
813 only say if a communication is finished, no matter whether it succeeded or
814 failed. You can call MSG_comm_get_status() to know the status of a finished
816 * New function MSG_task_dsend() to send a task and detach it. When a
817 communication is detached, you are never notified of its success or failure
818 and the memory is released automatically once it is finished. This function
819 is useful when you don't care about the end nor the success of a
821 * Change the prototypes of action replay. Sorry for inconvenience,
822 but this is really more efficient this way (and to adapt your code,
823 you just have to fix the initialization, that shouldn't be too long)
824 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
825 ever managed to do anything useful with it.
826 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
827 takes a dynar of communications. It returns immediately and gives the
828 index of a finished communication (if any).
829 * New example: a basic implementation of the Chord P2P algorithm.
832 * New model for multi-core CPUs. You can now use the core attribute to
833 precise the number of cores of a host. This is a basic model. Every
834 process running on the host receives at most the power provided in
835 the DTD (throughput<=power). Total throughput of process cannot exceed
837 * New peer tag. This peer tag creates a tiny AS comprising a host and a
838 router linked by an up-link and a down-link (possibly asymmetrical).
839 This kind of pattern allows to easily build last-mile model style platforms.
840 Aggregating such patterns in a rule-based AS is thus the technique of
841 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
842 * New model for Vivaldi routing. We transformed the Vivaldi network model
843 into a Vivaldi routing model (based on the rule-based model). This allows to
844 combine Vivaldi based latencies with last-mile platforms.
847 * Added a check for NaN of IEEE754 infinite in the double entries of
849 * Introduce a new context factory "raw", highly inspirated from the
850 ucontext factory, but using manually crafted functions in assembly to
851 do the work in an efficient manner.
852 * Allow to change the used context factory at run time, not only at
853 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
854 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
855 contexts at runtime (only with raw contexts or ucontexts).
856 * Completely rewrote this module to allow parallel execution of user
857 processes. Use --cfg=contexts/nthreads:N to execute user processes
858 with N parallel threads (the default is 1, meaning no parallelism).
859 * Allow to decide dynamically between sequential and parallel modes.
860 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
861 processes in parallel only when their number is greater than or equal to P
863 * Added a check for NaN of IEEE754 infinite in the double entries of
867 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
868 won't output the state of processes when interrupted with Ctrl-C
869 * Add a new function xbt_dynar_to_array that transforms a dynar into a
870 NULL-terminated array. This may solve backward compatibility issues
871 due to the change to return type of SD_simulate. See also:
872 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
873 * Add new macros with variable number of arguments.
874 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
875 - in xbt/asserts.h: xbt_assert
876 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
877 - in xbt/ex.h: THROWF and RETHROWF.
878 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
880 * Change xbt_die() to accept a format string with arguments, just like printf.
881 * New data structure: xbt_lib_t, like a dict but more general and with better
885 * New configuration options
886 Options triva/categorized and triva/uncategorized can be used to generate
887 graph configuration files for Triva visualization tool.
888 * Configuration option tracing/platform is renamed to tracing/categorized
889 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
890 * New TRACE_user_link_variable interface:
891 User provides the name of the link and the tracing variable to attach to it
892 * the declaration of tracing categories must be done after the environment
894 * simpler tracing interface, just one way to declare categories
895 TRACE_category or TRACE_category_with_color, it is up to you
896 * links in the trace file are again identified by their names
897 * trace contains the full platform hierarchy exactly as declared using the ASes
898 * Options tracing/msg/[task|process]:1 groups the process by hosts
899 for both cases, tasks and processes must have names that are unique during the simulation
900 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
901 * The experimental option tracing/msg/volume is deprecated
902 its functionality may be reincorporated if needed
904 The tracing generates a trace file with unordered timestamped events,
905 because of the way the core simulator (surf) works. A script available
906 at the tools directory (fix-paje-trace.sh) can be used to put the events
907 in order. We have changed the tracing so it can generate ordered timestamped
908 events in the final trace, but depending on the simulator (and how much time
909 is simulated) that can lead to a huge memory utilization. It is deactivated
910 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
913 * Define a SIMGRID_VERSION macro in simgrid_config.h.
914 - We are trying hard to keep the API stable, but it may happen that
915 some things change (we're a research project after all, not a
916 nuclear plant operating system). If such things should happen, you
917 could rely on that macro to adapt.
918 - current value: 30600 for 3.06.00, aka 3.6
919 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
920 a number that can be compared with SIMGRID_VERSION.
921 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
922 assertions and verbose/debug logging events are disabled at compile time.
924 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
926 SimGrid (3.5) stable; urgency=medium
929 * New feature to any SimGrid-based simulator: Model-Checking
930 Check SIN#1 for more details.
933 * New Model SMPI (three-interval linear regression for correction factors)
934 See RR-7426, available at http://hal.inria.fr/inria-00527150
935 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
936 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
937 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
938 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
939 * See also src/smpi/README
944 - Tracing API changes: TRACE_start and TRACE_end should not be called
945 by user-code. They are automatically called by simulators created
946 with SimDAG, MSG and SMPI if the toolkit is compiled with
947 tracing_enabled=ON. Categories declaration and utilization remain the
948 same for MSG and SimDag.
949 - A function was added to the tracing API to declare categories with
951 - TRACE_category_with_color (char *category, char *color)
952 where color must be in the following format
953 "%f %f %f", red, green, blue
954 and red, green, blue are float values in the interval [0, 1]
955 - User can specify NULL as color parameter, or continue calling
957 On that case, the tracing system will define random colors
958 - The following command-line options are supported:
959 --cfg=tracing/filename:msg.trace
960 --cfg=tracing:1 (activate tracing, needed to use others)
961 --cfg=tracing/platform:1 (categorized resource use)
962 --cfg=tracing/uncategorized:1 (uncategorized resource use)
963 --cfg=tracing/msg/task:1 (task creation)
964 --cfg=tracing/msg/process:1 (process creation, migration)
965 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
966 --cfg=tracing/smpi:1 (SMPI interface tracing)
967 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
968 - examples of examples/msg/tracing updated
970 - DAXLoader and DOTLoader functions can generate tasks with categories
971 - A new function to attribute a category to SD tasks:
972 TRACE_sd_set_task_category (SD_task_t task, char *category)
973 * Tracing the MPI interface implemented by SMPI
974 - Collective operations are traced with states
975 - Point-to-Point operations are traced with states/links
976 - Tracing activated by a parameter "-trace filename" passed
977 to smpirun during execution (considering that simgrid
978 is compiled with tracing enabled)
979 - To run the simulation with gdb, the simulator
980 accepts --cfg=tracing/smpi:1 to trace SMPI
981 - tesh files to check if smpi tracing is ok
982 - See examples/smpi/NAS/DT-trace
983 * GTNetS tracing re-worked
984 - adaptation to the tracing system of GTNets to cope
985 with modifications regarding the fullduplex mode
986 - new tesh files to check if gtnets tracing is ok
989 * Asynchronous communications through the functions:
990 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
991 * New function: MSG_load_platform_script()
992 to make possible using a lua script instead of XML files to set up platforms
993 * New function: MSG_set_function
994 to associate functions to processes, used when bypassing the parser
995 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
997 Platforms: Add some more examples in examples/platforms
998 * Grid'5000: see www.grid5000.fr
999 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1002 * Change the XML format. This is a very important modification. SimGrid
1003 3.5 introduces a new hierarchical format based on the notion of
1004 Autonomous Systems. Compatibility with old format is ensured
1005 through the perl script provided in the install bin directory
1006 bin/simgrid_update_xml.
1007 It is now possible to build platforms with specific routing
1008 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1009 connect several platforms together. We will try to provide soon
1010 set of realistic platforms exploiting these properties (have a
1011 look at examples/platforms/ for the moment).
1012 * Take the opportunity of the XML format change to be a good XML citizen:
1013 rename link:ctn to link_ctn and similar changes (also dealt with by
1015 * Add a new routing scheme (rule-based) using regular expressions. It
1016 enables to have an extremely low memory footprint when the
1017 underlying routing is simple and can be compactly described. You
1018 need to have libpcre4-dev (perl regular expressions) installed if
1019 you want to use this routing scheme.
1020 * Revive the cluster TAG and allow to easily and efficiently (both in
1021 term of memory and speed) connect clusters together. Have a look
1022 at teshsuite/simdag/platforms/ to see how this can be done. With
1023 this tag, you can create clusters with thousands of tasks at no
1024 cost (have a look at examples/platforms/).
1025 Note: clusters are implemented as ASes, so there is no need for an
1026 enclosing AS tag if you have only one cluster in your platform.
1027 * Add new generic functions in the public interface that allows the user
1028 to call SURF 'create_resource' methods from your code (same
1029 functionality as the XML bypass mechanism but with a much lighter
1031 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1032 that uses a piecewise linear approximation to produce better
1033 results when exchanging small messages.
1034 * Add a new parameter to handle correctly full duplex link and account
1035 for interferences between uplink and downlink communications
1036 (activate with --cfg=fullduplex:1).
1039 * Rename the SD_READY (all dependencies are satisfied and task is
1040 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1041 dependencies are satisfied) state.
1042 This prevents a confusion between the notion of "ready to schedule"
1043 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1044 simulated" (SD_RUNNABLE) used by the simulation kernel.
1045 * Change the way a task is considered as ready. Instead of removing
1046 dependencies when a task is done, a counter is decreased. This way,
1047 it is always possible to reach ancestors thanks to the
1048 SD_taks_get_parents function (even after the end of the simulation.)
1049 * Change the return type of SD_Simulate from (SD_task_t*) into
1050 xbt_dynar_t. This function was in handling a dynar internally and
1051 converted it into a NULL terminated array for historical reasons.
1052 * New function SD_dotload(char*) to load a DAG described in dot
1053 format. This loader and the corresponding examples require the
1054 installation of the graphviz library.
1055 * Fix a bug in the management of tasks of size 0 in the surf network
1056 models. This problem was only visible with SIMDAG and you should
1057 thus disregard results produced with earlier versions if you
1058 relied on this feature (some tasks were blocked because of this).
1059 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1060 even though your applications were doing only point-to-point
1061 communications and sequential computations. Now you can really use any
1062 model you want (of course, if you create real parallel tasks, which are
1063 not implemented in most models beside ptaskL07, this will abort).
1064 * Add an example that schedules a DAX on an heterogeneous platform
1065 using a Min-Min strategy.
1066 * New function SD_workstation_get_current_task() that returns the kind
1067 of task currently running on a workstation in the sequential access
1069 * Raise some warnings when unexecuted tasks remains at the end of the
1070 simulation. This is usually caused by cycles in the DAG.
1073 * New function: SIMIX_process_set_function() called by MSG_set_function
1074 * Change the underlying waiting queue in semaphores so that a process
1075 can wait on several of them simultaneously (as in waitany).
1076 * Fix the way to handle tokens in semaphores so that all access patterns
1077 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1078 * kill the dirty pimple SIMIX_message_sizes_output()
1079 Please use (proper) visualization instead
1082 * New data container: setset (set of sets of elements)
1083 * New module: mmalloc (mapped malloc, allowing to have several
1084 independent segments of malloc)
1085 * New function: xbt_dict_cursor_set_data()
1086 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1087 * New function: xbt_dynar_is_empty()
1088 * New function: xbt_fifo_get_last_item()
1089 * Fix xbt_dynar_shrink(): use the right element size.
1090 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1091 uninitialized areas during expand.
1092 * Fix semaphores: previous implementation was severely broken.
1093 * Use library init/fini functions for our initialization.
1094 - you can use logs and other feature as soon as you want in your
1095 code (even before the xbt_init / MSG_init)
1096 - xbt_exit is now a no-op and produce a warning when used.
1099 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1100 benefit from the latest improvement to the simulation kernel.
1101 * Kill measurement sockets for now. If you rely on them, sorry. This
1102 release is not for you. This feature will be reintroduced in the
1103 future, but we cannot delay the release any further.
1104 * New function: gras_msgtype_get_name().
1105 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1106 * Fix (at last) the pmm example: it should not randomly fail anymore.
1108 Build chain: bug fixes and overall polishing
1109 * Cmake is now stable enough. Hence, we killed the autotools.
1110 * Port to windows ( TM :)
1111 * Fix the 'make install' target.
1112 No need to use 'make install-simgrid' anymore
1113 * Introduce a 'make dist' target compiling a *source* archive
1114 'make package' compiles a binary archive
1115 * Compile java files only on need
1116 * Add --cd and --setenv command line options to tesh
1117 * Out of source builds are not fully supported yet, but we are close
1118 * Enable supernovae and optimization flags by default for our users
1121 * Add layer to set up environment directly from lua, without XML.
1122 * The effect of gras_stub_generator can be achieved through
1123 lua too (check examples/gras/console/ping_generator.lua)
1125 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1127 SimGrid (3.4.1) stable; urgency=low
1129 The "Polishing easter eggs is probably a good idea" release.
1130 This is a bug fixes release only.
1134 * Fix a bug preventing the tasks from begin garbage collected.
1137 * Fix a bug occuring when a host involved in a communication fails.
1138 This was not detected properly by the other peer involved in the
1139 communication. Now, it's reported as a network error.
1142 * Warn the user about loop dependencies in data flow of DAX files
1143 * Obey the control-flow dependencies of DAX files
1146 * Add option "enable_smpi" allowing to not compile SMPI.
1147 Probably useful for the (Mac) users experiencing a build error here
1148 * Improve the detection of lua5.1 and ruby1.8
1150 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1152 SimGrid (3.4) stable; urgency=low
1154 The "Easter in Cargese" release. Also known as (major changes):
1156 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1157 ~> bindings were greatly improved
1158 ~> new tracing infrastructure for better visualization introduced
1160 * the "Welcome to configury modernity" release.
1161 ~> we switched from autotools to cmake, and improved our cdash
1164 A more detailled list of changes follow (full detail in svn log).
1166 Java Bindings: Various Cleanups
1167 * (install java-gcj-compat-dev on debian-like to use them)
1168 * Remove put/get: no need to export deprecated interface in Java
1169 Use send/receive instead.
1170 * Cleanup the examples and add a README per directory
1171 * Remove example autoDestination (that's the only way to go now)
1172 * Remove example explicitDestination (was a plain copy of basic)
1173 * Make JniException a runtime exception, so that there is no need to
1174 declare the fact that you may encounter such a beast. I guess that
1175 nobody will ever want to survive such error.
1176 * Create specific errors for each MSG case of failure:
1177 host failure, transfer failure, timeout, task cancelled
1178 * Cleanup the exceptions that may get thrown by each function
1179 * Other internal cleanups in Java bindings. Performance still bad :/
1180 Ruby and Lua Bindings: create them
1181 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1182 * That's new and great, you should try them out.
1183 Same functionalities than Java bindings, only even less polished
1185 * Kill the useless "rate" argument of SD_task_get_execution_time()
1186 Everyone used to provide -1 as a value, it was not used, and the
1187 semantic of a possible use wasn't even clear.
1188 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1189 either as comm costs or compute costs to mean that there is no
1190 such thing for that specific task.
1191 * Add a SD_task_set_name() function
1192 * Fix SD_task_unschedule() on typed tasks
1193 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1194 * In DAX loader, accept useless 'level' attributes to <job> since
1195 LIGO DAGs have them (seem to be to ease graphical representation).
1197 * Add an example masterslave_mailbox.c using send/receive and not
1198 the deprecated put/get interface.
1199 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1200 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1202 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1203 (declare MSG_USE_DEPRECATED to still have the old name)
1204 * Add a MSG_task_set_data() function
1205 * About trace replay (see examples/msg/actions):
1207 - Allow to work with splitted trace files for each process
1208 Give the specific trace file as argument of each process,
1209 and call MSG_action_trace_run(NULL)
1210 You can still have one merged file for all processes.
1211 - Fix implementation of collective operations
1212 * Allow task_execute() on 0-sized tasks (closes #10063)
1214 * This is the first release of SimGrid where SMPI is not considered
1215 beta anymore (even if some corners should still be improved)
1216 * Port over the new SIMIX_network submodule (internal refactoring)
1217 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1218 * Implement more missing elements of the standard:
1220 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1222 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1223 MPI_Scan() MPI_Get_processor_name()
1224 - Added implementation of missing case for Alltoall (warning: it's
1225 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1226 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1227 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1228 SMPI_MPI_Allgatherv()
1229 * Bug fixes include:
1230 - MPI_Waitsome() was broken
1231 - Allow relative includes in smpicc
1232 - Command line cfg argument 'reference_speed' was ignored...
1233 - Some functions did not properly lead to auto-benching of user code
1234 - smpicc passes -O2 by default (just like openmpi one)
1236 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1237 * Bug fixes about timeouts during communications
1238 * add SIMIX_message_sizes_output() as a pimple to write to file the
1239 amount of messages per size. Use gnuplot to get histogram.
1240 Pimple because that's the only user-visible function of simix,
1241 defined directly in xbt.h (irk, sorry)
1243 - Add a SIMIX_sem_get_capacity() function
1244 - Fix interactions with processe resume/suspende
1245 - release_forever() was stupidly broken
1246 - Fix SIMIX_display_process_status() for processes in a semaphore
1247 - Make SIMIX_sem_block_onto() user-visible
1248 * Refactoring context stuff:
1249 - Use pseudo-OOP for better modularity
1250 - reimplement SIMIX_process_kill() without process_schedule() so
1251 that the latter can take as invariant that it is called from
1253 - Merge context_start into context_new for sake of simplicity
1255 * Add a Vivaldi network model, coded live during SUD'10 ;)
1256 * Rename configuration variables to start a hierarchy:
1257 o cpu_model -> cpu/model
1258 o network_model -> network/model
1259 o workstation_model -> workstation/model
1260 * New configuration variables:
1261 o network/bandwidth_factor: correction to bandwith
1262 o network/latency_factor: correction to latency
1263 o netwotk/weight_S: correction to the weight of competing streams
1264 * Add a long description to the models, that users can see with such
1265 argument on the command line: --cfg=cpu/model:help
1266 * --help-models display the long description of all known models
1268 * config: add the ability to set a default value after registration
1269 Does not override any previously set value (e.g. from cmd line)
1270 * dict: allow to have integer key and data.
1271 When so, you need to use the following functions
1272 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1273 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1274 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1275 In contrary to regular dicts, the key is not malloced before copy.
1276 Mixing scalar and regular elements in the same dict is not tested
1278 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1279 Tracing for Visualization:
1280 * SimGrid is now instrumented in order to generate a trace file for
1281 visualization analysis: to use it, need to compile SimGrid with the
1282 "tracing" option enabled, and instrument the program using SimGrid with
1283 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1284 (among other functions).
1285 * The instrumentation only traces the platform utilization for now
1286 * Documentation to use the tracing functions and how to analyze the
1287 traces with the Triva tool is written.
1288 * More information about: SimGrid FAQ (in the section Tracing Simulations
1291 * We moved to cmake as default build system. Autotools support will
1292 be dropped soon. Check the FAQ for more info about how to use it.
1293 * Greatly improved our cdash/ctest interactions
1294 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1295 * Added memory checking tests with valgrind; lot of memleak fixing.
1296 This may be the first release of simgrid with so few memory issues
1297 * Added code coverage tests.
1298 Our coverage is still improvable, but at least we see it on cdash.
1300 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1302 SimGrid (3.3.4) stable; urgency=low
1304 The "Desktop Grid needs love too" release (also called Xmas release).
1306 Models improvements:
1307 * Major speedup in the maxmin system solving by using lazy evaluation
1308 Instead of solving completely the maxmin system at each iteration,
1309 only invalidate (and recompute) the modified parts.
1310 This new feature is enabled in default models but you can try to
1311 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1312 * Cas01 IMproved as default CPU model
1313 This CPU model is the same Cas01 model, but it uses the
1314 maxmin-selective-update flag and a heap structure to manage
1315 actions on SURF kernel.
1316 It reduces the complexity to find the next action to finish and,
1317 consequently, it's faster than the old Cas01.
1318 This is the new default CPU model (Cas01).
1319 * Rename the old Cas01 model to Cas01_fullupdate
1320 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1321 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1322 * CpuTI (CPU Trace Integration)
1323 A new CPU model whose objective is simulate faster when using
1324 availability trace files.
1325 Instead of using a full featured, over engineered maxmin system for
1326 CPU modeling, this model does the pre-integration of traces files
1327 to calculate the amount of CPU power available, and so, executes
1328 faster than the old CPU models.
1329 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1330 * Use LV08 as default network model since it gives better accuracy
1331 for small messages and shouldn't change things for big ones.
1332 Use --cfg=network_model:CM02 to get the previous behavior.
1335 ******************************************
1336 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1337 ******************************************
1338 * The new CPU model may changes simulations!
1339 The point is that events occurring at the exact same timestamp
1340 are not scheduled in the same order with the old and new
1341 version. This may be enough to completely change the execution
1342 of simulations in some cases.
1343 * The new network model will change simulations!
1344 This new model is more realistic than the previous one, so you
1345 should consider redoing your old experiments with this model.
1346 Sorry for the inconvenience.
1349 * Introduce the supernovae compilation mode
1350 When compiled that way, the whole SimGrid (or almost) is put in a
1351 single compilation unit and compiled in one shoot.
1352 This is to help gcc which has difficulties to inline stuff from one
1354 The speedup seem to be above 15%, althrough more tests are needed on
1355 amd64 to confirm that gain.
1358 * Port of MSG's mailbox on top of SIMIX network
1359 The put/get mechanism was greatly simplified on the way.
1362 * New SIMIX network module. Provides:
1363 - Mailbox: rendez-vous mecanism to find with who you want to speak
1364 - Synchronous send/recv: easier and hopefully faster since the
1365 logic is handled in the maestro process directly now
1366 - Asynchronous send/recv: you dreamt of it? It's here now
1367 Too bad that nobody cared enough to propagate the change to MSG.
1368 * Add semaphores as SIMIX synchronization mechanism.
1371 * new function SD_daxload(char*) to load a DAX file
1372 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1373 * Introduce typed tasks. Specify its kind and cost at creation.
1374 At scheduling, just give where it should be placed, and the cost
1375 for each involved resource is automatically computed.
1376 Existing constructors so far (more to come of course):
1377 - SD_task_create_comm_e2e() for end-to-end communication
1378 - SD_task_create_comp_seq() for sequential computation
1379 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1380 * new function SD_task_dump() for debuging display
1381 * new function SD_task_dotty(task,FILE*) writing to file the info
1382 about the task in dotty format
1383 * SD_task_dependency_exists() can now cope with having one of its
1384 arguments NULL. If so, it tests whether the other argument has any
1386 * Add getters on list of preceding/following tasks:
1387 SD_task_get_parents(task) and SD_task_get_children(task)
1388 * Add getters on amount of workstations and list:
1389 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1390 * Add getter on task kind: SD_task_get_kind(task)
1391 * Update the start_time and finish_time of tasks on completion/failure
1392 * Bugfix: Remove task from state swags when destroyed
1395 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1398 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1399 dynar with one separate thread per value of the dynar.
1400 * Change the prototype of xbt_thread_create(), sorry.
1401 Added a boolean parameter indicating whether we want to join this
1402 thread (used in SG only for now)
1403 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1406 * GTNetS wrappers should now be usable again (and betterly tested too)
1407 * Fix a major regression from 3.2 where the timeout provided to
1408 MSG_task_put_with_timeout() was used as absolute time before which
1409 the comm should be done.
1410 * Start to fix the <cluster> tag.
1411 - Internal links should be good now (beside of the loopback, which
1412 use the private link instead)
1413 - paths to the external world is still rather broken
1414 - the <route:multi> tag is just broken. Actually that's brain-dead.
1415 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1416 to make it less stupid
1417 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1418 * Fix a source-level compatibility glitch from 3.2: after defining
1419 MSG_USE_DEPRECATED, you can use the old name
1420 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1421 * Allow to compile from the SVN with automake 1.11
1422 * Fix some problems when using the "start_time" tag in deployment XMLs.
1423 * Fix #8569: XBT/synchro.h has redundant declarations
1424 * Fix #8563: MSG return values and exceptions
1425 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1426 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1427 Thanks to Robson Peixoto.
1428 * Fix a vicious bug in dictionaries inducing that some elements were
1429 not freed on xbt_dict_free()
1431 Portability report of this version:
1432 * Main portability targets:
1433 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1434 on (amd64/i386/ia64)
1435 - mac leopard on i386
1436 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1437 but nothing critical.
1438 * Other platforms: windows, AIX and others were not tested for this release
1440 Timing report of this version:
1441 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1442 scenario parameters). From 8h to a few seconds in desktop grid settings.
1443 * Supernovae brings about 25% speedup on i386.
1445 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1447 SimGrid (3.3.3) stable; urgency=low
1449 The "Need for Speed" release.
1451 The timings done to validate the 3.3.2 were faulty.
1452 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1454 The problem was a conversion from a manually handled vector to
1455 xbt_dynar_t on the critical path.
1456 xbt_dynar_foreach calls functions, inducing stack management crap.
1458 We inlined these functions and xbt_dynar_foreach is now breath taking.
1459 We also inlined xbt_swag_belong on the way.
1461 Here are some approximate speedup measurements (on master/slaves
1462 simulations lasting between 10s and 20s each):
1463 3.3.1 -> 3.3.2: about same performance
1464 3.3.2 -> 3.3.3: 40% speedup
1465 3.3.1 -> 3.3.3: 40% speedup
1466 3.3.1 with inline patch -> 3.3.3: 30% speedup
1468 Our reading is that the refactoring which occurred in 3.3.2 made us
1469 suffer much more from the xbt_dynar_foreach low performance, but
1470 once we solved this, this refactoring proved to be very performance
1471 effective. From the 40% speedup, somehow, 10% are due to the
1472 inlining and 30% to the refactoring.
1474 That's a pitty that gcc cannot inline functions placed in other files
1475 alone. We have to choose between:
1476 - break the encapsulation (by putting private data structures and
1477 accessors in headers files to help gcc)
1478 - live with low performance
1479 - switch to a decent compiler such as icc (not quite possible).
1481 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1483 SimGrid (3.3.2) stable; urgency=low
1485 The "Simplicity does not preceed complexity, but follows it" release.
1487 The main contributors of this release were (lexical order):
1488 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1491 * Extract the routing logic into its own object.
1492 (was dupplicated in network.c and workstation_LV07.c;
1493 Allows to implement other ways of storing that info)
1494 => kill now useless network_card concept
1495 - Use dynar to represent routes (instead of void** + int*)
1496 - kill link_set (use surf_network_model->resource_set instead)
1497 - Add a command-line option to choose the routing schema to use
1498 - Add three new models:
1499 * Floyd (shortest path computed at initialization)
1500 * Dijikstra (shortest path recomputed all the time)
1501 * Cached Dijikstra (shortest path computed on need)
1502 All these models where contributed by Silas De Munck, and are
1503 described in his ICCS09 paper.
1505 * Simplify model declaration
1506 (less redirections, less function to write when defining a model)
1507 - Factorize stuff between models:
1510 surf_model_resource_set(model)
1511 surf_model_resource_by_name(model, name)
1512 - Unify the types of models in s_surf_model_t (using an union)
1513 - Embeed fields of common_public directly into s_surf_model_t
1514 - Rename model methods:
1515 action_free ~> action_unref
1516 action_change_state ~> action_state_set
1517 action_get_state ~> action_state_get
1518 - Change model methods into functions :
1519 (model)->common_public->action_use ~> surf_action_ref
1521 * Implement a generic resource; use it as ancestor to specific ones
1522 (allows to kill duplicated code in models)
1523 Drawback: timer command don't need no name nor properties;
1524 workstation_CLM03 don't need no properties
1525 (but I guess we can live with those few bytes wasted)
1527 * Improve the action object model
1528 - implement a constructor avoiding dupplicated code about field
1529 initialization in generic_action part.
1531 * Kill the SDP model: it has an external dependency, is deprecated
1532 in flavor of modern lmm models, and didn't compile since a while
1535 * Relocation of the context module from XBT to SIMIX.
1536 (the context were decoupled from the simix processes, duplicating a lot of code)
1537 => a lot of code was factorized
1538 - less overhead is introduced during scheduling
1539 - simpler API for the context factory
1540 - the logic for process creation,destruction and manipulation was simplified
1541 * Simplification of the s_smx_process_t data structure.
1542 => accesing the simix level data associated to a process is faster now,
1543 and the code is a lot more readable.
1546 * Implement some more MPI primitives:
1547 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1548 -implementation: Bcast: flat or 2-ary tree (default),
1549 Barrier: 4-ary tree,
1551 Allreduce: Reduce then Bcast
1552 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1553 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1554 Alltoallv: flat tree, like ompi
1556 * Add support for optimized collectives (Bcast is now binomial by default)
1557 * Port smpirun and smpicc to OS X
1560 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1561 Shout out if you used it.
1564 * Display the list of still queued messages in SG mode when existing
1568 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1569 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1570 * Add xbt_str_from_file(FILE*)
1571 * Add xbt_dict_get_key achieving a linear reverse search
1572 * Remove the context module
1574 Portability report of this version:
1575 * Main portability targets:
1576 - Linux(debian)/x86/context
1577 - Linux(debian)/x86/pthreads
1578 - Linux(debian)/amd64/context
1579 - Linux(debian)/amd64/pthreads
1580 On these, we still have the eratic breakages of gras/pmm and
1581 amok/saturate_sg reported in previous version. We still think
1582 that the tests are the cause of the fault, not the tested code.
1584 - Mac OSX Leopard/x86/context
1585 Still false negative in tesh autotesting.
1586 Smpi still fails, but this time because readlink does not accept -f
1587 Everything seems to work properly beside of that.
1590 - AIX version 5.3 (only tested contexts this time)
1591 Smpi still fails there because mktemp is not installed.
1592 Everything seems to work properly beside of that.
1594 I managed to compile it for the first time, but several breakages.
1595 Won't delay the release for this exotic platform.
1597 * Windows: it's still lagging behind. If you want to help, please
1600 Timing report of this version:
1601 This version seem to be more than 5% faster than 3.3.1 (on linux
1602 64bits with contextes). The gain is less than expected, we are
1603 investigating this for next release.
1605 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1607 SimGrid (3.3.1) stable; urgency=low
1610 * Implement a --cfg-help to show existing configuration variables
1611 * Build chain do not require doxygen in maintainer mode
1614 * fix a bug on struct sizeof computation, which prevented the
1615 exchange of arrays of structs in some conditions
1616 - added a regression test about this in datadesc_usage
1617 * Allow the exchange of 0-long dynamic vectors.
1618 - for that, use -1 as indicator of dynamic size instead of 0
1619 - This implied to change any size from unsigned long to long,
1620 reducing a bit communication abilities, but I guess that with
1621 64bits being quite common, this is more than enough.
1622 - This also induce a protocol change, thus bumping network protocol
1623 version from 0 to 1 (if we have external users, we have to get
1624 clean on that point too ;)
1625 - added two regression tests about this in datadesc_usage
1626 * Be more verbose when propagating local exceptions
1627 This helps debugging.
1628 * Display the status of simulated processes when receiving SIGINT in
1632 * Allow to control the simulation from a trace file.
1633 New functions MSG_action_register() and MSG_action_trace_run()
1634 The first one allows to associate a function execution to each
1635 kind of action while the second one parses a trace file and
1636 triggers the corresponding actions within the system.
1637 For now, only a toy example is provided in examples/msg/actions
1638 * Add an exemple of process migration in examples/msg/migration
1639 * Fix a bug in task exchange which broke MSG_task_get_sender()
1640 Add a teshsuite regression test for that.
1641 [Bug: if MSG_task_get_sender() is called after sender exit,
1643 * Fix a bug which prevented suspend/resume to work properly
1644 * Display the status of simulated processes when receiving SIGINT
1645 This fixes a regression of v3.3. due to the introduction of SIMIX
1646 * Bug fixing in failure management:
1647 - trace could not start by a failure at time 0
1648 - failure during communications were not working
1651 * Add SIMIX_process_set_name() to change the name of the current
1652 process in the log messages.
1653 * Store smx_hosts in a dict since we only retrieve them by name
1654 * Move the configuration infrastructure to surf
1657 * Move the configuration infrastructure to surf
1660 * Massive internal cleanups:
1661 - Store internal structures on processes instead of hosts (allows
1662 to have more than one process per host, in addition of being more
1664 - Cleanup the initialization/finalization process
1665 - Kill a whole bunch of unneeded synchronization:
1666 processes run in exclusive manner within the simulator
1667 - Move queues from global tables to process data fields
1669 - now accept -platform and -hostfile arguments
1670 - Pass the right rank value to processes according to the hostfile
1671 * Compile the examples by default, and use them as regression tests
1672 * Implement MPI_Wtime()
1673 * Change the reference speed to a command line option
1676 * TCP_gamma can now be specified as command line option using
1677 --cfg=TCP_gamma:10000000.0
1678 * Change the --surf-path cmd line option into --cfg=path:
1681 * Also include strbuff from xbt.h public header
1682 * xbt_ex_display(): do not free the exception after displaying
1683 This allows to do more with the given exception afterward.
1684 Users should call xbt_ex_free() themselves.
1688 Portability report of this version:
1689 * Main portability targets:
1690 - Linux(debian)/x86/context
1691 - Linux(debian)/x86/pthreads
1692 - Linux(debian)/amd64/context
1693 - Linux(debian)/amd64/pthreads
1694 These targets fail about 1/10 of times on gras/pmm, but we believe
1695 that this is because of the test, not because of simgrid.
1696 amok/saturate_sg fails even more rarely, and the test may not be
1699 - Mac OSX Leopard/x86/context
1700 The test suite still spits tons of errors because some obscure
1701 force prevents us from removing the temporary directories
1702 arguing that they still contain some metadata I've never heard of.
1703 Smpi fails because seq is not installed.
1704 Everything seems to work properly beside of that.
1707 - AIX version 5.3 (both contexts and pthread)
1708 Smpi still fails there because mktemp is not installed.
1709 XML inclusions seems rosty on AIX.
1711 * Windows: it's still lagging behind. If you want to help, please
1714 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1716 SimGrid (3.3) stable; urgency=high
1720 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1721 [Malek Cherier & Mt]
1723 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1726 Until now, GRAS were using MSG as an interface to SURF. It was
1727 quite difficult because both interface have several differences
1728 (MSG channels vs GRAS sockets were the most notable point).
1730 This also opens the gate to SMPI (which should occur soon) and speed
1731 up simulations by to 40% (even if it were not the main goal).
1733 **************************************
1734 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1735 **************************************
1736 The point is that events occuring at the exact same timestamp are
1737 not scheduled in the same order with the old and new version. This
1738 may be enough to completely change the execution of simulations in
1739 some cases. Sorry for the inconvenience.
1741 * Cleanup and upgrade the XML format to push further scalability
1742 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1744 * Improve the testing infrastructure with tesh. Now a very large part of
1745 the code is tested not only by being run but also by checking that the
1746 output match an expected output [Mt].
1748 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1749 is really less memory-demanding, which should allow you to use
1750 larger files in SimGrid [AL].
1752 * Inform valgrind about our contextes, so that it becomes usable
1753 with the default (and more effecient) version of SimGrid
1754 [contributed by Sékou Diakite, many thanks]
1757 * Introduce a listener thread in charge of receiving incoming
1758 messages from the network. It allows to overlap communication and
1759 computation but most notably, it removes some stupid deadlocks due
1760 to the fact that so far, a process could not send and receive at
1761 the same time. This made most non trivial communication schema
1763 * Convert the PIDs from long int to int to match the MSG ones (and
1764 linux ones too) [Mt]
1765 * New function: gras_agent_spawn() to launch a new process on
1766 current host. Only working in simulation for now. [Mt]
1767 * New function: gras_os_hostport() returning a constant form (ie,
1768 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1771 * Make the backtrace of exceptions more human readable [Mt]
1772 * New module: xbt/str [Mt]
1773 a ton of string utility functions (split, join, printf to a newly
1774 allocated buffer, trim, etc)
1775 * New module: xbt/hash [Mt]
1776 SHA1 hashing algorithm (more to come if needed)
1777 * New module: xbt/synchro [Mt]
1778 synchronization tools (mutex and conditions) working the same way
1779 in simulation and in real life (mainly useful for GRAS, but not
1781 * New module: xbt/queue [Mt]
1782 classical producer/consumer synchronization scheme
1783 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1784 (using the classical functions will get serialized) [Mt]
1785 * Make dictionary internal table dynamic. No need to specify its size
1786 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1788 * Make sure the log channels are organized as a tree under windows
1789 (because of ANSI C compatibility issue, any channel were child of
1790 root directly) [Mt].
1793 * Cleaned many thing in surf and fixed a few bugs [AL].
1794 * Add a nice command line configuration mechanism to compose models [AL].
1795 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1796 the previous one (KCCFLN05). It relies on something that looks like
1797 a max-min sharing mechanism but cannot be written as such. A new solver
1798 was thus designed [AL].
1799 * Add a new solver to lmm. Based on Lagrange optimization and
1800 gradient-based descent, it enables to efficiently maximise systems s.a
1802 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1804 This solver enables to propose two new network models for TCP Reno and
1805 TCP Vegas based on Low's work. These models still need to be fully
1806 tested though [Pedro Velho].
1809 * Bug fix in SD_simulate. Now the time bound given as argument is
1811 * Use the new parallel task model (ptask_L07) as default.
1812 * Use the SURF command line configuration mechanism.
1813 * 0-size tasks (for synchronization) should now work.
1815 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1817 SimGrid (3.2) stable; urgency=high
1821 We still experience issues on this platform, but we believe that at
1822 least MSG is usable.
1824 GRAS API BREAKAGE (for simplification purpose, sorry):
1825 * the gras_msgtype_by_name is not used anymore. Instead of
1826 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1827 you can write (and must)
1828 gras_msg_send(toserver, "request", &request);
1829 - If you still want to pass a gras_msgtype_t to the function (to cache
1830 the type and avoid the lookup time), use the gras_msg_send_() variant.
1831 - Impacted functions:
1832 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1833 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1834 * The callbacks are now expected to return 0 when everything went well
1835 (just like the main() function)
1837 GRAS new features and improvements:
1838 * New module mecanism where user code can use per process globals [Mt]
1839 This is similar to gras_userdata_*() functions, but for libraries. It
1840 factorize some code developped over and over in the examples and AMOK.
1841 It has still to be documented and used (only amok/peermanagement is
1843 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1844 loss when they were small enough to fit into the buffer and sent quickly
1845 enough so that they can all get received in one shoot.
1846 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1847 if not found. Use the *_or_null() variant for the old semantic.
1848 * In gras_msg_handle, do not discard messages without callback.
1849 They are probably messages to be explicitly awaited later (ie, proofs of
1850 mis-synchronization in userland since they are sent before being awaited)
1851 No big deal usually.
1852 * gras_socket_meas_send/recv: semantic changed!
1853 The numerical arguments used to be (1) the total amount of data to send
1854 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1855 messages. This was need for the fool willing to send more than MAXINT
1856 bytes on quite fat pipes.
1859 * Do really rename the hostmanagement module to peermanagement. [Mt]
1860 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1861 but this is rather new and this was documented in the module
1862 documentation (poor excuses, I admit)
1863 * Bandwidth measurement semantic changed! This follows the changes to
1864 gras_socket_meas_send/recv explained above.
1867 * A sequential mode has been added to the workstations. When a workstation
1868 is in sequential mode, it can execute only one task, and the other tasks
1869 are waiting in a FIFO. [Christophe Thiery]
1872 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1873 model for SIMDAG. [Christophe Thiery]
1874 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1875 equal to 0, which caused some bad behaviors in
1876 saturated_constraint_set_update. I now use a threshold mechanism like in
1880 * When running manually src/testall, you select specific units [Mt]
1881 testall is the result of our cunit mecanism, and should replace all
1882 the scripty thingy around since bash don't run easily on billware.
1884 * A mallocator system has been added. [Christophe Thiery]
1885 Mallocators allow you to recycle your unused objects instead of freeing them
1886 and allocating new ones.
1888 Documentation update:
1889 * FAQ reworking + New FAQs:
1890 - "Valgrind spits tons of errors!" [Mt]
1891 - "How to repport bugs" [Mt]
1892 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1893 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1894 - Communication time measurement within MSG [AL]
1895 - I experience weird communication times when I change the latency [AL]
1896 * GRAS tutorial [Mt]
1898 - an introduction to the framework and to the used communication model
1899 - an initiatic tour introducing the most proheminent features:
1901 . Lesson 0: Installing GRAS
1902 . Lesson 1: Setting up your own project
1903 o Part 2: Message passing
1904 . Lesson 2: Exchanging simple messages
1905 . Lesson 3: Passing arguments to the processes (in SG)
1906 . Lesson 4: Attaching callbacks to messages
1907 . Lesson 5: Using globals in processes
1908 . Lesson 6: Logging informations properly
1909 . Lesson 7: Using internal timers
1910 . Lesson 8: Handling errors through exceptions
1911 . Lesson 9: Exchanging simple data
1912 . Lesson 10: Remote Procedure Calling (RPC)
1913 . Lesson 11: Explicitely waiting for messages
1914 . Recapping of message passing features in GRAS
1915 - A HOWTO section containing:
1916 o HOWTO design a GRAS application
1917 More are due, of course. They will come latter. In the meanwhile, you can
1918 check the examples which are still here.
1920 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1922 SimGrid (3.1) stable; urgency=high
1926 There was a stack corruption somewhere, visible only when optimizing
1927 with these versions. [Vince]
1930 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1931 module that enabled to play with Directed Acyclic Graphs. It is built
1932 directly on top of SURF and provides an API rather close to the old
1933 SG. Some old codes using SG are currently under rewrite to check that
1934 all needful functions are provided. [Christophe Thiery]
1937 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1938 extension of the classical CLM03 model that gracefully handles
1939 failures. This is now the default model for MSG and GRAS. It doesn't
1940 handle parallel tasks yet though. [AL]
1941 * Bug fix: Weights were not correctly set in the network part.
1942 WARNING: This may have resulted in incorrect results with simulations
1943 where there are more than one flow on a given link. [AL]
1946 * After a (long ?) discussion on simgrid-devel, we have decided that the
1947 convention we had on units was stupid. That is why it has been decided
1948 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1949 WARNING : This means that all previous platform files will not work as
1950 such with this version! A warning is issued to ask users to update
1952 A conversion script can be found in the contrib module of the CVS, under
1953 the name contrib/platform_generation/surfxml_update.pl [MQ]
1956 * Bug fix: Processes were started in reverse order, wrt deployment file.
1957 WARNING: if your code relies on this bug, please fix it. [AL]
1958 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1959 being executed on two different locations. [AL]
1960 * Bug fix: Failures are now better supported thanks to Derrick's tests
1961 (there was many failure situations I hadn't thought of and that weren't
1962 correctly handled). [AL]
1963 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1967 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1968 Exception occurring on server-side are propagated back to client (!).
1970 API CHANGE: the callback changed their prototype. Change:
1971 int my_handler(gras_socket_t expeditor, void *payload_data) {
1973 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1974 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1976 * New! function: gras_msg_handleall to deal with all messages arriving
1977 within a given period.
1978 * New! function: gras_socket_server_range to get a server socket in a
1979 range of port numbers (ease to avoid port number conflicts) [MQ]
1980 * New! gras processes display their backtrace when they get a SIGUSR1
1981 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1982 Sweet to debug RL processes [MQ]
1986 - Do not force experiment sizes to be expressed in kb, or it becomes
1987 impossible to measure the latency this way (needs one byte-long tests)
1988 WARNING: this changes the amok_bw_* function semantic. [MQ]
1989 - Implements the link saturation stuff. [MQ]
1990 * Peer management module:
1991 New! module factorizing code that we wrote over and over [MQ].
1994 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1995 - Test units are placed directly into the library code, they get extracted
1996 automatically and placed into the src/testall binary.
1997 - Convert most of the XBT tests to this system.
1998 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1999 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2000 live in the same C code. This required to change a little bit the API
2001 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2002 * New module: sparse graph structure with basic algorithms (this is work
2003 in progress and the API is not considered to be frozen yet). [AL]
2004 * Display more information on backtraces: source line & function names are
2005 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2006 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2007 * Reworked a little bit some #include statements to load only required
2008 headers. Some user code that relied on SimGrid to include stdlib or
2009 stdio may need to include it by themselves. [AL]
2010 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2012 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2013 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2014 but I guess that almost nobody used those parts. [MQ]
2016 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2018 SimGrid (3.0.1) stable; urgency=low
2021 * Unfortunately, I had missed 5 misnamed functions:
2022 xbt_fifo_item_t xbt_fifo_newitem(void);
2023 void xbt_fifo_freeitem(xbt_fifo_item_t);
2024 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2025 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2026 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2027 They're now deprecated. Please use their new versions:
2028 xbt_fifo_item_t xbt_fifo_new_item(void);
2029 void xbt_fifo_free_item(xbt_fifo_item_t);
2030 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2031 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2032 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2034 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2035 * Documentation: xbt_log module unmercifully reworked [MQ]
2036 * Bugfix: there was a problem with the ending of contexts with
2037 the pthread backend. It caused some weird deadlock or behavior
2038 depending on the pthread implementation. [AL]
2039 * Bugfix: get the exceptions raised in the simulator repport where
2040 and why they come from when they are not catched in time [AL, MQ]
2043 * Bugfix: Do repport the error when two non-connected hosts try to
2044 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2047 * Add additionnal checkings on communications. Assert that two
2048 communicating hosts are connected by a set of links... [AL]
2051 * Add additionnal checkings on channel values in communication [AL]
2052 * New: MSG_task_get_source to see on which host a task was generated [HC]
2053 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2054 the number of tasks waiting to be received on channel and sent
2056 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2057 waits for the first task coming from a given host.. [AL]
2059 GRAS new functionnalities: [MQ]
2060 * Enhance the parsing macro to allow the size of multidimentional objects
2061 to be given thru annotations.
2062 * New example (and documentation): Matrix Multiplication a la RPC
2063 (as when I was young!) and fix a bunch of bugs found on the way.
2065 GRAS performance improvements: [MQ]
2067 * Reduce the amount of cbps creation/destruction by making it static to
2068 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2071 * Change libdata to a set so that we can search for stuff by ID (and thus
2072 reduce the insane amount of dict lookups)
2075 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2076 RL communication on top of SG since it's so uneffective.
2077 It may also be used for inter-thread communication in RL, one day.
2078 * Use gras_datadesc_copy() to exchange messages on top of SG
2080 - improve message exchange performance on top of SG
2081 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2082 * Don't exchange on the network the size of the used part of buffer,
2083 instead, specify the possible buffer size to read().
2085 - reduces the amount of read/write calls (one pair per exchange)
2086 - reduces the amount of exchanged data (the size)
2087 - allows to retrieve all arrived data on receiver side, if we don't need
2088 it right now (subsequent read will peek the buffer)
2089 - allows the receiver to proceed with the begining of the stream before
2090 everything is arrived
2091 - make it possible to build an iov transport (using readv/writev)
2093 - take care of the data with non-stable storage (like stacked data),
2095 * If possible, TCP send uses vector I/O (when writev() is here)
2096 - Don't use it for receive since we send data sizes and data on the
2097 same stream, so we wouldn't be able to chain large amount of chunks
2098 before having to flush the stuff to read the size.
2099 * Rework the transport plugin mecanism to simplify it and reduce the
2100 amount of pointer dereferencement when searching for the right function
2103 * I guess that now, we do almost as few system calls as possible while
2104 doing as few data copy as possible.
2106 To improve it further, we could try to send all the sizes first and then
2107 all the data (to use iov on receiving size), but it's only a partial
2108 solution: when you have 2 dimensional data, the sizes of the second
2109 dimension is data of the first dimension, so you need 3 streams.
2111 I'm not sure the potential performance gains justify the coding burden.
2113 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2115 SimGrid (3.00) stable; urgency=high
2118 * New! Give the possibility to hijack the surf parser and thus bypass
2119 MSG_create_environment and MSG_launch_application. Have a look at
2120 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2122 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2124 SimGrid (2.96) unstable; urgency=low
2129 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2130 This deprecates the xbt_error_t mecanisms.
2131 It modifies (simplifies) all XBT and GRAS API.
2132 MSG API keeps unchanged (exceptions raised by XBT are catched from
2133 within MSG and masked with existing error handling facilities)
2136 * New! Add a FATPIPE model. [AL]
2137 * New! Add a parallel task model. [AL]
2138 * New! Add automatically a loopback interface (in the default
2139 network model) if none was precised.
2142 * Bugfix: MSG_process_resume now works with the current running process.
2144 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2145 * Modification of MSG_task_get_compute_duration. Once a task has been
2146 processed, the value returned by this function is now equal to 0. [AL]
2147 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2148 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2149 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2150 of a m_task. For now, MSG_task_cancel only works with computation
2152 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2153 of the processor (in Mflop/s) regardless of the current load on the
2155 * API Change: use proper naming convention for MSG_getClock and
2156 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2158 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2159 This function changes the priority of a computation task. This priority
2160 doesn't affect the transfer rate. A priority of 2 will make a task
2161 receive two times more cpu power than the other ones. This function
2162 has been added to suit the needs of Nguyen The Loc and hasn't been that
2163 much tested yet. So if it fails, please report it and send me your code.
2165 * API Change: removed all functions and types that were marked "deprecated"
2166 since many months. Renamed MSG_global_init_args to MSG_global_init.
2168 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2170 SimGrid (2.95) unstable; urgency=low
2173 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2174 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2175 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2178 * Bugfix: Allow absolute paths for platform description files [MQ]
2179 * Bugfix: do free the variables after use. Leads to drastic performance
2181 * Implement max_duration (ie, timeouts) on resources [AL]
2184 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2186 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2187 the needed support (timeouts) [AL]
2190 * Implement measurement sockets. You can now get the bandwidth between two
2191 hosts thanks to AMOK (see below). [MQ]
2192 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2193 dynar over the network (yeah) [MQ]
2194 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2195 * simple/static token-ring example. [Alexandre Colucci and MQ]
2196 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2197 in place before [MQ]
2200 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2201 type parsing. "short" and "long" modifiers were also ignored; other
2202 modifier (such as reference level) are still ignored. [MQ]
2203 * Embeed the buffer size within the buffer itself on SG. [MQ]
2204 That way, send() are atomic and cannot get intermixed anymore (at least
2205 the ones which are less than 100k; bigger messages still have the issue)
2206 * Array size pushed by the field, not by the field type (or each
2207 and every long int will push stuff to the cbps) [MQ]
2208 * use select() to sleep since it allows to portably sleep less than one
2211 GRAS (minor cleanups)
2212 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2215 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2216 * type category 'ignored' killed as it was never used and were difficult
2218 * whether a type can cycle or not is now a flag, leaving room for more
2219 flags (as "ignored", if we feel the need one day ;)
2220 * Rename raw sockets to measurement sockets since "raw" has another
2221 meaning in networking community.
2224 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2225 and offers features not belonging to GRAS but that most applications
2226 need. One day, it may be a set of plugins loadable at runtime.
2227 * New module: bandwidth
2228 bandwidth measurement between arbitrary nodes running this module. [MQ]
2230 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2232 SimGrid (2.94) unstable; urgency=low
2234 The first beta release of SimGrid 3 !
2238 * Update the main page and the FAQ. Adding references to gforge.
2241 * Add a gras_os_getpid function.
2244 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2245 * Extend the logs so that they also print PID, hostname, date, ... if
2247 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2248 and kill the old version which were in testsuite/
2249 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2253 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2254 for users. Instead, each programming environment (GRAS, MSG,...) use it
2255 when needed to provide such a feature to users.
2256 Don't shortcut the mecanism or you will also shortcut the virtualization
2257 you need on the simulator.
2261 * Cleanups in configury with regard to compile optimization/warning flags.
2262 Also add -fno-loop-optimize to any powerpc since it's the optimization
2263 killing gcc (< 3.4.0).
2264 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2265 needed by MSG examples complications
2266 * Borrow configury beautifications from PHP
2269 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2270 warning (thanks loris for stumbling into this one).
2271 * Bugfix: stop loading private headers (gras_config.h) from the public
2275 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2276 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2277 RL would imply a DNS resolver, which is *hard* to do in a portable way
2278 (and therefore delayed).
2279 * Implement a real timer mecanism and use it in timing macros. This allows
2280 to avoid rounding errors and get a 0.000005 sec precision in timing
2281 macros. While I was at it, various cleanups:
2282 - allow to declare more than one timed section per file (fix a stupid bug)
2283 - move some private declaration to the right place
2284 - merge conditional execution and timing macros into emulation module
2285 - document the module
2286 - make sure the module cleanups its mess on gras_exit
2287 * Documentation improvements:
2288 - (new) how to compile applications using GRAS
2289 - (new) emulation support (timing macros)
2291 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2293 SimGrid (2.93) unstable; urgency=low
2295 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2298 - Use Paje properly where used. Still to be sanitized properly.
2299 - Portability fix: Add an implementation of the contexts using pthread
2303 - Add xbt_procname(): returns the name of the current process.
2304 Use it to show the current process's name in all logging.
2306 - fix detection of older flex version and the reaction, since we do
2307 depend on modern ones (we use lex_destroy)
2308 - Better separation of SG and RL in the libs: remove all simulation code
2309 from libgras. As a result, this lib is now only 200k when stripped.
2310 Some of the xbt modules may also be duplicated (two sets and such) and
2311 should be cleaned/killed before SG3.
2312 - Insist on using xlC on AIX because of weird problems involving gcc there.
2313 - Cleanup the make remote stuff. This is now done by scripts
2314 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2315 mainly for our private use, but we're working on changing them to user
2318 - Bugfix: flush the socket on close only if there is some *output*.
2319 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2321 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2324 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2326 SimGrid (2.92) unstable; urgency=low
2328 Alpha 3 on the path to SimGrid 3
2332 - New! First try of benchmarking macros.
2333 - New! First try so that gras_stub_generator generate deployment and
2334 remote compilation helpers.
2336 - Bugfix: Initialization fix in msg_test.
2340 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2342 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2344 - API Change: gras_os_sleep to take the amount of seconds as a double.
2345 Accepting an int was error prone since it was the only location where
2346 seconds were coded as such. It leaded to damn rounding errors.
2347 - Bugfix: Hard to belive that timers ever worked before this.
2349 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2351 SimGrid (2.91) unstable; urgency=low
2353 Alpha 2 on the path to SimGrid 3
2357 - Bug fix in the lmm_solver.
2359 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2360 through the function MSG_paje_output.
2361 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2362 - It is possible to bound the rate of a communication in MSG with
2363 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2364 to write it in the changelog).
2365 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2370 - Various cleanups to the autotools stuff
2371 - Begin to move Gras examples to examples/gras/
2372 - Let make distcheck work again (yeah!)
2374 - documentation overhauled using doxygen.
2375 gtk-doc-tools is dead in SimGrid now.
2376 - Automatically extract all existing logging categories, and add the list
2377 to the documentation (long standing one, to say the less)
2379 - Cleanup the known architecture table. Reorder the entries to group what
2380 should be, and use a more consistent naming scheme.
2381 (some of the test dataset are still to be regenerated)
2382 - New! Allow library to register globals on each process just as userdata
2384 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2385 lookup time (for now).
2386 Use it in msg and trp.
2387 This cleans a lot the internals and helps enforcing privacy of the
2388 headers between the gras components.
2389 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2390 - Bugfix: gras_os_time was delirious in RL.
2391 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2393 - Reenable GRAS now that it works.
2395 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2397 SimGrid (2.90) unstable; urgency=low
2399 Alpha 1 on the path to SimGrid 3
2401 * It is a long time since the last release of SimGrid. I'm sorry about
2402 that but as I had told you, I was rewriting a lot of things. I apologize
2403 to those who had been reporting bugs to me and that I had not answered.
2404 If your bug is still in the new version, please tell me. Here is a
2405 summary of the main changes.
2407 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2408 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2409 portability, speed, and a lot more but you'll figure it out later.
2410 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2411 I wanted to release the new MSG as soon as possible and I have
2412 broken GRAS, which is the reason why, for now, only MSG is fully
2413 functional. A laconic description of these projects is available
2414 in the documentation.
2416 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2417 kernel optimized for our needs (called SURF but only the developers
2418 should use it). Hence, MSG is now roughly 30 times faster and I think
2419 that by rewriting a little bit MSG, I could event speed it up a little
2420 bit more. Beside the gain in speed, it is also much easier to encode a
2421 new platform model with SURF than it was with SG. More to come...
2423 * REVOLUTION 3: I have tried to change a little as possible the API of
2424 MSG but a few things really had to disappear. The main differences
2425 with the previous version are :
2426 1) no more m_links_t and the corresponding functions. Platforms are
2427 directly read from a XML description and cannot be hard-coded
2428 anymore. The same format is used for application deployment
2429 description. The new format is described in the documentation.
2430 Have a look in tools/platform_generation. There is a tiny script
2431 that converts from the old platform format to the new one. Concerning
2432 the application deployment format, parsing the old one is tricky.
2433 I think most of you should however be able to convert your files. If
2434 it is really an issue, I can write a C code that does the conversion.
2436 2) the toolbox tbx does not exist anymore. We now have a library
2437 with much more data-structures but without the hash-tables (we have
2438 dictionaries that are much faster).
2440 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2442 *****************************************************************************
2443 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2444 * don't feel like converting the oldies. (Mt) *
2445 *****************************************************************************
2448 Version 2.90: "the long awaited one"
2449 - Finished rewriting and debugging MSG. Rewrote the documentation.
2450 - disable GRAS for now since it needs to be ported to the newest SG
2453 - Finish the port to windows (using mingw32 for cross-compile)
2456 - Main loop and datastructures of SURF. A cpu resource object is
2457 functional. Surf can thus be used to create cpu's with variable
2458 performance on which you can execute some actions.
2460 2004-11-15 Martin Quinson
2461 - Port to ARM. Simply added the alignment and size descriptions. Should
2462 work, but the ARM machines are so slow that I didn't had the opportunity
2463 to 'make check' over there yet.
2465 2004-11-15 Arnaud Legrand
2466 - Trace manager now written. It uses a heap structure and is therefore
2467 expected to be efficient. It may however be speeded up (particularly
2468 when many events occur at the same date) by using red and black
2469 trees. One day maybe...
2470 - Max-min linear system solver written. It uses a sparse matrix
2471 structure taking advantage of its expected use. Most operations are
2472 O(1) and free/calloc are called as few as possible. The computation of
2473 the minimum could however be improved by using a red and black tree
2476 2004-11-03 Arnaud Legrand
2477 - Rename every gras_* function that was in xbt/ to its xbt_
2479 - Add a heap and a doubly-linked list to xbt
2480 - Added a dichotomy to the dictionaries. make check works as well before
2481 so I assume that the patch is correct. I do not know however if things
2482 run effectively faster than before now. :)
2484 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2485 SimGrid, and the version number is bumped to 2.x
2487 2004-10-29 Martin Quinson
2488 - Introduction of the remote errors.
2489 They are the result of a RMI/RPC on the remote machine.
2490 ErrCodes being scalar values, you can't get the host on which those
2491 errors did happen. Extending the error mechanism as in Gnome is possible.
2492 No idea yet whether it is a good idea.
2494 2004-10-28 Martin Quinson
2495 - Interface revolution: the Starred Structure Eradication.
2496 I used to do typedef struct {} toto_t; and then handle *toto_t.
2497 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2498 - ???_t is a valid type (builded with typedef)
2499 - s_toto_t is a structure (access to fields with .)
2500 - s_toto is a structure needing 'struct' keyword to be used
2501 - e_toto_t is an enum
2502 - toto_t is an 'object' (struct*)
2504 typedef struct s_toto {} s_toto_t, *toto_t;
2505 typedef enum {} e_toto_t;
2506 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2507 s_toto_t) is private.
2509 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2510 so that it can set the variable to NULL. It was so for dicts and sets,
2511 it changed for dynars.
2513 - Fix a bunch of memleaks in dict_remove
2514 - Fix a bug in sg/server_socket opening: it failed all the time.
2516 2004-10-07 Martin Quinson
2517 - Speed up dynar lookup operation a bit.
2519 gras_dynar_get is dead.
2521 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2522 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2523 (which gives you the address of the stored data).
2525 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2526 copy of the data using an affectation to do the job and not a memcpy.
2528 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2529 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2531 It does not really speedup the dynar test because they are
2532 setting elements all the time (and look them seldom). But the dict does
2533 far more lookup than setting.
2535 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2537 2004-10-05 Martin Quinson
2538 - Allow to (en/dis)able the cycle detection at run time.
2540 Whether we should check for cycle or not is now a property of each
2541 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2542 datadesc_cycle_unset allow to remove this property when previously set.
2544 Note that the cycle detection is off by default since it impacts the
2545 performance. Watch the data you feed GRAS with ;)
2547 This property is hereditary. Any element embedded in a structure having it
2548 set have it set for the time of this data exchange.
2550 You should set it both on sender and receiver side. If you don't set it on
2551 sender side, it will enter an endless loop. If you forget on receiver
2552 side, the cycles won't be recreated after communication.
2554 - Header reorganization.
2555 Kill gras_private.h, each submodule must load the headers it needs.
2557 2004-10-04 Martin Quinson
2558 - Interface revolution: do not try to survive to malloc failure.
2560 Now, gras_malloc and friends call gras_abort() on failure.
2561 As a conclusion, malloc_error is not a valid error anymore, and all
2562 functions for which it was the only gras_error_t return value are
2563 changed. They now return void, or there result directly.
2564 This simplify the API a lot.
2566 2004-09-29 Martin Quinson
2567 - Re-enable raw sockets.
2568 Created by gras_socket_{client,server}_ext;
2569 Used with gras_raw_{send,recv}
2572 It should allow to kill the last bits of gras first version soon.
2574 This is not completely satisfactory yet (duplicate code with
2575 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2578 - Simplify transport plugin (internal) interface by not passing any
2579 argument to _server and _client, but embedding them in the socket
2582 2004-09-28 Martin Quinson
2583 - Finish the port to AIX.
2584 autoconf was my problem (segfault within the malloc replacement
2585 function. No idea why)
2587 2004-09-16 Martin Quinson
2588 - Fix some size_t madness on 64bit architectures.
2590 2004-09-08 Martin Quinson
2591 - Reduce the number of system headers loaded, overload some more system
2592 calls (such as malloc to cast the result of the system one, and work
2594 - Fix and reintroduce the config support
2596 2004-09-07 Martin Quinson
2597 - Source code reorganization to allow Arnaud to surf all over there.
2598 - Allow to document the logging categories.
2599 - Remove all uppercase from logging categories and useless cleanup in names.
2601 2004-08-18 Martin Quinson
2602 Version 0.6.2 (protocol not changed; API changed)
2603 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2604 gras_error_t), and NULL when not found. Functions expecting a msgtype
2605 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2606 hopefully usefull message.
2607 - Portability to prehistoric sparcs again
2609 2004-08-17 Martin Quinson
2610 Version 0.6.1 (protocol not changed; ABI not changed)
2611 - prealloc some buffers to speed things up
2613 2004-08-11 Martin Quinson
2614 Version 0.6 (protocol not changed; ABI expended)
2615 - The parsing macro can deal with the references, provided that you add
2616 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2618 2004-08-09 Martin Quinson
2619 Version 0.5 (protocol not changed; ABI changed)
2620 - Allow to off turn the cycle detection code in data exchange at
2621 compilation time. It should be at run time, but I'm short of time (and
2622 the config stuff is still broken). That way, we keep dict out of the
2623 critical path, which is good because the performance is poor:
2624 - search not dichotomial yet
2625 - dynar give no way to access their content and memcpy everytime
2626 - In composed data description (struct, ref and so on), stop foolness of
2627 keeping the subtype's ID, but store the type itself. This keeps sets out
2628 of the critical path, which is good since they rely on dynar and
2629 dictionnaries. The only loose of that is that we cannot detect the
2630 redeclaration of a structure/union with another content (but I'm not sure
2631 the code detected well this error before anyway). We still can detect
2632 the redefinition discrepancy for the other types.
2633 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2634 it breaks the code because of type-punning used all over the place).
2635 This breaks on all non-gcc architectures (for now).
2637 All those changes (plus the buffer of last time) allow me to gain 2 order
2638 of magnitude on cruel tests consisting of 800000 array of integers on two
2639 level of a hierarchical structure (200 secondes -> 4 secondes)
2642 - the selector of reference must now return the type it points to, not
2643 the ID of this type.
2645 2004-08-06 Martin Quinson
2646 Version 0.4 (protocol changed; ABI not changed)
2647 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2648 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2649 - Implement a buffer transport, and use it by default (it relies on tcp in
2650 real life and on sg in simulation).
2651 That's a bit hackish since I had a new field to the structure to store
2652 its data without interfering with the subtype ones. Inheritance
2653 is tricky in C. And that's a kind of reverse inheritance with one class
2654 derivating two classes. Or maybe a game with java interfaces. Anyway,
2655 that's damn hard in C (at least).
2656 Moreover, I got tired while trying to ensure plugin separation and
2657 genericity in SG mode. MSG wants me to do weird things, so let's go for
2658 cruel hacks (temporarily of course ;).
2659 See comment in transport_private.h:71
2660 - do not include all the _interface headers in private but in the files
2661 which really need them (to cut the compilation time when they are
2664 2004-07-26 Martin Quinson
2665 Version 0.3 (protocol not changed; ABI changed)
2666 - Major overhault of the datadesc interface to simplify it:
2667 - shorted the function names:
2668 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2669 - add a trivial way to push/pop integers into the cbps without malloc.
2670 This allows to make really generic sub_type description, which simply
2671 pop their size of the stack.
2672 - add a function gras_datadesc_ref_pop_arr() which does what users want
2673 most of the time: Declare a dynamic array (which pops its size of the
2674 stack) and declare a reference to it. Poor name, but anyway.
2675 - kill the post-send callback, add a post-receive one
2677 2004-07-23 Martin Quinson
2678 Version 0.2 (protocol changed; ABI changed)
2679 - add some testing for cpbs in the test cases, and fix some more bugs.
2680 This invalidate again the little64 data file, since I cannot regenerate
2682 - remove an awfull optimization in the logging stuff, allowing me to:
2683 - understand it again
2684 - learn gcc how to check that the argument match the provided format
2685 - fix all errors revealed by gcc after that
2686 - internal keys of dict are not \0 terminated. Deal with it properly in
2687 loggings instead of segfaulting when the user want to see the logs :-/
2689 2004-07-22 Martin Quinson
2690 - Fix some stupid bug preventing cbps (callback postit) from working
2692 2004-07-21 Martin Quinson
2693 - Some documentation cleanups
2694 - remove the useless last argument of msgtype_declare
2695 - rename the Virtu functions to fit into the 'os' namespace
2696 - move headers src/include -> src/include/gras/ and stop fooling with
2697 gras -> . symbolic link
2698 - make distcheck is now successful
2700 2004-07-19 Martin Quinson
2702 - Build shared library also
2703 - Install html doc to the right location
2704 - stop removing maintainer files in make clean
2705 - build tests only on make check
2707 2004-07-13 Martin Quinson
2709 - No major issue in previous version => change versionning schema
2710 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2711 corresponding dataset.
2713 2004-07-11 Martin Quinson
2715 - Get it working with any kind of structure (we can compute the padding
2716 bytes remotely for all the architectures I have access to)
2717 - Implement the structure parsing macro (still not quite robust/complete)
2718 - Improvement to the remote testing toysuite
2720 2004-07-10 Martin Quinson
2721 [autoconf mechanism]
2722 - get ride of a bunch of deprecated macros
2723 - actually run the test for two-compliment, not only compile it :-/
2724 - test whether the structures get packed (and bail out if yes. Damn.
2725 Alignment is a serious matter)
2726 - test whether the structures get compacted (but respecting the alignment
2727 constraints of each types)
2728 - test whether the array fields of structures can straddle alignment boundaries
2730 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2731 (took me 2 hours to find that bug, looking at the wrong place)
2732 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2733 before sending/receiving and are used to compute the offsets of fields
2734 - Given that padding size depend even on compiler options, keep track of
2735 alignment and aligned_size only for the current architecture. Not a big
2736 deal since we send structure fields one after the other (seems
2738 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2739 program, let it work on linux/gcc/little32. portability todo.
2741 2004-07-08 Martin Quinson
2742 - import and improve remote compilation support from FAST
2743 - make sure make check works on half a dozen of machines out there
2745 2004-07-07 Martin Quinson
2746 Let's say it's version 0.0.3 ;)
2747 - Implement conversions (yuhu!)
2748 - Let it work on solaris (beside conversion, of course)
2749 - Stupid me, using rand() to generate the conversion datatests in not wise.
2751 2004-07-06 Martin Quinson
2752 - Let make dist work, since I'm gonna need it to compile on remote hosts
2753 - Let Tests/datadesc_usage write the architecture on which the file was
2754 generated as first byte.
2755 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2756 architecture descriptions.
2757 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2758 on those architectures.
2759 - Optimization: send/recv array of scalar in one shoot
2761 2004-07-05 Martin Quinson
2762 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2764 - Plug a whole bunch of memleaks
2765 - each process now have to call gras_{init,exit}. One day, their log
2766 settings will be separated
2767 - Continue the code factorisation between SG, RL and common in Transport.
2769 2004-07-04 Martin Quinson
2771 - Redistribution between SG and RL.
2772 We wanna have to accept in SG, so move accepted related parts of RL in
2773 the common part. (more precisely, the dynar of all known sockets is no
2774 more a static in transport.c, but part of the process_data)
2776 [gras_stub_generator]
2777 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2779 2004-07-03 Martin Quinson
2780 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2781 them all up in one shot)
2783 - Fix the ugly case of reference to dynamic array.
2784 - New (semi-public) function gras_datadesc_size to allow the messaging
2785 layer to malloc the needed space for the buffer.
2787 - gras_socket_close now expect the socket to close (and not its address to
2788 put NULL in it after it). This is because the socket passed to handlers
2789 is one of their argument (=> not writable).
2791 - propagate the interface cleanup from last week in datadesc, ie remove a
2792 superfluous level of indirection. User pass adress of variable
2793 containing data (both when sending and receiving), and not of a variable
2794 being a pointer to the data. Let's say that I like it better ;)
2795 The price for that is constructs like "int msg=*(int*)payload" in
2796 handlers, but it's a fine price, IMHO.
2798 - Let it work in RL (yuhu)
2800 2004-06-21 Martin Quinson
2802 - porting SG plugin and SG select to new standards (works almost).
2803 - plug memleaks and fix bugs around.
2806 - cleanup the prototype of data recv and force users to specify when they
2807 want to handle references to objects. Test case working even for cycles.
2808 - plug memleaks. Valgrind is perfectly ok with this.
2810 2004-06-12 Martin Quinson
2812 - cleanup the separation between plugin and main code in plugin creation
2814 2004-06-11 Martin Quinson
2816 - Reput hook for raw sockets, needed for BW experiments
2817 - kill a few lines of dead code
2818 [Data description] Interface cleanup
2819 - gras_datadesc_by_name returns the searched type or NULL.
2820 That way, no variable is needed to use a type desc once, which makes
2822 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2823 two parameters were strings (field name, type name), leading to
2825 [Dicos] Interface cleanup
2826 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2827 This is consistant with the dynar API.
2829 2004-04-21 Martin Quinson
2831 - Porting to new standards.
2833 - interface cleanup.
2834 There is no bag anymore, no need to take extra provision to mask the
2835 pointers behind "ID".
2836 Better splitup of functions between files create/exchange/convert.
2837 This is still a bit artificial since convert and receive are so
2838 interleaved, but anyway.
2840 - add a queued message list to procdata (the ones not matching criteria
2842 - factorize some more code between SG and RL wrt procdata
2844 - use gras_exit in example to track memleaks
2845 - get rid of gs_example now that GS is properly integrated into gras
2846 - update run_test to integrate the latest tests (datadesc)
2848 - rename WARNINGn macros to WARNn since it prooved error-prone
2850 2004-04-19 Martin Quinson
2852 - register init/exit functions within gras module mechanism
2853 - send/receive function.
2854 Convertion is not implemented, but short-cutted if not needed.
2855 struct/array elements are sent one by one (instead of block-wise), but
2856 nobody really cares (yet). Get a prototype before optimizing.
2857 - tests (using a file socket) for DD send/receive on:
2858 - base types: int, float
2859 - array: fixed size, string (ie ref to dynamic string)
2860 - structure: homogeneous, heterogeneous
2861 - chained list, graph with cycle
2862 Believe it or not, valgrind is not too unhappy with the results. The
2863 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2867 - Bugfix when using a filename explicitely (instead of '-')
2869 2004-04-09 Martin Quinson
2871 - factorize more code between RL and SG in socket creation
2872 - Complete the implementation and tests of:
2874 o file (only in RL, and mainly for debugging)
2876 I lost 3 days to design a portable address resolver, and then decided
2877 that the prototype mainly have to run on my box.
2878 Addressing portability too early may be like optimizing too early :-/
2880 - use gras_init in the Tests instead of the crappy parse_log_opt
2881 (the latter function is removed)
2882 [Conditional execution]
2883 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2884 [Code reorganisation]
2885 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2886 Build examples against the RL library, since there is no way to disable
2887 its creation for now.
2889 For information, the beginning of coding on GRAS was back in june
2890 2003. I guess that every line has been rewritten at least twice since