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 * Java must be version 7 at least when activated.
7 * Tracing is now always enabled (no way to turn it out)
8 * Remove GTNetS. It was not working anyway.
9 * Move headers around to sort them out on installed systems:
10 - instr/instr.h -> simgrid/instr.h
11 - instr/jedule/* -> simgrid/jedule
12 - simdag/datatypes.h was removed
13 - simdag/simdag.h -> simgrid/simdag.h
14 - msg/datatypes.h was removed
15 - msg/msg.h -> simgrid/msg.h
18 * Interface improvement:
19 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
20 - Sanitize the interface in MSG_task_ module:
21 - Merge two functions that were close enough but misleading:
22 set_compute_duration(t) -> set_flops_amount(t)
23 get_remaining_computation(t) -> get_flops_amount(t)
24 - set_data_size(t) -> set_bytes_amount(t)
25 get_data_size(t) -> get_bytes_amount(t)
26 - Massive cleanups in the functions related to the energy
27 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
28 - New: MSG_host_get_pstate()
30 - msg/energy/onoff: switching hosts on and off
32 * Interface improvement:
33 - Rename Host.isAvail() to Host.isOn()
34 - Rename Process.currentProcess() to Process.getCurrentProcess()
35 - Rename Task.setDataSize() to Task.setBytesAmount()
36 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
38 - #18874: Actually allows the GC to reclaim tasks
42 - SIMIX_process_throw: raises an exception in a remote process
43 * Refactoring: Separate sub-modules
44 - libsmx: the public interface, as libc in a real system
45 - popping: the strange dance that converts a user request into a kernel handling
46 - smx_context_*: the virtualization mechanisms that embed the user code
47 - smx_*: the handling of each simcalls
49 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
50 * Rename smx_action_t into smx_synchro_t, making explicit that these
51 things are used to synchronize processes with their environment.
52 For example, a communication is a sort of synchronization involving
53 the communicating processes (that may block until the exchange) and
54 the platform. The same can be said from computations, etc.
56 - #18888: segfault when a process ends before its kill_time
60 - 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.
61 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
62 - MPI_Type_set_name, MPI_Type_get_name
63 - MPI_*_c2f and MPI_*_f2c functions
64 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
65 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
66 - Activate a lot of new tests from the mpich 3 testsuite
68 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
69 - InfiniBand network model added : Based on the works of Jerome Vienne
70 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
71 - When smpi/display_timing is set, also display global simulation time and application times
72 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
73 * Collective communications
74 - 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
75 - 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/).
76 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
78 - Replay now uses algorithms from wanted collective selector
79 - Replay can be used with SMP-aware algorithms
80 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
81 - 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.
83 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
84 - Correctly use loopback on fat-tree clusters
85 - Asynchronous small messages shouldn't trigger deadlocks anymore
86 * Energy/DVFS cleanup and improvement
87 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
88 - new: smpi_get_host_pstate()
91 - "Full" network optimization flag was broken since Surf++
92 - Better handling of precision flags in maxmin
93 - Fix bug causing sometimes "Impossible" errors
94 - Properly pass cluster properties to included hosts
95 * Improvement of the Energy plugin.
96 - Always update the consumption before returning that value
97 - New property: watt_off to denote the dissipation when the host is off
98 - New functions getWattMinAt and getWattMaxAt to retrieve the
99 dissipation of pstates that we are not currently at.
100 * Java: class NetworkLink renamed to Link
101 * New function: simcall_process_get_kill_time()
102 * Massive rename s/workstation/host/
103 - That's intrusive, but that's good for the project consistency. Sorry.
104 - Change config option "workstation/model" into "host/model"
107 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
108 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
109 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
110 - xbt_dirname and xbt_basename for non-POSIX systems
112 * The model checker now runs as a separate process.
113 * The model checker runs must now be launched with the new simgrid-mc program.
114 * Record/Replay: the MC can display a textual representation of a path in the
115 execution graph. It can then be replayed outside of the model checker.
117 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
119 SimGrid (3.11) stable; urgency=low
124 * Normalizing pointers addresses tool for better diff between logs
127 * Add cloud examples using new VMs
128 - examples/msg/cloud/two_tasks_vm.tesh
129 - examples/msg/cloud/simple_vm.tesh
130 - examples/java/cloud/cloud.tesh
131 - examples/java/cloud/migration/migration.tesh
132 * Add java surf examples:
133 - examples/java/surfPlugin/surf_plugin.tesh
134 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
135 - examples/java/surfCpuModel/surf_cpu_model.tesh
136 * Add SMPI+MSG example:
137 - examples/smpi/smpi_msg_masterslave/
142 - msg task destroy cancel test
143 - msg_host on/off test
144 * Move all tests in testsuite to teshsuite (adding tesh files)
145 * Restructure teshsuites
146 - one folder for each kind of test
147 * Restructure AddTests.cmake
149 - structure the order of tests (with sections)
152 * Add virtual machine
153 - creation of a VM on a PM
154 - migration of a VM from a PM to another PM
156 - MSG_process_join(msg_process_t process, double timeout)
157 - msg_bar_t MSG_barrier_init(unsigned int count)
158 - int MSG_barrier_wait(msg_bar_t barrier)
159 - void MSG_barrier_destroy(msg_bar_t barrier)
160 - msg_as_t MSG_environment_as_get_by_name(const char * name)
161 * New option "msg/debug_multiple_use" to help debugging when a task is used
164 - Improvements and finalization of MSG_storage, MSG_file APIs and their
166 - Increase code coverage in test suites
170 * Protect context stacks against stack overflow. The number of protected memory
171 pages allocated on the top of each stack (1 by default) can be configured
172 with the new command line option --cfg=contexts/guard_size:<value>.
173 * Simcalls are now generated by a python script that
174 - generates files included by SimGrid
175 - checks that all the functions exist, and proposes prototypes
177 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
178 host_set_data, host_get_data
180 - simcall_process_join(smx_process_t process, double timeout)
181 * Fix bug where sleeping processing could not be suspended.
184 * Translate surf models from C to C++
185 - Generic classes for all models: Model, Resource, Action
186 - A generic interface for each kind of model (CPU, Network, Storage
187 Workstation, WorkstationVM)
189 * Translate surf routings from C to C++
190 * Add callbacks using sigc++ or boost::signals2
191 - Add callback functions for resource creation/destruction
192 - Add callback functions for action state change
193 - Handle Energy as a plugin
194 * Replace swag by boost::intrusive
195 * Add new routing models for clusters. For documentation, see
196 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
197 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
198 parameters for cluster tag
199 - Fat trees, with topology="FAT_TREE" and
200 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
201 - see examples/platforms/torus_cluster.xml and
202 examples/platforms/fat_tree_cluster.xml
206 * Hostfiles support host:nb_processes construct to deploy several processes on
208 * Collective communication algorithms should not crash if used with
209 improper number of nodes and report the error.
210 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
211 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
212 MPI_Dims_create, MPI_Cart_sub are supported.
213 * New interface to use SMPI programmatically (still depends on MSG for
214 some parts, see examples/smpi/smpi_msg_masterslave) :
215 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
219 * Global variables privatization in MPI executables is now performed at runtime
220 with the option smpi/privatize_global_variables (default:no).
221 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
222 dynamic libraries loaded by the application are not privatized (static
223 linking with these libraries is advised in this case)
226 * Options defined in XML work correctly now.
229 * New cmake option, enable_lib_in_jar, to control whether native libraries are
230 copied into simgrid.jar or not (ON by default). Use this option if you want
231 to reduce the size of the installed simgrid.jar, *and* the native libraries
232 are kept installed elsewhere.
233 * Surf binding with SWIG (code generated in maintainer mode only):
234 - plugin to handle callbacks
235 - CPU model only for the moment
238 * Supernovae build mode is definitively removed. It was used to improve
239 inlining and inter-module optimizations. It is nowadays superseded by
240 link-time optimizations commonly available in compilers.
241 * Update ns-3 find lib. Bindings for ns-3 should work again now.
242 * Add boost dependency for surf++
243 * Add new macro for tests
244 - ADD_TESH(name <tesh_args>)
245 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
249 - xbt_bar_t XBT_barrier_init(unsigned int count)
250 - int XBT_barrier_wait(xbt_bar_t barrier)
251 - void XBT_barrier_destroy(xbt_bar_t barrier)
252 * Make the xbt_os_time module public
254 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
256 SimGrid (3.10) stable; urgency=low
258 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
261 * Reintegrate Java to the main archive as desynchronizing these
262 package is not acceptable anymore (Java is now considered stable)
263 * Add explicit synchronization facilities through semaphores
264 * Bug fix: Task.setDataSize() only changed the C world, not the value
265 cached in the Java world
268 * Dramatically change the way files are handled. API and internals changed, but
269 this part of MSG was not considered as production grade either.
270 * Add explicit synchronization facilities through semaphores
271 * Add a new function MSG_host_get_process_list()
272 * Preliminary DVFS support (see examples/msg/energy/ for details)
275 * SMPI is now included directly in the libsimgrid as the windows
276 linker doesn't force us on splitting it anymore.
277 * Improvements of the SMPI replay tool:
278 - Most of the collective communications are now rooted in the same process as
279 in the original application.
280 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
281 used until now). Multiple data types can now be used in a trace.
282 - The replay tool now supports traces produce either by TAU or a modified
284 - Bug Fix: the compute part of the reduce action is now taken into account.
285 - Gatherv collective is now supported
286 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
287 of smpirun outputs time independent traces for the current run. One file
288 is created per process. If too many processes are simulated, this behavior
289 can be changed to one file for all processes by using the
290 tracing/smpi/format/ti_one_file flag
291 * smpirun generates the host file if needed (with given host count and platform)
292 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
293 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
294 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
295 --cfg=smpi/coll_selector:(mpich/ompi)
296 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
298 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
299 happening inside a collective SMPI call.
300 * Fix the behavior of complex data types handling.
301 * Make MPI_Wtime another synchronization point to take computations into
303 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
304 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
305 * Add all missing Fortran bindings, SMPI should work with Fortran 90
306 (no privatization of global variables yet)
307 * Preliminary DVFS support (see examples/smpi/energy/ for details)
310 * Verification of liveness properties is now available for SMPI applications
311 (in addition to MSG applications)
312 * Bugged examples using SMPI in examples/smpi/mc/
313 * Add --cfg=model-check/visited option. Allows the verification of infinite
314 programs. Detection of loops in the execution thanks to the system state
315 comparison and reduction of the state space to explore. Can be combined with
316 DPOR for safety properties.
319 * Allow to change SimGrid configuration (see --help) within the code
320 thanks to SD_config() as it can be done in MSG.
321 * Add a new function SD_task_set_amount() upon user request.
324 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
325 * Remove rule based routing (no more PCRE dependency)
326 * Add a limiter_link option to cluster tag, to specify a maximum reachable
327 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
328 * Add a loopback_bw and loopback_lat options to cluster tag.
329 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
330 routing. Now peers are encapsulated in an AS and have their own private
331 router but this is transparent.
334 * Our own implementation of getline is renamed xbt_getline, and gets
335 used even if the OS provide a getline(). This should reduce the
336 configuration complexity by using the same code on all platforms.
337 * New type: xbt_cfg_elm_boolean.
338 * Allow to use yes/no for boolean configuration options in the command line.
339 * Allow to disable SimGrid cleanups at exit from command line option.
340 There are situations where one may want a simulation to end with an exit.
341 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
342 annoying when scripting around the simulator. Adding a
343 --cfg=clean_atexit:no allows to circumvent this issue.
346 * Lots of memory leaks were corrected in this release.
347 There is no known memory leaks anymore, in all of our 600+ tests.
348 * New command line option --version, to get SimGrid version information.
349 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
351 * Supernovae builds are deprecated, and expected to be removed in the next
354 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
356 SimGrid (3.9) stable; urgency=low
358 The Grasgory release: GRAS is really dead now.
360 * Complete overhaul of the internal host structures scheme.
363 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
364 was considered as experimental and badly maintained since then.
365 * Keeping it was thus a trap to our potential users, that could take
366 it instead of MSG or SMPI by mistake despite is pity state.
367 * GRAS seems to have very few users (if any), and no one volunteered
368 to maintain it further. It also induces a lot of XBT code (for
369 portability sake), that must be maintained too.
370 * For all these reasons, we killed GRAS. If someone wants to revive it
371 in the future, don't cry, our git history still remembers of GRAS.
374 * Major overhaul. Merge our documentation again as time proved that
375 splitting it was really not helping our users.
376 * Further improve the developer documentation to help newcomers
377 hacking on SimGrid itself. The user documentation (and in
378 particular, the beginner documentation) is still in a sorry state.
381 * Now works on Windows too!
382 * Much more extensive test suite, from MPICH
385 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
386 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
387 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
388 * Bug fix: let task be scheduled when the last dependency to be solved is
389 a control dependency.
390 * Remove SD_load_environment_script function.
391 Use the C sg_platf function if you want to declare a platform
395 * New function: MSG_process_get_number()
396 * Old function documented: MSG_config()
397 * Remove MSG_load_platform_script function
398 Use the C sg_platf function if you want to declare a platform
402 * Change the default value of the TCP_gamma constant (maximal size of TCP
403 congestion window) to a more realistic 4MiB value. If you notice changes in
404 your simulation results, you can fall back to the previous 20k tiny window
405 by adding --cfg=network/TCP_gamma:20000 on command line.
406 * (Hopefully) fix a bug wrt periodic availability and state traces
407 * Bug fix: use default values at start when first event in availability/state
408 trace is not at time 0.
411 * remove the "new_" part of function name sg_platf_new_trace_connect
412 (resulting in sg_platf_trace_connect), since it does not create
416 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
417 use case where it's really mandatory, and maintaining it was a pain in
419 * New: xbt_fifo_search(), search an item with a user-provided
420 comparison function instead of dumb pointer comparison.
423 * Fix the lua deployment:
424 Use `simgrid.init_application()` before deployment instead of
425 `simgrid.msg_register_application()` after.
428 * Transfer the tracing files into the corresponding modules.
430 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
432 SimGrid (3.8.1) stable; urgency=low
434 The "we are told that some people want to also *install* the simgrid
437 * Add missing file "tesh.1" to the archive.
439 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
441 SimGrid (3.8) stable; urgency=low
443 The Psssshiiiit release: SimGrid jumps into the Cloud.
446 * Add an experimental interface to manipulate VMs. They are mainly
447 process groups with very few intrinsic semantic, but they should
448 allow you to build the semantic you want easily.
449 * New function: MSG_host_set_property_value()
450 * New function: MSG_process_on_exit(). To clean memory in all cases.
451 * Bug fixes that made the host (and link) failures unusable.
452 * Add a way to auto-restart process when the host in which they are
453 executing comes back (ON_FAILURE="RESTART" on deployment file,
454 MSG_process_auto_restart_set).
455 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
456 please stop using the old ones, they are DEPRECATED.
458 * Deprecate functions MSG_global_init() / MSG_global_init_args()
459 Please use MSG_init() instead. (reducing the amount of entry
460 points in the library helps us).
461 * Make it impossible to link against the wrong version of the lib
462 * Deprecate MSG_clean(). No need to call it anymore.
463 * Function MSG_get_host_number() is not deprecated anymore.
466 * Split the doc into a user guide and a reference guide.
467 * Start a developper guide to help people hacking on SimGrid.
470 * Enable tracing by default. This modules rocks you should use it.
471 * Remove option custom_flags. Now use environment variables CFLAGS
473 * Use default cmake things to detect lua instead of home grown ones.
474 * New option "enable_mallocators" to disable mallocators, for debugging
475 purpose ("on" by default).
478 * Bug fixes around the resource failures: don't let the processes
479 survive the host they are running onto.
480 * Add an interface to auto-restart processes when the host in which they are
481 executing comes back.
482 * Ensures that SIMIX_clean is called automatically. It's not part of
483 the public interface anymore (bindings should be updated).
486 * Bug fix for when SD_Simulate is called with a positive value: be careful
487 when comparing doubles. Sometimes they are different for non significant
489 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
490 parallel task whose initial work is distributed among host according
491 to the Amdahl's law. Such tasks are created with a parameter alpha
492 that corresponds to the non-parallelizable part of the computation.
493 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
494 between two sets of workstations assuming a 1D block distribution (each
495 workstation owns a similar share of data) on both sides.
497 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
498 Data redistribution will be automatically scheduled once parent and child
499 are both scheduled. The filling of computation_amount and
500 communication_amount structures is now done seamlessly thanks to the chosen
502 * New function SD_workstation_dump to display various information
503 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
504 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
505 bandwidth on the route onto which the task is scheduled and the amount of
507 To divide the nominal bandwidth by 2, the rate then has to be :
508 rate = bandwidth/(2*amount)
509 * Compute tasks that have failed can now be rescheduled and executed again
510 (from their beginning)
511 * Increasing source code coverage (src/simdag is now covered at 95.8%
515 * Re-implement time-independent trace replay using SMPI (at the
516 smpi_smp_* level) instead of MSG. This should replace
517 examples/msg/actions/actions.c
518 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
519 hindexed and structs)
520 * Implement the exchange of non-contiguous data.
521 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
522 * Correct behavior of smpi/sender_gap and set its default value to 0
523 * Add option to asynchronously send small messages to allow better
524 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
525 specifies the size in bytes under which messages will be asynchronously sent.
526 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
527 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
530 * New C interface to define a platform: XML is now optional.
531 For more info, please check include/simgrid/platf.h.
532 * New interface to define random platforms from the C:
533 For more info, please check include/simgrid/platf_generator.h and
534 examples/msg/masterslave/masterslave_platfgen.c
535 * Export a sg_cmdline dynar containing all the arguments we got from
539 * Two new tracing options for adding comments to trace file so you
540 can track your experiments (see --help-tracing for details).
541 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
542 * Adding the SimGrid version that generated the trace file as a comment.
543 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
544 * Fix to avoid key clashes on Paje links
545 * Other minor fixes related to the Paje specification
548 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
549 and renamed to xbt_str_hash() and xbt_str_hash_ext().
550 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
551 * Greatly improve the robustness of mmalloc to user errors (such as
552 using an area after freeing it, or freeing it twice)
554 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
556 SimGrid (3.7.1) stable; urgency=low
559 * Restore the prototype of MSG_process_create_with_environment() to
560 the pre-3.7 situation by removing the kill_time argument.
561 * Add a MSG_process_set_kill_time() function instead.
564 * Fix weird behaviors when dealing with parallel tasks.
567 * Simgrid is now built as a dll.
568 * Simgrid-java now works on Windows.
569 * Simgrid-Java is now included into Windows package.
572 * First pre-build package for MacOSX.
575 * Fix compilation when using MSG_USE_DEPRECATED.
576 * Fix some compilation issues on Macs and Windows.
577 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
578 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
580 We discovered that the Lua console is broken, but we are missing the
581 manpower to fix it right now. The problem existed in 3.7 too, so we
582 are not blocking the release for that. Sorry if you depended on this
583 feature, any help would be really welcome.
585 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
587 SimGrid (3.7) stable; urgency=low
589 The "spring cleanups (before the next Big Project kicks in)" release.
592 * We can specify the SMPI latency/bandwidth factor with command line
593 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
594 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
595 You can also use the "config tag" from platform file by adding this line
596 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
597 (see "example/platforms/tag_config.xml" to use "config tag").
598 Note that the command line supersedes the platform file configuration.
599 * Change the correction factors used in LMM model, according to
600 the latest experiments described in INRIA RR-7821.
601 Accuracy should be improved this way.
602 * Use the partial invalidation optimization by default for the
603 network too. Should produce the exact same results, only faster.
604 * Major cleanup in surf to merge models and split some optimization
605 mechanisms from the core of the models. As a result you can now
606 specify which model to use (e.g., --cfg=network/model:LV08
607 --cfg=cpu/model:Cas01) and which optimization mode to use
608 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
609 Incompatible combinations should err at initialization. See
610 --help-models for the list of all models and optimization modes.
611 * The CLM03 workstation model was dropped for simplicity because it
612 used the deprecated CM02 network model. Use default instead.
613 * Rename the TCP_gamma configuration option to network/TCP_gamma
614 * Rename the coordinates configuration option to
615 network/coordinates, and document it
616 * Use now crosstraffic keyword instead of the terribly misleading
617 fullduplex keyword. It is activated by default now in the current
618 default model, use --cfg=network/crosstraffic:0 to turn it off.
619 * Ongoing refactoring the model parsing to make XML files optional
620 See include/simgrid/platf.h for details (still to be completed)
623 * Major overhaul of the documentation. Almost instructive now :/
624 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
625 functions and friends. This interface was considered as
626 deprecated since over 2 years, it's time to inform our users that it is.
627 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
628 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
629 use these (crappy) functions in your code.
630 These functions will be removed soon. Stop using them now.
631 * Deprecate MSG_get_host_{table,number}
632 Implement MSG_hosts_as_dynar() instead.
633 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
634 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
637 * Stabilize the parallel execution mode of user contexts
638 * Introduce configuration variables to control parallel execution:
639 - contexts/synchro: Synchronization mode to use when running
640 contexts in parallel (either futex, posix or busy_wait)
641 - contexts/parallel_threshold: Minimal number of user contexts
642 that must be part of a scheduling round to switch to parallel
643 execution mode (raw contexts only)
644 * Fix bugs that prevented to use suspend/resume along with
645 synchronization structures.
646 * Fix bugs in process termination that lead to invalid memory access
647 in very specific conditions.
650 * Introduce a parallel mode for the models (controlled by surf/nthreads
651 configuration item). In our tests, running the models in parallel
652 never lead to any speedups because they are so fast that the gain
653 of computing each model in parallel does not amortizes the
654 synchronization costs, even when ultra fast futexes are used.
655 This is released anyway because YMMV.
658 * Performance boost by using a swag internally to compute the set of
659 tasks that are finished and should constitute the return value of
663 * Enable it by default now that it is considered rather stable.
666 * Documentation of the tracing functions.
667 * Performance gains when tracing categorized/uncategorized resource
668 utilization by avoiding calls to get route when updating resource
669 variables. LMM constraints are being used instead.
670 * API changed to set task categories. Use MSG_task_set_category instead
671 of TRACE_msg_set_task_category, and SD_task_set_category instead
672 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
673 * Bugfix for graphicator, routes not correctly obtained, memory leaks
674 * Examples for link user variables added (see at examples/msg/tracing/)
675 * Deprecated function TRACE_msg_set_process_category completely removed
676 * Trace header updated according to the latest Paje file format
677 * Tracing network lazy updates, no longer obligate users to use full updates
678 * --cfg=tracing/platform:1 also registers power/bandwidth variables
679 * Experimental: let user code declare/set/push/pop application states for hosts
680 * API changed to allow the manual creation of graph configuration files
681 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
684 * Improve the API of Lua MSG bindings, using the Lua spirit.
685 * Each simulated process now lives in its own Lua world (globals are
686 automatically duplicated). It helps writing simulators. Will allow
687 to run Splay programs within SimGrid in the future.
688 * Add a Chord example in Lua, equivalent to the MSG one.
691 * Start the implementation of a solution to express temporal
692 properties, not only local assertions. This is still an
693 experimental work in progress, stay clear from it to be safe.
697 - Add new runtime parameters --help-logs and --help-log-categories
698 to display informations about supported logging parameters and
700 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
701 don't exists anymore.
702 * Mallocators: allow value NULL for the reset function.
704 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
705 create homogeneous dictionaries, where all the elements share the
706 same free function. Non homogeneous dictionaries will be
707 deprecated in the next release.
708 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
709 - Multi-level dictionaries are deprecated.
711 - new function xbt_dynar_search_or_negative() that is useful when
712 you have less than 2 million elements in your dynar and don't
713 want of the extra complexity of catching exceptions when the
714 element is not found.
716 - Make xbt_os_thread module (for thread portability) public.
717 Documentation is still to come, sorry.
719 - Cleanups and simplifications to make it maintainable again.
720 - Exotic features (such as memalign and valloc) were removed.
721 - The metadata were extended and improved so that the
722 model-checker becomes able to explore and inspect the heaps.
723 - This may induce a performance drop when enable_model-checking is
724 ON in cmake (even if it's not used in the simulation), but it is
725 necessary at this point to get MC working.
727 Turn model-checking OFF if simulation performance matters to you.
728 Not enabling it at runtime is not enough, disable it in cmake.
730 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
732 SimGrid (3.6.2) stable; urgency=low
734 The "Not coding new stuff allows to polish old things" release.
737 * New bindings to the NS3 packet level simulator (experimental)
738 * Use the raw (efficient) execution contexts instead of the sysv
739 (portable) ones when possible.
740 * libpcre is now mandatory in any cases since not using it led to
741 severe performance loss and possibly other issues
742 * Update the XML platforms:
743 - G5K: include the latest machine in Nancy
744 - GridPP and LCG: new platforms
745 * Documentation was partially updated, at least (more to come)
747 Bug fixes, cosmetics and small improvements
748 * Free terminated processes before the end of the simulation to avoid
749 exhausting the memory of users having very dynamic amount of
751 * Bug fix and cosmetics about canceling non-running tasks
752 * Bug fix about the dot loader's issues when using libcgraph
755 * Create an installer for windows with nsis (amd64 and win32)
756 - Add an hello world project to illustrate simgrid project creation.
757 - Embed libpcre into the Simgrid installer to avoid
758 its compilation burden
759 * The raw execution contexts should work on Apple now
760 * Port to Windows 64 bits
761 - Sysv contexts now have an implementation for this arch
762 - GRAS communication features now support this arch
763 * Drop support for borland compiler on windows
764 - this code was not maintained, and we kinda depend on gcc nowadays
765 * Fix portability issues on kfreebsd/gnu: build error about semaphores
766 * Fix portability issue on unstable ubuntu: linker became picky on
769 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
771 SimGrid (3.6.1) stable; urgency=low
773 The "Oops, we broke Macs too" release
776 * Fixed contexts detection so that raw ones are used when possible
777 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
778 leads to a strange error, with user code segfaulting sometimes when
779 the generated code is not perfectly aligned (which is not
780 controllable from the user side, depends on the amount of code)
783 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
785 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
787 SimGrid (3.6) stable; urgency=medium
789 The Summer Release, also known as the "OMG! They Killed Kenny!" version
792 * Bindings now constitute their own package, separated from the main one.
793 Rationale: reduce our maintenance nightmare by reducing the module coupling
794 They will soon be released on their own on gforge.
796 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
797 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
799 GRAS: It is not considered as stable anymore, but experimental. Sorry.
800 * It's not quite deprecated for now because we have no replacement,
801 but it may soon become the case.
804 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
805 * Fortran: New user-level cache variable to store the rank of the running
806 process. This improves performance by an order of magnitude.
807 * C: New coccinelle script to automatically locate and modify global and
808 local static variables.
809 * Improved SMPI network model with a sender-side gap to account for multiple
813 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
814 only say if a communication is finished, no matter whether it succeeded or
815 failed. You can call MSG_comm_get_status() to know the status of a finished
817 * New function MSG_task_dsend() to send a task and detach it. When a
818 communication is detached, you are never notified of its success or failure
819 and the memory is released automatically once it is finished. This function
820 is useful when you don't care about the end nor the success of a
822 * Change the prototypes of action replay. Sorry for inconvenience,
823 but this is really more efficient this way (and to adapt your code,
824 you just have to fix the initialization, that shouldn't be too long)
825 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
826 ever managed to do anything useful with it.
827 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
828 takes a dynar of communications. It returns immediately and gives the
829 index of a finished communication (if any).
830 * New example: a basic implementation of the Chord P2P algorithm.
833 * New model for multi-core CPUs. You can now use the core attribute to
834 precise the number of cores of a host. This is a basic model. Every
835 process running on the host receives at most the power provided in
836 the DTD (throughput<=power). Total throughput of process cannot exceed
838 * New peer tag. This peer tag creates a tiny AS comprising a host and a
839 router linked by an up-link and a down-link (possibly asymmetrical).
840 This kind of pattern allows to easily build last-mile model style platforms.
841 Aggregating such patterns in a rule-based AS is thus the technique of
842 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
843 * New model for Vivaldi routing. We transformed the Vivaldi network model
844 into a Vivaldi routing model (based on the rule-based model). This allows to
845 combine Vivaldi based latencies with last-mile platforms.
848 * Added a check for NaN of IEEE754 infinite in the double entries of
850 * Introduce a new context factory "raw", highly inspirated from the
851 ucontext factory, but using manually crafted functions in assembly to
852 do the work in an efficient manner.
853 * Allow to change the used context factory at run time, not only at
854 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
855 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
856 contexts at runtime (only with raw contexts or ucontexts).
857 * Completely rewrote this module to allow parallel execution of user
858 processes. Use --cfg=contexts/nthreads:N to execute user processes
859 with N parallel threads (the default is 1, meaning no parallelism).
860 * Allow to decide dynamically between sequential and parallel modes.
861 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
862 processes in parallel only when their number is greater than or equal to P
864 * Added a check for NaN of IEEE754 infinite in the double entries of
868 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
869 won't output the state of processes when interrupted with Ctrl-C
870 * Add a new function xbt_dynar_to_array that transforms a dynar into a
871 NULL-terminated array. This may solve backward compatibility issues
872 due to the change to return type of SD_simulate. See also:
873 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
874 * Add new macros with variable number of arguments.
875 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
876 - in xbt/asserts.h: xbt_assert
877 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
878 - in xbt/ex.h: THROWF and RETHROWF.
879 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
881 * Change xbt_die() to accept a format string with arguments, just like printf.
882 * New data structure: xbt_lib_t, like a dict but more general and with better
886 * New configuration options
887 Options triva/categorized and triva/uncategorized can be used to generate
888 graph configuration files for Triva visualization tool.
889 * Configuration option tracing/platform is renamed to tracing/categorized
890 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
891 * New TRACE_user_link_variable interface:
892 User provides the name of the link and the tracing variable to attach to it
893 * the declaration of tracing categories must be done after the environment
895 * simpler tracing interface, just one way to declare categories
896 TRACE_category or TRACE_category_with_color, it is up to you
897 * links in the trace file are again identified by their names
898 * trace contains the full platform hierarchy exactly as declared using the ASes
899 * Options tracing/msg/[task|process]:1 groups the process by hosts
900 for both cases, tasks and processes must have names that are unique during the simulation
901 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
902 * The experimental option tracing/msg/volume is deprecated
903 its functionality may be reincorporated if needed
905 The tracing generates a trace file with unordered timestamped events,
906 because of the way the core simulator (surf) works. A script available
907 at the tools directory (fix-paje-trace.sh) can be used to put the events
908 in order. We have changed the tracing so it can generate ordered timestamped
909 events in the final trace, but depending on the simulator (and how much time
910 is simulated) that can lead to a huge memory utilization. It is deactivated
911 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
914 * Define a SIMGRID_VERSION macro in simgrid_config.h.
915 - We are trying hard to keep the API stable, but it may happen that
916 some things change (we're a research project after all, not a
917 nuclear plant operating system). If such things should happen, you
918 could rely on that macro to adapt.
919 - current value: 30600 for 3.06.00, aka 3.6
920 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
921 a number that can be compared with SIMGRID_VERSION.
922 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
923 assertions and verbose/debug logging events are disabled at compile time.
925 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
927 SimGrid (3.5) stable; urgency=medium
930 * New feature to any SimGrid-based simulator: Model-Checking
931 Check SIN#1 for more details.
934 * New Model SMPI (three-interval linear regression for correction factors)
935 See RR-7426, available at http://hal.inria.fr/inria-00527150
936 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
937 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
938 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
939 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
940 * See also src/smpi/README
945 - Tracing API changes: TRACE_start and TRACE_end should not be called
946 by user-code. They are automatically called by simulators created
947 with SimDAG, MSG and SMPI if the toolkit is compiled with
948 tracing_enabled=ON. Categories declaration and utilization remain the
949 same for MSG and SimDag.
950 - A function was added to the tracing API to declare categories with
952 - TRACE_category_with_color (char *category, char *color)
953 where color must be in the following format
954 "%f %f %f", red, green, blue
955 and red, green, blue are float values in the interval [0, 1]
956 - User can specify NULL as color parameter, or continue calling
958 On that case, the tracing system will define random colors
959 - The following command-line options are supported:
960 --cfg=tracing/filename:msg.trace
961 --cfg=tracing:1 (activate tracing, needed to use others)
962 --cfg=tracing/platform:1 (categorized resource use)
963 --cfg=tracing/uncategorized:1 (uncategorized resource use)
964 --cfg=tracing/msg/task:1 (task creation)
965 --cfg=tracing/msg/process:1 (process creation, migration)
966 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
967 --cfg=tracing/smpi:1 (SMPI interface tracing)
968 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
969 - examples of examples/msg/tracing updated
971 - DAXLoader and DOTLoader functions can generate tasks with categories
972 - A new function to attribute a category to SD tasks:
973 TRACE_sd_set_task_category (SD_task_t task, char *category)
974 * Tracing the MPI interface implemented by SMPI
975 - Collective operations are traced with states
976 - Point-to-Point operations are traced with states/links
977 - Tracing activated by a parameter "-trace filename" passed
978 to smpirun during execution (considering that simgrid
979 is compiled with tracing enabled)
980 - To run the simulation with gdb, the simulator
981 accepts --cfg=tracing/smpi:1 to trace SMPI
982 - tesh files to check if smpi tracing is ok
983 - See examples/smpi/NAS/DT-trace
984 * GTNetS tracing re-worked
985 - adaptation to the tracing system of GTNets to cope
986 with modifications regarding the fullduplex mode
987 - new tesh files to check if gtnets tracing is ok
990 * Asynchronous communications through the functions:
991 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
992 * New function: MSG_load_platform_script()
993 to make possible using a lua script instead of XML files to set up platforms
994 * New function: MSG_set_function
995 to associate functions to processes, used when bypassing the parser
996 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
998 Platforms: Add some more examples in examples/platforms
999 * Grid'5000: see www.grid5000.fr
1000 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1003 * Change the XML format. This is a very important modification. SimGrid
1004 3.5 introduces a new hierarchical format based on the notion of
1005 Autonomous Systems. Compatibility with old format is ensured
1006 through the perl script provided in the install bin directory
1007 bin/simgrid_update_xml.
1008 It is now possible to build platforms with specific routing
1009 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1010 connect several platforms together. We will try to provide soon
1011 set of realistic platforms exploiting these properties (have a
1012 look at examples/platforms/ for the moment).
1013 * Take the opportunity of the XML format change to be a good XML citizen:
1014 rename link:ctn to link_ctn and similar changes (also dealt with by
1016 * Add a new routing scheme (rule-based) using regular expressions. It
1017 enables to have an extremely low memory footprint when the
1018 underlying routing is simple and can be compactly described. You
1019 need to have libpcre4-dev (perl regular expressions) installed if
1020 you want to use this routing scheme.
1021 * Revive the cluster TAG and allow to easily and efficiently (both in
1022 term of memory and speed) connect clusters together. Have a look
1023 at teshsuite/simdag/platforms/ to see how this can be done. With
1024 this tag, you can create clusters with thousands of tasks at no
1025 cost (have a look at examples/platforms/).
1026 Note: clusters are implemented as ASes, so there is no need for an
1027 enclosing AS tag if you have only one cluster in your platform.
1028 * Add new generic functions in the public interface that allows the user
1029 to call SURF 'create_resource' methods from your code (same
1030 functionality as the XML bypass mechanism but with a much lighter
1032 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1033 that uses a piecewise linear approximation to produce better
1034 results when exchanging small messages.
1035 * Add a new parameter to handle correctly full duplex link and account
1036 for interferences between uplink and downlink communications
1037 (activate with --cfg=fullduplex:1).
1040 * Rename the SD_READY (all dependencies are satisfied and task is
1041 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1042 dependencies are satisfied) state.
1043 This prevents a confusion between the notion of "ready to schedule"
1044 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1045 simulated" (SD_RUNNABLE) used by the simulation kernel.
1046 * Change the way a task is considered as ready. Instead of removing
1047 dependencies when a task is done, a counter is decreased. This way,
1048 it is always possible to reach ancestors thanks to the
1049 SD_taks_get_parents function (even after the end of the simulation.)
1050 * Change the return type of SD_Simulate from (SD_task_t*) into
1051 xbt_dynar_t. This function was in handling a dynar internally and
1052 converted it into a NULL terminated array for historical reasons.
1053 * New function SD_dotload(char*) to load a DAG described in dot
1054 format. This loader and the corresponding examples require the
1055 installation of the graphviz library.
1056 * Fix a bug in the management of tasks of size 0 in the surf network
1057 models. This problem was only visible with SIMDAG and you should
1058 thus disregard results produced with earlier versions if you
1059 relied on this feature (some tasks were blocked because of this).
1060 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1061 even though your applications were doing only point-to-point
1062 communications and sequential computations. Now you can really use any
1063 model you want (of course, if you create real parallel tasks, which are
1064 not implemented in most models beside ptaskL07, this will abort).
1065 * Add an example that schedules a DAX on an heterogeneous platform
1066 using a Min-Min strategy.
1067 * New function SD_workstation_get_current_task() that returns the kind
1068 of task currently running on a workstation in the sequential access
1070 * Raise some warnings when unexecuted tasks remains at the end of the
1071 simulation. This is usually caused by cycles in the DAG.
1074 * New function: SIMIX_process_set_function() called by MSG_set_function
1075 * Change the underlying waiting queue in semaphores so that a process
1076 can wait on several of them simultaneously (as in waitany).
1077 * Fix the way to handle tokens in semaphores so that all access patterns
1078 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1079 * kill the dirty pimple SIMIX_message_sizes_output()
1080 Please use (proper) visualization instead
1083 * New data container: setset (set of sets of elements)
1084 * New module: mmalloc (mapped malloc, allowing to have several
1085 independent segments of malloc)
1086 * New function: xbt_dict_cursor_set_data()
1087 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1088 * New function: xbt_dynar_is_empty()
1089 * New function: xbt_fifo_get_last_item()
1090 * Fix xbt_dynar_shrink(): use the right element size.
1091 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1092 uninitialized areas during expand.
1093 * Fix semaphores: previous implementation was severely broken.
1094 * Use library init/fini functions for our initialization.
1095 - you can use logs and other feature as soon as you want in your
1096 code (even before the xbt_init / MSG_init)
1097 - xbt_exit is now a no-op and produce a warning when used.
1100 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1101 benefit from the latest improvement to the simulation kernel.
1102 * Kill measurement sockets for now. If you rely on them, sorry. This
1103 release is not for you. This feature will be reintroduced in the
1104 future, but we cannot delay the release any further.
1105 * New function: gras_msgtype_get_name().
1106 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1107 * Fix (at last) the pmm example: it should not randomly fail anymore.
1109 Build chain: bug fixes and overall polishing
1110 * Cmake is now stable enough. Hence, we killed the autotools.
1111 * Port to windows ( TM :)
1112 * Fix the 'make install' target.
1113 No need to use 'make install-simgrid' anymore
1114 * Introduce a 'make dist' target compiling a *source* archive
1115 'make package' compiles a binary archive
1116 * Compile java files only on need
1117 * Add --cd and --setenv command line options to tesh
1118 * Out of source builds are not fully supported yet, but we are close
1119 * Enable supernovae and optimization flags by default for our users
1122 * Add layer to set up environment directly from lua, without XML.
1123 * The effect of gras_stub_generator can be achieved through
1124 lua too (check examples/gras/console/ping_generator.lua)
1126 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1128 SimGrid (3.4.1) stable; urgency=low
1130 The "Polishing easter eggs is probably a good idea" release.
1131 This is a bug fixes release only.
1135 * Fix a bug preventing the tasks from begin garbage collected.
1138 * Fix a bug occuring when a host involved in a communication fails.
1139 This was not detected properly by the other peer involved in the
1140 communication. Now, it's reported as a network error.
1143 * Warn the user about loop dependencies in data flow of DAX files
1144 * Obey the control-flow dependencies of DAX files
1147 * Add option "enable_smpi" allowing to not compile SMPI.
1148 Probably useful for the (Mac) users experiencing a build error here
1149 * Improve the detection of lua5.1 and ruby1.8
1151 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1153 SimGrid (3.4) stable; urgency=low
1155 The "Easter in Cargese" release. Also known as (major changes):
1157 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1158 ~> bindings were greatly improved
1159 ~> new tracing infrastructure for better visualization introduced
1161 * the "Welcome to configury modernity" release.
1162 ~> we switched from autotools to cmake, and improved our cdash
1165 A more detailled list of changes follow (full detail in svn log).
1167 Java Bindings: Various Cleanups
1168 * (install java-gcj-compat-dev on debian-like to use them)
1169 * Remove put/get: no need to export deprecated interface in Java
1170 Use send/receive instead.
1171 * Cleanup the examples and add a README per directory
1172 * Remove example autoDestination (that's the only way to go now)
1173 * Remove example explicitDestination (was a plain copy of basic)
1174 * Make JniException a runtime exception, so that there is no need to
1175 declare the fact that you may encounter such a beast. I guess that
1176 nobody will ever want to survive such error.
1177 * Create specific errors for each MSG case of failure:
1178 host failure, transfer failure, timeout, task cancelled
1179 * Cleanup the exceptions that may get thrown by each function
1180 * Other internal cleanups in Java bindings. Performance still bad :/
1181 Ruby and Lua Bindings: create them
1182 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1183 * That's new and great, you should try them out.
1184 Same functionalities than Java bindings, only even less polished
1186 * Kill the useless "rate" argument of SD_task_get_execution_time()
1187 Everyone used to provide -1 as a value, it was not used, and the
1188 semantic of a possible use wasn't even clear.
1189 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1190 either as comm costs or compute costs to mean that there is no
1191 such thing for that specific task.
1192 * Add a SD_task_set_name() function
1193 * Fix SD_task_unschedule() on typed tasks
1194 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1195 * In DAX loader, accept useless 'level' attributes to <job> since
1196 LIGO DAGs have them (seem to be to ease graphical representation).
1198 * Add an example masterslave_mailbox.c using send/receive and not
1199 the deprecated put/get interface.
1200 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1201 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1203 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1204 (declare MSG_USE_DEPRECATED to still have the old name)
1205 * Add a MSG_task_set_data() function
1206 * About trace replay (see examples/msg/actions):
1208 - Allow to work with splitted trace files for each process
1209 Give the specific trace file as argument of each process,
1210 and call MSG_action_trace_run(NULL)
1211 You can still have one merged file for all processes.
1212 - Fix implementation of collective operations
1213 * Allow task_execute() on 0-sized tasks (closes #10063)
1215 * This is the first release of SimGrid where SMPI is not considered
1216 beta anymore (even if some corners should still be improved)
1217 * Port over the new SIMIX_network submodule (internal refactoring)
1218 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1219 * Implement more missing elements of the standard:
1221 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1223 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1224 MPI_Scan() MPI_Get_processor_name()
1225 - Added implementation of missing case for Alltoall (warning: it's
1226 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1227 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1228 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1229 SMPI_MPI_Allgatherv()
1230 * Bug fixes include:
1231 - MPI_Waitsome() was broken
1232 - Allow relative includes in smpicc
1233 - Command line cfg argument 'reference_speed' was ignored...
1234 - Some functions did not properly lead to auto-benching of user code
1235 - smpicc passes -O2 by default (just like openmpi one)
1237 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1238 * Bug fixes about timeouts during communications
1239 * add SIMIX_message_sizes_output() as a pimple to write to file the
1240 amount of messages per size. Use gnuplot to get histogram.
1241 Pimple because that's the only user-visible function of simix,
1242 defined directly in xbt.h (irk, sorry)
1244 - Add a SIMIX_sem_get_capacity() function
1245 - Fix interactions with processe resume/suspende
1246 - release_forever() was stupidly broken
1247 - Fix SIMIX_display_process_status() for processes in a semaphore
1248 - Make SIMIX_sem_block_onto() user-visible
1249 * Refactoring context stuff:
1250 - Use pseudo-OOP for better modularity
1251 - reimplement SIMIX_process_kill() without process_schedule() so
1252 that the latter can take as invariant that it is called from
1254 - Merge context_start into context_new for sake of simplicity
1256 * Add a Vivaldi network model, coded live during SUD'10 ;)
1257 * Rename configuration variables to start a hierarchy:
1258 o cpu_model -> cpu/model
1259 o network_model -> network/model
1260 o workstation_model -> workstation/model
1261 * New configuration variables:
1262 o network/bandwidth_factor: correction to bandwith
1263 o network/latency_factor: correction to latency
1264 o netwotk/weight_S: correction to the weight of competing streams
1265 * Add a long description to the models, that users can see with such
1266 argument on the command line: --cfg=cpu/model:help
1267 * --help-models display the long description of all known models
1269 * config: add the ability to set a default value after registration
1270 Does not override any previously set value (e.g. from cmd line)
1271 * dict: allow to have integer key and data.
1272 When so, you need to use the following functions
1273 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1274 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1275 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1276 In contrary to regular dicts, the key is not malloced before copy.
1277 Mixing scalar and regular elements in the same dict is not tested
1279 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1280 Tracing for Visualization:
1281 * SimGrid is now instrumented in order to generate a trace file for
1282 visualization analysis: to use it, need to compile SimGrid with the
1283 "tracing" option enabled, and instrument the program using SimGrid with
1284 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1285 (among other functions).
1286 * The instrumentation only traces the platform utilization for now
1287 * Documentation to use the tracing functions and how to analyze the
1288 traces with the Triva tool is written.
1289 * More information about: SimGrid FAQ (in the section Tracing Simulations
1292 * We moved to cmake as default build system. Autotools support will
1293 be dropped soon. Check the FAQ for more info about how to use it.
1294 * Greatly improved our cdash/ctest interactions
1295 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1296 * Added memory checking tests with valgrind; lot of memleak fixing.
1297 This may be the first release of simgrid with so few memory issues
1298 * Added code coverage tests.
1299 Our coverage is still improvable, but at least we see it on cdash.
1301 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1303 SimGrid (3.3.4) stable; urgency=low
1305 The "Desktop Grid needs love too" release (also called Xmas release).
1307 Models improvements:
1308 * Major speedup in the maxmin system solving by using lazy evaluation
1309 Instead of solving completely the maxmin system at each iteration,
1310 only invalidate (and recompute) the modified parts.
1311 This new feature is enabled in default models but you can try to
1312 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1313 * Cas01 IMproved as default CPU model
1314 This CPU model is the same Cas01 model, but it uses the
1315 maxmin-selective-update flag and a heap structure to manage
1316 actions on SURF kernel.
1317 It reduces the complexity to find the next action to finish and,
1318 consequently, it's faster than the old Cas01.
1319 This is the new default CPU model (Cas01).
1320 * Rename the old Cas01 model to Cas01_fullupdate
1321 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1322 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1323 * CpuTI (CPU Trace Integration)
1324 A new CPU model whose objective is simulate faster when using
1325 availability trace files.
1326 Instead of using a full featured, over engineered maxmin system for
1327 CPU modeling, this model does the pre-integration of traces files
1328 to calculate the amount of CPU power available, and so, executes
1329 faster than the old CPU models.
1330 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1331 * Use LV08 as default network model since it gives better accuracy
1332 for small messages and shouldn't change things for big ones.
1333 Use --cfg=network_model:CM02 to get the previous behavior.
1336 ******************************************
1337 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1338 ******************************************
1339 * The new CPU model may changes simulations!
1340 The point is that events occurring at the exact same timestamp
1341 are not scheduled in the same order with the old and new
1342 version. This may be enough to completely change the execution
1343 of simulations in some cases.
1344 * The new network model will change simulations!
1345 This new model is more realistic than the previous one, so you
1346 should consider redoing your old experiments with this model.
1347 Sorry for the inconvenience.
1350 * Introduce the supernovae compilation mode
1351 When compiled that way, the whole SimGrid (or almost) is put in a
1352 single compilation unit and compiled in one shoot.
1353 This is to help gcc which has difficulties to inline stuff from one
1355 The speedup seem to be above 15%, althrough more tests are needed on
1356 amd64 to confirm that gain.
1359 * Port of MSG's mailbox on top of SIMIX network
1360 The put/get mechanism was greatly simplified on the way.
1363 * New SIMIX network module. Provides:
1364 - Mailbox: rendez-vous mecanism to find with who you want to speak
1365 - Synchronous send/recv: easier and hopefully faster since the
1366 logic is handled in the maestro process directly now
1367 - Asynchronous send/recv: you dreamt of it? It's here now
1368 Too bad that nobody cared enough to propagate the change to MSG.
1369 * Add semaphores as SIMIX synchronization mechanism.
1372 * new function SD_daxload(char*) to load a DAX file
1373 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1374 * Introduce typed tasks. Specify its kind and cost at creation.
1375 At scheduling, just give where it should be placed, and the cost
1376 for each involved resource is automatically computed.
1377 Existing constructors so far (more to come of course):
1378 - SD_task_create_comm_e2e() for end-to-end communication
1379 - SD_task_create_comp_seq() for sequential computation
1380 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1381 * new function SD_task_dump() for debuging display
1382 * new function SD_task_dotty(task,FILE*) writing to file the info
1383 about the task in dotty format
1384 * SD_task_dependency_exists() can now cope with having one of its
1385 arguments NULL. If so, it tests whether the other argument has any
1387 * Add getters on list of preceding/following tasks:
1388 SD_task_get_parents(task) and SD_task_get_children(task)
1389 * Add getters on amount of workstations and list:
1390 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1391 * Add getter on task kind: SD_task_get_kind(task)
1392 * Update the start_time and finish_time of tasks on completion/failure
1393 * Bugfix: Remove task from state swags when destroyed
1396 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1399 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1400 dynar with one separate thread per value of the dynar.
1401 * Change the prototype of xbt_thread_create(), sorry.
1402 Added a boolean parameter indicating whether we want to join this
1403 thread (used in SG only for now)
1404 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1407 * GTNetS wrappers should now be usable again (and betterly tested too)
1408 * Fix a major regression from 3.2 where the timeout provided to
1409 MSG_task_put_with_timeout() was used as absolute time before which
1410 the comm should be done.
1411 * Start to fix the <cluster> tag.
1412 - Internal links should be good now (beside of the loopback, which
1413 use the private link instead)
1414 - paths to the external world is still rather broken
1415 - the <route:multi> tag is just broken. Actually that's brain-dead.
1416 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1417 to make it less stupid
1418 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1419 * Fix a source-level compatibility glitch from 3.2: after defining
1420 MSG_USE_DEPRECATED, you can use the old name
1421 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1422 * Allow to compile from the SVN with automake 1.11
1423 * Fix some problems when using the "start_time" tag in deployment XMLs.
1424 * Fix #8569: XBT/synchro.h has redundant declarations
1425 * Fix #8563: MSG return values and exceptions
1426 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1427 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1428 Thanks to Robson Peixoto.
1429 * Fix a vicious bug in dictionaries inducing that some elements were
1430 not freed on xbt_dict_free()
1432 Portability report of this version:
1433 * Main portability targets:
1434 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1435 on (amd64/i386/ia64)
1436 - mac leopard on i386
1437 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1438 but nothing critical.
1439 * Other platforms: windows, AIX and others were not tested for this release
1441 Timing report of this version:
1442 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1443 scenario parameters). From 8h to a few seconds in desktop grid settings.
1444 * Supernovae brings about 25% speedup on i386.
1446 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1448 SimGrid (3.3.3) stable; urgency=low
1450 The "Need for Speed" release.
1452 The timings done to validate the 3.3.2 were faulty.
1453 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1455 The problem was a conversion from a manually handled vector to
1456 xbt_dynar_t on the critical path.
1457 xbt_dynar_foreach calls functions, inducing stack management crap.
1459 We inlined these functions and xbt_dynar_foreach is now breath taking.
1460 We also inlined xbt_swag_belong on the way.
1462 Here are some approximate speedup measurements (on master/slaves
1463 simulations lasting between 10s and 20s each):
1464 3.3.1 -> 3.3.2: about same performance
1465 3.3.2 -> 3.3.3: 40% speedup
1466 3.3.1 -> 3.3.3: 40% speedup
1467 3.3.1 with inline patch -> 3.3.3: 30% speedup
1469 Our reading is that the refactoring which occurred in 3.3.2 made us
1470 suffer much more from the xbt_dynar_foreach low performance, but
1471 once we solved this, this refactoring proved to be very performance
1472 effective. From the 40% speedup, somehow, 10% are due to the
1473 inlining and 30% to the refactoring.
1475 That's a pitty that gcc cannot inline functions placed in other files
1476 alone. We have to choose between:
1477 - break the encapsulation (by putting private data structures and
1478 accessors in headers files to help gcc)
1479 - live with low performance
1480 - switch to a decent compiler such as icc (not quite possible).
1482 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1484 SimGrid (3.3.2) stable; urgency=low
1486 The "Simplicity does not preceed complexity, but follows it" release.
1488 The main contributors of this release were (lexical order):
1489 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1492 * Extract the routing logic into its own object.
1493 (was dupplicated in network.c and workstation_LV07.c;
1494 Allows to implement other ways of storing that info)
1495 => kill now useless network_card concept
1496 - Use dynar to represent routes (instead of void** + int*)
1497 - kill link_set (use surf_network_model->resource_set instead)
1498 - Add a command-line option to choose the routing schema to use
1499 - Add three new models:
1500 * Floyd (shortest path computed at initialization)
1501 * Dijikstra (shortest path recomputed all the time)
1502 * Cached Dijikstra (shortest path computed on need)
1503 All these models where contributed by Silas De Munck, and are
1504 described in his ICCS09 paper.
1506 * Simplify model declaration
1507 (less redirections, less function to write when defining a model)
1508 - Factorize stuff between models:
1511 surf_model_resource_set(model)
1512 surf_model_resource_by_name(model, name)
1513 - Unify the types of models in s_surf_model_t (using an union)
1514 - Embeed fields of common_public directly into s_surf_model_t
1515 - Rename model methods:
1516 action_free ~> action_unref
1517 action_change_state ~> action_state_set
1518 action_get_state ~> action_state_get
1519 - Change model methods into functions :
1520 (model)->common_public->action_use ~> surf_action_ref
1522 * Implement a generic resource; use it as ancestor to specific ones
1523 (allows to kill duplicated code in models)
1524 Drawback: timer command don't need no name nor properties;
1525 workstation_CLM03 don't need no properties
1526 (but I guess we can live with those few bytes wasted)
1528 * Improve the action object model
1529 - implement a constructor avoiding dupplicated code about field
1530 initialization in generic_action part.
1532 * Kill the SDP model: it has an external dependency, is deprecated
1533 in flavor of modern lmm models, and didn't compile since a while
1536 * Relocation of the context module from XBT to SIMIX.
1537 (the context were decoupled from the simix processes, duplicating a lot of code)
1538 => a lot of code was factorized
1539 - less overhead is introduced during scheduling
1540 - simpler API for the context factory
1541 - the logic for process creation,destruction and manipulation was simplified
1542 * Simplification of the s_smx_process_t data structure.
1543 => accesing the simix level data associated to a process is faster now,
1544 and the code is a lot more readable.
1547 * Implement some more MPI primitives:
1548 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1549 -implementation: Bcast: flat or 2-ary tree (default),
1550 Barrier: 4-ary tree,
1552 Allreduce: Reduce then Bcast
1553 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1554 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1555 Alltoallv: flat tree, like ompi
1557 * Add support for optimized collectives (Bcast is now binomial by default)
1558 * Port smpirun and smpicc to OS X
1561 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1562 Shout out if you used it.
1565 * Display the list of still queued messages in SG mode when existing
1569 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1570 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1571 * Add xbt_str_from_file(FILE*)
1572 * Add xbt_dict_get_key achieving a linear reverse search
1573 * Remove the context module
1575 Portability report of this version:
1576 * Main portability targets:
1577 - Linux(debian)/x86/context
1578 - Linux(debian)/x86/pthreads
1579 - Linux(debian)/amd64/context
1580 - Linux(debian)/amd64/pthreads
1581 On these, we still have the eratic breakages of gras/pmm and
1582 amok/saturate_sg reported in previous version. We still think
1583 that the tests are the cause of the fault, not the tested code.
1585 - Mac OSX Leopard/x86/context
1586 Still false negative in tesh autotesting.
1587 Smpi still fails, but this time because readlink does not accept -f
1588 Everything seems to work properly beside of that.
1591 - AIX version 5.3 (only tested contexts this time)
1592 Smpi still fails there because mktemp is not installed.
1593 Everything seems to work properly beside of that.
1595 I managed to compile it for the first time, but several breakages.
1596 Won't delay the release for this exotic platform.
1598 * Windows: it's still lagging behind. If you want to help, please
1601 Timing report of this version:
1602 This version seem to be more than 5% faster than 3.3.1 (on linux
1603 64bits with contextes). The gain is less than expected, we are
1604 investigating this for next release.
1606 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1608 SimGrid (3.3.1) stable; urgency=low
1611 * Implement a --cfg-help to show existing configuration variables
1612 * Build chain do not require doxygen in maintainer mode
1615 * fix a bug on struct sizeof computation, which prevented the
1616 exchange of arrays of structs in some conditions
1617 - added a regression test about this in datadesc_usage
1618 * Allow the exchange of 0-long dynamic vectors.
1619 - for that, use -1 as indicator of dynamic size instead of 0
1620 - This implied to change any size from unsigned long to long,
1621 reducing a bit communication abilities, but I guess that with
1622 64bits being quite common, this is more than enough.
1623 - This also induce a protocol change, thus bumping network protocol
1624 version from 0 to 1 (if we have external users, we have to get
1625 clean on that point too ;)
1626 - added two regression tests about this in datadesc_usage
1627 * Be more verbose when propagating local exceptions
1628 This helps debugging.
1629 * Display the status of simulated processes when receiving SIGINT in
1633 * Allow to control the simulation from a trace file.
1634 New functions MSG_action_register() and MSG_action_trace_run()
1635 The first one allows to associate a function execution to each
1636 kind of action while the second one parses a trace file and
1637 triggers the corresponding actions within the system.
1638 For now, only a toy example is provided in examples/msg/actions
1639 * Add an exemple of process migration in examples/msg/migration
1640 * Fix a bug in task exchange which broke MSG_task_get_sender()
1641 Add a teshsuite regression test for that.
1642 [Bug: if MSG_task_get_sender() is called after sender exit,
1644 * Fix a bug which prevented suspend/resume to work properly
1645 * Display the status of simulated processes when receiving SIGINT
1646 This fixes a regression of v3.3. due to the introduction of SIMIX
1647 * Bug fixing in failure management:
1648 - trace could not start by a failure at time 0
1649 - failure during communications were not working
1652 * Add SIMIX_process_set_name() to change the name of the current
1653 process in the log messages.
1654 * Store smx_hosts in a dict since we only retrieve them by name
1655 * Move the configuration infrastructure to surf
1658 * Move the configuration infrastructure to surf
1661 * Massive internal cleanups:
1662 - Store internal structures on processes instead of hosts (allows
1663 to have more than one process per host, in addition of being more
1665 - Cleanup the initialization/finalization process
1666 - Kill a whole bunch of unneeded synchronization:
1667 processes run in exclusive manner within the simulator
1668 - Move queues from global tables to process data fields
1670 - now accept -platform and -hostfile arguments
1671 - Pass the right rank value to processes according to the hostfile
1672 * Compile the examples by default, and use them as regression tests
1673 * Implement MPI_Wtime()
1674 * Change the reference speed to a command line option
1677 * TCP_gamma can now be specified as command line option using
1678 --cfg=TCP_gamma:10000000.0
1679 * Change the --surf-path cmd line option into --cfg=path:
1682 * Also include strbuff from xbt.h public header
1683 * xbt_ex_display(): do not free the exception after displaying
1684 This allows to do more with the given exception afterward.
1685 Users should call xbt_ex_free() themselves.
1689 Portability report of this version:
1690 * Main portability targets:
1691 - Linux(debian)/x86/context
1692 - Linux(debian)/x86/pthreads
1693 - Linux(debian)/amd64/context
1694 - Linux(debian)/amd64/pthreads
1695 These targets fail about 1/10 of times on gras/pmm, but we believe
1696 that this is because of the test, not because of simgrid.
1697 amok/saturate_sg fails even more rarely, and the test may not be
1700 - Mac OSX Leopard/x86/context
1701 The test suite still spits tons of errors because some obscure
1702 force prevents us from removing the temporary directories
1703 arguing that they still contain some metadata I've never heard of.
1704 Smpi fails because seq is not installed.
1705 Everything seems to work properly beside of that.
1708 - AIX version 5.3 (both contexts and pthread)
1709 Smpi still fails there because mktemp is not installed.
1710 XML inclusions seems rosty on AIX.
1712 * Windows: it's still lagging behind. If you want to help, please
1715 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1717 SimGrid (3.3) stable; urgency=high
1721 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1722 [Malek Cherier & Mt]
1724 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1727 Until now, GRAS were using MSG as an interface to SURF. It was
1728 quite difficult because both interface have several differences
1729 (MSG channels vs GRAS sockets were the most notable point).
1731 This also opens the gate to SMPI (which should occur soon) and speed
1732 up simulations by to 40% (even if it were not the main goal).
1734 **************************************
1735 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1736 **************************************
1737 The point is that events occuring at the exact same timestamp are
1738 not scheduled in the same order with the old and new version. This
1739 may be enough to completely change the execution of simulations in
1740 some cases. Sorry for the inconvenience.
1742 * Cleanup and upgrade the XML format to push further scalability
1743 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1745 * Improve the testing infrastructure with tesh. Now a very large part of
1746 the code is tested not only by being run but also by checking that the
1747 output match an expected output [Mt].
1749 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1750 is really less memory-demanding, which should allow you to use
1751 larger files in SimGrid [AL].
1753 * Inform valgrind about our contextes, so that it becomes usable
1754 with the default (and more effecient) version of SimGrid
1755 [contributed by Sékou Diakite, many thanks]
1758 * Introduce a listener thread in charge of receiving incoming
1759 messages from the network. It allows to overlap communication and
1760 computation but most notably, it removes some stupid deadlocks due
1761 to the fact that so far, a process could not send and receive at
1762 the same time. This made most non trivial communication schema
1764 * Convert the PIDs from long int to int to match the MSG ones (and
1765 linux ones too) [Mt]
1766 * New function: gras_agent_spawn() to launch a new process on
1767 current host. Only working in simulation for now. [Mt]
1768 * New function: gras_os_hostport() returning a constant form (ie,
1769 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1772 * Make the backtrace of exceptions more human readable [Mt]
1773 * New module: xbt/str [Mt]
1774 a ton of string utility functions (split, join, printf to a newly
1775 allocated buffer, trim, etc)
1776 * New module: xbt/hash [Mt]
1777 SHA1 hashing algorithm (more to come if needed)
1778 * New module: xbt/synchro [Mt]
1779 synchronization tools (mutex and conditions) working the same way
1780 in simulation and in real life (mainly useful for GRAS, but not
1782 * New module: xbt/queue [Mt]
1783 classical producer/consumer synchronization scheme
1784 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1785 (using the classical functions will get serialized) [Mt]
1786 * Make dictionary internal table dynamic. No need to specify its size
1787 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1789 * Make sure the log channels are organized as a tree under windows
1790 (because of ANSI C compatibility issue, any channel were child of
1791 root directly) [Mt].
1794 * Cleaned many thing in surf and fixed a few bugs [AL].
1795 * Add a nice command line configuration mechanism to compose models [AL].
1796 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1797 the previous one (KCCFLN05). It relies on something that looks like
1798 a max-min sharing mechanism but cannot be written as such. A new solver
1799 was thus designed [AL].
1800 * Add a new solver to lmm. Based on Lagrange optimization and
1801 gradient-based descent, it enables to efficiently maximise systems s.a
1803 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1805 This solver enables to propose two new network models for TCP Reno and
1806 TCP Vegas based on Low's work. These models still need to be fully
1807 tested though [Pedro Velho].
1810 * Bug fix in SD_simulate. Now the time bound given as argument is
1812 * Use the new parallel task model (ptask_L07) as default.
1813 * Use the SURF command line configuration mechanism.
1814 * 0-size tasks (for synchronization) should now work.
1816 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1818 SimGrid (3.2) stable; urgency=high
1822 We still experience issues on this platform, but we believe that at
1823 least MSG is usable.
1825 GRAS API BREAKAGE (for simplification purpose, sorry):
1826 * the gras_msgtype_by_name is not used anymore. Instead of
1827 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1828 you can write (and must)
1829 gras_msg_send(toserver, "request", &request);
1830 - If you still want to pass a gras_msgtype_t to the function (to cache
1831 the type and avoid the lookup time), use the gras_msg_send_() variant.
1832 - Impacted functions:
1833 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1834 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1835 * The callbacks are now expected to return 0 when everything went well
1836 (just like the main() function)
1838 GRAS new features and improvements:
1839 * New module mecanism where user code can use per process globals [Mt]
1840 This is similar to gras_userdata_*() functions, but for libraries. It
1841 factorize some code developped over and over in the examples and AMOK.
1842 It has still to be documented and used (only amok/peermanagement is
1844 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1845 loss when they were small enough to fit into the buffer and sent quickly
1846 enough so that they can all get received in one shoot.
1847 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1848 if not found. Use the *_or_null() variant for the old semantic.
1849 * In gras_msg_handle, do not discard messages without callback.
1850 They are probably messages to be explicitly awaited later (ie, proofs of
1851 mis-synchronization in userland since they are sent before being awaited)
1852 No big deal usually.
1853 * gras_socket_meas_send/recv: semantic changed!
1854 The numerical arguments used to be (1) the total amount of data to send
1855 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1856 messages. This was need for the fool willing to send more than MAXINT
1857 bytes on quite fat pipes.
1860 * Do really rename the hostmanagement module to peermanagement. [Mt]
1861 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1862 but this is rather new and this was documented in the module
1863 documentation (poor excuses, I admit)
1864 * Bandwidth measurement semantic changed! This follows the changes to
1865 gras_socket_meas_send/recv explained above.
1868 * A sequential mode has been added to the workstations. When a workstation
1869 is in sequential mode, it can execute only one task, and the other tasks
1870 are waiting in a FIFO. [Christophe Thiery]
1873 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1874 model for SIMDAG. [Christophe Thiery]
1875 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1876 equal to 0, which caused some bad behaviors in
1877 saturated_constraint_set_update. I now use a threshold mechanism like in
1881 * When running manually src/testall, you select specific units [Mt]
1882 testall is the result of our cunit mecanism, and should replace all
1883 the scripty thingy around since bash don't run easily on billware.
1885 * A mallocator system has been added. [Christophe Thiery]
1886 Mallocators allow you to recycle your unused objects instead of freeing them
1887 and allocating new ones.
1889 Documentation update:
1890 * FAQ reworking + New FAQs:
1891 - "Valgrind spits tons of errors!" [Mt]
1892 - "How to repport bugs" [Mt]
1893 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1894 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1895 - Communication time measurement within MSG [AL]
1896 - I experience weird communication times when I change the latency [AL]
1897 * GRAS tutorial [Mt]
1899 - an introduction to the framework and to the used communication model
1900 - an initiatic tour introducing the most proheminent features:
1902 . Lesson 0: Installing GRAS
1903 . Lesson 1: Setting up your own project
1904 o Part 2: Message passing
1905 . Lesson 2: Exchanging simple messages
1906 . Lesson 3: Passing arguments to the processes (in SG)
1907 . Lesson 4: Attaching callbacks to messages
1908 . Lesson 5: Using globals in processes
1909 . Lesson 6: Logging informations properly
1910 . Lesson 7: Using internal timers
1911 . Lesson 8: Handling errors through exceptions
1912 . Lesson 9: Exchanging simple data
1913 . Lesson 10: Remote Procedure Calling (RPC)
1914 . Lesson 11: Explicitely waiting for messages
1915 . Recapping of message passing features in GRAS
1916 - A HOWTO section containing:
1917 o HOWTO design a GRAS application
1918 More are due, of course. They will come latter. In the meanwhile, you can
1919 check the examples which are still here.
1921 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1923 SimGrid (3.1) stable; urgency=high
1927 There was a stack corruption somewhere, visible only when optimizing
1928 with these versions. [Vince]
1931 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1932 module that enabled to play with Directed Acyclic Graphs. It is built
1933 directly on top of SURF and provides an API rather close to the old
1934 SG. Some old codes using SG are currently under rewrite to check that
1935 all needful functions are provided. [Christophe Thiery]
1938 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1939 extension of the classical CLM03 model that gracefully handles
1940 failures. This is now the default model for MSG and GRAS. It doesn't
1941 handle parallel tasks yet though. [AL]
1942 * Bug fix: Weights were not correctly set in the network part.
1943 WARNING: This may have resulted in incorrect results with simulations
1944 where there are more than one flow on a given link. [AL]
1947 * After a (long ?) discussion on simgrid-devel, we have decided that the
1948 convention we had on units was stupid. That is why it has been decided
1949 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1950 WARNING : This means that all previous platform files will not work as
1951 such with this version! A warning is issued to ask users to update
1953 A conversion script can be found in the contrib module of the CVS, under
1954 the name contrib/platform_generation/surfxml_update.pl [MQ]
1957 * Bug fix: Processes were started in reverse order, wrt deployment file.
1958 WARNING: if your code relies on this bug, please fix it. [AL]
1959 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1960 being executed on two different locations. [AL]
1961 * Bug fix: Failures are now better supported thanks to Derrick's tests
1962 (there was many failure situations I hadn't thought of and that weren't
1963 correctly handled). [AL]
1964 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1968 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1969 Exception occurring on server-side are propagated back to client (!).
1971 API CHANGE: the callback changed their prototype. Change:
1972 int my_handler(gras_socket_t expeditor, void *payload_data) {
1974 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1975 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1977 * New! function: gras_msg_handleall to deal with all messages arriving
1978 within a given period.
1979 * New! function: gras_socket_server_range to get a server socket in a
1980 range of port numbers (ease to avoid port number conflicts) [MQ]
1981 * New! gras processes display their backtrace when they get a SIGUSR1
1982 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1983 Sweet to debug RL processes [MQ]
1987 - Do not force experiment sizes to be expressed in kb, or it becomes
1988 impossible to measure the latency this way (needs one byte-long tests)
1989 WARNING: this changes the amok_bw_* function semantic. [MQ]
1990 - Implements the link saturation stuff. [MQ]
1991 * Peer management module:
1992 New! module factorizing code that we wrote over and over [MQ].
1995 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1996 - Test units are placed directly into the library code, they get extracted
1997 automatically and placed into the src/testall binary.
1998 - Convert most of the XBT tests to this system.
1999 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2000 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2001 live in the same C code. This required to change a little bit the API
2002 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2003 * New module: sparse graph structure with basic algorithms (this is work
2004 in progress and the API is not considered to be frozen yet). [AL]
2005 * Display more information on backtraces: source line & function names are
2006 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2007 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2008 * Reworked a little bit some #include statements to load only required
2009 headers. Some user code that relied on SimGrid to include stdlib or
2010 stdio may need to include it by themselves. [AL]
2011 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2013 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2014 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2015 but I guess that almost nobody used those parts. [MQ]
2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2019 SimGrid (3.0.1) stable; urgency=low
2022 * Unfortunately, I had missed 5 misnamed functions:
2023 xbt_fifo_item_t xbt_fifo_newitem(void);
2024 void xbt_fifo_freeitem(xbt_fifo_item_t);
2025 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2026 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2027 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2028 They're now deprecated. Please use their new versions:
2029 xbt_fifo_item_t xbt_fifo_new_item(void);
2030 void xbt_fifo_free_item(xbt_fifo_item_t);
2031 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2032 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2033 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2035 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2036 * Documentation: xbt_log module unmercifully reworked [MQ]
2037 * Bugfix: there was a problem with the ending of contexts with
2038 the pthread backend. It caused some weird deadlock or behavior
2039 depending on the pthread implementation. [AL]
2040 * Bugfix: get the exceptions raised in the simulator repport where
2041 and why they come from when they are not catched in time [AL, MQ]
2044 * Bugfix: Do repport the error when two non-connected hosts try to
2045 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2048 * Add additionnal checkings on communications. Assert that two
2049 communicating hosts are connected by a set of links... [AL]
2052 * Add additionnal checkings on channel values in communication [AL]
2053 * New: MSG_task_get_source to see on which host a task was generated [HC]
2054 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2055 the number of tasks waiting to be received on channel and sent
2057 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2058 waits for the first task coming from a given host.. [AL]
2060 GRAS new functionnalities: [MQ]
2061 * Enhance the parsing macro to allow the size of multidimentional objects
2062 to be given thru annotations.
2063 * New example (and documentation): Matrix Multiplication a la RPC
2064 (as when I was young!) and fix a bunch of bugs found on the way.
2066 GRAS performance improvements: [MQ]
2068 * Reduce the amount of cbps creation/destruction by making it static to
2069 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2072 * Change libdata to a set so that we can search for stuff by ID (and thus
2073 reduce the insane amount of dict lookups)
2076 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2077 RL communication on top of SG since it's so uneffective.
2078 It may also be used for inter-thread communication in RL, one day.
2079 * Use gras_datadesc_copy() to exchange messages on top of SG
2081 - improve message exchange performance on top of SG
2082 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2083 * Don't exchange on the network the size of the used part of buffer,
2084 instead, specify the possible buffer size to read().
2086 - reduces the amount of read/write calls (one pair per exchange)
2087 - reduces the amount of exchanged data (the size)
2088 - allows to retrieve all arrived data on receiver side, if we don't need
2089 it right now (subsequent read will peek the buffer)
2090 - allows the receiver to proceed with the begining of the stream before
2091 everything is arrived
2092 - make it possible to build an iov transport (using readv/writev)
2094 - take care of the data with non-stable storage (like stacked data),
2096 * If possible, TCP send uses vector I/O (when writev() is here)
2097 - Don't use it for receive since we send data sizes and data on the
2098 same stream, so we wouldn't be able to chain large amount of chunks
2099 before having to flush the stuff to read the size.
2100 * Rework the transport plugin mecanism to simplify it and reduce the
2101 amount of pointer dereferencement when searching for the right function
2104 * I guess that now, we do almost as few system calls as possible while
2105 doing as few data copy as possible.
2107 To improve it further, we could try to send all the sizes first and then
2108 all the data (to use iov on receiving size), but it's only a partial
2109 solution: when you have 2 dimensional data, the sizes of the second
2110 dimension is data of the first dimension, so you need 3 streams.
2112 I'm not sure the potential performance gains justify the coding burden.
2114 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2116 SimGrid (3.00) stable; urgency=high
2119 * New! Give the possibility to hijack the surf parser and thus bypass
2120 MSG_create_environment and MSG_launch_application. Have a look at
2121 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2123 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2125 SimGrid (2.96) unstable; urgency=low
2130 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2131 This deprecates the xbt_error_t mecanisms.
2132 It modifies (simplifies) all XBT and GRAS API.
2133 MSG API keeps unchanged (exceptions raised by XBT are catched from
2134 within MSG and masked with existing error handling facilities)
2137 * New! Add a FATPIPE model. [AL]
2138 * New! Add a parallel task model. [AL]
2139 * New! Add automatically a loopback interface (in the default
2140 network model) if none was precised.
2143 * Bugfix: MSG_process_resume now works with the current running process.
2145 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2146 * Modification of MSG_task_get_compute_duration. Once a task has been
2147 processed, the value returned by this function is now equal to 0. [AL]
2148 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2149 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2150 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2151 of a m_task. For now, MSG_task_cancel only works with computation
2153 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2154 of the processor (in Mflop/s) regardless of the current load on the
2156 * API Change: use proper naming convention for MSG_getClock and
2157 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2159 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2160 This function changes the priority of a computation task. This priority
2161 doesn't affect the transfer rate. A priority of 2 will make a task
2162 receive two times more cpu power than the other ones. This function
2163 has been added to suit the needs of Nguyen The Loc and hasn't been that
2164 much tested yet. So if it fails, please report it and send me your code.
2166 * API Change: removed all functions and types that were marked "deprecated"
2167 since many months. Renamed MSG_global_init_args to MSG_global_init.
2169 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2171 SimGrid (2.95) unstable; urgency=low
2174 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2175 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2176 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2179 * Bugfix: Allow absolute paths for platform description files [MQ]
2180 * Bugfix: do free the variables after use. Leads to drastic performance
2182 * Implement max_duration (ie, timeouts) on resources [AL]
2185 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2187 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2188 the needed support (timeouts) [AL]
2191 * Implement measurement sockets. You can now get the bandwidth between two
2192 hosts thanks to AMOK (see below). [MQ]
2193 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2194 dynar over the network (yeah) [MQ]
2195 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2196 * simple/static token-ring example. [Alexandre Colucci and MQ]
2197 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2198 in place before [MQ]
2201 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2202 type parsing. "short" and "long" modifiers were also ignored; other
2203 modifier (such as reference level) are still ignored. [MQ]
2204 * Embeed the buffer size within the buffer itself on SG. [MQ]
2205 That way, send() are atomic and cannot get intermixed anymore (at least
2206 the ones which are less than 100k; bigger messages still have the issue)
2207 * Array size pushed by the field, not by the field type (or each
2208 and every long int will push stuff to the cbps) [MQ]
2209 * use select() to sleep since it allows to portably sleep less than one
2212 GRAS (minor cleanups)
2213 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2216 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2217 * type category 'ignored' killed as it was never used and were difficult
2219 * whether a type can cycle or not is now a flag, leaving room for more
2220 flags (as "ignored", if we feel the need one day ;)
2221 * Rename raw sockets to measurement sockets since "raw" has another
2222 meaning in networking community.
2225 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2226 and offers features not belonging to GRAS but that most applications
2227 need. One day, it may be a set of plugins loadable at runtime.
2228 * New module: bandwidth
2229 bandwidth measurement between arbitrary nodes running this module. [MQ]
2231 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2233 SimGrid (2.94) unstable; urgency=low
2235 The first beta release of SimGrid 3 !
2239 * Update the main page and the FAQ. Adding references to gforge.
2242 * Add a gras_os_getpid function.
2245 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2246 * Extend the logs so that they also print PID, hostname, date, ... if
2248 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2249 and kill the old version which were in testsuite/
2250 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2254 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2255 for users. Instead, each programming environment (GRAS, MSG,...) use it
2256 when needed to provide such a feature to users.
2257 Don't shortcut the mecanism or you will also shortcut the virtualization
2258 you need on the simulator.
2262 * Cleanups in configury with regard to compile optimization/warning flags.
2263 Also add -fno-loop-optimize to any powerpc since it's the optimization
2264 killing gcc (< 3.4.0).
2265 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2266 needed by MSG examples complications
2267 * Borrow configury beautifications from PHP
2270 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2271 warning (thanks loris for stumbling into this one).
2272 * Bugfix: stop loading private headers (gras_config.h) from the public
2276 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2277 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2278 RL would imply a DNS resolver, which is *hard* to do in a portable way
2279 (and therefore delayed).
2280 * Implement a real timer mecanism and use it in timing macros. This allows
2281 to avoid rounding errors and get a 0.000005 sec precision in timing
2282 macros. While I was at it, various cleanups:
2283 - allow to declare more than one timed section per file (fix a stupid bug)
2284 - move some private declaration to the right place
2285 - merge conditional execution and timing macros into emulation module
2286 - document the module
2287 - make sure the module cleanups its mess on gras_exit
2288 * Documentation improvements:
2289 - (new) how to compile applications using GRAS
2290 - (new) emulation support (timing macros)
2292 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2294 SimGrid (2.93) unstable; urgency=low
2296 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2299 - Use Paje properly where used. Still to be sanitized properly.
2300 - Portability fix: Add an implementation of the contexts using pthread
2304 - Add xbt_procname(): returns the name of the current process.
2305 Use it to show the current process's name in all logging.
2307 - fix detection of older flex version and the reaction, since we do
2308 depend on modern ones (we use lex_destroy)
2309 - Better separation of SG and RL in the libs: remove all simulation code
2310 from libgras. As a result, this lib is now only 200k when stripped.
2311 Some of the xbt modules may also be duplicated (two sets and such) and
2312 should be cleaned/killed before SG3.
2313 - Insist on using xlC on AIX because of weird problems involving gcc there.
2314 - Cleanup the make remote stuff. This is now done by scripts
2315 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2316 mainly for our private use, but we're working on changing them to user
2319 - Bugfix: flush the socket on close only if there is some *output*.
2320 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2322 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2325 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2327 SimGrid (2.92) unstable; urgency=low
2329 Alpha 3 on the path to SimGrid 3
2333 - New! First try of benchmarking macros.
2334 - New! First try so that gras_stub_generator generate deployment and
2335 remote compilation helpers.
2337 - Bugfix: Initialization fix in msg_test.
2341 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2343 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2345 - API Change: gras_os_sleep to take the amount of seconds as a double.
2346 Accepting an int was error prone since it was the only location where
2347 seconds were coded as such. It leaded to damn rounding errors.
2348 - Bugfix: Hard to belive that timers ever worked before this.
2350 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2352 SimGrid (2.91) unstable; urgency=low
2354 Alpha 2 on the path to SimGrid 3
2358 - Bug fix in the lmm_solver.
2360 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2361 through the function MSG_paje_output.
2362 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2363 - It is possible to bound the rate of a communication in MSG with
2364 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2365 to write it in the changelog).
2366 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2371 - Various cleanups to the autotools stuff
2372 - Begin to move Gras examples to examples/gras/
2373 - Let make distcheck work again (yeah!)
2375 - documentation overhauled using doxygen.
2376 gtk-doc-tools is dead in SimGrid now.
2377 - Automatically extract all existing logging categories, and add the list
2378 to the documentation (long standing one, to say the less)
2380 - Cleanup the known architecture table. Reorder the entries to group what
2381 should be, and use a more consistent naming scheme.
2382 (some of the test dataset are still to be regenerated)
2383 - New! Allow library to register globals on each process just as userdata
2385 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2386 lookup time (for now).
2387 Use it in msg and trp.
2388 This cleans a lot the internals and helps enforcing privacy of the
2389 headers between the gras components.
2390 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2391 - Bugfix: gras_os_time was delirious in RL.
2392 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2394 - Reenable GRAS now that it works.
2396 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2398 SimGrid (2.90) unstable; urgency=low
2400 Alpha 1 on the path to SimGrid 3
2402 * It is a long time since the last release of SimGrid. I'm sorry about
2403 that but as I had told you, I was rewriting a lot of things. I apologize
2404 to those who had been reporting bugs to me and that I had not answered.
2405 If your bug is still in the new version, please tell me. Here is a
2406 summary of the main changes.
2408 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2409 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2410 portability, speed, and a lot more but you'll figure it out later.
2411 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2412 I wanted to release the new MSG as soon as possible and I have
2413 broken GRAS, which is the reason why, for now, only MSG is fully
2414 functional. A laconic description of these projects is available
2415 in the documentation.
2417 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2418 kernel optimized for our needs (called SURF but only the developers
2419 should use it). Hence, MSG is now roughly 30 times faster and I think
2420 that by rewriting a little bit MSG, I could event speed it up a little
2421 bit more. Beside the gain in speed, it is also much easier to encode a
2422 new platform model with SURF than it was with SG. More to come...
2424 * REVOLUTION 3: I have tried to change a little as possible the API of
2425 MSG but a few things really had to disappear. The main differences
2426 with the previous version are :
2427 1) no more m_links_t and the corresponding functions. Platforms are
2428 directly read from a XML description and cannot be hard-coded
2429 anymore. The same format is used for application deployment
2430 description. The new format is described in the documentation.
2431 Have a look in tools/platform_generation. There is a tiny script
2432 that converts from the old platform format to the new one. Concerning
2433 the application deployment format, parsing the old one is tricky.
2434 I think most of you should however be able to convert your files. If
2435 it is really an issue, I can write a C code that does the conversion.
2437 2) the toolbox tbx does not exist anymore. We now have a library
2438 with much more data-structures but without the hash-tables (we have
2439 dictionaries that are much faster).
2441 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2443 *****************************************************************************
2444 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2445 * don't feel like converting the oldies. (Mt) *
2446 *****************************************************************************
2449 Version 2.90: "the long awaited one"
2450 - Finished rewriting and debugging MSG. Rewrote the documentation.
2451 - disable GRAS for now since it needs to be ported to the newest SG
2454 - Finish the port to windows (using mingw32 for cross-compile)
2457 - Main loop and datastructures of SURF. A cpu resource object is
2458 functional. Surf can thus be used to create cpu's with variable
2459 performance on which you can execute some actions.
2461 2004-11-15 Martin Quinson
2462 - Port to ARM. Simply added the alignment and size descriptions. Should
2463 work, but the ARM machines are so slow that I didn't had the opportunity
2464 to 'make check' over there yet.
2466 2004-11-15 Arnaud Legrand
2467 - Trace manager now written. It uses a heap structure and is therefore
2468 expected to be efficient. It may however be speeded up (particularly
2469 when many events occur at the same date) by using red and black
2470 trees. One day maybe...
2471 - Max-min linear system solver written. It uses a sparse matrix
2472 structure taking advantage of its expected use. Most operations are
2473 O(1) and free/calloc are called as few as possible. The computation of
2474 the minimum could however be improved by using a red and black tree
2477 2004-11-03 Arnaud Legrand
2478 - Rename every gras_* function that was in xbt/ to its xbt_
2480 - Add a heap and a doubly-linked list to xbt
2481 - Added a dichotomy to the dictionaries. make check works as well before
2482 so I assume that the patch is correct. I do not know however if things
2483 run effectively faster than before now. :)
2485 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2486 SimGrid, and the version number is bumped to 2.x
2488 2004-10-29 Martin Quinson
2489 - Introduction of the remote errors.
2490 They are the result of a RMI/RPC on the remote machine.
2491 ErrCodes being scalar values, you can't get the host on which those
2492 errors did happen. Extending the error mechanism as in Gnome is possible.
2493 No idea yet whether it is a good idea.
2495 2004-10-28 Martin Quinson
2496 - Interface revolution: the Starred Structure Eradication.
2497 I used to do typedef struct {} toto_t; and then handle *toto_t.
2498 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2499 - ???_t is a valid type (builded with typedef)
2500 - s_toto_t is a structure (access to fields with .)
2501 - s_toto is a structure needing 'struct' keyword to be used
2502 - e_toto_t is an enum
2503 - toto_t is an 'object' (struct*)
2505 typedef struct s_toto {} s_toto_t, *toto_t;
2506 typedef enum {} e_toto_t;
2507 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2508 s_toto_t) is private.
2510 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2511 so that it can set the variable to NULL. It was so for dicts and sets,
2512 it changed for dynars.
2514 - Fix a bunch of memleaks in dict_remove
2515 - Fix a bug in sg/server_socket opening: it failed all the time.
2517 2004-10-07 Martin Quinson
2518 - Speed up dynar lookup operation a bit.
2520 gras_dynar_get is dead.
2522 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2523 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2524 (which gives you the address of the stored data).
2526 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2527 copy of the data using an affectation to do the job and not a memcpy.
2529 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2530 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2532 It does not really speedup the dynar test because they are
2533 setting elements all the time (and look them seldom). But the dict does
2534 far more lookup than setting.
2536 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2538 2004-10-05 Martin Quinson
2539 - Allow to (en/dis)able the cycle detection at run time.
2541 Whether we should check for cycle or not is now a property of each
2542 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2543 datadesc_cycle_unset allow to remove this property when previously set.
2545 Note that the cycle detection is off by default since it impacts the
2546 performance. Watch the data you feed GRAS with ;)
2548 This property is hereditary. Any element embedded in a structure having it
2549 set have it set for the time of this data exchange.
2551 You should set it both on sender and receiver side. If you don't set it on
2552 sender side, it will enter an endless loop. If you forget on receiver
2553 side, the cycles won't be recreated after communication.
2555 - Header reorganization.
2556 Kill gras_private.h, each submodule must load the headers it needs.
2558 2004-10-04 Martin Quinson
2559 - Interface revolution: do not try to survive to malloc failure.
2561 Now, gras_malloc and friends call gras_abort() on failure.
2562 As a conclusion, malloc_error is not a valid error anymore, and all
2563 functions for which it was the only gras_error_t return value are
2564 changed. They now return void, or there result directly.
2565 This simplify the API a lot.
2567 2004-09-29 Martin Quinson
2568 - Re-enable raw sockets.
2569 Created by gras_socket_{client,server}_ext;
2570 Used with gras_raw_{send,recv}
2573 It should allow to kill the last bits of gras first version soon.
2575 This is not completely satisfactory yet (duplicate code with
2576 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2579 - Simplify transport plugin (internal) interface by not passing any
2580 argument to _server and _client, but embedding them in the socket
2583 2004-09-28 Martin Quinson
2584 - Finish the port to AIX.
2585 autoconf was my problem (segfault within the malloc replacement
2586 function. No idea why)
2588 2004-09-16 Martin Quinson
2589 - Fix some size_t madness on 64bit architectures.
2591 2004-09-08 Martin Quinson
2592 - Reduce the number of system headers loaded, overload some more system
2593 calls (such as malloc to cast the result of the system one, and work
2595 - Fix and reintroduce the config support
2597 2004-09-07 Martin Quinson
2598 - Source code reorganization to allow Arnaud to surf all over there.
2599 - Allow to document the logging categories.
2600 - Remove all uppercase from logging categories and useless cleanup in names.
2602 2004-08-18 Martin Quinson
2603 Version 0.6.2 (protocol not changed; API changed)
2604 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2605 gras_error_t), and NULL when not found. Functions expecting a msgtype
2606 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2607 hopefully usefull message.
2608 - Portability to prehistoric sparcs again
2610 2004-08-17 Martin Quinson
2611 Version 0.6.1 (protocol not changed; ABI not changed)
2612 - prealloc some buffers to speed things up
2614 2004-08-11 Martin Quinson
2615 Version 0.6 (protocol not changed; ABI expended)
2616 - The parsing macro can deal with the references, provided that you add
2617 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2619 2004-08-09 Martin Quinson
2620 Version 0.5 (protocol not changed; ABI changed)
2621 - Allow to off turn the cycle detection code in data exchange at
2622 compilation time. It should be at run time, but I'm short of time (and
2623 the config stuff is still broken). That way, we keep dict out of the
2624 critical path, which is good because the performance is poor:
2625 - search not dichotomial yet
2626 - dynar give no way to access their content and memcpy everytime
2627 - In composed data description (struct, ref and so on), stop foolness of
2628 keeping the subtype's ID, but store the type itself. This keeps sets out
2629 of the critical path, which is good since they rely on dynar and
2630 dictionnaries. The only loose of that is that we cannot detect the
2631 redeclaration of a structure/union with another content (but I'm not sure
2632 the code detected well this error before anyway). We still can detect
2633 the redefinition discrepancy for the other types.
2634 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2635 it breaks the code because of type-punning used all over the place).
2636 This breaks on all non-gcc architectures (for now).
2638 All those changes (plus the buffer of last time) allow me to gain 2 order
2639 of magnitude on cruel tests consisting of 800000 array of integers on two
2640 level of a hierarchical structure (200 secondes -> 4 secondes)
2643 - the selector of reference must now return the type it points to, not
2644 the ID of this type.
2646 2004-08-06 Martin Quinson
2647 Version 0.4 (protocol changed; ABI not changed)
2648 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2649 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2650 - Implement a buffer transport, and use it by default (it relies on tcp in
2651 real life and on sg in simulation).
2652 That's a bit hackish since I had a new field to the structure to store
2653 its data without interfering with the subtype ones. Inheritance
2654 is tricky in C. And that's a kind of reverse inheritance with one class
2655 derivating two classes. Or maybe a game with java interfaces. Anyway,
2656 that's damn hard in C (at least).
2657 Moreover, I got tired while trying to ensure plugin separation and
2658 genericity in SG mode. MSG wants me to do weird things, so let's go for
2659 cruel hacks (temporarily of course ;).
2660 See comment in transport_private.h:71
2661 - do not include all the _interface headers in private but in the files
2662 which really need them (to cut the compilation time when they are
2665 2004-07-26 Martin Quinson
2666 Version 0.3 (protocol not changed; ABI changed)
2667 - Major overhault of the datadesc interface to simplify it:
2668 - shorted the function names:
2669 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2670 - add a trivial way to push/pop integers into the cbps without malloc.
2671 This allows to make really generic sub_type description, which simply
2672 pop their size of the stack.
2673 - add a function gras_datadesc_ref_pop_arr() which does what users want
2674 most of the time: Declare a dynamic array (which pops its size of the
2675 stack) and declare a reference to it. Poor name, but anyway.
2676 - kill the post-send callback, add a post-receive one
2678 2004-07-23 Martin Quinson
2679 Version 0.2 (protocol changed; ABI changed)
2680 - add some testing for cpbs in the test cases, and fix some more bugs.
2681 This invalidate again the little64 data file, since I cannot regenerate
2683 - remove an awfull optimization in the logging stuff, allowing me to:
2684 - understand it again
2685 - learn gcc how to check that the argument match the provided format
2686 - fix all errors revealed by gcc after that
2687 - internal keys of dict are not \0 terminated. Deal with it properly in
2688 loggings instead of segfaulting when the user want to see the logs :-/
2690 2004-07-22 Martin Quinson
2691 - Fix some stupid bug preventing cbps (callback postit) from working
2693 2004-07-21 Martin Quinson
2694 - Some documentation cleanups
2695 - remove the useless last argument of msgtype_declare
2696 - rename the Virtu functions to fit into the 'os' namespace
2697 - move headers src/include -> src/include/gras/ and stop fooling with
2698 gras -> . symbolic link
2699 - make distcheck is now successful
2701 2004-07-19 Martin Quinson
2703 - Build shared library also
2704 - Install html doc to the right location
2705 - stop removing maintainer files in make clean
2706 - build tests only on make check
2708 2004-07-13 Martin Quinson
2710 - No major issue in previous version => change versionning schema
2711 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2712 corresponding dataset.
2714 2004-07-11 Martin Quinson
2716 - Get it working with any kind of structure (we can compute the padding
2717 bytes remotely for all the architectures I have access to)
2718 - Implement the structure parsing macro (still not quite robust/complete)
2719 - Improvement to the remote testing toysuite
2721 2004-07-10 Martin Quinson
2722 [autoconf mechanism]
2723 - get ride of a bunch of deprecated macros
2724 - actually run the test for two-compliment, not only compile it :-/
2725 - test whether the structures get packed (and bail out if yes. Damn.
2726 Alignment is a serious matter)
2727 - test whether the structures get compacted (but respecting the alignment
2728 constraints of each types)
2729 - test whether the array fields of structures can straddle alignment boundaries
2731 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2732 (took me 2 hours to find that bug, looking at the wrong place)
2733 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2734 before sending/receiving and are used to compute the offsets of fields
2735 - Given that padding size depend even on compiler options, keep track of
2736 alignment and aligned_size only for the current architecture. Not a big
2737 deal since we send structure fields one after the other (seems
2739 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2740 program, let it work on linux/gcc/little32. portability todo.
2742 2004-07-08 Martin Quinson
2743 - import and improve remote compilation support from FAST
2744 - make sure make check works on half a dozen of machines out there
2746 2004-07-07 Martin Quinson
2747 Let's say it's version 0.0.3 ;)
2748 - Implement conversions (yuhu!)
2749 - Let it work on solaris (beside conversion, of course)
2750 - Stupid me, using rand() to generate the conversion datatests in not wise.
2752 2004-07-06 Martin Quinson
2753 - Let make dist work, since I'm gonna need it to compile on remote hosts
2754 - Let Tests/datadesc_usage write the architecture on which the file was
2755 generated as first byte.
2756 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2757 architecture descriptions.
2758 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2759 on those architectures.
2760 - Optimization: send/recv array of scalar in one shoot
2762 2004-07-05 Martin Quinson
2763 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2765 - Plug a whole bunch of memleaks
2766 - each process now have to call gras_{init,exit}. One day, their log
2767 settings will be separated
2768 - Continue the code factorisation between SG, RL and common in Transport.
2770 2004-07-04 Martin Quinson
2772 - Redistribution between SG and RL.
2773 We wanna have to accept in SG, so move accepted related parts of RL in
2774 the common part. (more precisely, the dynar of all known sockets is no
2775 more a static in transport.c, but part of the process_data)
2777 [gras_stub_generator]
2778 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2780 2004-07-03 Martin Quinson
2781 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2782 them all up in one shot)
2784 - Fix the ugly case of reference to dynamic array.
2785 - New (semi-public) function gras_datadesc_size to allow the messaging
2786 layer to malloc the needed space for the buffer.
2788 - gras_socket_close now expect the socket to close (and not its address to
2789 put NULL in it after it). This is because the socket passed to handlers
2790 is one of their argument (=> not writable).
2792 - propagate the interface cleanup from last week in datadesc, ie remove a
2793 superfluous level of indirection. User pass adress of variable
2794 containing data (both when sending and receiving), and not of a variable
2795 being a pointer to the data. Let's say that I like it better ;)
2796 The price for that is constructs like "int msg=*(int*)payload" in
2797 handlers, but it's a fine price, IMHO.
2799 - Let it work in RL (yuhu)
2801 2004-06-21 Martin Quinson
2803 - porting SG plugin and SG select to new standards (works almost).
2804 - plug memleaks and fix bugs around.
2807 - cleanup the prototype of data recv and force users to specify when they
2808 want to handle references to objects. Test case working even for cycles.
2809 - plug memleaks. Valgrind is perfectly ok with this.
2811 2004-06-12 Martin Quinson
2813 - cleanup the separation between plugin and main code in plugin creation
2815 2004-06-11 Martin Quinson
2817 - Reput hook for raw sockets, needed for BW experiments
2818 - kill a few lines of dead code
2819 [Data description] Interface cleanup
2820 - gras_datadesc_by_name returns the searched type or NULL.
2821 That way, no variable is needed to use a type desc once, which makes
2823 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2824 two parameters were strings (field name, type name), leading to
2826 [Dicos] Interface cleanup
2827 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2828 This is consistant with the dynar API.
2830 2004-04-21 Martin Quinson
2832 - Porting to new standards.
2834 - interface cleanup.
2835 There is no bag anymore, no need to take extra provision to mask the
2836 pointers behind "ID".
2837 Better splitup of functions between files create/exchange/convert.
2838 This is still a bit artificial since convert and receive are so
2839 interleaved, but anyway.
2841 - add a queued message list to procdata (the ones not matching criteria
2843 - factorize some more code between SG and RL wrt procdata
2845 - use gras_exit in example to track memleaks
2846 - get rid of gs_example now that GS is properly integrated into gras
2847 - update run_test to integrate the latest tests (datadesc)
2849 - rename WARNINGn macros to WARNn since it prooved error-prone
2851 2004-04-19 Martin Quinson
2853 - register init/exit functions within gras module mechanism
2854 - send/receive function.
2855 Convertion is not implemented, but short-cutted if not needed.
2856 struct/array elements are sent one by one (instead of block-wise), but
2857 nobody really cares (yet). Get a prototype before optimizing.
2858 - tests (using a file socket) for DD send/receive on:
2859 - base types: int, float
2860 - array: fixed size, string (ie ref to dynamic string)
2861 - structure: homogeneous, heterogeneous
2862 - chained list, graph with cycle
2863 Believe it or not, valgrind is not too unhappy with the results. The
2864 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2868 - Bugfix when using a filename explicitely (instead of '-')
2870 2004-04-09 Martin Quinson
2872 - factorize more code between RL and SG in socket creation
2873 - Complete the implementation and tests of:
2875 o file (only in RL, and mainly for debugging)
2877 I lost 3 days to design a portable address resolver, and then decided
2878 that the prototype mainly have to run on my box.
2879 Addressing portability too early may be like optimizing too early :-/
2881 - use gras_init in the Tests instead of the crappy parse_log_opt
2882 (the latter function is removed)
2883 [Conditional execution]
2884 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2885 [Code reorganisation]
2886 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2887 Build examples against the RL library, since there is no way to disable
2888 its creation for now.
2890 For information, the beginning of coding on GRAS was back in june
2891 2003. I guess that every line has been rewritten at least twice since