1 SimGrid (3.12) NOT RELEASED; urgency=low
4 * Interface improvement:
5 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
7 * Interface improvement:
8 - Rename Host.isAvail() to Host.isOn()
11 - SIMIX_process_throw: raises an exception in a remote process
14 - 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.
15 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
16 - MPI_Type_set_name, MPI_Type_get_name
17 - MPI_*_c2f and MPI_*_f2c functions
18 - Activate a lot of new tests from the mpich 3 testsuite
20 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
21 - InfiniBand network model added : Based on the works of Jerome Vienne (http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf )
22 - When smpi/display_timing is set, also display global simulation time and application times
23 * Collective communications
24 - 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
25 - 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/).
26 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
28 - Replay now uses algorithms from wanted collective selector
29 - Replay can be used with SMP-aware algorithms
30 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
31 - 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.
33 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
34 - Correctly use loopback on fat-tree clusters
35 - Asynchronous small messages shouldn't trigger deadlocks anymore
38 - "Full" network optimization flag was broken since Surf++
39 - Better handling of precision flags in maxmin
40 - Fix bug causing sometimes "Impossible" errors
43 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
44 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
47 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
49 SimGrid (3.11) stable; urgency=low
54 * Normalizing pointers addresses tool for better diff between logs
57 * Add cloud examples using new VMs
58 - examples/msg/cloud/two_tasks_vm.tesh
59 - examples/msg/cloud/simple_vm.tesh
60 - examples/java/cloud/cloud.tesh
61 - examples/java/cloud/migration/migration.tesh
62 * Add java surf examples:
63 - examples/java/surfPlugin/surf_plugin.tesh
64 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
65 - examples/java/surfCpuModel/surf_cpu_model.tesh
66 * Add SMPI+MSG example:
67 - examples/smpi/smpi_msg_masterslave/
72 - msg task destroy cancel test
73 - msg_host on/off test
74 * Move all tests in testsuite to teshsuite (adding tesh files)
75 * Restructure teshsuites
76 - one folder for each kind of test
77 * Restructure AddTests.cmake
79 - structure the order of tests (with sections)
83 - creation of a VM on a PM
84 - migration of a VM from a PM to another PM
86 - MSG_process_join(msg_process_t process, double timeout)
87 - msg_bar_t MSG_barrier_init(unsigned int count)
88 - int MSG_barrier_wait(msg_bar_t barrier)
89 - void MSG_barrier_destroy(msg_bar_t barrier)
90 - msg_as_t MSG_environment_as_get_by_name(const char * name)
91 * New option "msg/debug_multiple_use" to help debugging when a task is used
94 - Improvements and finalization of MSG_storage, MSG_file APIs and their
96 - Increase code coverage in test suites
100 * Protect context stacks against stack overflow. The number of protected memory
101 pages allocated on the top of each stack (1 by default) can be configured
102 with the new command line option --cfg=contexts/guard_size:<value>.
103 * Simcalls are now generated by a python script that
104 - generates files included by SimGrid
105 - checks that all the functions exist, and proposes prototypes
107 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
108 host_set_data, host_get_data
110 - simcall_process_join(smx_process_t process, double timeout)
111 * Fix bug where sleeping processing could not be suspended.
114 * Translate surf models from C to C++
115 - Generic classes for all models: Model, Resource, Action
116 - A generic interface for each kind of model (CPU, Network, Storage
117 Workstation, WorkstationVM)
119 * Translate surf routings from C to C++
120 * Add callbacks using sigc++ or boost::signals2
121 - Add callback functions for resource creation/destruction
122 - Add callback functions for action state change
123 - Handle Energy as a plugin
124 * Replace swag by boost::intrusive
125 * Add new routing models for clusters. For documentation, see
126 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
127 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
128 parameters for cluster tag
129 - Fat trees, with topology="FAT_TREE" and
130 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
131 - see examples/platforms/torus_cluster.xml and
132 examples/platforms/fat_tree_cluster.xml
136 * Hostfiles support host:nb_processes construct to deploy several processes on
138 * Collective communication algorithms should not crash if used with
139 improper number of nodes and report the error.
140 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
141 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
142 MPI_Dims_create, MPI_Cart_sub are supported.
143 * New interface to use SMPI programmatically (still depends on MSG for
144 some parts, see examples/smpi/smpi_msg_masterslave) :
145 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
149 * Global variables privatization in MPI executables is now performed at runtime
150 with the option smpi/privatize_global_variables (default:no).
151 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
152 dynamic libraries loaded by the application are not privatized (static
153 linking with these libraries is advised in this case)
156 * Options defined in XML work correctly now.
159 * New cmake option, enable_lib_in_jar, to control whether native libraries are
160 copied into simgrid.jar or not (ON by default). Use this option if you want
161 to reduce the size of the installed simgrid.jar, *and* the native libraries
162 are kept installed elsewhere.
163 * Surf binding with SWIG (code generated in maintainer mode only):
164 - plugin to handle callbacks
165 - CPU model only for the moment
168 * Supernovae build mode is definitively removed. It was used to improve
169 inlining and inter-module optimizations. It is nowadays superseded by
170 link-time optimizations commonly available in compilers.
171 * Update ns-3 find lib. Bindings for ns-3 should work again now.
172 * Add boost dependency for surf++
173 * Add new macro for tests
174 - ADD_TESH(name <tesh_args>)
175 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
179 - xbt_bar_t XBT_barrier_init(unsigned int count)
180 - int XBT_barrier_wait(xbt_bar_t barrier)
181 - void XBT_barrier_destroy(xbt_bar_t barrier)
182 * Make the xbt_os_time module public
184 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
186 SimGrid (3.10) stable; urgency=low
188 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
191 * Reintegrate Java to the main archive as desynchronizing these
192 package is not acceptable anymore (Java is now considered stable)
193 * Add explicit synchronization facilities through semaphores
194 * Bug fix: Task.setDataSize() only changed the C world, not the value
195 cached in the Java world
198 * Dramatically change the way files are handled. API and internals changed, but
199 this part of MSG was not considered as production grade either.
200 * Add explicit synchronization facilities through semaphores
201 * Add a new function MSG_host_get_process_list()
202 * Preliminary DVFS support (see examples/msg/energy/ for details)
205 * SMPI is now included directly in the libsimgrid as the windows
206 linker doesn't force us on splitting it anymore.
207 * Improvements of the SMPI replay tool:
208 - Most of the collective communications are now rooted in the same process as
209 in the original application.
210 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
211 used until now). Multiple data types can now be used in a trace.
212 - The replay tool now supports traces produce either by TAU or a modified
214 - Bug Fix: the compute part of the reduce action is now taken into account.
215 - Gatherv collective is now supported
216 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
217 of smpirun outputs time independent traces for the current run. One file
218 is created per process. If too many processes are simulated, this behavior
219 can be changed to one file for all processes by using the
220 tracing/smpi/format/ti_one_file flag
221 * smpirun generates the host file if needed (with given host count and platform)
222 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
223 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
224 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
225 --cfg=smpi/coll_selector:(mpich/ompi)
226 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
228 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
229 happening inside a collective SMPI call.
230 * Fix the behavior of complex data types handling.
231 * Make MPI_Wtime another synchronization point to take computations into
233 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
234 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
235 * Add all missing Fortran bindings, SMPI should work with Fortran 90
236 (no privatization of global variables yet)
237 * Preliminary DVFS support (see examples/smpi/energy/ for details)
240 * Verification of liveness properties is now available for SMPI applications
241 (in addition to MSG applications)
242 * Bugged examples using SMPI in examples/smpi/mc/
243 * Add --cfg=model-check/visited option. Allows the verification of infinite
244 programs. Detection of loops in the execution thanks to the system state
245 comparison and reduction of the state space to explore. Can be combined with
246 DPOR for safety properties.
249 * Allow to change SimGrid configuration (see --help) within the code
250 thanks to SD_config() as it can be done in MSG.
251 * Add a new function SD_task_set_amount() upon user request.
254 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
255 * Remove rule based routing (no more PCRE dependency)
256 * Add a limiter_link option to cluster tag, to specify a maximum reachable
257 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
258 * Add a loopback_bw and loopback_lat options to cluster tag.
259 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
260 routing. Now peers are encapsulated in an AS and have their own private
261 router but this is transparent.
264 * Our own implementation of getline is renamed xbt_getline, and gets
265 used even if the OS provide a getline(). This should reduce the
266 configuration complexity by using the same code on all platforms.
267 * New type: xbt_cfg_elm_boolean.
268 * Allow to use yes/no for boolean configuration options in the command line.
269 * Allow to disable SimGrid cleanups at exit from command line option.
270 There are situations where one may want a simulation to end with an exit.
271 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
272 annoying when scripting around the simulator. Adding a
273 --cfg=clean_atexit:no allows to circumvent this issue.
276 * Lots of memory leaks were corrected in this release.
277 There is no known memory leaks anymore, in all of our 600+ tests.
278 * New command line option --version, to get SimGrid version information.
279 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
281 * Supernovae builds are deprecated, and expected to be removed in the next
284 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
286 SimGrid (3.9) stable; urgency=low
288 The Grasgory release: GRAS is really dead now.
290 * Complete overhaul of the internal host structures scheme.
293 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
294 was considered as experimental and badly maintained since then.
295 * Keeping it was thus a trap to our potential users, that could take
296 it instead of MSG or SMPI by mistake despite is pity state.
297 * GRAS seems to have very few users (if any), and no one volunteered
298 to maintain it further. It also induces a lot of XBT code (for
299 portability sake), that must be maintained too.
300 * For all these reasons, we killed GRAS. If someone wants to revive it
301 in the future, don't cry, our git history still remembers of GRAS.
304 * Major overhaul. Merge our documentation again as time proved that
305 splitting it was really not helping our users.
306 * Further improve the developer documentation to help newcomers
307 hacking on SimGrid itself. The user documentation (and in
308 particular, the beginner documentation) is still in a sorry state.
311 * Now works on Windows too!
312 * Much more extensive test suite, from MPICH
315 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
316 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
317 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
318 * Bug fix: let task be scheduled when the last dependency to be solved is
319 a control dependency.
320 * Remove SD_load_environment_script function.
321 Use the C sg_platf function if you want to declare a platform
325 * New function: MSG_process_get_number()
326 * Old function documented: MSG_config()
327 * Remove MSG_load_platform_script function
328 Use the C sg_platf function if you want to declare a platform
332 * Change the default value of the TCP_gamma constant (maximal size of TCP
333 congestion window) to a more realistic 4MiB value. If you notice changes in
334 your simulation results, you can fall back to the previous 20k tiny window
335 by adding --cfg=network/TCP_gamma:20000 on command line.
336 * (Hopefully) fix a bug wrt periodic availability and state traces
337 * Bug fix: use default values at start when first event in availability/state
338 trace is not at time 0.
341 * remove the "new_" part of function name sg_platf_new_trace_connect
342 (resulting in sg_platf_trace_connect), since it does not create
346 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
347 use case where it's really mandatory, and maintaining it was a pain in
349 * New: xbt_fifo_search(), search an item with a user-provided
350 comparison function instead of dumb pointer comparison.
353 * Fix the lua deployment:
354 Use `simgrid.init_application()` before deployment instead of
355 `simgrid.msg_register_application()` after.
358 * Transfer the tracing files into the corresponding modules.
360 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
362 SimGrid (3.8.1) stable; urgency=low
364 The "we are told that some people want to also *install* the simgrid
367 * Add missing file "tesh.1" to the archive.
369 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
371 SimGrid (3.8) stable; urgency=low
373 The Psssshiiiit release: SimGrid jumps into the Cloud.
376 * Add an experimental interface to manipulate VMs. They are mainly
377 process groups with very few intrinsic semantic, but they should
378 allow you to build the semantic you want easily.
379 * New function: MSG_host_set_property_value()
380 * New function: MSG_process_on_exit(). To clean memory in all cases.
381 * Bug fixes that made the host (and link) failures unusable.
382 * Add a way to auto-restart process when the host in which they are
383 executing comes back (ON_FAILURE="RESTART" on deployment file,
384 MSG_process_auto_restart_set).
385 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
386 please stop using the old ones, they are DEPRECATED.
388 * Deprecate functions MSG_global_init() / MSG_global_init_args()
389 Please use MSG_init() instead. (reducing the amount of entry
390 points in the library helps us).
391 * Make it impossible to link against the wrong version of the lib
392 * Deprecate MSG_clean(). No need to call it anymore.
393 * Function MSG_get_host_number() is not deprecated anymore.
396 * Split the doc into a user guide and a reference guide.
397 * Start a developper guide to help people hacking on SimGrid.
400 * Enable tracing by default. This modules rocks you should use it.
401 * Remove option custom_flags. Now use environment variables CFLAGS
403 * Use default cmake things to detect lua instead of home grown ones.
404 * New option "enable_mallocators" to disable mallocators, for debugging
405 purpose ("on" by default).
408 * Bug fixes around the resource failures: don't let the processes
409 survive the host they are running onto.
410 * Add an interface to auto-restart processes when the host in which they are
411 executing comes back.
412 * Ensures that SIMIX_clean is called automatically. It's not part of
413 the public interface anymore (bindings should be updated).
416 * Bug fix for when SD_Simulate is called with a positive value: be careful
417 when comparing doubles. Sometimes they are different for non significant
419 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
420 parallel task whose initial work is distributed among host according
421 to the Amdahl's law. Such tasks are created with a parameter alpha
422 that corresponds to the non-parallelizable part of the computation.
423 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
424 between two sets of workstations assuming a 1D block distribution (each
425 workstation owns a similar share of data) on both sides.
427 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
428 Data redistribution will be automatically scheduled once parent and child
429 are both scheduled. The filling of computation_amount and
430 communication_amount structures is now done seamlessly thanks to the chosen
432 * New function SD_workstation_dump to display various information
433 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
434 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
435 bandwidth on the route onto which the task is scheduled and the amount of
437 To divide the nominal bandwidth by 2, the rate then has to be :
438 rate = bandwidth/(2*amount)
439 * Compute tasks that have failed can now be rescheduled and executed again
440 (from their beginning)
441 * Increasing source code coverage (src/simdag is now covered at 95.8%
445 * Re-implement time-independent trace replay using SMPI (at the
446 smpi_smp_* level) instead of MSG. This should replace
447 examples/msg/actions/actions.c
448 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
449 hindexed and structs)
450 * Implement the exchange of non-contiguous data.
451 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
452 * Correct behavior of smpi/sender_gap and set its default value to 0
453 * Add option to asynchronously send small messages to allow better
454 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
455 specifies the size in bytes under which messages will be asynchronously sent.
456 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
457 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
460 * New C interface to define a platform: XML is now optional.
461 For more info, please check include/simgrid/platf.h.
462 * New interface to define random platforms from the C:
463 For more info, please check include/simgrid/platf_generator.h and
464 examples/msg/masterslave/masterslave_platfgen.c
465 * Export a sg_cmdline dynar containing all the arguments we got from
469 * Two new tracing options for adding comments to trace file so you
470 can track your experiments (see --help-tracing for details).
471 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
472 * Adding the SimGrid version that generated the trace file as a comment.
473 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
474 * Fix to avoid key clashes on Paje links
475 * Other minor fixes related to the Paje specification
478 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
479 and renamed to xbt_str_hash() and xbt_str_hash_ext().
480 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
481 * Greatly improve the robustness of mmalloc to user errors (such as
482 using an area after freeing it, or freeing it twice)
484 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
486 SimGrid (3.7.1) stable; urgency=low
489 * Restore the prototype of MSG_process_create_with_environment() to
490 the pre-3.7 situation by removing the kill_time argument.
491 * Add a MSG_process_set_kill_time() function instead.
494 * Fix weird behaviors when dealing with parallel tasks.
497 * Simgrid is now built as a dll.
498 * Simgrid-java now works on Windows.
499 * Simgrid-Java is now included into Windows package.
502 * First pre-build package for MacOSX.
505 * Fix compilation when using MSG_USE_DEPRECATED.
506 * Fix some compilation issues on Macs and Windows.
507 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
508 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
510 We discovered that the Lua console is broken, but we are missing the
511 manpower to fix it right now. The problem existed in 3.7 too, so we
512 are not blocking the release for that. Sorry if you depended on this
513 feature, any help would be really welcome.
515 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
517 SimGrid (3.7) stable; urgency=low
519 The "spring cleanups (before the next Big Project kicks in)" release.
522 * We can specify the SMPI latency/bandwidth factor with command line
523 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
524 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
525 You can also use the "config tag" from platform file by adding this line
526 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
527 (see "example/platforms/tag_config.xml" to use "config tag").
528 Note that the command line supersedes the platform file configuration.
529 * Change the correction factors used in LMM model, according to
530 the latest experiments described in INRIA RR-7821.
531 Accuracy should be improved this way.
532 * Use the partial invalidation optimization by default for the
533 network too. Should produce the exact same results, only faster.
534 * Major cleanup in surf to merge models and split some optimization
535 mechanisms from the core of the models. As a result you can now
536 specify which model to use (e.g., --cfg=network/model:LV08
537 --cfg=cpu/model:Cas01) and which optimization mode to use
538 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
539 Incompatible combinations should err at initialization. See
540 --help-models for the list of all models and optimization modes.
541 * The CLM03 workstation model was dropped for simplicity because it
542 used the deprecated CM02 network model. Use default instead.
543 * Rename the TCP_gamma configuration option to network/TCP_gamma
544 * Rename the coordinates configuration option to
545 network/coordinates, and document it
546 * Use now crosstraffic keyword instead of the terribly misleading
547 fullduplex keyword. It is activated by default now in the current
548 default model, use --cfg=network/crosstraffic:0 to turn it off.
549 * Ongoing refactoring the model parsing to make XML files optional
550 See include/simgrid/platf.h for details (still to be completed)
553 * Major overhaul of the documentation. Almost instructive now :/
554 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
555 functions and friends. This interface was considered as
556 deprecated since over 2 years, it's time to inform our users that it is.
557 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
558 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
559 use these (crappy) functions in your code.
560 These functions will be removed soon. Stop using them now.
561 * Deprecate MSG_get_host_{table,number}
562 Implement MSG_hosts_as_dynar() instead.
563 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
564 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
567 * Stabilize the parallel execution mode of user contexts
568 * Introduce configuration variables to control parallel execution:
569 - contexts/synchro: Synchronization mode to use when running
570 contexts in parallel (either futex, posix or busy_wait)
571 - contexts/parallel_threshold: Minimal number of user contexts
572 that must be part of a scheduling round to switch to parallel
573 execution mode (raw contexts only)
574 * Fix bugs that prevented to use suspend/resume along with
575 synchronization structures.
576 * Fix bugs in process termination that lead to invalid memory access
577 in very specific conditions.
580 * Introduce a parallel mode for the models (controlled by surf/nthreads
581 configuration item). In our tests, running the models in parallel
582 never lead to any speedups because they are so fast that the gain
583 of computing each model in parallel does not amortizes the
584 synchronization costs, even when ultra fast futexes are used.
585 This is released anyway because YMMV.
588 * Performance boost by using a swag internally to compute the set of
589 tasks that are finished and should constitute the return value of
593 * Enable it by default now that it is considered rather stable.
596 * Documentation of the tracing functions.
597 * Performance gains when tracing categorized/uncategorized resource
598 utilization by avoiding calls to get route when updating resource
599 variables. LMM constraints are being used instead.
600 * API changed to set task categories. Use MSG_task_set_category instead
601 of TRACE_msg_set_task_category, and SD_task_set_category instead
602 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
603 * Bugfix for graphicator, routes not correctly obtained, memory leaks
604 * Examples for link user variables added (see at examples/msg/tracing/)
605 * Deprecated function TRACE_msg_set_process_category completely removed
606 * Trace header updated according to the latest Paje file format
607 * Tracing network lazy updates, no longer obligate users to use full updates
608 * --cfg=tracing/platform:1 also registers power/bandwidth variables
609 * Experimental: let user code declare/set/push/pop application states for hosts
610 * API changed to allow the manual creation of graph configuration files
611 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
614 * Improve the API of Lua MSG bindings, using the Lua spirit.
615 * Each simulated process now lives in its own Lua world (globals are
616 automatically duplicated). It helps writing simulators. Will allow
617 to run Splay programs within SimGrid in the future.
618 * Add a Chord example in Lua, equivalent to the MSG one.
621 * Start the implementation of a solution to express temporal
622 properties, not only local assertions. This is still an
623 experimental work in progress, stay clear from it to be safe.
627 - Add new runtime parameters --help-logs and --help-log-categories
628 to display informations about supported logging parameters and
630 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
631 don't exists anymore.
632 * Mallocators: allow value NULL for the reset function.
634 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
635 create homogeneous dictionaries, where all the elements share the
636 same free function. Non homogeneous dictionaries will be
637 deprecated in the next release.
638 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
639 - Multi-level dictionaries are deprecated.
641 - new function xbt_dynar_search_or_negative() that is useful when
642 you have less than 2 million elements in your dynar and don't
643 want of the extra complexity of catching exceptions when the
644 element is not found.
646 - Make xbt_os_thread module (for thread portability) public.
647 Documentation is still to come, sorry.
649 - Cleanups and simplifications to make it maintainable again.
650 - Exotic features (such as memalign and valloc) were removed.
651 - The metadata were extended and improved so that the
652 model-checker becomes able to explore and inspect the heaps.
653 - This may induce a performance drop when enable_model-checking is
654 ON in cmake (even if it's not used in the simulation), but it is
655 necessary at this point to get MC working.
657 Turn model-checking OFF if simulation performance matters to you.
658 Not enabling it at runtime is not enough, disable it in cmake.
660 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
662 SimGrid (3.6.2) stable; urgency=low
664 The "Not coding new stuff allows to polish old things" release.
667 * New bindings to the NS3 packet level simulator (experimental)
668 * Use the raw (efficient) execution contexts instead of the sysv
669 (portable) ones when possible.
670 * libpcre is now mandatory in any cases since not using it led to
671 severe performance loss and possibly other issues
672 * Update the XML platforms:
673 - G5K: include the latest machine in Nancy
674 - GridPP and LCG: new platforms
675 * Documentation was partially updated, at least (more to come)
677 Bug fixes, cosmetics and small improvements
678 * Free terminated processes before the end of the simulation to avoid
679 exhausting the memory of users having very dynamic amount of
681 * Bug fix and cosmetics about canceling non-running tasks
682 * Bug fix about the dot loader's issues when using libcgraph
685 * Create an installer for windows with nsis (amd64 and win32)
686 - Add an hello world project to illustrate simgrid project creation.
687 - Embed libpcre into the Simgrid installer to avoid
688 its compilation burden
689 * The raw execution contexts should work on Apple now
690 * Port to Windows 64 bits
691 - Sysv contexts now have an implementation for this arch
692 - GRAS communication features now support this arch
693 * Drop support for borland compiler on windows
694 - this code was not maintained, and we kinda depend on gcc nowadays
695 * Fix portability issues on kfreebsd/gnu: build error about semaphores
696 * Fix portability issue on unstable ubuntu: linker became picky on
699 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
701 SimGrid (3.6.1) stable; urgency=low
703 The "Oops, we broke Macs too" release
706 * Fixed contexts detection so that raw ones are used when possible
707 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
708 leads to a strange error, with user code segfaulting sometimes when
709 the generated code is not perfectly aligned (which is not
710 controllable from the user side, depends on the amount of code)
713 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
715 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
717 SimGrid (3.6) stable; urgency=medium
719 The Summer Release, also known as the "OMG! They Killed Kenny!" version
722 * Bindings now constitute their own package, separated from the main one.
723 Rationale: reduce our maintenance nightmare by reducing the module coupling
724 They will soon be released on their own on gforge.
726 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
727 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
729 GRAS: It is not considered as stable anymore, but experimental. Sorry.
730 * It's not quite deprecated for now because we have no replacement,
731 but it may soon become the case.
734 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
735 * Fortran: New user-level cache variable to store the rank of the running
736 process. This improves performance by an order of magnitude.
737 * C: New coccinelle script to automatically locate and modify global and
738 local static variables.
739 * Improved SMPI network model with a sender-side gap to account for multiple
743 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
744 only say if a communication is finished, no matter whether it succeeded or
745 failed. You can call MSG_comm_get_status() to know the status of a finished
747 * New function MSG_task_dsend() to send a task and detach it. When a
748 communication is detached, you are never notified of its success or failure
749 and the memory is released automatically once it is finished. This function
750 is useful when you don't care about the end nor the success of a
752 * Change the prototypes of action replay. Sorry for inconvenience,
753 but this is really more efficient this way (and to adapt your code,
754 you just have to fix the initialization, that shouldn't be too long)
755 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
756 ever managed to do anything useful with it.
757 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
758 takes a dynar of communications. It returns immediately and gives the
759 index of a finished communication (if any).
760 * New example: a basic implementation of the Chord P2P algorithm.
763 * New model for multi-core CPUs. You can now use the core attribute to
764 precise the number of cores of a host. This is a basic model. Every
765 process running on the host receives at most the power provided in
766 the DTD (throughput<=power). Total throughput of process cannot exceed
768 * New peer tag. This peer tag creates a tiny AS comprising a host and a
769 router linked by an up-link and a down-link (possibly asymmetrical).
770 This kind of pattern allows to easily build last-mile model style platforms.
771 Aggregating such patterns in a rule-based AS is thus the technique of
772 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
773 * New model for Vivaldi routing. We transformed the Vivaldi network model
774 into a Vivaldi routing model (based on the rule-based model). This allows to
775 combine Vivaldi based latencies with last-mile platforms.
778 * Added a check for NaN of IEEE754 infinite in the double entries of
780 * Introduce a new context factory "raw", highly inspirated from the
781 ucontext factory, but using manually crafted functions in assembly to
782 do the work in an efficient manner.
783 * Allow to change the used context factory at run time, not only at
784 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
785 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
786 contexts at runtime (only with raw contexts or ucontexts).
787 * Completely rewrote this module to allow parallel execution of user
788 processes. Use --cfg=contexts/nthreads:N to execute user processes
789 with N parallel threads (the default is 1, meaning no parallelism).
790 * Allow to decide dynamically between sequential and parallel modes.
791 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
792 processes in parallel only when their number is greater than or equal to P
794 * Added a check for NaN of IEEE754 infinite in the double entries of
798 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
799 won't output the state of processes when interrupted with Ctrl-C
800 * Add a new function xbt_dynar_to_array that transforms a dynar into a
801 NULL-terminated array. This may solve backward compatibility issues
802 due to the change to return type of SD_simulate. See also:
803 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
804 * Add new macros with variable number of arguments.
805 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
806 - in xbt/asserts.h: xbt_assert
807 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
808 - in xbt/ex.h: THROWF and RETHROWF.
809 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
811 * Change xbt_die() to accept a format string with arguments, just like printf.
812 * New data structure: xbt_lib_t, like a dict but more general and with better
816 * New configuration options
817 Options triva/categorized and triva/uncategorized can be used to generate
818 graph configuration files for Triva visualization tool.
819 * Configuration option tracing/platform is renamed to tracing/categorized
820 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
821 * New TRACE_user_link_variable interface:
822 User provides the name of the link and the tracing variable to attach to it
823 * the declaration of tracing categories must be done after the environment
825 * simpler tracing interface, just one way to declare categories
826 TRACE_category or TRACE_category_with_color, it is up to you
827 * links in the trace file are again identified by their names
828 * trace contains the full platform hierarchy exactly as declared using the ASes
829 * Options tracing/msg/[task|process]:1 groups the process by hosts
830 for both cases, tasks and processes must have names that are unique during the simulation
831 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
832 * The experimental option tracing/msg/volume is deprecated
833 its functionality may be reincorporated if needed
835 The tracing generates a trace file with unordered timestamped events,
836 because of the way the core simulator (surf) works. A script available
837 at the tools directory (fix-paje-trace.sh) can be used to put the events
838 in order. We have changed the tracing so it can generate ordered timestamped
839 events in the final trace, but depending on the simulator (and how much time
840 is simulated) that can lead to a huge memory utilization. It is deactivated
841 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
844 * Define a SIMGRID_VERSION macro in simgrid_config.h.
845 - We are trying hard to keep the API stable, but it may happen that
846 some things change (we're a research project after all, not a
847 nuclear plant operating system). If such things should happen, you
848 could rely on that macro to adapt.
849 - current value: 30600 for 3.06.00, aka 3.6
850 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
851 a number that can be compared with SIMGRID_VERSION.
852 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
853 assertions and verbose/debug logging events are disabled at compile time.
855 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
857 SimGrid (3.5) stable; urgency=medium
860 * New feature to any SimGrid-based simulator: Model-Checking
861 Check SIN#1 for more details.
864 * New Model SMPI (three-interval linear regression for correction factors)
865 See RR-7426, available at http://hal.inria.fr/inria-00527150
866 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
867 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
868 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
869 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
870 * See also src/smpi/README
875 - Tracing API changes: TRACE_start and TRACE_end should not be called
876 by user-code. They are automatically called by simulators created
877 with SimDAG, MSG and SMPI if the toolkit is compiled with
878 tracing_enabled=ON. Categories declaration and utilization remain the
879 same for MSG and SimDag.
880 - A function was added to the tracing API to declare categories with
882 - TRACE_category_with_color (char *category, char *color)
883 where color must be in the following format
884 "%f %f %f", red, green, blue
885 and red, green, blue are float values in the interval [0, 1]
886 - User can specify NULL as color parameter, or continue calling
888 On that case, the tracing system will define random colors
889 - The following command-line options are supported:
890 --cfg=tracing/filename:msg.trace
891 --cfg=tracing:1 (activate tracing, needed to use others)
892 --cfg=tracing/platform:1 (categorized resource use)
893 --cfg=tracing/uncategorized:1 (uncategorized resource use)
894 --cfg=tracing/msg/task:1 (task creation)
895 --cfg=tracing/msg/process:1 (process creation, migration)
896 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
897 --cfg=tracing/smpi:1 (SMPI interface tracing)
898 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
899 - examples of examples/msg/tracing updated
901 - DAXLoader and DOTLoader functions can generate tasks with categories
902 - A new function to attribute a category to SD tasks:
903 TRACE_sd_set_task_category (SD_task_t task, char *category)
904 * Tracing the MPI interface implemented by SMPI
905 - Collective operations are traced with states
906 - Point-to-Point operations are traced with states/links
907 - Tracing activated by a parameter "-trace filename" passed
908 to smpirun during execution (considering that simgrid
909 is compiled with tracing enabled)
910 - To run the simulation with gdb, the simulator
911 accepts --cfg=tracing/smpi:1 to trace SMPI
912 - tesh files to check if smpi tracing is ok
913 - See examples/smpi/NAS/DT-trace
914 * GTNetS tracing re-worked
915 - adaptation to the tracing system of GTNets to cope
916 with modifications regarding the fullduplex mode
917 - new tesh files to check if gtnets tracing is ok
920 * Asynchronous communications through the functions:
921 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
922 * New function: MSG_load_platform_script()
923 to make possible using a lua script instead of XML files to set up platforms
924 * New function: MSG_set_function
925 to associate functions to processes, used when bypassing the parser
926 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
928 Platforms: Add some more examples in examples/platforms
929 * Grid'5000: see www.grid5000.fr
930 * *_30000_hosts.xml: various huge files [mainly scalability testing]
933 * Change the XML format. This is a very important modification. SimGrid
934 3.5 introduces a new hierarchical format based on the notion of
935 Autonomous Systems. Compatibility with old format is ensured
936 through the perl script provided in the install bin directory
937 bin/simgrid_update_xml.
938 It is now possible to build platforms with specific routing
939 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
940 connect several platforms together. We will try to provide soon
941 set of realistic platforms exploiting these properties (have a
942 look at examples/platforms/ for the moment).
943 * Take the opportunity of the XML format change to be a good XML citizen:
944 rename link:ctn to link_ctn and similar changes (also dealt with by
946 * Add a new routing scheme (rule-based) using regular expressions. It
947 enables to have an extremely low memory footprint when the
948 underlying routing is simple and can be compactly described. You
949 need to have libpcre4-dev (perl regular expressions) installed if
950 you want to use this routing scheme.
951 * Revive the cluster TAG and allow to easily and efficiently (both in
952 term of memory and speed) connect clusters together. Have a look
953 at teshsuite/simdag/platforms/ to see how this can be done. With
954 this tag, you can create clusters with thousands of tasks at no
955 cost (have a look at examples/platforms/).
956 Note: clusters are implemented as ASes, so there is no need for an
957 enclosing AS tag if you have only one cluster in your platform.
958 * Add new generic functions in the public interface that allows the user
959 to call SURF 'create_resource' methods from your code (same
960 functionality as the XML bypass mechanism but with a much lighter
962 * Add a new model (enabled through command line --cfg=network/model:SMPI)
963 that uses a piecewise linear approximation to produce better
964 results when exchanging small messages.
965 * Add a new parameter to handle correctly full duplex link and account
966 for interferences between uplink and downlink communications
967 (activate with --cfg=fullduplex:1).
970 * Rename the SD_READY (all dependencies are satisfied and task is
971 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
972 dependencies are satisfied) state.
973 This prevents a confusion between the notion of "ready to schedule"
974 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
975 simulated" (SD_RUNNABLE) used by the simulation kernel.
976 * Change the way a task is considered as ready. Instead of removing
977 dependencies when a task is done, a counter is decreased. This way,
978 it is always possible to reach ancestors thanks to the
979 SD_taks_get_parents function (even after the end of the simulation.)
980 * Change the return type of SD_Simulate from (SD_task_t*) into
981 xbt_dynar_t. This function was in handling a dynar internally and
982 converted it into a NULL terminated array for historical reasons.
983 * New function SD_dotload(char*) to load a DAG described in dot
984 format. This loader and the corresponding examples require the
985 installation of the graphviz library.
986 * Fix a bug in the management of tasks of size 0 in the surf network
987 models. This problem was only visible with SIMDAG and you should
988 thus disregard results produced with earlier versions if you
989 relied on this feature (some tasks were blocked because of this).
990 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
991 even though your applications were doing only point-to-point
992 communications and sequential computations. Now you can really use any
993 model you want (of course, if you create real parallel tasks, which are
994 not implemented in most models beside ptaskL07, this will abort).
995 * Add an example that schedules a DAX on an heterogeneous platform
996 using a Min-Min strategy.
997 * New function SD_workstation_get_current_task() that returns the kind
998 of task currently running on a workstation in the sequential access
1000 * Raise some warnings when unexecuted tasks remains at the end of the
1001 simulation. This is usually caused by cycles in the DAG.
1004 * New function: SIMIX_process_set_function() called by MSG_set_function
1005 * Change the underlying waiting queue in semaphores so that a process
1006 can wait on several of them simultaneously (as in waitany).
1007 * Fix the way to handle tokens in semaphores so that all access patterns
1008 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1009 * kill the dirty pimple SIMIX_message_sizes_output()
1010 Please use (proper) visualization instead
1013 * New data container: setset (set of sets of elements)
1014 * New module: mmalloc (mapped malloc, allowing to have several
1015 independent segments of malloc)
1016 * New function: xbt_dict_cursor_set_data()
1017 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1018 * New function: xbt_dynar_is_empty()
1019 * New function: xbt_fifo_get_last_item()
1020 * Fix xbt_dynar_shrink(): use the right element size.
1021 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1022 uninitialized areas during expand.
1023 * Fix semaphores: previous implementation was severely broken.
1024 * Use library init/fini functions for our initialization.
1025 - you can use logs and other feature as soon as you want in your
1026 code (even before the xbt_init / MSG_init)
1027 - xbt_exit is now a no-op and produce a warning when used.
1030 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1031 benefit from the latest improvement to the simulation kernel.
1032 * Kill measurement sockets for now. If you rely on them, sorry. This
1033 release is not for you. This feature will be reintroduced in the
1034 future, but we cannot delay the release any further.
1035 * New function: gras_msgtype_get_name().
1036 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1037 * Fix (at last) the pmm example: it should not randomly fail anymore.
1039 Build chain: bug fixes and overall polishing
1040 * Cmake is now stable enough. Hence, we killed the autotools.
1041 * Port to windows ( TM :)
1042 * Fix the 'make install' target.
1043 No need to use 'make install-simgrid' anymore
1044 * Introduce a 'make dist' target compiling a *source* archive
1045 'make package' compiles a binary archive
1046 * Compile java files only on need
1047 * Add --cd and --setenv command line options to tesh
1048 * Out of source builds are not fully supported yet, but we are close
1049 * Enable supernovae and optimization flags by default for our users
1052 * Add layer to set up environment directly from lua, without XML.
1053 * The effect of gras_stub_generator can be achieved through
1054 lua too (check examples/gras/console/ping_generator.lua)
1056 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1058 SimGrid (3.4.1) stable; urgency=low
1060 The "Polishing easter eggs is probably a good idea" release.
1061 This is a bug fixes release only.
1065 * Fix a bug preventing the tasks from begin garbage collected.
1068 * Fix a bug occuring when a host involved in a communication fails.
1069 This was not detected properly by the other peer involved in the
1070 communication. Now, it's reported as a network error.
1073 * Warn the user about loop dependencies in data flow of DAX files
1074 * Obey the control-flow dependencies of DAX files
1077 * Add option "enable_smpi" allowing to not compile SMPI.
1078 Probably useful for the (Mac) users experiencing a build error here
1079 * Improve the detection of lua5.1 and ruby1.8
1081 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1083 SimGrid (3.4) stable; urgency=low
1085 The "Easter in Cargese" release. Also known as (major changes):
1087 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1088 ~> bindings were greatly improved
1089 ~> new tracing infrastructure for better visualization introduced
1091 * the "Welcome to configury modernity" release.
1092 ~> we switched from autotools to cmake, and improved our cdash
1095 A more detailled list of changes follow (full detail in svn log).
1097 Java Bindings: Various Cleanups
1098 * (install java-gcj-compat-dev on debian-like to use them)
1099 * Remove put/get: no need to export deprecated interface in Java
1100 Use send/receive instead.
1101 * Cleanup the examples and add a README per directory
1102 * Remove example autoDestination (that's the only way to go now)
1103 * Remove example explicitDestination (was a plain copy of basic)
1104 * Make JniException a runtime exception, so that there is no need to
1105 declare the fact that you may encounter such a beast. I guess that
1106 nobody will ever want to survive such error.
1107 * Create specific errors for each MSG case of failure:
1108 host failure, transfer failure, timeout, task cancelled
1109 * Cleanup the exceptions that may get thrown by each function
1110 * Other internal cleanups in Java bindings. Performance still bad :/
1111 Ruby and Lua Bindings: create them
1112 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1113 * That's new and great, you should try them out.
1114 Same functionalities than Java bindings, only even less polished
1116 * Kill the useless "rate" argument of SD_task_get_execution_time()
1117 Everyone used to provide -1 as a value, it was not used, and the
1118 semantic of a possible use wasn't even clear.
1119 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1120 either as comm costs or compute costs to mean that there is no
1121 such thing for that specific task.
1122 * Add a SD_task_set_name() function
1123 * Fix SD_task_unschedule() on typed tasks
1124 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1125 * In DAX loader, accept useless 'level' attributes to <job> since
1126 LIGO DAGs have them (seem to be to ease graphical representation).
1128 * Add an example masterslave_mailbox.c using send/receive and not
1129 the deprecated put/get interface.
1130 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1131 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1133 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1134 (declare MSG_USE_DEPRECATED to still have the old name)
1135 * Add a MSG_task_set_data() function
1136 * About trace replay (see examples/msg/actions):
1138 - Allow to work with splitted trace files for each process
1139 Give the specific trace file as argument of each process,
1140 and call MSG_action_trace_run(NULL)
1141 You can still have one merged file for all processes.
1142 - Fix implementation of collective operations
1143 * Allow task_execute() on 0-sized tasks (closes #10063)
1145 * This is the first release of SimGrid where SMPI is not considered
1146 beta anymore (even if some corners should still be improved)
1147 * Port over the new SIMIX_network submodule (internal refactoring)
1148 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1149 * Implement more missing elements of the standard:
1151 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1153 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1154 MPI_Scan() MPI_Get_processor_name()
1155 - Added implementation of missing case for Alltoall (warning: it's
1156 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1157 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1158 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1159 SMPI_MPI_Allgatherv()
1160 * Bug fixes include:
1161 - MPI_Waitsome() was broken
1162 - Allow relative includes in smpicc
1163 - Command line cfg argument 'reference_speed' was ignored...
1164 - Some functions did not properly lead to auto-benching of user code
1165 - smpicc passes -O2 by default (just like openmpi one)
1167 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1168 * Bug fixes about timeouts during communications
1169 * add SIMIX_message_sizes_output() as a pimple to write to file the
1170 amount of messages per size. Use gnuplot to get histogram.
1171 Pimple because that's the only user-visible function of simix,
1172 defined directly in xbt.h (irk, sorry)
1174 - Add a SIMIX_sem_get_capacity() function
1175 - Fix interactions with processe resume/suspende
1176 - release_forever() was stupidly broken
1177 - Fix SIMIX_display_process_status() for processes in a semaphore
1178 - Make SIMIX_sem_block_onto() user-visible
1179 * Refactoring context stuff:
1180 - Use pseudo-OOP for better modularity
1181 - reimplement SIMIX_process_kill() without process_schedule() so
1182 that the latter can take as invariant that it is called from
1184 - Merge context_start into context_new for sake of simplicity
1186 * Add a Vivaldi network model, coded live during SUD'10 ;)
1187 * Rename configuration variables to start a hierarchy:
1188 o cpu_model -> cpu/model
1189 o network_model -> network/model
1190 o workstation_model -> workstation/model
1191 * New configuration variables:
1192 o network/bandwidth_factor: correction to bandwith
1193 o network/latency_factor: correction to latency
1194 o netwotk/weight_S: correction to the weight of competing streams
1195 * Add a long description to the models, that users can see with such
1196 argument on the command line: --cfg=cpu/model:help
1197 * --help-models display the long description of all known models
1199 * config: add the ability to set a default value after registration
1200 Does not override any previously set value (e.g. from cmd line)
1201 * dict: allow to have integer key and data.
1202 When so, you need to use the following functions
1203 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1204 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1205 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1206 In contrary to regular dicts, the key is not malloced before copy.
1207 Mixing scalar and regular elements in the same dict is not tested
1209 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1210 Tracing for Visualization:
1211 * SimGrid is now instrumented in order to generate a trace file for
1212 visualization analysis: to use it, need to compile SimGrid with the
1213 "tracing" option enabled, and instrument the program using SimGrid with
1214 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1215 (among other functions).
1216 * The instrumentation only traces the platform utilization for now
1217 * Documentation to use the tracing functions and how to analyze the
1218 traces with the Triva tool is written.
1219 * More information about: SimGrid FAQ (in the section Tracing Simulations
1222 * We moved to cmake as default build system. Autotools support will
1223 be dropped soon. Check the FAQ for more info about how to use it.
1224 * Greatly improved our cdash/ctest interactions
1225 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1226 * Added memory checking tests with valgrind; lot of memleak fixing.
1227 This may be the first release of simgrid with so few memory issues
1228 * Added code coverage tests.
1229 Our coverage is still improvable, but at least we see it on cdash.
1231 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1233 SimGrid (3.3.4) stable; urgency=low
1235 The "Desktop Grid needs love too" release (also called Xmas release).
1237 Models improvements:
1238 * Major speedup in the maxmin system solving by using lazy evaluation
1239 Instead of solving completely the maxmin system at each iteration,
1240 only invalidate (and recompute) the modified parts.
1241 This new feature is enabled in default models but you can try to
1242 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1243 * Cas01 IMproved as default CPU model
1244 This CPU model is the same Cas01 model, but it uses the
1245 maxmin-selective-update flag and a heap structure to manage
1246 actions on SURF kernel.
1247 It reduces the complexity to find the next action to finish and,
1248 consequently, it's faster than the old Cas01.
1249 This is the new default CPU model (Cas01).
1250 * Rename the old Cas01 model to Cas01_fullupdate
1251 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1252 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1253 * CpuTI (CPU Trace Integration)
1254 A new CPU model whose objective is simulate faster when using
1255 availability trace files.
1256 Instead of using a full featured, over engineered maxmin system for
1257 CPU modeling, this model does the pre-integration of traces files
1258 to calculate the amount of CPU power available, and so, executes
1259 faster than the old CPU models.
1260 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1261 * Use LV08 as default network model since it gives better accuracy
1262 for small messages and shouldn't change things for big ones.
1263 Use --cfg=network_model:CM02 to get the previous behavior.
1266 ******************************************
1267 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1268 ******************************************
1269 * The new CPU model may changes simulations!
1270 The point is that events occurring at the exact same timestamp
1271 are not scheduled in the same order with the old and new
1272 version. This may be enough to completely change the execution
1273 of simulations in some cases.
1274 * The new network model will change simulations!
1275 This new model is more realistic than the previous one, so you
1276 should consider redoing your old experiments with this model.
1277 Sorry for the inconvenience.
1280 * Introduce the supernovae compilation mode
1281 When compiled that way, the whole SimGrid (or almost) is put in a
1282 single compilation unit and compiled in one shoot.
1283 This is to help gcc which has difficulties to inline stuff from one
1285 The speedup seem to be above 15%, althrough more tests are needed on
1286 amd64 to confirm that gain.
1289 * Port of MSG's mailbox on top of SIMIX network
1290 The put/get mechanism was greatly simplified on the way.
1293 * New SIMIX network module. Provides:
1294 - Mailbox: rendez-vous mecanism to find with who you want to speak
1295 - Synchronous send/recv: easier and hopefully faster since the
1296 logic is handled in the maestro process directly now
1297 - Asynchronous send/recv: you dreamt of it? It's here now
1298 Too bad that nobody cared enough to propagate the change to MSG.
1299 * Add semaphores as SIMIX synchronization mechanism.
1302 * new function SD_daxload(char*) to load a DAX file
1303 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1304 * Introduce typed tasks. Specify its kind and cost at creation.
1305 At scheduling, just give where it should be placed, and the cost
1306 for each involved resource is automatically computed.
1307 Existing constructors so far (more to come of course):
1308 - SD_task_create_comm_e2e() for end-to-end communication
1309 - SD_task_create_comp_seq() for sequential computation
1310 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1311 * new function SD_task_dump() for debuging display
1312 * new function SD_task_dotty(task,FILE*) writing to file the info
1313 about the task in dotty format
1314 * SD_task_dependency_exists() can now cope with having one of its
1315 arguments NULL. If so, it tests whether the other argument has any
1317 * Add getters on list of preceding/following tasks:
1318 SD_task_get_parents(task) and SD_task_get_children(task)
1319 * Add getters on amount of workstations and list:
1320 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1321 * Add getter on task kind: SD_task_get_kind(task)
1322 * Update the start_time and finish_time of tasks on completion/failure
1323 * Bugfix: Remove task from state swags when destroyed
1326 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1329 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1330 dynar with one separate thread per value of the dynar.
1331 * Change the prototype of xbt_thread_create(), sorry.
1332 Added a boolean parameter indicating whether we want to join this
1333 thread (used in SG only for now)
1334 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1337 * GTNetS wrappers should now be usable again (and betterly tested too)
1338 * Fix a major regression from 3.2 where the timeout provided to
1339 MSG_task_put_with_timeout() was used as absolute time before which
1340 the comm should be done.
1341 * Start to fix the <cluster> tag.
1342 - Internal links should be good now (beside of the loopback, which
1343 use the private link instead)
1344 - paths to the external world is still rather broken
1345 - the <route:multi> tag is just broken. Actually that's brain-dead.
1346 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1347 to make it less stupid
1348 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1349 * Fix a source-level compatibility glitch from 3.2: after defining
1350 MSG_USE_DEPRECATED, you can use the old name
1351 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1352 * Allow to compile from the SVN with automake 1.11
1353 * Fix some problems when using the "start_time" tag in deployment XMLs.
1354 * Fix #8569: XBT/synchro.h has redundant declarations
1355 * Fix #8563: MSG return values and exceptions
1356 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1357 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1358 Thanks to Robson Peixoto.
1359 * Fix a vicious bug in dictionaries inducing that some elements were
1360 not freed on xbt_dict_free()
1362 Portability report of this version:
1363 * Main portability targets:
1364 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1365 on (amd64/i386/ia64)
1366 - mac leopard on i386
1367 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1368 but nothing critical.
1369 * Other platforms: windows, AIX and others were not tested for this release
1371 Timing report of this version:
1372 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1373 scenario parameters). From 8h to a few seconds in desktop grid settings.
1374 * Supernovae brings about 25% speedup on i386.
1376 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1378 SimGrid (3.3.3) stable; urgency=low
1380 The "Need for Speed" release.
1382 The timings done to validate the 3.3.2 were faulty.
1383 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1385 The problem was a conversion from a manually handled vector to
1386 xbt_dynar_t on the critical path.
1387 xbt_dynar_foreach calls functions, inducing stack management crap.
1389 We inlined these functions and xbt_dynar_foreach is now breath taking.
1390 We also inlined xbt_swag_belong on the way.
1392 Here are some approximate speedup measurements (on master/slaves
1393 simulations lasting between 10s and 20s each):
1394 3.3.1 -> 3.3.2: about same performance
1395 3.3.2 -> 3.3.3: 40% speedup
1396 3.3.1 -> 3.3.3: 40% speedup
1397 3.3.1 with inline patch -> 3.3.3: 30% speedup
1399 Our reading is that the refactoring which occurred in 3.3.2 made us
1400 suffer much more from the xbt_dynar_foreach low performance, but
1401 once we solved this, this refactoring proved to be very performance
1402 effective. From the 40% speedup, somehow, 10% are due to the
1403 inlining and 30% to the refactoring.
1405 That's a pitty that gcc cannot inline functions placed in other files
1406 alone. We have to choose between:
1407 - break the encapsulation (by putting private data structures and
1408 accessors in headers files to help gcc)
1409 - live with low performance
1410 - switch to a decent compiler such as icc (not quite possible).
1412 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1414 SimGrid (3.3.2) stable; urgency=low
1416 The "Simplicity does not preceed complexity, but follows it" release.
1418 The main contributors of this release were (lexical order):
1419 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1422 * Extract the routing logic into its own object.
1423 (was dupplicated in network.c and workstation_LV07.c;
1424 Allows to implement other ways of storing that info)
1425 => kill now useless network_card concept
1426 - Use dynar to represent routes (instead of void** + int*)
1427 - kill link_set (use surf_network_model->resource_set instead)
1428 - Add a command-line option to choose the routing schema to use
1429 - Add three new models:
1430 * Floyd (shortest path computed at initialization)
1431 * Dijikstra (shortest path recomputed all the time)
1432 * Cached Dijikstra (shortest path computed on need)
1433 All these models where contributed by Silas De Munck, and are
1434 described in his ICCS09 paper.
1436 * Simplify model declaration
1437 (less redirections, less function to write when defining a model)
1438 - Factorize stuff between models:
1441 surf_model_resource_set(model)
1442 surf_model_resource_by_name(model, name)
1443 - Unify the types of models in s_surf_model_t (using an union)
1444 - Embeed fields of common_public directly into s_surf_model_t
1445 - Rename model methods:
1446 action_free ~> action_unref
1447 action_change_state ~> action_state_set
1448 action_get_state ~> action_state_get
1449 - Change model methods into functions :
1450 (model)->common_public->action_use ~> surf_action_ref
1452 * Implement a generic resource; use it as ancestor to specific ones
1453 (allows to kill duplicated code in models)
1454 Drawback: timer command don't need no name nor properties;
1455 workstation_CLM03 don't need no properties
1456 (but I guess we can live with those few bytes wasted)
1458 * Improve the action object model
1459 - implement a constructor avoiding dupplicated code about field
1460 initialization in generic_action part.
1462 * Kill the SDP model: it has an external dependency, is deprecated
1463 in flavor of modern lmm models, and didn't compile since a while
1466 * Relocation of the context module from XBT to SIMIX.
1467 (the context were decoupled from the simix processes, duplicating a lot of code)
1468 => a lot of code was factorized
1469 - less overhead is introduced during scheduling
1470 - simpler API for the context factory
1471 - the logic for process creation,destruction and manipulation was simplified
1472 * Simplification of the s_smx_process_t data structure.
1473 => accesing the simix level data associated to a process is faster now,
1474 and the code is a lot more readable.
1477 * Implement some more MPI primitives:
1478 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1479 -implementation: Bcast: flat or 2-ary tree (default),
1480 Barrier: 4-ary tree,
1482 Allreduce: Reduce then Bcast
1483 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1484 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1485 Alltoallv: flat tree, like ompi
1487 * Add support for optimized collectives (Bcast is now binomial by default)
1488 * Port smpirun and smpicc to OS X
1491 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1492 Shout out if you used it.
1495 * Display the list of still queued messages in SG mode when existing
1499 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1500 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1501 * Add xbt_str_from_file(FILE*)
1502 * Add xbt_dict_get_key achieving a linear reverse search
1503 * Remove the context module
1505 Portability report of this version:
1506 * Main portability targets:
1507 - Linux(debian)/x86/context
1508 - Linux(debian)/x86/pthreads
1509 - Linux(debian)/amd64/context
1510 - Linux(debian)/amd64/pthreads
1511 On these, we still have the eratic breakages of gras/pmm and
1512 amok/saturate_sg reported in previous version. We still think
1513 that the tests are the cause of the fault, not the tested code.
1515 - Mac OSX Leopard/x86/context
1516 Still false negative in tesh autotesting.
1517 Smpi still fails, but this time because readlink does not accept -f
1518 Everything seems to work properly beside of that.
1521 - AIX version 5.3 (only tested contexts this time)
1522 Smpi still fails there because mktemp is not installed.
1523 Everything seems to work properly beside of that.
1525 I managed to compile it for the first time, but several breakages.
1526 Won't delay the release for this exotic platform.
1528 * Windows: it's still lagging behind. If you want to help, please
1531 Timing report of this version:
1532 This version seem to be more than 5% faster than 3.3.1 (on linux
1533 64bits with contextes). The gain is less than expected, we are
1534 investigating this for next release.
1536 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1538 SimGrid (3.3.1) stable; urgency=low
1541 * Implement a --cfg-help to show existing configuration variables
1542 * Build chain do not require doxygen in maintainer mode
1545 * fix a bug on struct sizeof computation, which prevented the
1546 exchange of arrays of structs in some conditions
1547 - added a regression test about this in datadesc_usage
1548 * Allow the exchange of 0-long dynamic vectors.
1549 - for that, use -1 as indicator of dynamic size instead of 0
1550 - This implied to change any size from unsigned long to long,
1551 reducing a bit communication abilities, but I guess that with
1552 64bits being quite common, this is more than enough.
1553 - This also induce a protocol change, thus bumping network protocol
1554 version from 0 to 1 (if we have external users, we have to get
1555 clean on that point too ;)
1556 - added two regression tests about this in datadesc_usage
1557 * Be more verbose when propagating local exceptions
1558 This helps debugging.
1559 * Display the status of simulated processes when receiving SIGINT in
1563 * Allow to control the simulation from a trace file.
1564 New functions MSG_action_register() and MSG_action_trace_run()
1565 The first one allows to associate a function execution to each
1566 kind of action while the second one parses a trace file and
1567 triggers the corresponding actions within the system.
1568 For now, only a toy example is provided in examples/msg/actions
1569 * Add an exemple of process migration in examples/msg/migration
1570 * Fix a bug in task exchange which broke MSG_task_get_sender()
1571 Add a teshsuite regression test for that.
1572 [Bug: if MSG_task_get_sender() is called after sender exit,
1574 * Fix a bug which prevented suspend/resume to work properly
1575 * Display the status of simulated processes when receiving SIGINT
1576 This fixes a regression of v3.3. due to the introduction of SIMIX
1577 * Bug fixing in failure management:
1578 - trace could not start by a failure at time 0
1579 - failure during communications were not working
1582 * Add SIMIX_process_set_name() to change the name of the current
1583 process in the log messages.
1584 * Store smx_hosts in a dict since we only retrieve them by name
1585 * Move the configuration infrastructure to surf
1588 * Move the configuration infrastructure to surf
1591 * Massive internal cleanups:
1592 - Store internal structures on processes instead of hosts (allows
1593 to have more than one process per host, in addition of being more
1595 - Cleanup the initialization/finalization process
1596 - Kill a whole bunch of unneeded synchronization:
1597 processes run in exclusive manner within the simulator
1598 - Move queues from global tables to process data fields
1600 - now accept -platform and -hostfile arguments
1601 - Pass the right rank value to processes according to the hostfile
1602 * Compile the examples by default, and use them as regression tests
1603 * Implement MPI_Wtime()
1604 * Change the reference speed to a command line option
1607 * TCP_gamma can now be specified as command line option using
1608 --cfg=TCP_gamma:10000000.0
1609 * Change the --surf-path cmd line option into --cfg=path:
1612 * Also include strbuff from xbt.h public header
1613 * xbt_ex_display(): do not free the exception after displaying
1614 This allows to do more with the given exception afterward.
1615 Users should call xbt_ex_free() themselves.
1619 Portability report of this version:
1620 * Main portability targets:
1621 - Linux(debian)/x86/context
1622 - Linux(debian)/x86/pthreads
1623 - Linux(debian)/amd64/context
1624 - Linux(debian)/amd64/pthreads
1625 These targets fail about 1/10 of times on gras/pmm, but we believe
1626 that this is because of the test, not because of simgrid.
1627 amok/saturate_sg fails even more rarely, and the test may not be
1630 - Mac OSX Leopard/x86/context
1631 The test suite still spits tons of errors because some obscure
1632 force prevents us from removing the temporary directories
1633 arguing that they still contain some metadata I've never heard of.
1634 Smpi fails because seq is not installed.
1635 Everything seems to work properly beside of that.
1638 - AIX version 5.3 (both contexts and pthread)
1639 Smpi still fails there because mktemp is not installed.
1640 XML inclusions seems rosty on AIX.
1642 * Windows: it's still lagging behind. If you want to help, please
1645 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1647 SimGrid (3.3) stable; urgency=high
1651 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1652 [Malek Cherier & Mt]
1654 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1657 Until now, GRAS were using MSG as an interface to SURF. It was
1658 quite difficult because both interface have several differences
1659 (MSG channels vs GRAS sockets were the most notable point).
1661 This also opens the gate to SMPI (which should occur soon) and speed
1662 up simulations by to 40% (even if it were not the main goal).
1664 **************************************
1665 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1666 **************************************
1667 The point is that events occuring at the exact same timestamp are
1668 not scheduled in the same order with the old and new version. This
1669 may be enough to completely change the execution of simulations in
1670 some cases. Sorry for the inconvenience.
1672 * Cleanup and upgrade the XML format to push further scalability
1673 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1675 * Improve the testing infrastructure with tesh. Now a very large part of
1676 the code is tested not only by being run but also by checking that the
1677 output match an expected output [Mt].
1679 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1680 is really less memory-demanding, which should allow you to use
1681 larger files in SimGrid [AL].
1683 * Inform valgrind about our contextes, so that it becomes usable
1684 with the default (and more effecient) version of SimGrid
1685 [contributed by Sékou Diakite, many thanks]
1688 * Introduce a listener thread in charge of receiving incoming
1689 messages from the network. It allows to overlap communication and
1690 computation but most notably, it removes some stupid deadlocks due
1691 to the fact that so far, a process could not send and receive at
1692 the same time. This made most non trivial communication schema
1694 * Convert the PIDs from long int to int to match the MSG ones (and
1695 linux ones too) [Mt]
1696 * New function: gras_agent_spawn() to launch a new process on
1697 current host. Only working in simulation for now. [Mt]
1698 * New function: gras_os_hostport() returning a constant form (ie,
1699 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1702 * Make the backtrace of exceptions more human readable [Mt]
1703 * New module: xbt/str [Mt]
1704 a ton of string utility functions (split, join, printf to a newly
1705 allocated buffer, trim, etc)
1706 * New module: xbt/hash [Mt]
1707 SHA1 hashing algorithm (more to come if needed)
1708 * New module: xbt/synchro [Mt]
1709 synchronization tools (mutex and conditions) working the same way
1710 in simulation and in real life (mainly useful for GRAS, but not
1712 * New module: xbt/queue [Mt]
1713 classical producer/consumer synchronization scheme
1714 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1715 (using the classical functions will get serialized) [Mt]
1716 * Make dictionary internal table dynamic. No need to specify its size
1717 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1719 * Make sure the log channels are organized as a tree under windows
1720 (because of ANSI C compatibility issue, any channel were child of
1721 root directly) [Mt].
1724 * Cleaned many thing in surf and fixed a few bugs [AL].
1725 * Add a nice command line configuration mechanism to compose models [AL].
1726 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1727 the previous one (KCCFLN05). It relies on something that looks like
1728 a max-min sharing mechanism but cannot be written as such. A new solver
1729 was thus designed [AL].
1730 * Add a new solver to lmm. Based on Lagrange optimization and
1731 gradient-based descent, it enables to efficiently maximise systems s.a
1733 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1735 This solver enables to propose two new network models for TCP Reno and
1736 TCP Vegas based on Low's work. These models still need to be fully
1737 tested though [Pedro Velho].
1740 * Bug fix in SD_simulate. Now the time bound given as argument is
1742 * Use the new parallel task model (ptask_L07) as default.
1743 * Use the SURF command line configuration mechanism.
1744 * 0-size tasks (for synchronization) should now work.
1746 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1748 SimGrid (3.2) stable; urgency=high
1752 We still experience issues on this platform, but we believe that at
1753 least MSG is usable.
1755 GRAS API BREAKAGE (for simplification purpose, sorry):
1756 * the gras_msgtype_by_name is not used anymore. Instead of
1757 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1758 you can write (and must)
1759 gras_msg_send(toserver, "request", &request);
1760 - If you still want to pass a gras_msgtype_t to the function (to cache
1761 the type and avoid the lookup time), use the gras_msg_send_() variant.
1762 - Impacted functions:
1763 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1764 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1765 * The callbacks are now expected to return 0 when everything went well
1766 (just like the main() function)
1768 GRAS new features and improvements:
1769 * New module mecanism where user code can use per process globals [Mt]
1770 This is similar to gras_userdata_*() functions, but for libraries. It
1771 factorize some code developped over and over in the examples and AMOK.
1772 It has still to be documented and used (only amok/peermanagement is
1774 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1775 loss when they were small enough to fit into the buffer and sent quickly
1776 enough so that they can all get received in one shoot.
1777 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1778 if not found. Use the *_or_null() variant for the old semantic.
1779 * In gras_msg_handle, do not discard messages without callback.
1780 They are probably messages to be explicitly awaited later (ie, proofs of
1781 mis-synchronization in userland since they are sent before being awaited)
1782 No big deal usually.
1783 * gras_socket_meas_send/recv: semantic changed!
1784 The numerical arguments used to be (1) the total amount of data to send
1785 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1786 messages. This was need for the fool willing to send more than MAXINT
1787 bytes on quite fat pipes.
1790 * Do really rename the hostmanagement module to peermanagement. [Mt]
1791 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1792 but this is rather new and this was documented in the module
1793 documentation (poor excuses, I admit)
1794 * Bandwidth measurement semantic changed! This follows the changes to
1795 gras_socket_meas_send/recv explained above.
1798 * A sequential mode has been added to the workstations. When a workstation
1799 is in sequential mode, it can execute only one task, and the other tasks
1800 are waiting in a FIFO. [Christophe Thiery]
1803 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1804 model for SIMDAG. [Christophe Thiery]
1805 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1806 equal to 0, which caused some bad behaviors in
1807 saturated_constraint_set_update. I now use a threshold mechanism like in
1811 * When running manually src/testall, you select specific units [Mt]
1812 testall is the result of our cunit mecanism, and should replace all
1813 the scripty thingy around since bash don't run easily on billware.
1815 * A mallocator system has been added. [Christophe Thiery]
1816 Mallocators allow you to recycle your unused objects instead of freeing them
1817 and allocating new ones.
1819 Documentation update:
1820 * FAQ reworking + New FAQs:
1821 - "Valgrind spits tons of errors!" [Mt]
1822 - "How to repport bugs" [Mt]
1823 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1824 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1825 - Communication time measurement within MSG [AL]
1826 - I experience weird communication times when I change the latency [AL]
1827 * GRAS tutorial [Mt]
1829 - an introduction to the framework and to the used communication model
1830 - an initiatic tour introducing the most proheminent features:
1832 . Lesson 0: Installing GRAS
1833 . Lesson 1: Setting up your own project
1834 o Part 2: Message passing
1835 . Lesson 2: Exchanging simple messages
1836 . Lesson 3: Passing arguments to the processes (in SG)
1837 . Lesson 4: Attaching callbacks to messages
1838 . Lesson 5: Using globals in processes
1839 . Lesson 6: Logging informations properly
1840 . Lesson 7: Using internal timers
1841 . Lesson 8: Handling errors through exceptions
1842 . Lesson 9: Exchanging simple data
1843 . Lesson 10: Remote Procedure Calling (RPC)
1844 . Lesson 11: Explicitely waiting for messages
1845 . Recapping of message passing features in GRAS
1846 - A HOWTO section containing:
1847 o HOWTO design a GRAS application
1848 More are due, of course. They will come latter. In the meanwhile, you can
1849 check the examples which are still here.
1851 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1853 SimGrid (3.1) stable; urgency=high
1857 There was a stack corruption somewhere, visible only when optimizing
1858 with these versions. [Vince]
1861 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1862 module that enabled to play with Directed Acyclic Graphs. It is built
1863 directly on top of SURF and provides an API rather close to the old
1864 SG. Some old codes using SG are currently under rewrite to check that
1865 all needful functions are provided. [Christophe Thiery]
1868 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1869 extension of the classical CLM03 model that gracefully handles
1870 failures. This is now the default model for MSG and GRAS. It doesn't
1871 handle parallel tasks yet though. [AL]
1872 * Bug fix: Weights were not correctly set in the network part.
1873 WARNING: This may have resulted in incorrect results with simulations
1874 where there are more than one flow on a given link. [AL]
1877 * After a (long ?) discussion on simgrid-devel, we have decided that the
1878 convention we had on units was stupid. That is why it has been decided
1879 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1880 WARNING : This means that all previous platform files will not work as
1881 such with this version! A warning is issued to ask users to update
1883 A conversion script can be found in the contrib module of the CVS, under
1884 the name contrib/platform_generation/surfxml_update.pl [MQ]
1887 * Bug fix: Processes were started in reverse order, wrt deployment file.
1888 WARNING: if your code relies on this bug, please fix it. [AL]
1889 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1890 being executed on two different locations. [AL]
1891 * Bug fix: Failures are now better supported thanks to Derrick's tests
1892 (there was many failure situations I hadn't thought of and that weren't
1893 correctly handled). [AL]
1894 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1898 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1899 Exception occurring on server-side are propagated back to client (!).
1901 API CHANGE: the callback changed their prototype. Change:
1902 int my_handler(gras_socket_t expeditor, void *payload_data) {
1904 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1905 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1907 * New! function: gras_msg_handleall to deal with all messages arriving
1908 within a given period.
1909 * New! function: gras_socket_server_range to get a server socket in a
1910 range of port numbers (ease to avoid port number conflicts) [MQ]
1911 * New! gras processes display their backtrace when they get a SIGUSR1
1912 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1913 Sweet to debug RL processes [MQ]
1917 - Do not force experiment sizes to be expressed in kb, or it becomes
1918 impossible to measure the latency this way (needs one byte-long tests)
1919 WARNING: this changes the amok_bw_* function semantic. [MQ]
1920 - Implements the link saturation stuff. [MQ]
1921 * Peer management module:
1922 New! module factorizing code that we wrote over and over [MQ].
1925 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1926 - Test units are placed directly into the library code, they get extracted
1927 automatically and placed into the src/testall binary.
1928 - Convert most of the XBT tests to this system.
1929 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1930 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1931 live in the same C code. This required to change a little bit the API
1932 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1933 * New module: sparse graph structure with basic algorithms (this is work
1934 in progress and the API is not considered to be frozen yet). [AL]
1935 * Display more information on backtraces: source line & function names are
1936 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1937 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1938 * Reworked a little bit some #include statements to load only required
1939 headers. Some user code that relied on SimGrid to include stdlib or
1940 stdio may need to include it by themselves. [AL]
1941 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1943 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1944 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1945 but I guess that almost nobody used those parts. [MQ]
1947 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1949 SimGrid (3.0.1) stable; urgency=low
1952 * Unfortunately, I had missed 5 misnamed functions:
1953 xbt_fifo_item_t xbt_fifo_newitem(void);
1954 void xbt_fifo_freeitem(xbt_fifo_item_t);
1955 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1956 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1957 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1958 They're now deprecated. Please use their new versions:
1959 xbt_fifo_item_t xbt_fifo_new_item(void);
1960 void xbt_fifo_free_item(xbt_fifo_item_t);
1961 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1962 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1963 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1965 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1966 * Documentation: xbt_log module unmercifully reworked [MQ]
1967 * Bugfix: there was a problem with the ending of contexts with
1968 the pthread backend. It caused some weird deadlock or behavior
1969 depending on the pthread implementation. [AL]
1970 * Bugfix: get the exceptions raised in the simulator repport where
1971 and why they come from when they are not catched in time [AL, MQ]
1974 * Bugfix: Do repport the error when two non-connected hosts try to
1975 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1978 * Add additionnal checkings on communications. Assert that two
1979 communicating hosts are connected by a set of links... [AL]
1982 * Add additionnal checkings on channel values in communication [AL]
1983 * New: MSG_task_get_source to see on which host a task was generated [HC]
1984 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1985 the number of tasks waiting to be received on channel and sent
1987 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1988 waits for the first task coming from a given host.. [AL]
1990 GRAS new functionnalities: [MQ]
1991 * Enhance the parsing macro to allow the size of multidimentional objects
1992 to be given thru annotations.
1993 * New example (and documentation): Matrix Multiplication a la RPC
1994 (as when I was young!) and fix a bunch of bugs found on the way.
1996 GRAS performance improvements: [MQ]
1998 * Reduce the amount of cbps creation/destruction by making it static to
1999 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2002 * Change libdata to a set so that we can search for stuff by ID (and thus
2003 reduce the insane amount of dict lookups)
2006 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2007 RL communication on top of SG since it's so uneffective.
2008 It may also be used for inter-thread communication in RL, one day.
2009 * Use gras_datadesc_copy() to exchange messages on top of SG
2011 - improve message exchange performance on top of SG
2012 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2013 * Don't exchange on the network the size of the used part of buffer,
2014 instead, specify the possible buffer size to read().
2016 - reduces the amount of read/write calls (one pair per exchange)
2017 - reduces the amount of exchanged data (the size)
2018 - allows to retrieve all arrived data on receiver side, if we don't need
2019 it right now (subsequent read will peek the buffer)
2020 - allows the receiver to proceed with the begining of the stream before
2021 everything is arrived
2022 - make it possible to build an iov transport (using readv/writev)
2024 - take care of the data with non-stable storage (like stacked data),
2026 * If possible, TCP send uses vector I/O (when writev() is here)
2027 - Don't use it for receive since we send data sizes and data on the
2028 same stream, so we wouldn't be able to chain large amount of chunks
2029 before having to flush the stuff to read the size.
2030 * Rework the transport plugin mecanism to simplify it and reduce the
2031 amount of pointer dereferencement when searching for the right function
2034 * I guess that now, we do almost as few system calls as possible while
2035 doing as few data copy as possible.
2037 To improve it further, we could try to send all the sizes first and then
2038 all the data (to use iov on receiving size), but it's only a partial
2039 solution: when you have 2 dimensional data, the sizes of the second
2040 dimension is data of the first dimension, so you need 3 streams.
2042 I'm not sure the potential performance gains justify the coding burden.
2044 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2046 SimGrid (3.00) stable; urgency=high
2049 * New! Give the possibility to hijack the surf parser and thus bypass
2050 MSG_create_environment and MSG_launch_application. Have a look at
2051 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2053 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2055 SimGrid (2.96) unstable; urgency=low
2060 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2061 This deprecates the xbt_error_t mecanisms.
2062 It modifies (simplifies) all XBT and GRAS API.
2063 MSG API keeps unchanged (exceptions raised by XBT are catched from
2064 within MSG and masked with existing error handling facilities)
2067 * New! Add a FATPIPE model. [AL]
2068 * New! Add a parallel task model. [AL]
2069 * New! Add automatically a loopback interface (in the default
2070 network model) if none was precised.
2073 * Bugfix: MSG_process_resume now works with the current running process.
2075 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2076 * Modification of MSG_task_get_compute_duration. Once a task has been
2077 processed, the value returned by this function is now equal to 0. [AL]
2078 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2079 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2080 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2081 of a m_task. For now, MSG_task_cancel only works with computation
2083 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2084 of the processor (in Mflop/s) regardless of the current load on the
2086 * API Change: use proper naming convention for MSG_getClock and
2087 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2089 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2090 This function changes the priority of a computation task. This priority
2091 doesn't affect the transfer rate. A priority of 2 will make a task
2092 receive two times more cpu power than the other ones. This function
2093 has been added to suit the needs of Nguyen The Loc and hasn't been that
2094 much tested yet. So if it fails, please report it and send me your code.
2096 * API Change: removed all functions and types that were marked "deprecated"
2097 since many months. Renamed MSG_global_init_args to MSG_global_init.
2099 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2101 SimGrid (2.95) unstable; urgency=low
2104 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2105 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2106 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2109 * Bugfix: Allow absolute paths for platform description files [MQ]
2110 * Bugfix: do free the variables after use. Leads to drastic performance
2112 * Implement max_duration (ie, timeouts) on resources [AL]
2115 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2117 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2118 the needed support (timeouts) [AL]
2121 * Implement measurement sockets. You can now get the bandwidth between two
2122 hosts thanks to AMOK (see below). [MQ]
2123 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2124 dynar over the network (yeah) [MQ]
2125 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2126 * simple/static token-ring example. [Alexandre Colucci and MQ]
2127 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2128 in place before [MQ]
2131 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2132 type parsing. "short" and "long" modifiers were also ignored; other
2133 modifier (such as reference level) are still ignored. [MQ]
2134 * Embeed the buffer size within the buffer itself on SG. [MQ]
2135 That way, send() are atomic and cannot get intermixed anymore (at least
2136 the ones which are less than 100k; bigger messages still have the issue)
2137 * Array size pushed by the field, not by the field type (or each
2138 and every long int will push stuff to the cbps) [MQ]
2139 * use select() to sleep since it allows to portably sleep less than one
2142 GRAS (minor cleanups)
2143 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2146 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2147 * type category 'ignored' killed as it was never used and were difficult
2149 * whether a type can cycle or not is now a flag, leaving room for more
2150 flags (as "ignored", if we feel the need one day ;)
2151 * Rename raw sockets to measurement sockets since "raw" has another
2152 meaning in networking community.
2155 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2156 and offers features not belonging to GRAS but that most applications
2157 need. One day, it may be a set of plugins loadable at runtime.
2158 * New module: bandwidth
2159 bandwidth measurement between arbitrary nodes running this module. [MQ]
2161 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2163 SimGrid (2.94) unstable; urgency=low
2165 The first beta release of SimGrid 3 !
2169 * Update the main page and the FAQ. Adding references to gforge.
2172 * Add a gras_os_getpid function.
2175 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2176 * Extend the logs so that they also print PID, hostname, date, ... if
2178 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2179 and kill the old version which were in testsuite/
2180 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2184 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2185 for users. Instead, each programming environment (GRAS, MSG,...) use it
2186 when needed to provide such a feature to users.
2187 Don't shortcut the mecanism or you will also shortcut the virtualization
2188 you need on the simulator.
2192 * Cleanups in configury with regard to compile optimization/warning flags.
2193 Also add -fno-loop-optimize to any powerpc since it's the optimization
2194 killing gcc (< 3.4.0).
2195 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2196 needed by MSG examples complications
2197 * Borrow configury beautifications from PHP
2200 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2201 warning (thanks loris for stumbling into this one).
2202 * Bugfix: stop loading private headers (gras_config.h) from the public
2206 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2207 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2208 RL would imply a DNS resolver, which is *hard* to do in a portable way
2209 (and therefore delayed).
2210 * Implement a real timer mecanism and use it in timing macros. This allows
2211 to avoid rounding errors and get a 0.000005 sec precision in timing
2212 macros. While I was at it, various cleanups:
2213 - allow to declare more than one timed section per file (fix a stupid bug)
2214 - move some private declaration to the right place
2215 - merge conditional execution and timing macros into emulation module
2216 - document the module
2217 - make sure the module cleanups its mess on gras_exit
2218 * Documentation improvements:
2219 - (new) how to compile applications using GRAS
2220 - (new) emulation support (timing macros)
2222 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2224 SimGrid (2.93) unstable; urgency=low
2226 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2229 - Use Paje properly where used. Still to be sanitized properly.
2230 - Portability fix: Add an implementation of the contexts using pthread
2234 - Add xbt_procname(): returns the name of the current process.
2235 Use it to show the current process's name in all logging.
2237 - fix detection of older flex version and the reaction, since we do
2238 depend on modern ones (we use lex_destroy)
2239 - Better separation of SG and RL in the libs: remove all simulation code
2240 from libgras. As a result, this lib is now only 200k when stripped.
2241 Some of the xbt modules may also be duplicated (two sets and such) and
2242 should be cleaned/killed before SG3.
2243 - Insist on using xlC on AIX because of weird problems involving gcc there.
2244 - Cleanup the make remote stuff. This is now done by scripts
2245 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2246 mainly for our private use, but we're working on changing them to user
2249 - Bugfix: flush the socket on close only if there is some *output*.
2250 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2252 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2255 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2257 SimGrid (2.92) unstable; urgency=low
2259 Alpha 3 on the path to SimGrid 3
2263 - New! First try of benchmarking macros.
2264 - New! First try so that gras_stub_generator generate deployment and
2265 remote compilation helpers.
2267 - Bugfix: Initialization fix in msg_test.
2271 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2273 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2275 - API Change: gras_os_sleep to take the amount of seconds as a double.
2276 Accepting an int was error prone since it was the only location where
2277 seconds were coded as such. It leaded to damn rounding errors.
2278 - Bugfix: Hard to belive that timers ever worked before this.
2280 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2282 SimGrid (2.91) unstable; urgency=low
2284 Alpha 2 on the path to SimGrid 3
2288 - Bug fix in the lmm_solver.
2290 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2291 through the function MSG_paje_output.
2292 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2293 - It is possible to bound the rate of a communication in MSG with
2294 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2295 to write it in the changelog).
2296 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2301 - Various cleanups to the autotools stuff
2302 - Begin to move Gras examples to examples/gras/
2303 - Let make distcheck work again (yeah!)
2305 - documentation overhauled using doxygen.
2306 gtk-doc-tools is dead in SimGrid now.
2307 - Automatically extract all existing logging categories, and add the list
2308 to the documentation (long standing one, to say the less)
2310 - Cleanup the known architecture table. Reorder the entries to group what
2311 should be, and use a more consistent naming scheme.
2312 (some of the test dataset are still to be regenerated)
2313 - New! Allow library to register globals on each process just as userdata
2315 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2316 lookup time (for now).
2317 Use it in msg and trp.
2318 This cleans a lot the internals and helps enforcing privacy of the
2319 headers between the gras components.
2320 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2321 - Bugfix: gras_os_time was delirious in RL.
2322 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2324 - Reenable GRAS now that it works.
2326 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2328 SimGrid (2.90) unstable; urgency=low
2330 Alpha 1 on the path to SimGrid 3
2332 * It is a long time since the last release of SimGrid. I'm sorry about
2333 that but as I had told you, I was rewriting a lot of things. I apologize
2334 to those who had been reporting bugs to me and that I had not answered.
2335 If your bug is still in the new version, please tell me. Here is a
2336 summary of the main changes.
2338 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2339 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2340 portability, speed, and a lot more but you'll figure it out later.
2341 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2342 I wanted to release the new MSG as soon as possible and I have
2343 broken GRAS, which is the reason why, for now, only MSG is fully
2344 functional. A laconic description of these projects is available
2345 in the documentation.
2347 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2348 kernel optimized for our needs (called SURF but only the developers
2349 should use it). Hence, MSG is now roughly 30 times faster and I think
2350 that by rewriting a little bit MSG, I could event speed it up a little
2351 bit more. Beside the gain in speed, it is also much easier to encode a
2352 new platform model with SURF than it was with SG. More to come...
2354 * REVOLUTION 3: I have tried to change a little as possible the API of
2355 MSG but a few things really had to disappear. The main differences
2356 with the previous version are :
2357 1) no more m_links_t and the corresponding functions. Platforms are
2358 directly read from a XML description and cannot be hard-coded
2359 anymore. The same format is used for application deployment
2360 description. The new format is described in the documentation.
2361 Have a look in tools/platform_generation. There is a tiny script
2362 that converts from the old platform format to the new one. Concerning
2363 the application deployment format, parsing the old one is tricky.
2364 I think most of you should however be able to convert your files. If
2365 it is really an issue, I can write a C code that does the conversion.
2367 2) the toolbox tbx does not exist anymore. We now have a library
2368 with much more data-structures but without the hash-tables (we have
2369 dictionaries that are much faster).
2371 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2373 *****************************************************************************
2374 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2375 * don't feel like converting the oldies. (Mt) *
2376 *****************************************************************************
2379 Version 2.90: "the long awaited one"
2380 - Finished rewriting and debugging MSG. Rewrote the documentation.
2381 - disable GRAS for now since it needs to be ported to the newest SG
2384 - Finish the port to windows (using mingw32 for cross-compile)
2387 - Main loop and datastructures of SURF. A cpu resource object is
2388 functional. Surf can thus be used to create cpu's with variable
2389 performance on which you can execute some actions.
2391 2004-11-15 Martin Quinson
2392 - Port to ARM. Simply added the alignment and size descriptions. Should
2393 work, but the ARM machines are so slow that I didn't had the opportunity
2394 to 'make check' over there yet.
2396 2004-11-15 Arnaud Legrand
2397 - Trace manager now written. It uses a heap structure and is therefore
2398 expected to be efficient. It may however be speeded up (particularly
2399 when many events occur at the same date) by using red and black
2400 trees. One day maybe...
2401 - Max-min linear system solver written. It uses a sparse matrix
2402 structure taking advantage of its expected use. Most operations are
2403 O(1) and free/calloc are called as few as possible. The computation of
2404 the minimum could however be improved by using a red and black tree
2407 2004-11-03 Arnaud Legrand
2408 - Rename every gras_* function that was in xbt/ to its xbt_
2410 - Add a heap and a doubly-linked list to xbt
2411 - Added a dichotomy to the dictionaries. make check works as well before
2412 so I assume that the patch is correct. I do not know however if things
2413 run effectively faster than before now. :)
2415 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2416 SimGrid, and the version number is bumped to 2.x
2418 2004-10-29 Martin Quinson
2419 - Introduction of the remote errors.
2420 They are the result of a RMI/RPC on the remote machine.
2421 ErrCodes being scalar values, you can't get the host on which those
2422 errors did happen. Extending the error mechanism as in Gnome is possible.
2423 No idea yet whether it is a good idea.
2425 2004-10-28 Martin Quinson
2426 - Interface revolution: the Starred Structure Eradication.
2427 I used to do typedef struct {} toto_t; and then handle *toto_t.
2428 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2429 - ???_t is a valid type (builded with typedef)
2430 - s_toto_t is a structure (access to fields with .)
2431 - s_toto is a structure needing 'struct' keyword to be used
2432 - e_toto_t is an enum
2433 - toto_t is an 'object' (struct*)
2435 typedef struct s_toto {} s_toto_t, *toto_t;
2436 typedef enum {} e_toto_t;
2437 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2438 s_toto_t) is private.
2440 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2441 so that it can set the variable to NULL. It was so for dicts and sets,
2442 it changed for dynars.
2444 - Fix a bunch of memleaks in dict_remove
2445 - Fix a bug in sg/server_socket opening: it failed all the time.
2447 2004-10-07 Martin Quinson
2448 - Speed up dynar lookup operation a bit.
2450 gras_dynar_get is dead.
2452 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2453 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2454 (which gives you the address of the stored data).
2456 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2457 copy of the data using an affectation to do the job and not a memcpy.
2459 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2460 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2462 It does not really speedup the dynar test because they are
2463 setting elements all the time (and look them seldom). But the dict does
2464 far more lookup than setting.
2466 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2468 2004-10-05 Martin Quinson
2469 - Allow to (en/dis)able the cycle detection at run time.
2471 Whether we should check for cycle or not is now a property of each
2472 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2473 datadesc_cycle_unset allow to remove this property when previously set.
2475 Note that the cycle detection is off by default since it impacts the
2476 performance. Watch the data you feed GRAS with ;)
2478 This property is hereditary. Any element embedded in a structure having it
2479 set have it set for the time of this data exchange.
2481 You should set it both on sender and receiver side. If you don't set it on
2482 sender side, it will enter an endless loop. If you forget on receiver
2483 side, the cycles won't be recreated after communication.
2485 - Header reorganization.
2486 Kill gras_private.h, each submodule must load the headers it needs.
2488 2004-10-04 Martin Quinson
2489 - Interface revolution: do not try to survive to malloc failure.
2491 Now, gras_malloc and friends call gras_abort() on failure.
2492 As a conclusion, malloc_error is not a valid error anymore, and all
2493 functions for which it was the only gras_error_t return value are
2494 changed. They now return void, or there result directly.
2495 This simplify the API a lot.
2497 2004-09-29 Martin Quinson
2498 - Re-enable raw sockets.
2499 Created by gras_socket_{client,server}_ext;
2500 Used with gras_raw_{send,recv}
2503 It should allow to kill the last bits of gras first version soon.
2505 This is not completely satisfactory yet (duplicate code with
2506 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2509 - Simplify transport plugin (internal) interface by not passing any
2510 argument to _server and _client, but embedding them in the socket
2513 2004-09-28 Martin Quinson
2514 - Finish the port to AIX.
2515 autoconf was my problem (segfault within the malloc replacement
2516 function. No idea why)
2518 2004-09-16 Martin Quinson
2519 - Fix some size_t madness on 64bit architectures.
2521 2004-09-08 Martin Quinson
2522 - Reduce the number of system headers loaded, overload some more system
2523 calls (such as malloc to cast the result of the system one, and work
2525 - Fix and reintroduce the config support
2527 2004-09-07 Martin Quinson
2528 - Source code reorganization to allow Arnaud to surf all over there.
2529 - Allow to document the logging categories.
2530 - Remove all uppercase from logging categories and useless cleanup in names.
2532 2004-08-18 Martin Quinson
2533 Version 0.6.2 (protocol not changed; API changed)
2534 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2535 gras_error_t), and NULL when not found. Functions expecting a msgtype
2536 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2537 hopefully usefull message.
2538 - Portability to prehistoric sparcs again
2540 2004-08-17 Martin Quinson
2541 Version 0.6.1 (protocol not changed; ABI not changed)
2542 - prealloc some buffers to speed things up
2544 2004-08-11 Martin Quinson
2545 Version 0.6 (protocol not changed; ABI expended)
2546 - The parsing macro can deal with the references, provided that you add
2547 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2549 2004-08-09 Martin Quinson
2550 Version 0.5 (protocol not changed; ABI changed)
2551 - Allow to off turn the cycle detection code in data exchange at
2552 compilation time. It should be at run time, but I'm short of time (and
2553 the config stuff is still broken). That way, we keep dict out of the
2554 critical path, which is good because the performance is poor:
2555 - search not dichotomial yet
2556 - dynar give no way to access their content and memcpy everytime
2557 - In composed data description (struct, ref and so on), stop foolness of
2558 keeping the subtype's ID, but store the type itself. This keeps sets out
2559 of the critical path, which is good since they rely on dynar and
2560 dictionnaries. The only loose of that is that we cannot detect the
2561 redeclaration of a structure/union with another content (but I'm not sure
2562 the code detected well this error before anyway). We still can detect
2563 the redefinition discrepancy for the other types.
2564 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2565 it breaks the code because of type-punning used all over the place).
2566 This breaks on all non-gcc architectures (for now).
2568 All those changes (plus the buffer of last time) allow me to gain 2 order
2569 of magnitude on cruel tests consisting of 800000 array of integers on two
2570 level of a hierarchical structure (200 secondes -> 4 secondes)
2573 - the selector of reference must now return the type it points to, not
2574 the ID of this type.
2576 2004-08-06 Martin Quinson
2577 Version 0.4 (protocol changed; ABI not changed)
2578 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2579 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2580 - Implement a buffer transport, and use it by default (it relies on tcp in
2581 real life and on sg in simulation).
2582 That's a bit hackish since I had a new field to the structure to store
2583 its data without interfering with the subtype ones. Inheritance
2584 is tricky in C. And that's a kind of reverse inheritance with one class
2585 derivating two classes. Or maybe a game with java interfaces. Anyway,
2586 that's damn hard in C (at least).
2587 Moreover, I got tired while trying to ensure plugin separation and
2588 genericity in SG mode. MSG wants me to do weird things, so let's go for
2589 cruel hacks (temporarily of course ;).
2590 See comment in transport_private.h:71
2591 - do not include all the _interface headers in private but in the files
2592 which really need them (to cut the compilation time when they are
2595 2004-07-26 Martin Quinson
2596 Version 0.3 (protocol not changed; ABI changed)
2597 - Major overhault of the datadesc interface to simplify it:
2598 - shorted the function names:
2599 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2600 - add a trivial way to push/pop integers into the cbps without malloc.
2601 This allows to make really generic sub_type description, which simply
2602 pop their size of the stack.
2603 - add a function gras_datadesc_ref_pop_arr() which does what users want
2604 most of the time: Declare a dynamic array (which pops its size of the
2605 stack) and declare a reference to it. Poor name, but anyway.
2606 - kill the post-send callback, add a post-receive one
2608 2004-07-23 Martin Quinson
2609 Version 0.2 (protocol changed; ABI changed)
2610 - add some testing for cpbs in the test cases, and fix some more bugs.
2611 This invalidate again the little64 data file, since I cannot regenerate
2613 - remove an awfull optimization in the logging stuff, allowing me to:
2614 - understand it again
2615 - learn gcc how to check that the argument match the provided format
2616 - fix all errors revealed by gcc after that
2617 - internal keys of dict are not \0 terminated. Deal with it properly in
2618 loggings instead of segfaulting when the user want to see the logs :-/
2620 2004-07-22 Martin Quinson
2621 - Fix some stupid bug preventing cbps (callback postit) from working
2623 2004-07-21 Martin Quinson
2624 - Some documentation cleanups
2625 - remove the useless last argument of msgtype_declare
2626 - rename the Virtu functions to fit into the 'os' namespace
2627 - move headers src/include -> src/include/gras/ and stop fooling with
2628 gras -> . symbolic link
2629 - make distcheck is now successful
2631 2004-07-19 Martin Quinson
2633 - Build shared library also
2634 - Install html doc to the right location
2635 - stop removing maintainer files in make clean
2636 - build tests only on make check
2638 2004-07-13 Martin Quinson
2640 - No major issue in previous version => change versionning schema
2641 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2642 corresponding dataset.
2644 2004-07-11 Martin Quinson
2646 - Get it working with any kind of structure (we can compute the padding
2647 bytes remotely for all the architectures I have access to)
2648 - Implement the structure parsing macro (still not quite robust/complete)
2649 - Improvement to the remote testing toysuite
2651 2004-07-10 Martin Quinson
2652 [autoconf mechanism]
2653 - get ride of a bunch of deprecated macros
2654 - actually run the test for two-compliment, not only compile it :-/
2655 - test whether the structures get packed (and bail out if yes. Damn.
2656 Alignment is a serious matter)
2657 - test whether the structures get compacted (but respecting the alignment
2658 constraints of each types)
2659 - test whether the array fields of structures can straddle alignment boundaries
2661 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2662 (took me 2 hours to find that bug, looking at the wrong place)
2663 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2664 before sending/receiving and are used to compute the offsets of fields
2665 - Given that padding size depend even on compiler options, keep track of
2666 alignment and aligned_size only for the current architecture. Not a big
2667 deal since we send structure fields one after the other (seems
2669 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2670 program, let it work on linux/gcc/little32. portability todo.
2672 2004-07-08 Martin Quinson
2673 - import and improve remote compilation support from FAST
2674 - make sure make check works on half a dozen of machines out there
2676 2004-07-07 Martin Quinson
2677 Let's say it's version 0.0.3 ;)
2678 - Implement conversions (yuhu!)
2679 - Let it work on solaris (beside conversion, of course)
2680 - Stupid me, using rand() to generate the conversion datatests in not wise.
2682 2004-07-06 Martin Quinson
2683 - Let make dist work, since I'm gonna need it to compile on remote hosts
2684 - Let Tests/datadesc_usage write the architecture on which the file was
2685 generated as first byte.
2686 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2687 architecture descriptions.
2688 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2689 on those architectures.
2690 - Optimization: send/recv array of scalar in one shoot
2692 2004-07-05 Martin Quinson
2693 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2695 - Plug a whole bunch of memleaks
2696 - each process now have to call gras_{init,exit}. One day, their log
2697 settings will be separated
2698 - Continue the code factorisation between SG, RL and common in Transport.
2700 2004-07-04 Martin Quinson
2702 - Redistribution between SG and RL.
2703 We wanna have to accept in SG, so move accepted related parts of RL in
2704 the common part. (more precisely, the dynar of all known sockets is no
2705 more a static in transport.c, but part of the process_data)
2707 [gras_stub_generator]
2708 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2710 2004-07-03 Martin Quinson
2711 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2712 them all up in one shot)
2714 - Fix the ugly case of reference to dynamic array.
2715 - New (semi-public) function gras_datadesc_size to allow the messaging
2716 layer to malloc the needed space for the buffer.
2718 - gras_socket_close now expect the socket to close (and not its address to
2719 put NULL in it after it). This is because the socket passed to handlers
2720 is one of their argument (=> not writable).
2722 - propagate the interface cleanup from last week in datadesc, ie remove a
2723 superfluous level of indirection. User pass adress of variable
2724 containing data (both when sending and receiving), and not of a variable
2725 being a pointer to the data. Let's say that I like it better ;)
2726 The price for that is constructs like "int msg=*(int*)payload" in
2727 handlers, but it's a fine price, IMHO.
2729 - Let it work in RL (yuhu)
2731 2004-06-21 Martin Quinson
2733 - porting SG plugin and SG select to new standards (works almost).
2734 - plug memleaks and fix bugs around.
2737 - cleanup the prototype of data recv and force users to specify when they
2738 want to handle references to objects. Test case working even for cycles.
2739 - plug memleaks. Valgrind is perfectly ok with this.
2741 2004-06-12 Martin Quinson
2743 - cleanup the separation between plugin and main code in plugin creation
2745 2004-06-11 Martin Quinson
2747 - Reput hook for raw sockets, needed for BW experiments
2748 - kill a few lines of dead code
2749 [Data description] Interface cleanup
2750 - gras_datadesc_by_name returns the searched type or NULL.
2751 That way, no variable is needed to use a type desc once, which makes
2753 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2754 two parameters were strings (field name, type name), leading to
2756 [Dicos] Interface cleanup
2757 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2758 This is consistant with the dynar API.
2760 2004-04-21 Martin Quinson
2762 - Porting to new standards.
2764 - interface cleanup.
2765 There is no bag anymore, no need to take extra provision to mask the
2766 pointers behind "ID".
2767 Better splitup of functions between files create/exchange/convert.
2768 This is still a bit artificial since convert and receive are so
2769 interleaved, but anyway.
2771 - add a queued message list to procdata (the ones not matching criteria
2773 - factorize some more code between SG and RL wrt procdata
2775 - use gras_exit in example to track memleaks
2776 - get rid of gs_example now that GS is properly integrated into gras
2777 - update run_test to integrate the latest tests (datadesc)
2779 - rename WARNINGn macros to WARNn since it prooved error-prone
2781 2004-04-19 Martin Quinson
2783 - register init/exit functions within gras module mechanism
2784 - send/receive function.
2785 Convertion is not implemented, but short-cutted if not needed.
2786 struct/array elements are sent one by one (instead of block-wise), but
2787 nobody really cares (yet). Get a prototype before optimizing.
2788 - tests (using a file socket) for DD send/receive on:
2789 - base types: int, float
2790 - array: fixed size, string (ie ref to dynamic string)
2791 - structure: homogeneous, heterogeneous
2792 - chained list, graph with cycle
2793 Believe it or not, valgrind is not too unhappy with the results. The
2794 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2798 - Bugfix when using a filename explicitely (instead of '-')
2800 2004-04-09 Martin Quinson
2802 - factorize more code between RL and SG in socket creation
2803 - Complete the implementation and tests of:
2805 o file (only in RL, and mainly for debugging)
2807 I lost 3 days to design a portable address resolver, and then decided
2808 that the prototype mainly have to run on my box.
2809 Addressing portability too early may be like optimizing too early :-/
2811 - use gras_init in the Tests instead of the crappy parse_log_opt
2812 (the latter function is removed)
2813 [Conditional execution]
2814 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2815 [Code reorganisation]
2816 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2817 Build examples against the RL library, since there is no way to disable
2818 its creation for now.
2820 For information, the beginning of coding on GRAS was back in june
2821 2003. I guess that every line has been rewritten at least twice since