1 SimGrid (3.11) NOT RELEASED; urgency=low
6 * Normalizing pointers addresses tool for better diff between logs
9 * Add cloud examples using new VMs
10 - examples/msg/cloud/two_tasks_vm.tesh
11 - examples/msg/cloud/simple_vm.tesh
12 - examples/java/cloud/cloud.tesh
13 - examples/java/cloud/migration/migration.tesh
14 * Add java surf examples:
15 - examples/java/surfPlugin/surf_plugin.tesh
16 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
17 - examples/java/surfCpuModel/surf_cpu_model.tesh
18 * Add SMPI+MSG example:
19 - examples/smpi/smpi_msg_masterslave/
24 - msg task destroy cancel test
25 - msg_host on/off test
26 * Move all tests in testsuite to teshsuite (adding tesh files)
27 * Restructure teshsuites
28 - one folder for each kind of test
29 * Restructure AddTests.cmake
31 - structure the order of tests (with sections)
35 - creation of a VM on a PM
36 - migration of a VM from a PM to another PM
38 - MSG_process_join(msg_process_t process, double timeout)
39 - msg_bar_t MSG_barrier_init(unsigned int count)
40 - int MSG_barrier_wait(msg_bar_t barrier)
41 - void MSG_barrier_destroy(msg_bar_t barrier)
42 - msg_as_t MSG_environment_as_get_by_name(const char * name)
43 * New option "msg/debug_multiple_use" to help debugging when a task is used
46 - Improvements and finalization of MSG_storage, MSG_file APIs and their
48 - Increase code coverage in test suites
52 * Protect context stacks against stack overflow. The number of protected memory
53 pages allocated on the top of each stack (1 by default) can be configured
54 with the new command line option --cfg=contexts/guard_size:<value>.
55 * Simcalls are now generated by a python script that
56 - generates files included by SimGrid
57 - checks that all the functions exist, and proposes prototypes
59 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
60 host_set_data, host_get_data
62 - simcall_process_join(smx_process_t process, double timeout)
63 * Fix bug where sleeping processing could not be suspended.
66 * Translate surf models from C to C++
67 - Generic classes for all models: Model, Resource, Action
68 - A generic interface for each kind of model (CPU, Network, Storage
69 Workstation, WorkstationVM)
71 * Translate surf routings from C to C++
72 * Add callbacks using sigc++ or boost::signals2
73 - Add callback functions for resource creation/destruction
74 - Add callback functions for action state change
75 - Handle Energy as a plugin
76 * Replace swag by boost::intrusive
77 * Add new routing models for clusters
78 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
79 parameters for cluster tag
80 - Fat trees, with topology="FAT_TREE" and
81 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
82 - see examples/platforms/torus_cluster.xml and
83 examples/platforms/fat_tree_cluster.xml
87 * Hostfiles support host:nb_processes construct to deploy several processes on
89 * Collective communication algorithms should not crash if used with
90 improper number of nodes and report the error.
91 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
92 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
93 MPI_Dims_create, MPI_Cart_sub are supported.
94 * New interface to use SMPI programmatically (still depends on MSG for
95 some parts, see examples/smpi/smpi_msg_masterslave) :
96 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
100 * Global variables privatization in MPI executables is now performed at runtime
101 with the option smpi/privatize_global_variables (default:no).
102 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
103 dynamic libraries loaded by the application are not privatized (static
104 linking with these libraries is advised in this case)
107 * Options defined in XML work correctly now.
110 * New cmake option, enable_lib_in_jar, to control whether native libraries are
111 copied into simgrid.jar or not (ON by default). Use this option if you want
112 to reduce the size of the installed simgrid.jar, *and* the native libraries
113 are kept installed elsewhere.
114 * Surf binding with SWIG (code generated in maintainer mode only):
115 - plugin to handle callbacks
116 - CPU model only for the moment
119 * Supernovae build mode is definitively removed. It was used to improve
120 inlining and inter-module optimizations. It is nowadays superseded by
121 link-time optimizations commonly available in compilers.
122 * Update ns-3 find lib. Bindings for ns-3 should work again now.
123 * Add boost dependency for surf++
124 * Add new macro for tests
125 - ADD_TESH(name <tesh_args>)
126 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
130 - xbt_bar_t XBT_barrier_init(unsigned int count)
131 - int XBT_barrier_wait(xbt_bar_t barrier)
132 - void XBT_barrier_destroy(xbt_bar_t barrier)
133 * Make the xbt_os_time module public
135 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
137 SimGrid (3.10) stable; urgency=low
139 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
142 * Reintegrate Java to the main archive as desynchronizing these
143 package is not acceptable anymore (Java is now considered stable)
144 * Add explicit synchronization facilities through semaphores
145 * Bug fix: Task.setDataSize() only changed the C world, not the value
146 cached in the Java world
149 * Dramatically change the way files are handled. API and internals changed, but
150 this part of MSG was not considered as production grade either.
151 * Add explicit synchronization facilities through semaphores
152 * Add a new function MSG_host_get_process_list()
153 * Preliminary DVFS support (see examples/msg/energy/ for details)
156 * SMPI is now included directly in the libsimgrid as the windows
157 linker doesn't force us on splitting it anymore.
158 * Improvements of the SMPI replay tool:
159 - Most of the collective communications are now rooted in the same process as
160 in the original application.
161 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
162 used until now). Multiple data types can now be used in a trace.
163 - The replay tool now supports traces produce either by TAU or a modified
165 - Bug Fix: the compute part of the reduce action is now taken into account.
166 - Gatherv collective is now supported
167 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
168 of smpirun outputs time independent traces for the current run. One file
169 is created per process. If too many processes are simulated, this behavior
170 can be changed to one file for all processes by using the
171 tracing/smpi/format/ti_one_file flag
172 * smpirun generates the host file if needed (with given host count and platform)
173 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
174 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
175 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
176 --cfg=smpi/coll_selector:(mpich/ompi)
177 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
179 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
180 happening inside a collective SMPI call.
181 * Fix the behavior of complex data types handling.
182 * Make MPI_Wtime another synchronization point to take computations into
184 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
185 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
186 * Add all missing Fortran bindings, SMPI should work with Fortran 90
187 (no privatization of global variables yet)
188 * Preliminary DVFS support (see examples/smpi/energy/ for details)
191 * Verification of liveness properties is now available for SMPI applications
192 (in addition to MSG applications)
193 * Bugged examples using SMPI in examples/smpi/mc/
194 * Add --cfg=model-check/visited option. Allows the verification of infinite
195 programs. Detection of loops in the execution thanks to the system state
196 comparison and reduction of the state space to explore. Can be combined with
197 DPOR for safety properties.
200 * Allow to change SimGrid configuration (see --help) within the code
201 thanks to SD_config() as it can be done in MSG.
202 * Add a new function SD_task_set_amount() upon user request.
205 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
206 * Remove rule based routing (no more PCRE dependency)
207 * Add a limiter_link option to cluster tag, to specify a maximum reachable
208 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
209 * Add a loopback_bw and loopback_lat options to cluster tag.
210 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
211 routing. Now peers are encapsulated in an AS and have their own private
212 router but this is transparent.
215 * Our own implementation of getline is renamed xbt_getline, and gets
216 used even if the OS provide a getline(). This should reduce the
217 configuration complexity by using the same code on all platforms.
218 * New type: xbt_cfg_elm_boolean.
219 * Allow to use yes/no for boolean configuration options in the command line.
220 * Allow to disable SimGrid cleanups at exit from command line option.
221 There are situations where one may want a simulation to end with an exit.
222 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
223 annoying when scripting around the simulator. Adding a
224 --cfg=clean_atexit:no allows to circumvent this issue.
227 * Lots of memory leaks were corrected in this release.
228 There is no known memory leaks anymore, in all of our 600+ tests.
229 * New command line option --version, to get SimGrid version information.
230 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
232 * Supernovae builds are deprecated, and expected to be removed in the next
235 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
237 SimGrid (3.9) stable; urgency=low
239 The Grasgory release: GRAS is really dead now.
241 * Complete overhaul of the internal host structures scheme.
244 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
245 was considered as experimental and badly maintained since then.
246 * Keeping it was thus a trap to our potential users, that could take
247 it instead of MSG or SMPI by mistake despite is pity state.
248 * GRAS seems to have very few users (if any), and no one volunteered
249 to maintain it further. It also induces a lot of XBT code (for
250 portability sake), that must be maintained too.
251 * For all these reasons, we killed GRAS. If someone wants to revive it
252 in the future, don't cry, our git history still remembers of GRAS.
255 * Major overhaul. Merge our documentation again as time proved that
256 splitting it was really not helping our users.
257 * Further improve the developer documentation to help newcomers
258 hacking on SimGrid itself. The user documentation (and in
259 particular, the beginner documentation) is still in a sorry state.
262 * Now works on Windows too!
263 * Much more extensive test suite, from MPICH
266 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
267 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
268 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
269 * Bug fix: let task be scheduled when the last dependency to be solved is
270 a control dependency.
271 * Remove SD_load_environment_script function.
272 Use the C sg_platf function if you want to declare a platform
276 * New function: MSG_process_get_number()
277 * Old function documented: MSG_config()
278 * Remove MSG_load_platform_script function
279 Use the C sg_platf function if you want to declare a platform
283 * Change the default value of the TCP_gamma constant (maximal size of TCP
284 congestion window) to a more realistic 4MiB value. If you notice changes in
285 your simulation results, you can fall back to the previous 20k tiny window
286 by adding --cfg=network/TCP_gamma:20000 on command line.
287 * (Hopefully) fix a bug wrt periodic availability and state traces
288 * Bug fix: use default values at start when first event in availability/state
289 trace is not at time 0.
292 * remove the "new_" part of function name sg_platf_new_trace_connect
293 (resulting in sg_platf_trace_connect), since it does not create
297 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
298 use case where it's really mandatory, and maintaining it was a pain in
300 * New: xbt_fifo_search(), search an item with a user-provided
301 comparison function instead of dumb pointer comparison.
304 * Fix the lua deployment:
305 Use `simgrid.init_application()` before deployment instead of
306 `simgrid.msg_register_application()` after.
309 * Transfer the tracing files into the corresponding modules.
311 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
313 SimGrid (3.8.1) stable; urgency=low
315 The "we are told that some people want to also *install* the simgrid
318 * Add missing file "tesh.1" to the archive.
320 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
322 SimGrid (3.8) stable; urgency=low
324 The Psssshiiiit release: SimGrid jumps into the Cloud.
327 * Add an experimental interface to manipulate VMs. They are mainly
328 process groups with very few intrinsic semantic, but they should
329 allow you to build the semantic you want easily.
330 * New function: MSG_host_set_property_value()
331 * New function: MSG_process_on_exit(). To clean memory in all cases.
332 * Bug fixes that made the host (and link) failures unusable.
333 * Add a way to auto-restart process when the host in which they are
334 executing comes back (ON_FAILURE="RESTART" on deployment file,
335 MSG_process_auto_restart_set).
336 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
337 please stop using the old ones, they are DEPRECATED.
339 * Deprecate functions MSG_global_init() / MSG_global_init_args()
340 Please use MSG_init() instead. (reducing the amount of entry
341 points in the library helps us).
342 * Make it impossible to link against the wrong version of the lib
343 * Deprecate MSG_clean(). No need to call it anymore.
344 * Function MSG_get_host_number() is not deprecated anymore.
347 * Split the doc into a user guide and a reference guide.
348 * Start a developper guide to help people hacking on SimGrid.
351 * Enable tracing by default. This modules rocks you should use it.
352 * Remove option custom_flags. Now use environment variables CFLAGS
354 * Use default cmake things to detect lua instead of home grown ones.
355 * New option "enable_mallocators" to disable mallocators, for debugging
356 purpose ("on" by default).
359 * Bug fixes around the resource failures: don't let the processes
360 survive the host they are running onto.
361 * Add an interface to auto-restart processes when the host in which they are
362 executing comes back.
363 * Ensures that SIMIX_clean is called automatically. It's not part of
364 the public interface anymore (bindings should be updated).
367 * Bug fix for when SD_Simulate is called with a positive value: be careful
368 when comparing doubles. Sometimes they are different for non significant
370 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
371 parallel task whose initial work is distributed among host according
372 to the Amdahl's law. Such tasks are created with a parameter alpha
373 that corresponds to the non-parallelizable part of the computation.
374 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
375 between two sets of workstations assuming a 1D block distribution (each
376 workstation owns a similar share of data) on both sides.
378 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
379 Data redistribution will be automatically scheduled once parent and child
380 are both scheduled. The filling of computation_amount and
381 communication_amount structures is now done seamlessly thanks to the chosen
383 * New function SD_workstation_dump to display various information
384 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
385 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
386 bandwidth on the route onto which the task is scheduled and the amount of
388 To divide the nominal bandwidth by 2, the rate then has to be :
389 rate = bandwidth/(2*amount)
390 * Compute tasks that have failed can now be rescheduled and executed again
391 (from their beginning)
392 * Increasing source code coverage (src/simdag is now covered at 95.8%
396 * Re-implement time-independent trace replay using SMPI (at the
397 smpi_smp_* level) instead of MSG. This should replace
398 examples/msg/actions/actions.c
399 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
400 hindexed and structs)
401 * Implement the exchange of non-contiguous data.
402 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
403 * Correct behavior of smpi/sender_gap and set its default value to 0
404 * Add option to asynchronously send small messages to allow better
405 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
406 specifies the size in bytes under which messages will be asynchronously sent.
407 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
408 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
411 * New C interface to define a platform: XML is now optional.
412 For more info, please check include/simgrid/platf.h.
413 * New interface to define random platforms from the C:
414 For more info, please check include/simgrid/platf_generator.h and
415 examples/msg/masterslave/masterslave_platfgen.c
416 * Export a sg_cmdline dynar containing all the arguments we got from
420 * Two new tracing options for adding comments to trace file so you
421 can track your experiments (see --help-tracing for details).
422 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
423 * Adding the SimGrid version that generated the trace file as a comment.
424 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
425 * Fix to avoid key clashes on Paje links
426 * Other minor fixes related to the Paje specification
429 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
430 and renamed to xbt_str_hash() and xbt_str_hash_ext().
431 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
432 * Greatly improve the robustness of mmalloc to user errors (such as
433 using an area after freeing it, or freeing it twice)
435 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
437 SimGrid (3.7.1) stable; urgency=low
440 * Restore the prototype of MSG_process_create_with_environment() to
441 the pre-3.7 situation by removing the kill_time argument.
442 * Add a MSG_process_set_kill_time() function instead.
445 * Fix weird behaviors when dealing with parallel tasks.
448 * Simgrid is now built as a dll.
449 * Simgrid-java now works on Windows.
450 * Simgrid-Java is now included into Windows package.
453 * First pre-build package for MacOSX.
456 * Fix compilation when using MSG_USE_DEPRECATED.
457 * Fix some compilation issues on Macs and Windows.
458 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
459 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
461 We discovered that the Lua console is broken, but we are missing the
462 manpower to fix it right now. The problem existed in 3.7 too, so we
463 are not blocking the release for that. Sorry if you depended on this
464 feature, any help would be really welcome.
466 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
468 SimGrid (3.7) stable; urgency=low
470 The "spring cleanups (before the next Big Project kicks in)" release.
473 * We can specify the SMPI latency/bandwidth factor with command line
474 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
475 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
476 You can also use the "config tag" from platform file by adding this line
477 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
478 (see "example/platforms/tag_config.xml" to use "config tag").
479 Note that the command line supersedes the platform file configuration.
480 * Change the correction factors used in LMM model, according to
481 the latest experiments described in INRIA RR-7821.
482 Accuracy should be improved this way.
483 * Use the partial invalidation optimization by default for the
484 network too. Should produce the exact same results, only faster.
485 * Major cleanup in surf to merge models and split some optimization
486 mechanisms from the core of the models. As a result you can now
487 specify which model to use (e.g., --cfg=network/model:LV08
488 --cfg=cpu/model:Cas01) and which optimization mode to use
489 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
490 Incompatible combinations should err at initialization. See
491 --help-models for the list of all models and optimization modes.
492 * The CLM03 workstation model was dropped for simplicity because it
493 used the deprecated CM02 network model. Use default instead.
494 * Rename the TCP_gamma configuration option to network/TCP_gamma
495 * Rename the coordinates configuration option to
496 network/coordinates, and document it
497 * Use now crosstraffic keyword instead of the terribly misleading
498 fullduplex keyword. It is activated by default now in the current
499 default model, use --cfg=network/crosstraffic:0 to turn it off.
500 * Ongoing refactoring the model parsing to make XML files optional
501 See include/simgrid/platf.h for details (still to be completed)
504 * Major overhaul of the documentation. Almost instructive now :/
505 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
506 functions and friends. This interface was considered as
507 deprecated since over 2 years, it's time to inform our users that it is.
508 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
509 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
510 use these (crappy) functions in your code.
511 These functions will be removed soon. Stop using them now.
512 * Deprecate MSG_get_host_{table,number}
513 Implement MSG_hosts_as_dynar() instead.
514 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
515 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
518 * Stabilize the parallel execution mode of user contexts
519 * Introduce configuration variables to control parallel execution:
520 - contexts/synchro: Synchronization mode to use when running
521 contexts in parallel (either futex, posix or busy_wait)
522 - contexts/parallel_threshold: Minimal number of user contexts
523 that must be part of a scheduling round to switch to parallel
524 execution mode (raw contexts only)
525 * Fix bugs that prevented to use suspend/resume along with
526 synchronization structures.
527 * Fix bugs in process termination that lead to invalid memory access
528 in very specific conditions.
531 * Introduce a parallel mode for the models (controlled by surf/nthreads
532 configuration item). In our tests, running the models in parallel
533 never lead to any speedups because they are so fast that the gain
534 of computing each model in parallel does not amortizes the
535 synchronization costs, even when ultra fast futexes are used.
536 This is released anyway because YMMV.
539 * Performance boost by using a swag internally to compute the set of
540 tasks that are finished and should constitute the return value of
544 * Enable it by default now that it is considered rather stable.
547 * Documentation of the tracing functions.
548 * Performance gains when tracing categorized/uncategorized resource
549 utilization by avoiding calls to get route when updating resource
550 variables. LMM constraints are being used instead.
551 * API changed to set task categories. Use MSG_task_set_category instead
552 of TRACE_msg_set_task_category, and SD_task_set_category instead
553 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
554 * Bugfix for graphicator, routes not correctly obtained, memory leaks
555 * Examples for link user variables added (see at examples/msg/tracing/)
556 * Deprecated function TRACE_msg_set_process_category completely removed
557 * Trace header updated according to the latest Paje file format
558 * Tracing network lazy updates, no longer obligate users to use full updates
559 * --cfg=tracing/platform:1 also registers power/bandwidth variables
560 * Experimental: let user code declare/set/push/pop application states for hosts
561 * API changed to allow the manual creation of graph configuration files
562 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
565 * Improve the API of Lua MSG bindings, using the Lua spirit.
566 * Each simulated process now lives in its own Lua world (globals are
567 automatically duplicated). It helps writing simulators. Will allow
568 to run Splay programs within SimGrid in the future.
569 * Add a Chord example in Lua, equivalent to the MSG one.
572 * Start the implementation of a solution to express temporal
573 properties, not only local assertions. This is still an
574 experimental work in progress, stay clear from it to be safe.
578 - Add new runtime parameters --help-logs and --help-log-categories
579 to display informations about supported logging parameters and
581 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
582 don't exists anymore.
583 * Mallocators: allow value NULL for the reset function.
585 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
586 create homogeneous dictionaries, where all the elements share the
587 same free function. Non homogeneous dictionaries will be
588 deprecated in the next release.
589 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
590 - Multi-level dictionaries are deprecated.
592 - new function xbt_dynar_search_or_negative() that is useful when
593 you have less than 2 million elements in your dynar and don't
594 want of the extra complexity of catching exceptions when the
595 element is not found.
597 - Make xbt_os_thread module (for thread portability) public.
598 Documentation is still to come, sorry.
600 - Cleanups and simplifications to make it maintainable again.
601 - Exotic features (such as memalign and valloc) were removed.
602 - The metadata were extended and improved so that the
603 model-checker becomes able to explore and inspect the heaps.
604 - This may induce a performance drop when enable_model-checking is
605 ON in cmake (even if it's not used in the simulation), but it is
606 necessary at this point to get MC working.
608 Turn model-checking OFF if simulation performance matters to you.
609 Not enabling it at runtime is not enough, disable it in cmake.
611 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
613 SimGrid (3.6.2) stable; urgency=low
615 The "Not coding new stuff allows to polish old things" release.
618 * New bindings to the NS3 packet level simulator (experimental)
619 * Use the raw (efficient) execution contexts instead of the sysv
620 (portable) ones when possible.
621 * libpcre is now mandatory in any cases since not using it led to
622 severe performance loss and possibly other issues
623 * Update the XML platforms:
624 - G5K: include the latest machine in Nancy
625 - GridPP and LCG: new platforms
626 * Documentation was partially updated, at least (more to come)
628 Bug fixes, cosmetics and small improvements
629 * Free terminated processes before the end of the simulation to avoid
630 exhausting the memory of users having very dynamic amount of
632 * Bug fix and cosmetics about canceling non-running tasks
633 * Bug fix about the dot loader's issues when using libcgraph
636 * Create an installer for windows with nsis (amd64 and win32)
637 - Add an hello world project to illustrate simgrid project creation.
638 - Embed libpcre into the Simgrid installer to avoid
639 its compilation burden
640 * The raw execution contexts should work on Apple now
641 * Port to Windows 64 bits
642 - Sysv contexts now have an implementation for this arch
643 - GRAS communication features now support this arch
644 * Drop support for borland compiler on windows
645 - this code was not maintained, and we kinda depend on gcc nowadays
646 * Fix portability issues on kfreebsd/gnu: build error about semaphores
647 * Fix portability issue on unstable ubuntu: linker became picky on
650 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
652 SimGrid (3.6.1) stable; urgency=low
654 The "Oops, we broke Macs too" release
657 * Fixed contexts detection so that raw ones are used when possible
658 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
659 leads to a strange error, with user code segfaulting sometimes when
660 the generated code is not perfectly aligned (which is not
661 controllable from the user side, depends on the amount of code)
664 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
666 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
668 SimGrid (3.6) stable; urgency=medium
670 The Summer Release, also known as the "OMG! They Killed Kenny!" version
673 * Bindings now constitute their own package, separated from the main one.
674 Rationale: reduce our maintenance nightmare by reducing the module coupling
675 They will soon be released on their own on gforge.
677 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
678 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
680 GRAS: It is not considered as stable anymore, but experimental. Sorry.
681 * It's not quite deprecated for now because we have no replacement,
682 but it may soon become the case.
685 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
686 * Fortran: New user-level cache variable to store the rank of the running
687 process. This improves performance by an order of magnitude.
688 * C: New coccinelle script to automatically locate and modify global and
689 local static variables.
690 * Improved SMPI network model with a sender-side gap to account for multiple
694 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
695 only say if a communication is finished, no matter whether it succeeded or
696 failed. You can call MSG_comm_get_status() to know the status of a finished
698 * New function MSG_task_dsend() to send a task and detach it. When a
699 communication is detached, you are never notified of its success or failure
700 and the memory is released automatically once it is finished. This function
701 is useful when you don't care about the end nor the success of a
703 * Change the prototypes of action replay. Sorry for inconvenience,
704 but this is really more efficient this way (and to adapt your code,
705 you just have to fix the initialization, that shouldn't be too long)
706 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
707 ever managed to do anything useful with it.
708 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
709 takes a dynar of communications. It returns immediately and gives the
710 index of a finished communication (if any).
711 * New example: a basic implementation of the Chord P2P algorithm.
714 * New model for multi-core CPUs. You can now use the core attribute to
715 precise the number of cores of a host. This is a basic model. Every
716 process running on the host receives at most the power provided in
717 the DTD (throughput<=power). Total throughput of process cannot exceed
719 * New peer tag. This peer tag creates a tiny AS comprising a host and a
720 router linked by an up-link and a down-link (possibly asymmetrical).
721 This kind of pattern allows to easily build last-mile model style platforms.
722 Aggregating such patterns in a rule-based AS is thus the technique of
723 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
724 * New model for Vivaldi routing. We transformed the Vivaldi network model
725 into a Vivaldi routing model (based on the rule-based model). This allows to
726 combine Vivaldi based latencies with last-mile platforms.
729 * Added a check for NaN of IEEE754 infinite in the double entries of
731 * Introduce a new context factory "raw", highly inspirated from the
732 ucontext factory, but using manually crafted functions in assembly to
733 do the work in an efficient manner.
734 * Allow to change the used context factory at run time, not only at
735 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
736 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
737 contexts at runtime (only with raw contexts or ucontexts).
738 * Completely rewrote this module to allow parallel execution of user
739 processes. Use --cfg=contexts/nthreads:N to execute user processes
740 with N parallel threads (the default is 1, meaning no parallelism).
741 * Allow to decide dynamically between sequential and parallel modes.
742 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
743 processes in parallel only when their number is greater than or equal to P
745 * Added a check for NaN of IEEE754 infinite in the double entries of
749 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
750 won't output the state of processes when interrupted with Ctrl-C
751 * Add a new function xbt_dynar_to_array that transforms a dynar into a
752 NULL-terminated array. This may solve backward compatibility issues
753 due to the change to return type of SD_simulate. See also:
754 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
755 * Add new macros with variable number of arguments.
756 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
757 - in xbt/asserts.h: xbt_assert
758 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
759 - in xbt/ex.h: THROWF and RETHROWF.
760 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
762 * Change xbt_die() to accept a format string with arguments, just like printf.
763 * New data structure: xbt_lib_t, like a dict but more general and with better
767 * New configuration options
768 Options triva/categorized and triva/uncategorized can be used to generate
769 graph configuration files for Triva visualization tool.
770 * Configuration option tracing/platform is renamed to tracing/categorized
771 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
772 * New TRACE_user_link_variable interface:
773 User provides the name of the link and the tracing variable to attach to it
774 * the declaration of tracing categories must be done after the environment
776 * simpler tracing interface, just one way to declare categories
777 TRACE_category or TRACE_category_with_color, it is up to you
778 * links in the trace file are again identified by their names
779 * trace contains the full platform hierarchy exactly as declared using the ASes
780 * Options tracing/msg/[task|process]:1 groups the process by hosts
781 for both cases, tasks and processes must have names that are unique during the simulation
782 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
783 * The experimental option tracing/msg/volume is deprecated
784 its functionality may be reincorporated if needed
786 The tracing generates a trace file with unordered timestamped events,
787 because of the way the core simulator (surf) works. A script available
788 at the tools directory (fix-paje-trace.sh) can be used to put the events
789 in order. We have changed the tracing so it can generate ordered timestamped
790 events in the final trace, but depending on the simulator (and how much time
791 is simulated) that can lead to a huge memory utilization. It is deactivated
792 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
795 * Define a SIMGRID_VERSION macro in simgrid_config.h.
796 - We are trying hard to keep the API stable, but it may happen that
797 some things change (we're a research project after all, not a
798 nuclear plant operating system). If such things should happen, you
799 could rely on that macro to adapt.
800 - current value: 30600 for 3.06.00, aka 3.6
801 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
802 a number that can be compared with SIMGRID_VERSION.
803 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
804 assertions and verbose/debug logging events are disabled at compile time.
806 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
808 SimGrid (3.5) stable; urgency=medium
811 * New feature to any SimGrid-based simulator: Model-Checking
812 Check SIN#1 for more details.
815 * New Model SMPI (three-interval linear regression for correction factors)
816 See RR-7426, available at http://hal.inria.fr/inria-00527150
817 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
818 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
819 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
820 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
821 * See also src/smpi/README
826 - Tracing API changes: TRACE_start and TRACE_end should not be called
827 by user-code. They are automatically called by simulators created
828 with SimDAG, MSG and SMPI if the toolkit is compiled with
829 tracing_enabled=ON. Categories declaration and utilization remain the
830 same for MSG and SimDag.
831 - A function was added to the tracing API to declare categories with
833 - TRACE_category_with_color (char *category, char *color)
834 where color must be in the following format
835 "%f %f %f", red, green, blue
836 and red, green, blue are float values in the interval [0, 1]
837 - User can specify NULL as color parameter, or continue calling
839 On that case, the tracing system will define random colors
840 - The following command-line options are supported:
841 --cfg=tracing/filename:msg.trace
842 --cfg=tracing:1 (activate tracing, needed to use others)
843 --cfg=tracing/platform:1 (categorized resource use)
844 --cfg=tracing/uncategorized:1 (uncategorized resource use)
845 --cfg=tracing/msg/task:1 (task creation)
846 --cfg=tracing/msg/process:1 (process creation, migration)
847 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
848 --cfg=tracing/smpi:1 (SMPI interface tracing)
849 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
850 - examples of examples/msg/tracing updated
852 - DAXLoader and DOTLoader functions can generate tasks with categories
853 - A new function to attribute a category to SD tasks:
854 TRACE_sd_set_task_category (SD_task_t task, char *category)
855 * Tracing the MPI interface implemented by SMPI
856 - Collective operations are traced with states
857 - Point-to-Point operations are traced with states/links
858 - Tracing activated by a parameter "-trace filename" passed
859 to smpirun during execution (considering that simgrid
860 is compiled with tracing enabled)
861 - To run the simulation with gdb, the simulator
862 accepts --cfg=tracing/smpi:1 to trace SMPI
863 - tesh files to check if smpi tracing is ok
864 - See examples/smpi/NAS/DT-trace
865 * GTNetS tracing re-worked
866 - adaptation to the tracing system of GTNets to cope
867 with modifications regarding the fullduplex mode
868 - new tesh files to check if gtnets tracing is ok
871 * Asynchronous communications through the functions:
872 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
873 * New function: MSG_load_platform_script()
874 to make possible using a lua script instead of XML files to set up platforms
875 * New function: MSG_set_function
876 to associate functions to processes, used when bypassing the parser
877 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
879 Platforms: Add some more examples in examples/platforms
880 * Grid'5000: see www.grid5000.fr
881 * *_30000_hosts.xml: various huge files [mainly scalability testing]
884 * Change the XML format. This is a very important modification. SimGrid
885 3.5 introduces a new hierarchical format based on the notion of
886 Autonomous Systems. Compatibility with old format is ensured
887 through the perl script provided in the install bin directory
888 bin/simgrid_update_xml.
889 It is now possible to build platforms with specific routing
890 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
891 connect several platforms together. We will try to provide soon
892 set of realistic platforms exploiting these properties (have a
893 look at examples/platforms/ for the moment).
894 * Take the opportunity of the XML format change to be a good XML citizen:
895 rename link:ctn to link_ctn and similar changes (also dealt with by
897 * Add a new routing scheme (rule-based) using regular expressions. It
898 enables to have an extremely low memory footprint when the
899 underlying routing is simple and can be compactly described. You
900 need to have libpcre4-dev (perl regular expressions) installed if
901 you want to use this routing scheme.
902 * Revive the cluster TAG and allow to easily and efficiently (both in
903 term of memory and speed) connect clusters together. Have a look
904 at teshsuite/simdag/platforms/ to see how this can be done. With
905 this tag, you can create clusters with thousands of tasks at no
906 cost (have a look at examples/platforms/).
907 Note: clusters are implemented as ASes, so there is no need for an
908 enclosing AS tag if you have only one cluster in your platform.
909 * Add new generic functions in the public interface that allows the user
910 to call SURF 'create_resource' methods from your code (same
911 functionality as the XML bypass mechanism but with a much lighter
913 * Add a new model (enabled through command line --cfg=network/model:SMPI)
914 that uses a piecewise linear approximation to produce better
915 results when exchanging small messages.
916 * Add a new parameter to handle correctly full duplex link and account
917 for interferences between uplink and downlink communications
918 (activate with --cfg=fullduplex:1).
921 * Rename the SD_READY (all dependencies are satisfied and task is
922 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
923 dependencies are satisfied) state.
924 This prevents a confusion between the notion of "ready to schedule"
925 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
926 simulated" (SD_RUNNABLE) used by the simulation kernel.
927 * Change the way a task is considered as ready. Instead of removing
928 dependencies when a task is done, a counter is decreased. This way,
929 it is always possible to reach ancestors thanks to the
930 SD_taks_get_parents function (even after the end of the simulation.)
931 * Change the return type of SD_Simulate from (SD_task_t*) into
932 xbt_dynar_t. This function was in handling a dynar internally and
933 converted it into a NULL terminated array for historical reasons.
934 * New function SD_dotload(char*) to load a DAG described in dot
935 format. This loader and the corresponding examples require the
936 installation of the graphviz library.
937 * Fix a bug in the management of tasks of size 0 in the surf network
938 models. This problem was only visible with SIMDAG and you should
939 thus disregard results produced with earlier versions if you
940 relied on this feature (some tasks were blocked because of this).
941 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
942 even though your applications were doing only point-to-point
943 communications and sequential computations. Now you can really use any
944 model you want (of course, if you create real parallel tasks, which are
945 not implemented in most models beside ptaskL07, this will abort).
946 * Add an example that schedules a DAX on an heterogeneous platform
947 using a Min-Min strategy.
948 * New function SD_workstation_get_current_task() that returns the kind
949 of task currently running on a workstation in the sequential access
951 * Raise some warnings when unexecuted tasks remains at the end of the
952 simulation. This is usually caused by cycles in the DAG.
955 * New function: SIMIX_process_set_function() called by MSG_set_function
956 * Change the underlying waiting queue in semaphores so that a process
957 can wait on several of them simultaneously (as in waitany).
958 * Fix the way to handle tokens in semaphores so that all access patterns
959 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
960 * kill the dirty pimple SIMIX_message_sizes_output()
961 Please use (proper) visualization instead
964 * New data container: setset (set of sets of elements)
965 * New module: mmalloc (mapped malloc, allowing to have several
966 independent segments of malloc)
967 * New function: xbt_dict_cursor_set_data()
968 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
969 * New function: xbt_dynar_is_empty()
970 * New function: xbt_fifo_get_last_item()
971 * Fix xbt_dynar_shrink(): use the right element size.
972 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
973 uninitialized areas during expand.
974 * Fix semaphores: previous implementation was severely broken.
975 * Use library init/fini functions for our initialization.
976 - you can use logs and other feature as soon as you want in your
977 code (even before the xbt_init / MSG_init)
978 - xbt_exit is now a no-op and produce a warning when used.
981 * Port GRAS to new SIMIX mechanisms. This allows gras users to
982 benefit from the latest improvement to the simulation kernel.
983 * Kill measurement sockets for now. If you rely on them, sorry. This
984 release is not for you. This feature will be reintroduced in the
985 future, but we cannot delay the release any further.
986 * New function: gras_msgtype_get_name().
987 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
988 * Fix (at last) the pmm example: it should not randomly fail anymore.
990 Build chain: bug fixes and overall polishing
991 * Cmake is now stable enough. Hence, we killed the autotools.
992 * Port to windows ( TM :)
993 * Fix the 'make install' target.
994 No need to use 'make install-simgrid' anymore
995 * Introduce a 'make dist' target compiling a *source* archive
996 'make package' compiles a binary archive
997 * Compile java files only on need
998 * Add --cd and --setenv command line options to tesh
999 * Out of source builds are not fully supported yet, but we are close
1000 * Enable supernovae and optimization flags by default for our users
1003 * Add layer to set up environment directly from lua, without XML.
1004 * The effect of gras_stub_generator can be achieved through
1005 lua too (check examples/gras/console/ping_generator.lua)
1007 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1009 SimGrid (3.4.1) stable; urgency=low
1011 The "Polishing easter eggs is probably a good idea" release.
1012 This is a bug fixes release only.
1016 * Fix a bug preventing the tasks from begin garbage collected.
1019 * Fix a bug occuring when a host involved in a communication fails.
1020 This was not detected properly by the other peer involved in the
1021 communication. Now, it's reported as a network error.
1024 * Warn the user about loop dependencies in data flow of DAX files
1025 * Obey the control-flow dependencies of DAX files
1028 * Add option "enable_smpi" allowing to not compile SMPI.
1029 Probably useful for the (Mac) users experiencing a build error here
1030 * Improve the detection of lua5.1 and ruby1.8
1032 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1034 SimGrid (3.4) stable; urgency=low
1036 The "Easter in Cargese" release. Also known as (major changes):
1038 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1039 ~> bindings were greatly improved
1040 ~> new tracing infrastructure for better visualization introduced
1042 * the "Welcome to configury modernity" release.
1043 ~> we switched from autotools to cmake, and improved our cdash
1046 A more detailled list of changes follow (full detail in svn log).
1048 Java Bindings: Various Cleanups
1049 * (install java-gcj-compat-dev on debian-like to use them)
1050 * Remove put/get: no need to export deprecated interface in Java
1051 Use send/receive instead.
1052 * Cleanup the examples and add a README per directory
1053 * Remove example autoDestination (that's the only way to go now)
1054 * Remove example explicitDestination (was a plain copy of basic)
1055 * Make JniException a runtime exception, so that there is no need to
1056 declare the fact that you may encounter such a beast. I guess that
1057 nobody will ever want to survive such error.
1058 * Create specific errors for each MSG case of failure:
1059 host failure, transfer failure, timeout, task cancelled
1060 * Cleanup the exceptions that may get thrown by each function
1061 * Other internal cleanups in Java bindings. Performance still bad :/
1062 Ruby and Lua Bindings: create them
1063 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1064 * That's new and great, you should try them out.
1065 Same functionalities than Java bindings, only even less polished
1067 * Kill the useless "rate" argument of SD_task_get_execution_time()
1068 Everyone used to provide -1 as a value, it was not used, and the
1069 semantic of a possible use wasn't even clear.
1070 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1071 either as comm costs or compute costs to mean that there is no
1072 such thing for that specific task.
1073 * Add a SD_task_set_name() function
1074 * Fix SD_task_unschedule() on typed tasks
1075 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1076 * In DAX loader, accept useless 'level' attributes to <job> since
1077 LIGO DAGs have them (seem to be to ease graphical representation).
1079 * Add an example masterslave_mailbox.c using send/receive and not
1080 the deprecated put/get interface.
1081 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1082 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1084 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1085 (declare MSG_USE_DEPRECATED to still have the old name)
1086 * Add a MSG_task_set_data() function
1087 * About trace replay (see examples/msg/actions):
1089 - Allow to work with splitted trace files for each process
1090 Give the specific trace file as argument of each process,
1091 and call MSG_action_trace_run(NULL)
1092 You can still have one merged file for all processes.
1093 - Fix implementation of collective operations
1094 * Allow task_execute() on 0-sized tasks (closes #10063)
1096 * This is the first release of SimGrid where SMPI is not considered
1097 beta anymore (even if some corners should still be improved)
1098 * Port over the new SIMIX_network submodule (internal refactoring)
1099 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1100 * Implement more missing elements of the standard:
1102 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1104 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1105 MPI_Scan() MPI_Get_processor_name()
1106 - Added implementation of missing case for Alltoall (warning: it's
1107 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1108 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1109 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1110 SMPI_MPI_Allgatherv()
1111 * Bug fixes include:
1112 - MPI_Waitsome() was broken
1113 - Allow relative includes in smpicc
1114 - Command line cfg argument 'reference_speed' was ignored...
1115 - Some functions did not properly lead to auto-benching of user code
1116 - smpicc passes -O2 by default (just like openmpi one)
1118 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1119 * Bug fixes about timeouts during communications
1120 * add SIMIX_message_sizes_output() as a pimple to write to file the
1121 amount of messages per size. Use gnuplot to get histogram.
1122 Pimple because that's the only user-visible function of simix,
1123 defined directly in xbt.h (irk, sorry)
1125 - Add a SIMIX_sem_get_capacity() function
1126 - Fix interactions with processe resume/suspende
1127 - release_forever() was stupidly broken
1128 - Fix SIMIX_display_process_status() for processes in a semaphore
1129 - Make SIMIX_sem_block_onto() user-visible
1130 * Refactoring context stuff:
1131 - Use pseudo-OOP for better modularity
1132 - reimplement SIMIX_process_kill() without process_schedule() so
1133 that the latter can take as invariant that it is called from
1135 - Merge context_start into context_new for sake of simplicity
1137 * Add a Vivaldi network model, coded live during SUD'10 ;)
1138 * Rename configuration variables to start a hierarchy:
1139 o cpu_model -> cpu/model
1140 o network_model -> network/model
1141 o workstation_model -> workstation/model
1142 * New configuration variables:
1143 o network/bandwidth_factor: correction to bandwith
1144 o network/latency_factor: correction to latency
1145 o netwotk/weight_S: correction to the weight of competing streams
1146 * Add a long description to the models, that users can see with such
1147 argument on the command line: --cfg=cpu/model:help
1148 * --help-models display the long description of all known models
1150 * config: add the ability to set a default value after registration
1151 Does not override any previously set value (e.g. from cmd line)
1152 * dict: allow to have integer key and data.
1153 When so, you need to use the following functions
1154 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1155 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1156 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1157 In contrary to regular dicts, the key is not malloced before copy.
1158 Mixing scalar and regular elements in the same dict is not tested
1160 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1161 Tracing for Visualization:
1162 * SimGrid is now instrumented in order to generate a trace file for
1163 visualization analysis: to use it, need to compile SimGrid with the
1164 "tracing" option enabled, and instrument the program using SimGrid with
1165 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1166 (among other functions).
1167 * The instrumentation only traces the platform utilization for now
1168 * Documentation to use the tracing functions and how to analyze the
1169 traces with the Triva tool is written.
1170 * More information about: SimGrid FAQ (in the section Tracing Simulations
1173 * We moved to cmake as default build system. Autotools support will
1174 be dropped soon. Check the FAQ for more info about how to use it.
1175 * Greatly improved our cdash/ctest interactions
1176 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1177 * Added memory checking tests with valgrind; lot of memleak fixing.
1178 This may be the first release of simgrid with so few memory issues
1179 * Added code coverage tests.
1180 Our coverage is still improvable, but at least we see it on cdash.
1182 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1184 SimGrid (3.3.4) stable; urgency=low
1186 The "Desktop Grid needs love too" release (also called Xmas release).
1188 Models improvements:
1189 * Major speedup in the maxmin system solving by using lazy evaluation
1190 Instead of solving completely the maxmin system at each iteration,
1191 only invalidate (and recompute) the modified parts.
1192 This new feature is enabled in default models but you can try to
1193 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1194 * Cas01 IMproved as default CPU model
1195 This CPU model is the same Cas01 model, but it uses the
1196 maxmin-selective-update flag and a heap structure to manage
1197 actions on SURF kernel.
1198 It reduces the complexity to find the next action to finish and,
1199 consequently, it's faster than the old Cas01.
1200 This is the new default CPU model (Cas01).
1201 * Rename the old Cas01 model to Cas01_fullupdate
1202 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1203 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1204 * CpuTI (CPU Trace Integration)
1205 A new CPU model whose objective is simulate faster when using
1206 availability trace files.
1207 Instead of using a full featured, over engineered maxmin system for
1208 CPU modeling, this model does the pre-integration of traces files
1209 to calculate the amount of CPU power available, and so, executes
1210 faster than the old CPU models.
1211 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1212 * Use LV08 as default network model since it gives better accuracy
1213 for small messages and shouldn't change things for big ones.
1214 Use --cfg=network_model:CM02 to get the previous behavior.
1217 ******************************************
1218 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1219 ******************************************
1220 * The new CPU model may changes simulations!
1221 The point is that events occurring at the exact same timestamp
1222 are not scheduled in the same order with the old and new
1223 version. This may be enough to completely change the execution
1224 of simulations in some cases.
1225 * The new network model will change simulations!
1226 This new model is more realistic than the previous one, so you
1227 should consider redoing your old experiments with this model.
1228 Sorry for the inconvenience.
1231 * Introduce the supernovae compilation mode
1232 When compiled that way, the whole SimGrid (or almost) is put in a
1233 single compilation unit and compiled in one shoot.
1234 This is to help gcc which has difficulties to inline stuff from one
1236 The speedup seem to be above 15%, althrough more tests are needed on
1237 amd64 to confirm that gain.
1240 * Port of MSG's mailbox on top of SIMIX network
1241 The put/get mechanism was greatly simplified on the way.
1244 * New SIMIX network module. Provides:
1245 - Mailbox: rendez-vous mecanism to find with who you want to speak
1246 - Synchronous send/recv: easier and hopefully faster since the
1247 logic is handled in the maestro process directly now
1248 - Asynchronous send/recv: you dreamt of it? It's here now
1249 Too bad that nobody cared enough to propagate the change to MSG.
1250 * Add semaphores as SIMIX synchronization mechanism.
1253 * new function SD_daxload(char*) to load a DAX file
1254 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1255 * Introduce typed tasks. Specify its kind and cost at creation.
1256 At scheduling, just give where it should be placed, and the cost
1257 for each involved resource is automatically computed.
1258 Existing constructors so far (more to come of course):
1259 - SD_task_create_comm_e2e() for end-to-end communication
1260 - SD_task_create_comp_seq() for sequential computation
1261 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1262 * new function SD_task_dump() for debuging display
1263 * new function SD_task_dotty(task,FILE*) writing to file the info
1264 about the task in dotty format
1265 * SD_task_dependency_exists() can now cope with having one of its
1266 arguments NULL. If so, it tests whether the other argument has any
1268 * Add getters on list of preceding/following tasks:
1269 SD_task_get_parents(task) and SD_task_get_children(task)
1270 * Add getters on amount of workstations and list:
1271 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1272 * Add getter on task kind: SD_task_get_kind(task)
1273 * Update the start_time and finish_time of tasks on completion/failure
1274 * Bugfix: Remove task from state swags when destroyed
1277 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1280 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1281 dynar with one separate thread per value of the dynar.
1282 * Change the prototype of xbt_thread_create(), sorry.
1283 Added a boolean parameter indicating whether we want to join this
1284 thread (used in SG only for now)
1285 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1288 * GTNetS wrappers should now be usable again (and betterly tested too)
1289 * Fix a major regression from 3.2 where the timeout provided to
1290 MSG_task_put_with_timeout() was used as absolute time before which
1291 the comm should be done.
1292 * Start to fix the <cluster> tag.
1293 - Internal links should be good now (beside of the loopback, which
1294 use the private link instead)
1295 - paths to the external world is still rather broken
1296 - the <route:multi> tag is just broken. Actually that's brain-dead.
1297 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1298 to make it less stupid
1299 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1300 * Fix a source-level compatibility glitch from 3.2: after defining
1301 MSG_USE_DEPRECATED, you can use the old name
1302 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1303 * Allow to compile from the SVN with automake 1.11
1304 * Fix some problems when using the "start_time" tag in deployment XMLs.
1305 * Fix #8569: XBT/synchro.h has redundant declarations
1306 * Fix #8563: MSG return values and exceptions
1307 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1308 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1309 Thanks to Robson Peixoto.
1310 * Fix a vicious bug in dictionaries inducing that some elements were
1311 not freed on xbt_dict_free()
1313 Portability report of this version:
1314 * Main portability targets:
1315 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1316 on (amd64/i386/ia64)
1317 - mac leopard on i386
1318 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1319 but nothing critical.
1320 * Other platforms: windows, AIX and others were not tested for this release
1322 Timing report of this version:
1323 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1324 scenario parameters). From 8h to a few seconds in desktop grid settings.
1325 * Supernovae brings about 25% speedup on i386.
1327 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1329 SimGrid (3.3.3) stable; urgency=low
1331 The "Need for Speed" release.
1333 The timings done to validate the 3.3.2 were faulty.
1334 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1336 The problem was a conversion from a manually handled vector to
1337 xbt_dynar_t on the critical path.
1338 xbt_dynar_foreach calls functions, inducing stack management crap.
1340 We inlined these functions and xbt_dynar_foreach is now breath taking.
1341 We also inlined xbt_swag_belong on the way.
1343 Here are some approximate speedup measurements (on master/slaves
1344 simulations lasting between 10s and 20s each):
1345 3.3.1 -> 3.3.2: about same performance
1346 3.3.2 -> 3.3.3: 40% speedup
1347 3.3.1 -> 3.3.3: 40% speedup
1348 3.3.1 with inline patch -> 3.3.3: 30% speedup
1350 Our reading is that the refactoring which occurred in 3.3.2 made us
1351 suffer much more from the xbt_dynar_foreach low performance, but
1352 once we solved this, this refactoring proved to be very performance
1353 effective. From the 40% speedup, somehow, 10% are due to the
1354 inlining and 30% to the refactoring.
1356 That's a pitty that gcc cannot inline functions placed in other files
1357 alone. We have to choose between:
1358 - break the encapsulation (by putting private data structures and
1359 accessors in headers files to help gcc)
1360 - live with low performance
1361 - switch to a decent compiler such as icc (not quite possible).
1363 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1365 SimGrid (3.3.2) stable; urgency=low
1367 The "Simplicity does not preceed complexity, but follows it" release.
1369 The main contributors of this release were (lexical order):
1370 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1373 * Extract the routing logic into its own object.
1374 (was dupplicated in network.c and workstation_LV07.c;
1375 Allows to implement other ways of storing that info)
1376 => kill now useless network_card concept
1377 - Use dynar to represent routes (instead of void** + int*)
1378 - kill link_set (use surf_network_model->resource_set instead)
1379 - Add a command-line option to choose the routing schema to use
1380 - Add three new models:
1381 * Floyd (shortest path computed at initialization)
1382 * Dijikstra (shortest path recomputed all the time)
1383 * Cached Dijikstra (shortest path computed on need)
1384 All these models where contributed by Silas De Munck, and are
1385 described in his ICCS09 paper.
1387 * Simplify model declaration
1388 (less redirections, less function to write when defining a model)
1389 - Factorize stuff between models:
1392 surf_model_resource_set(model)
1393 surf_model_resource_by_name(model, name)
1394 - Unify the types of models in s_surf_model_t (using an union)
1395 - Embeed fields of common_public directly into s_surf_model_t
1396 - Rename model methods:
1397 action_free ~> action_unref
1398 action_change_state ~> action_state_set
1399 action_get_state ~> action_state_get
1400 - Change model methods into functions :
1401 (model)->common_public->action_use ~> surf_action_ref
1403 * Implement a generic resource; use it as ancestor to specific ones
1404 (allows to kill duplicated code in models)
1405 Drawback: timer command don't need no name nor properties;
1406 workstation_CLM03 don't need no properties
1407 (but I guess we can live with those few bytes wasted)
1409 * Improve the action object model
1410 - implement a constructor avoiding dupplicated code about field
1411 initialization in generic_action part.
1413 * Kill the SDP model: it has an external dependency, is deprecated
1414 in flavor of modern lmm models, and didn't compile since a while
1417 * Relocation of the context module from XBT to SIMIX.
1418 (the context were decoupled from the simix processes, duplicating a lot of code)
1419 => a lot of code was factorized
1420 - less overhead is introduced during scheduling
1421 - simpler API for the context factory
1422 - the logic for process creation,destruction and manipulation was simplified
1423 * Simplification of the s_smx_process_t data structure.
1424 => accesing the simix level data associated to a process is faster now,
1425 and the code is a lot more readable.
1428 * Implement some more MPI primitives:
1429 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1430 -implementation: Bcast: flat or 2-ary tree (default),
1431 Barrier: 4-ary tree,
1433 Allreduce: Reduce then Bcast
1434 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1435 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1436 Alltoallv: flat tree, like ompi
1438 * Add support for optimized collectives (Bcast is now binomial by default)
1439 * Port smpirun and smpicc to OS X
1442 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1443 Shout out if you used it.
1446 * Display the list of still queued messages in SG mode when existing
1450 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1451 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1452 * Add xbt_str_from_file(FILE*)
1453 * Add xbt_dict_get_key achieving a linear reverse search
1454 * Remove the context module
1456 Portability report of this version:
1457 * Main portability targets:
1458 - Linux(debian)/x86/context
1459 - Linux(debian)/x86/pthreads
1460 - Linux(debian)/amd64/context
1461 - Linux(debian)/amd64/pthreads
1462 On these, we still have the eratic breakages of gras/pmm and
1463 amok/saturate_sg reported in previous version. We still think
1464 that the tests are the cause of the fault, not the tested code.
1466 - Mac OSX Leopard/x86/context
1467 Still false negative in tesh autotesting.
1468 Smpi still fails, but this time because readlink does not accept -f
1469 Everything seems to work properly beside of that.
1472 - AIX version 5.3 (only tested contexts this time)
1473 Smpi still fails there because mktemp is not installed.
1474 Everything seems to work properly beside of that.
1476 I managed to compile it for the first time, but several breakages.
1477 Won't delay the release for this exotic platform.
1479 * Windows: it's still lagging behind. If you want to help, please
1482 Timing report of this version:
1483 This version seem to be more than 5% faster than 3.3.1 (on linux
1484 64bits with contextes). The gain is less than expected, we are
1485 investigating this for next release.
1487 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1489 SimGrid (3.3.1) stable; urgency=low
1492 * Implement a --cfg-help to show existing configuration variables
1493 * Build chain do not require doxygen in maintainer mode
1496 * fix a bug on struct sizeof computation, which prevented the
1497 exchange of arrays of structs in some conditions
1498 - added a regression test about this in datadesc_usage
1499 * Allow the exchange of 0-long dynamic vectors.
1500 - for that, use -1 as indicator of dynamic size instead of 0
1501 - This implied to change any size from unsigned long to long,
1502 reducing a bit communication abilities, but I guess that with
1503 64bits being quite common, this is more than enough.
1504 - This also induce a protocol change, thus bumping network protocol
1505 version from 0 to 1 (if we have external users, we have to get
1506 clean on that point too ;)
1507 - added two regression tests about this in datadesc_usage
1508 * Be more verbose when propagating local exceptions
1509 This helps debugging.
1510 * Display the status of simulated processes when receiving SIGINT in
1514 * Allow to control the simulation from a trace file.
1515 New functions MSG_action_register() and MSG_action_trace_run()
1516 The first one allows to associate a function execution to each
1517 kind of action while the second one parses a trace file and
1518 triggers the corresponding actions within the system.
1519 For now, only a toy example is provided in examples/msg/actions
1520 * Add an exemple of process migration in examples/msg/migration
1521 * Fix a bug in task exchange which broke MSG_task_get_sender()
1522 Add a teshsuite regression test for that.
1523 [Bug: if MSG_task_get_sender() is called after sender exit,
1525 * Fix a bug which prevented suspend/resume to work properly
1526 * Display the status of simulated processes when receiving SIGINT
1527 This fixes a regression of v3.3. due to the introduction of SIMIX
1528 * Bug fixing in failure management:
1529 - trace could not start by a failure at time 0
1530 - failure during communications were not working
1533 * Add SIMIX_process_set_name() to change the name of the current
1534 process in the log messages.
1535 * Store smx_hosts in a dict since we only retrieve them by name
1536 * Move the configuration infrastructure to surf
1539 * Move the configuration infrastructure to surf
1542 * Massive internal cleanups:
1543 - Store internal structures on processes instead of hosts (allows
1544 to have more than one process per host, in addition of being more
1546 - Cleanup the initialization/finalization process
1547 - Kill a whole bunch of unneeded synchronization:
1548 processes run in exclusive manner within the simulator
1549 - Move queues from global tables to process data fields
1551 - now accept -platform and -hostfile arguments
1552 - Pass the right rank value to processes according to the hostfile
1553 * Compile the examples by default, and use them as regression tests
1554 * Implement MPI_Wtime()
1555 * Change the reference speed to a command line option
1558 * TCP_gamma can now be specified as command line option using
1559 --cfg=TCP_gamma:10000000.0
1560 * Change the --surf-path cmd line option into --cfg=path:
1563 * Also include strbuff from xbt.h public header
1564 * xbt_ex_display(): do not free the exception after displaying
1565 This allows to do more with the given exception afterward.
1566 Users should call xbt_ex_free() themselves.
1570 Portability report of this version:
1571 * Main portability targets:
1572 - Linux(debian)/x86/context
1573 - Linux(debian)/x86/pthreads
1574 - Linux(debian)/amd64/context
1575 - Linux(debian)/amd64/pthreads
1576 These targets fail about 1/10 of times on gras/pmm, but we believe
1577 that this is because of the test, not because of simgrid.
1578 amok/saturate_sg fails even more rarely, and the test may not be
1581 - Mac OSX Leopard/x86/context
1582 The test suite still spits tons of errors because some obscure
1583 force prevents us from removing the temporary directories
1584 arguing that they still contain some metadata I've never heard of.
1585 Smpi fails because seq is not installed.
1586 Everything seems to work properly beside of that.
1589 - AIX version 5.3 (both contexts and pthread)
1590 Smpi still fails there because mktemp is not installed.
1591 XML inclusions seems rosty on AIX.
1593 * Windows: it's still lagging behind. If you want to help, please
1596 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1598 SimGrid (3.3) stable; urgency=high
1602 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1603 [Malek Cherier & Mt]
1605 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1608 Until now, GRAS were using MSG as an interface to SURF. It was
1609 quite difficult because both interface have several differences
1610 (MSG channels vs GRAS sockets were the most notable point).
1612 This also opens the gate to SMPI (which should occur soon) and speed
1613 up simulations by to 40% (even if it were not the main goal).
1615 **************************************
1616 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1617 **************************************
1618 The point is that events occuring at the exact same timestamp are
1619 not scheduled in the same order with the old and new version. This
1620 may be enough to completely change the execution of simulations in
1621 some cases. Sorry for the inconvenience.
1623 * Cleanup and upgrade the XML format to push further scalability
1624 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1626 * Improve the testing infrastructure with tesh. Now a very large part of
1627 the code is tested not only by being run but also by checking that the
1628 output match an expected output [Mt].
1630 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1631 is really less memory-demanding, which should allow you to use
1632 larger files in SimGrid [AL].
1634 * Inform valgrind about our contextes, so that it becomes usable
1635 with the default (and more effecient) version of SimGrid
1636 [contributed by Sékou Diakite, many thanks]
1639 * Introduce a listener thread in charge of receiving incoming
1640 messages from the network. It allows to overlap communication and
1641 computation but most notably, it removes some stupid deadlocks due
1642 to the fact that so far, a process could not send and receive at
1643 the same time. This made most non trivial communication schema
1645 * Convert the PIDs from long int to int to match the MSG ones (and
1646 linux ones too) [Mt]
1647 * New function: gras_agent_spawn() to launch a new process on
1648 current host. Only working in simulation for now. [Mt]
1649 * New function: gras_os_hostport() returning a constant form (ie,
1650 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1653 * Make the backtrace of exceptions more human readable [Mt]
1654 * New module: xbt/str [Mt]
1655 a ton of string utility functions (split, join, printf to a newly
1656 allocated buffer, trim, etc)
1657 * New module: xbt/hash [Mt]
1658 SHA1 hashing algorithm (more to come if needed)
1659 * New module: xbt/synchro [Mt]
1660 synchronization tools (mutex and conditions) working the same way
1661 in simulation and in real life (mainly useful for GRAS, but not
1663 * New module: xbt/queue [Mt]
1664 classical producer/consumer synchronization scheme
1665 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1666 (using the classical functions will get serialized) [Mt]
1667 * Make dictionary internal table dynamic. No need to specify its size
1668 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1670 * Make sure the log channels are organized as a tree under windows
1671 (because of ANSI C compatibility issue, any channel were child of
1672 root directly) [Mt].
1675 * Cleaned many thing in surf and fixed a few bugs [AL].
1676 * Add a nice command line configuration mechanism to compose models [AL].
1677 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1678 the previous one (KCCFLN05). It relies on something that looks like
1679 a max-min sharing mechanism but cannot be written as such. A new solver
1680 was thus designed [AL].
1681 * Add a new solver to lmm. Based on Lagrange optimization and
1682 gradient-based descent, it enables to efficiently maximise systems s.a
1684 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1686 This solver enables to propose two new network models for TCP Reno and
1687 TCP Vegas based on Low's work. These models still need to be fully
1688 tested though [Pedro Velho].
1691 * Bug fix in SD_simulate. Now the time bound given as argument is
1693 * Use the new parallel task model (ptask_L07) as default.
1694 * Use the SURF command line configuration mechanism.
1695 * 0-size tasks (for synchronization) should now work.
1697 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1699 SimGrid (3.2) stable; urgency=high
1703 We still experience issues on this platform, but we believe that at
1704 least MSG is usable.
1706 GRAS API BREAKAGE (for simplification purpose, sorry):
1707 * the gras_msgtype_by_name is not used anymore. Instead of
1708 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1709 you can write (and must)
1710 gras_msg_send(toserver, "request", &request);
1711 - If you still want to pass a gras_msgtype_t to the function (to cache
1712 the type and avoid the lookup time), use the gras_msg_send_() variant.
1713 - Impacted functions:
1714 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1715 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1716 * The callbacks are now expected to return 0 when everything went well
1717 (just like the main() function)
1719 GRAS new features and improvements:
1720 * New module mecanism where user code can use per process globals [Mt]
1721 This is similar to gras_userdata_*() functions, but for libraries. It
1722 factorize some code developped over and over in the examples and AMOK.
1723 It has still to be documented and used (only amok/peermanagement is
1725 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1726 loss when they were small enough to fit into the buffer and sent quickly
1727 enough so that they can all get received in one shoot.
1728 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1729 if not found. Use the *_or_null() variant for the old semantic.
1730 * In gras_msg_handle, do not discard messages without callback.
1731 They are probably messages to be explicitly awaited later (ie, proofs of
1732 mis-synchronization in userland since they are sent before being awaited)
1733 No big deal usually.
1734 * gras_socket_meas_send/recv: semantic changed!
1735 The numerical arguments used to be (1) the total amount of data to send
1736 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1737 messages. This was need for the fool willing to send more than MAXINT
1738 bytes on quite fat pipes.
1741 * Do really rename the hostmanagement module to peermanagement. [Mt]
1742 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1743 but this is rather new and this was documented in the module
1744 documentation (poor excuses, I admit)
1745 * Bandwidth measurement semantic changed! This follows the changes to
1746 gras_socket_meas_send/recv explained above.
1749 * A sequential mode has been added to the workstations. When a workstation
1750 is in sequential mode, it can execute only one task, and the other tasks
1751 are waiting in a FIFO. [Christophe Thiery]
1754 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1755 model for SIMDAG. [Christophe Thiery]
1756 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1757 equal to 0, which caused some bad behaviors in
1758 saturated_constraint_set_update. I now use a threshold mechanism like in
1762 * When running manually src/testall, you select specific units [Mt]
1763 testall is the result of our cunit mecanism, and should replace all
1764 the scripty thingy around since bash don't run easily on billware.
1766 * A mallocator system has been added. [Christophe Thiery]
1767 Mallocators allow you to recycle your unused objects instead of freeing them
1768 and allocating new ones.
1770 Documentation update:
1771 * FAQ reworking + New FAQs:
1772 - "Valgrind spits tons of errors!" [Mt]
1773 - "How to repport bugs" [Mt]
1774 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1775 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1776 - Communication time measurement within MSG [AL]
1777 - I experience weird communication times when I change the latency [AL]
1778 * GRAS tutorial [Mt]
1780 - an introduction to the framework and to the used communication model
1781 - an initiatic tour introducing the most proheminent features:
1783 . Lesson 0: Installing GRAS
1784 . Lesson 1: Setting up your own project
1785 o Part 2: Message passing
1786 . Lesson 2: Exchanging simple messages
1787 . Lesson 3: Passing arguments to the processes (in SG)
1788 . Lesson 4: Attaching callbacks to messages
1789 . Lesson 5: Using globals in processes
1790 . Lesson 6: Logging informations properly
1791 . Lesson 7: Using internal timers
1792 . Lesson 8: Handling errors through exceptions
1793 . Lesson 9: Exchanging simple data
1794 . Lesson 10: Remote Procedure Calling (RPC)
1795 . Lesson 11: Explicitely waiting for messages
1796 . Recapping of message passing features in GRAS
1797 - A HOWTO section containing:
1798 o HOWTO design a GRAS application
1799 More are due, of course. They will come latter. In the meanwhile, you can
1800 check the examples which are still here.
1802 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1804 SimGrid (3.1) stable; urgency=high
1808 There was a stack corruption somewhere, visible only when optimizing
1809 with these versions. [Vince]
1812 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1813 module that enabled to play with Directed Acyclic Graphs. It is built
1814 directly on top of SURF and provides an API rather close to the old
1815 SG. Some old codes using SG are currently under rewrite to check that
1816 all needful functions are provided. [Christophe Thiery]
1819 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1820 extension of the classical CLM03 model that gracefully handles
1821 failures. This is now the default model for MSG and GRAS. It doesn't
1822 handle parallel tasks yet though. [AL]
1823 * Bug fix: Weights were not correctly set in the network part.
1824 WARNING: This may have resulted in incorrect results with simulations
1825 where there are more than one flow on a given link. [AL]
1828 * After a (long ?) discussion on simgrid-devel, we have decided that the
1829 convention we had on units was stupid. That is why it has been decided
1830 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1831 WARNING : This means that all previous platform files will not work as
1832 such with this version! A warning is issued to ask users to update
1834 A conversion script can be found in the contrib module of the CVS, under
1835 the name contrib/platform_generation/surfxml_update.pl [MQ]
1838 * Bug fix: Processes were started in reverse order, wrt deployment file.
1839 WARNING: if your code relies on this bug, please fix it. [AL]
1840 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1841 being executed on two different locations. [AL]
1842 * Bug fix: Failures are now better supported thanks to Derrick's tests
1843 (there was many failure situations I hadn't thought of and that weren't
1844 correctly handled). [AL]
1845 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1849 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1850 Exception occurring on server-side are propagated back to client (!).
1852 API CHANGE: the callback changed their prototype. Change:
1853 int my_handler(gras_socket_t expeditor, void *payload_data) {
1855 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1856 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1858 * New! function: gras_msg_handleall to deal with all messages arriving
1859 within a given period.
1860 * New! function: gras_socket_server_range to get a server socket in a
1861 range of port numbers (ease to avoid port number conflicts) [MQ]
1862 * New! gras processes display their backtrace when they get a SIGUSR1
1863 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1864 Sweet to debug RL processes [MQ]
1868 - Do not force experiment sizes to be expressed in kb, or it becomes
1869 impossible to measure the latency this way (needs one byte-long tests)
1870 WARNING: this changes the amok_bw_* function semantic. [MQ]
1871 - Implements the link saturation stuff. [MQ]
1872 * Peer management module:
1873 New! module factorizing code that we wrote over and over [MQ].
1876 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1877 - Test units are placed directly into the library code, they get extracted
1878 automatically and placed into the src/testall binary.
1879 - Convert most of the XBT tests to this system.
1880 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1881 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1882 live in the same C code. This required to change a little bit the API
1883 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1884 * New module: sparse graph structure with basic algorithms (this is work
1885 in progress and the API is not considered to be frozen yet). [AL]
1886 * Display more information on backtraces: source line & function names are
1887 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1888 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1889 * Reworked a little bit some #include statements to load only required
1890 headers. Some user code that relied on SimGrid to include stdlib or
1891 stdio may need to include it by themselves. [AL]
1892 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1894 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1895 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1896 but I guess that almost nobody used those parts. [MQ]
1898 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1900 SimGrid (3.0.1) stable; urgency=low
1903 * Unfortunately, I had missed 5 misnamed functions:
1904 xbt_fifo_item_t xbt_fifo_newitem(void);
1905 void xbt_fifo_freeitem(xbt_fifo_item_t);
1906 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1907 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1908 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1909 They're now deprecated. Please use their new versions:
1910 xbt_fifo_item_t xbt_fifo_new_item(void);
1911 void xbt_fifo_free_item(xbt_fifo_item_t);
1912 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1913 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1914 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1916 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1917 * Documentation: xbt_log module unmercifully reworked [MQ]
1918 * Bugfix: there was a problem with the ending of contexts with
1919 the pthread backend. It caused some weird deadlock or behavior
1920 depending on the pthread implementation. [AL]
1921 * Bugfix: get the exceptions raised in the simulator repport where
1922 and why they come from when they are not catched in time [AL, MQ]
1925 * Bugfix: Do repport the error when two non-connected hosts try to
1926 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1929 * Add additionnal checkings on communications. Assert that two
1930 communicating hosts are connected by a set of links... [AL]
1933 * Add additionnal checkings on channel values in communication [AL]
1934 * New: MSG_task_get_source to see on which host a task was generated [HC]
1935 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1936 the number of tasks waiting to be received on channel and sent
1938 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1939 waits for the first task coming from a given host.. [AL]
1941 GRAS new functionnalities: [MQ]
1942 * Enhance the parsing macro to allow the size of multidimentional objects
1943 to be given thru annotations.
1944 * New example (and documentation): Matrix Multiplication a la RPC
1945 (as when I was young!) and fix a bunch of bugs found on the way.
1947 GRAS performance improvements: [MQ]
1949 * Reduce the amount of cbps creation/destruction by making it static to
1950 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1953 * Change libdata to a set so that we can search for stuff by ID (and thus
1954 reduce the insane amount of dict lookups)
1957 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1958 RL communication on top of SG since it's so uneffective.
1959 It may also be used for inter-thread communication in RL, one day.
1960 * Use gras_datadesc_copy() to exchange messages on top of SG
1962 - improve message exchange performance on top of SG
1963 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1964 * Don't exchange on the network the size of the used part of buffer,
1965 instead, specify the possible buffer size to read().
1967 - reduces the amount of read/write calls (one pair per exchange)
1968 - reduces the amount of exchanged data (the size)
1969 - allows to retrieve all arrived data on receiver side, if we don't need
1970 it right now (subsequent read will peek the buffer)
1971 - allows the receiver to proceed with the begining of the stream before
1972 everything is arrived
1973 - make it possible to build an iov transport (using readv/writev)
1975 - take care of the data with non-stable storage (like stacked data),
1977 * If possible, TCP send uses vector I/O (when writev() is here)
1978 - Don't use it for receive since we send data sizes and data on the
1979 same stream, so we wouldn't be able to chain large amount of chunks
1980 before having to flush the stuff to read the size.
1981 * Rework the transport plugin mecanism to simplify it and reduce the
1982 amount of pointer dereferencement when searching for the right function
1985 * I guess that now, we do almost as few system calls as possible while
1986 doing as few data copy as possible.
1988 To improve it further, we could try to send all the sizes first and then
1989 all the data (to use iov on receiving size), but it's only a partial
1990 solution: when you have 2 dimensional data, the sizes of the second
1991 dimension is data of the first dimension, so you need 3 streams.
1993 I'm not sure the potential performance gains justify the coding burden.
1995 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1997 SimGrid (3.00) stable; urgency=high
2000 * New! Give the possibility to hijack the surf parser and thus bypass
2001 MSG_create_environment and MSG_launch_application. Have a look at
2002 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2004 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2006 SimGrid (2.96) unstable; urgency=low
2011 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2012 This deprecates the xbt_error_t mecanisms.
2013 It modifies (simplifies) all XBT and GRAS API.
2014 MSG API keeps unchanged (exceptions raised by XBT are catched from
2015 within MSG and masked with existing error handling facilities)
2018 * New! Add a FATPIPE model. [AL]
2019 * New! Add a parallel task model. [AL]
2020 * New! Add automatically a loopback interface (in the default
2021 network model) if none was precised.
2024 * Bugfix: MSG_process_resume now works with the current running process.
2026 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2027 * Modification of MSG_task_get_compute_duration. Once a task has been
2028 processed, the value returned by this function is now equal to 0. [AL]
2029 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2030 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2031 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2032 of a m_task. For now, MSG_task_cancel only works with computation
2034 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2035 of the processor (in Mflop/s) regardless of the current load on the
2037 * API Change: use proper naming convention for MSG_getClock and
2038 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2040 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2041 This function changes the priority of a computation task. This priority
2042 doesn't affect the transfer rate. A priority of 2 will make a task
2043 receive two times more cpu power than the other ones. This function
2044 has been added to suit the needs of Nguyen The Loc and hasn't been that
2045 much tested yet. So if it fails, please report it and send me your code.
2047 * API Change: removed all functions and types that were marked "deprecated"
2048 since many months. Renamed MSG_global_init_args to MSG_global_init.
2050 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2052 SimGrid (2.95) unstable; urgency=low
2055 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2056 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2057 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2060 * Bugfix: Allow absolute paths for platform description files [MQ]
2061 * Bugfix: do free the variables after use. Leads to drastic performance
2063 * Implement max_duration (ie, timeouts) on resources [AL]
2066 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2068 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2069 the needed support (timeouts) [AL]
2072 * Implement measurement sockets. You can now get the bandwidth between two
2073 hosts thanks to AMOK (see below). [MQ]
2074 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2075 dynar over the network (yeah) [MQ]
2076 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2077 * simple/static token-ring example. [Alexandre Colucci and MQ]
2078 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2079 in place before [MQ]
2082 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2083 type parsing. "short" and "long" modifiers were also ignored; other
2084 modifier (such as reference level) are still ignored. [MQ]
2085 * Embeed the buffer size within the buffer itself on SG. [MQ]
2086 That way, send() are atomic and cannot get intermixed anymore (at least
2087 the ones which are less than 100k; bigger messages still have the issue)
2088 * Array size pushed by the field, not by the field type (or each
2089 and every long int will push stuff to the cbps) [MQ]
2090 * use select() to sleep since it allows to portably sleep less than one
2093 GRAS (minor cleanups)
2094 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2097 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2098 * type category 'ignored' killed as it was never used and were difficult
2100 * whether a type can cycle or not is now a flag, leaving room for more
2101 flags (as "ignored", if we feel the need one day ;)
2102 * Rename raw sockets to measurement sockets since "raw" has another
2103 meaning in networking community.
2106 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2107 and offers features not belonging to GRAS but that most applications
2108 need. One day, it may be a set of plugins loadable at runtime.
2109 * New module: bandwidth
2110 bandwidth measurement between arbitrary nodes running this module. [MQ]
2112 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2114 SimGrid (2.94) unstable; urgency=low
2116 The first beta release of SimGrid 3 !
2120 * Update the main page and the FAQ. Adding references to gforge.
2123 * Add a gras_os_getpid function.
2126 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2127 * Extend the logs so that they also print PID, hostname, date, ... if
2129 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2130 and kill the old version which were in testsuite/
2131 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2135 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2136 for users. Instead, each programming environment (GRAS, MSG,...) use it
2137 when needed to provide such a feature to users.
2138 Don't shortcut the mecanism or you will also shortcut the virtualization
2139 you need on the simulator.
2143 * Cleanups in configury with regard to compile optimization/warning flags.
2144 Also add -fno-loop-optimize to any powerpc since it's the optimization
2145 killing gcc (< 3.4.0).
2146 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2147 needed by MSG examples complications
2148 * Borrow configury beautifications from PHP
2151 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2152 warning (thanks loris for stumbling into this one).
2153 * Bugfix: stop loading private headers (gras_config.h) from the public
2157 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2158 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2159 RL would imply a DNS resolver, which is *hard* to do in a portable way
2160 (and therefore delayed).
2161 * Implement a real timer mecanism and use it in timing macros. This allows
2162 to avoid rounding errors and get a 0.000005 sec precision in timing
2163 macros. While I was at it, various cleanups:
2164 - allow to declare more than one timed section per file (fix a stupid bug)
2165 - move some private declaration to the right place
2166 - merge conditional execution and timing macros into emulation module
2167 - document the module
2168 - make sure the module cleanups its mess on gras_exit
2169 * Documentation improvements:
2170 - (new) how to compile applications using GRAS
2171 - (new) emulation support (timing macros)
2173 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2175 SimGrid (2.93) unstable; urgency=low
2177 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2180 - Use Paje properly where used. Still to be sanitized properly.
2181 - Portability fix: Add an implementation of the contexts using pthread
2185 - Add xbt_procname(): returns the name of the current process.
2186 Use it to show the current process's name in all logging.
2188 - fix detection of older flex version and the reaction, since we do
2189 depend on modern ones (we use lex_destroy)
2190 - Better separation of SG and RL in the libs: remove all simulation code
2191 from libgras. As a result, this lib is now only 200k when stripped.
2192 Some of the xbt modules may also be duplicated (two sets and such) and
2193 should be cleaned/killed before SG3.
2194 - Insist on using xlC on AIX because of weird problems involving gcc there.
2195 - Cleanup the make remote stuff. This is now done by scripts
2196 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2197 mainly for our private use, but we're working on changing them to user
2200 - Bugfix: flush the socket on close only if there is some *output*.
2201 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2203 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2206 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2208 SimGrid (2.92) unstable; urgency=low
2210 Alpha 3 on the path to SimGrid 3
2214 - New! First try of benchmarking macros.
2215 - New! First try so that gras_stub_generator generate deployment and
2216 remote compilation helpers.
2218 - Bugfix: Initialization fix in msg_test.
2222 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2224 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2226 - API Change: gras_os_sleep to take the amount of seconds as a double.
2227 Accepting an int was error prone since it was the only location where
2228 seconds were coded as such. It leaded to damn rounding errors.
2229 - Bugfix: Hard to belive that timers ever worked before this.
2231 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2233 SimGrid (2.91) unstable; urgency=low
2235 Alpha 2 on the path to SimGrid 3
2239 - Bug fix in the lmm_solver.
2241 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2242 through the function MSG_paje_output.
2243 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2244 - It is possible to bound the rate of a communication in MSG with
2245 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2246 to write it in the changelog).
2247 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2252 - Various cleanups to the autotools stuff
2253 - Begin to move Gras examples to examples/gras/
2254 - Let make distcheck work again (yeah!)
2256 - documentation overhauled using doxygen.
2257 gtk-doc-tools is dead in SimGrid now.
2258 - Automatically extract all existing logging categories, and add the list
2259 to the documentation (long standing one, to say the less)
2261 - Cleanup the known architecture table. Reorder the entries to group what
2262 should be, and use a more consistent naming scheme.
2263 (some of the test dataset are still to be regenerated)
2264 - New! Allow library to register globals on each process just as userdata
2266 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2267 lookup time (for now).
2268 Use it in msg and trp.
2269 This cleans a lot the internals and helps enforcing privacy of the
2270 headers between the gras components.
2271 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2272 - Bugfix: gras_os_time was delirious in RL.
2273 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2275 - Reenable GRAS now that it works.
2277 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2279 SimGrid (2.90) unstable; urgency=low
2281 Alpha 1 on the path to SimGrid 3
2283 * It is a long time since the last release of SimGrid. I'm sorry about
2284 that but as I had told you, I was rewriting a lot of things. I apologize
2285 to those who had been reporting bugs to me and that I had not answered.
2286 If your bug is still in the new version, please tell me. Here is a
2287 summary of the main changes.
2289 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2290 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2291 portability, speed, and a lot more but you'll figure it out later.
2292 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2293 I wanted to release the new MSG as soon as possible and I have
2294 broken GRAS, which is the reason why, for now, only MSG is fully
2295 functional. A laconic description of these projects is available
2296 in the documentation.
2298 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2299 kernel optimized for our needs (called SURF but only the developers
2300 should use it). Hence, MSG is now roughly 30 times faster and I think
2301 that by rewriting a little bit MSG, I could event speed it up a little
2302 bit more. Beside the gain in speed, it is also much easier to encode a
2303 new platform model with SURF than it was with SG. More to come...
2305 * REVOLUTION 3: I have tried to change a little as possible the API of
2306 MSG but a few things really had to disappear. The main differences
2307 with the previous version are :
2308 1) no more m_links_t and the corresponding functions. Platforms are
2309 directly read from a XML description and cannot be hard-coded
2310 anymore. The same format is used for application deployment
2311 description. The new format is described in the documentation.
2312 Have a look in tools/platform_generation. There is a tiny script
2313 that converts from the old platform format to the new one. Concerning
2314 the application deployment format, parsing the old one is tricky.
2315 I think most of you should however be able to convert your files. If
2316 it is really an issue, I can write a C code that does the conversion.
2318 2) the toolbox tbx does not exist anymore. We now have a library
2319 with much more data-structures but without the hash-tables (we have
2320 dictionaries that are much faster).
2322 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2324 *****************************************************************************
2325 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2326 * don't feel like converting the oldies. (Mt) *
2327 *****************************************************************************
2330 Version 2.90: "the long awaited one"
2331 - Finished rewriting and debugging MSG. Rewrote the documentation.
2332 - disable GRAS for now since it needs to be ported to the newest SG
2335 - Finish the port to windows (using mingw32 for cross-compile)
2338 - Main loop and datastructures of SURF. A cpu resource object is
2339 functional. Surf can thus be used to create cpu's with variable
2340 performance on which you can execute some actions.
2342 2004-11-15 Martin Quinson
2343 - Port to ARM. Simply added the alignment and size descriptions. Should
2344 work, but the ARM machines are so slow that I didn't had the opportunity
2345 to 'make check' over there yet.
2347 2004-11-15 Arnaud Legrand
2348 - Trace manager now written. It uses a heap structure and is therefore
2349 expected to be efficient. It may however be speeded up (particularly
2350 when many events occur at the same date) by using red and black
2351 trees. One day maybe...
2352 - Max-min linear system solver written. It uses a sparse matrix
2353 structure taking advantage of its expected use. Most operations are
2354 O(1) and free/calloc are called as few as possible. The computation of
2355 the minimum could however be improved by using a red and black tree
2358 2004-11-03 Arnaud Legrand
2359 - Rename every gras_* function that was in xbt/ to its xbt_
2361 - Add a heap and a doubly-linked list to xbt
2362 - Added a dichotomy to the dictionaries. make check works as well before
2363 so I assume that the patch is correct. I do not know however if things
2364 run effectively faster than before now. :)
2366 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2367 SimGrid, and the version number is bumped to 2.x
2369 2004-10-29 Martin Quinson
2370 - Introduction of the remote errors.
2371 They are the result of a RMI/RPC on the remote machine.
2372 ErrCodes being scalar values, you can't get the host on which those
2373 errors did happen. Extending the error mechanism as in Gnome is possible.
2374 No idea yet whether it is a good idea.
2376 2004-10-28 Martin Quinson
2377 - Interface revolution: the Starred Structure Eradication.
2378 I used to do typedef struct {} toto_t; and then handle *toto_t.
2379 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2380 - ???_t is a valid type (builded with typedef)
2381 - s_toto_t is a structure (access to fields with .)
2382 - s_toto is a structure needing 'struct' keyword to be used
2383 - e_toto_t is an enum
2384 - toto_t is an 'object' (struct*)
2386 typedef struct s_toto {} s_toto_t, *toto_t;
2387 typedef enum {} e_toto_t;
2388 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2389 s_toto_t) is private.
2391 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2392 so that it can set the variable to NULL. It was so for dicts and sets,
2393 it changed for dynars.
2395 - Fix a bunch of memleaks in dict_remove
2396 - Fix a bug in sg/server_socket opening: it failed all the time.
2398 2004-10-07 Martin Quinson
2399 - Speed up dynar lookup operation a bit.
2401 gras_dynar_get is dead.
2403 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2404 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2405 (which gives you the address of the stored data).
2407 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2408 copy of the data using an affectation to do the job and not a memcpy.
2410 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2411 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2413 It does not really speedup the dynar test because they are
2414 setting elements all the time (and look them seldom). But the dict does
2415 far more lookup than setting.
2417 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2419 2004-10-05 Martin Quinson
2420 - Allow to (en/dis)able the cycle detection at run time.
2422 Whether we should check for cycle or not is now a property of each
2423 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2424 datadesc_cycle_unset allow to remove this property when previously set.
2426 Note that the cycle detection is off by default since it impacts the
2427 performance. Watch the data you feed GRAS with ;)
2429 This property is hereditary. Any element embedded in a structure having it
2430 set have it set for the time of this data exchange.
2432 You should set it both on sender and receiver side. If you don't set it on
2433 sender side, it will enter an endless loop. If you forget on receiver
2434 side, the cycles won't be recreated after communication.
2436 - Header reorganization.
2437 Kill gras_private.h, each submodule must load the headers it needs.
2439 2004-10-04 Martin Quinson
2440 - Interface revolution: do not try to survive to malloc failure.
2442 Now, gras_malloc and friends call gras_abort() on failure.
2443 As a conclusion, malloc_error is not a valid error anymore, and all
2444 functions for which it was the only gras_error_t return value are
2445 changed. They now return void, or there result directly.
2446 This simplify the API a lot.
2448 2004-09-29 Martin Quinson
2449 - Re-enable raw sockets.
2450 Created by gras_socket_{client,server}_ext;
2451 Used with gras_raw_{send,recv}
2454 It should allow to kill the last bits of gras first version soon.
2456 This is not completely satisfactory yet (duplicate code with
2457 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2460 - Simplify transport plugin (internal) interface by not passing any
2461 argument to _server and _client, but embedding them in the socket
2464 2004-09-28 Martin Quinson
2465 - Finish the port to AIX.
2466 autoconf was my problem (segfault within the malloc replacement
2467 function. No idea why)
2469 2004-09-16 Martin Quinson
2470 - Fix some size_t madness on 64bit architectures.
2472 2004-09-08 Martin Quinson
2473 - Reduce the number of system headers loaded, overload some more system
2474 calls (such as malloc to cast the result of the system one, and work
2476 - Fix and reintroduce the config support
2478 2004-09-07 Martin Quinson
2479 - Source code reorganization to allow Arnaud to surf all over there.
2480 - Allow to document the logging categories.
2481 - Remove all uppercase from logging categories and useless cleanup in names.
2483 2004-08-18 Martin Quinson
2484 Version 0.6.2 (protocol not changed; API changed)
2485 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2486 gras_error_t), and NULL when not found. Functions expecting a msgtype
2487 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2488 hopefully usefull message.
2489 - Portability to prehistoric sparcs again
2491 2004-08-17 Martin Quinson
2492 Version 0.6.1 (protocol not changed; ABI not changed)
2493 - prealloc some buffers to speed things up
2495 2004-08-11 Martin Quinson
2496 Version 0.6 (protocol not changed; ABI expended)
2497 - The parsing macro can deal with the references, provided that you add
2498 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2500 2004-08-09 Martin Quinson
2501 Version 0.5 (protocol not changed; ABI changed)
2502 - Allow to off turn the cycle detection code in data exchange at
2503 compilation time. It should be at run time, but I'm short of time (and
2504 the config stuff is still broken). That way, we keep dict out of the
2505 critical path, which is good because the performance is poor:
2506 - search not dichotomial yet
2507 - dynar give no way to access their content and memcpy everytime
2508 - In composed data description (struct, ref and so on), stop foolness of
2509 keeping the subtype's ID, but store the type itself. This keeps sets out
2510 of the critical path, which is good since they rely on dynar and
2511 dictionnaries. The only loose of that is that we cannot detect the
2512 redeclaration of a structure/union with another content (but I'm not sure
2513 the code detected well this error before anyway). We still can detect
2514 the redefinition discrepancy for the other types.
2515 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2516 it breaks the code because of type-punning used all over the place).
2517 This breaks on all non-gcc architectures (for now).
2519 All those changes (plus the buffer of last time) allow me to gain 2 order
2520 of magnitude on cruel tests consisting of 800000 array of integers on two
2521 level of a hierarchical structure (200 secondes -> 4 secondes)
2524 - the selector of reference must now return the type it points to, not
2525 the ID of this type.
2527 2004-08-06 Martin Quinson
2528 Version 0.4 (protocol changed; ABI not changed)
2529 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2530 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2531 - Implement a buffer transport, and use it by default (it relies on tcp in
2532 real life and on sg in simulation).
2533 That's a bit hackish since I had a new field to the structure to store
2534 its data without interfering with the subtype ones. Inheritance
2535 is tricky in C. And that's a kind of reverse inheritance with one class
2536 derivating two classes. Or maybe a game with java interfaces. Anyway,
2537 that's damn hard in C (at least).
2538 Moreover, I got tired while trying to ensure plugin separation and
2539 genericity in SG mode. MSG wants me to do weird things, so let's go for
2540 cruel hacks (temporarily of course ;).
2541 See comment in transport_private.h:71
2542 - do not include all the _interface headers in private but in the files
2543 which really need them (to cut the compilation time when they are
2546 2004-07-26 Martin Quinson
2547 Version 0.3 (protocol not changed; ABI changed)
2548 - Major overhault of the datadesc interface to simplify it:
2549 - shorted the function names:
2550 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2551 - add a trivial way to push/pop integers into the cbps without malloc.
2552 This allows to make really generic sub_type description, which simply
2553 pop their size of the stack.
2554 - add a function gras_datadesc_ref_pop_arr() which does what users want
2555 most of the time: Declare a dynamic array (which pops its size of the
2556 stack) and declare a reference to it. Poor name, but anyway.
2557 - kill the post-send callback, add a post-receive one
2559 2004-07-23 Martin Quinson
2560 Version 0.2 (protocol changed; ABI changed)
2561 - add some testing for cpbs in the test cases, and fix some more bugs.
2562 This invalidate again the little64 data file, since I cannot regenerate
2564 - remove an awfull optimization in the logging stuff, allowing me to:
2565 - understand it again
2566 - learn gcc how to check that the argument match the provided format
2567 - fix all errors revealed by gcc after that
2568 - internal keys of dict are not \0 terminated. Deal with it properly in
2569 loggings instead of segfaulting when the user want to see the logs :-/
2571 2004-07-22 Martin Quinson
2572 - Fix some stupid bug preventing cbps (callback postit) from working
2574 2004-07-21 Martin Quinson
2575 - Some documentation cleanups
2576 - remove the useless last argument of msgtype_declare
2577 - rename the Virtu functions to fit into the 'os' namespace
2578 - move headers src/include -> src/include/gras/ and stop fooling with
2579 gras -> . symbolic link
2580 - make distcheck is now successful
2582 2004-07-19 Martin Quinson
2584 - Build shared library also
2585 - Install html doc to the right location
2586 - stop removing maintainer files in make clean
2587 - build tests only on make check
2589 2004-07-13 Martin Quinson
2591 - No major issue in previous version => change versionning schema
2592 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2593 corresponding dataset.
2595 2004-07-11 Martin Quinson
2597 - Get it working with any kind of structure (we can compute the padding
2598 bytes remotely for all the architectures I have access to)
2599 - Implement the structure parsing macro (still not quite robust/complete)
2600 - Improvement to the remote testing toysuite
2602 2004-07-10 Martin Quinson
2603 [autoconf mechanism]
2604 - get ride of a bunch of deprecated macros
2605 - actually run the test for two-compliment, not only compile it :-/
2606 - test whether the structures get packed (and bail out if yes. Damn.
2607 Alignment is a serious matter)
2608 - test whether the structures get compacted (but respecting the alignment
2609 constraints of each types)
2610 - test whether the array fields of structures can straddle alignment boundaries
2612 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2613 (took me 2 hours to find that bug, looking at the wrong place)
2614 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2615 before sending/receiving and are used to compute the offsets of fields
2616 - Given that padding size depend even on compiler options, keep track of
2617 alignment and aligned_size only for the current architecture. Not a big
2618 deal since we send structure fields one after the other (seems
2620 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2621 program, let it work on linux/gcc/little32. portability todo.
2623 2004-07-08 Martin Quinson
2624 - import and improve remote compilation support from FAST
2625 - make sure make check works on half a dozen of machines out there
2627 2004-07-07 Martin Quinson
2628 Let's say it's version 0.0.3 ;)
2629 - Implement conversions (yuhu!)
2630 - Let it work on solaris (beside conversion, of course)
2631 - Stupid me, using rand() to generate the conversion datatests in not wise.
2633 2004-07-06 Martin Quinson
2634 - Let make dist work, since I'm gonna need it to compile on remote hosts
2635 - Let Tests/datadesc_usage write the architecture on which the file was
2636 generated as first byte.
2637 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2638 architecture descriptions.
2639 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2640 on those architectures.
2641 - Optimization: send/recv array of scalar in one shoot
2643 2004-07-05 Martin Quinson
2644 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2646 - Plug a whole bunch of memleaks
2647 - each process now have to call gras_{init,exit}. One day, their log
2648 settings will be separated
2649 - Continue the code factorisation between SG, RL and common in Transport.
2651 2004-07-04 Martin Quinson
2653 - Redistribution between SG and RL.
2654 We wanna have to accept in SG, so move accepted related parts of RL in
2655 the common part. (more precisely, the dynar of all known sockets is no
2656 more a static in transport.c, but part of the process_data)
2658 [gras_stub_generator]
2659 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2661 2004-07-03 Martin Quinson
2662 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2663 them all up in one shot)
2665 - Fix the ugly case of reference to dynamic array.
2666 - New (semi-public) function gras_datadesc_size to allow the messaging
2667 layer to malloc the needed space for the buffer.
2669 - gras_socket_close now expect the socket to close (and not its address to
2670 put NULL in it after it). This is because the socket passed to handlers
2671 is one of their argument (=> not writable).
2673 - propagate the interface cleanup from last week in datadesc, ie remove a
2674 superfluous level of indirection. User pass adress of variable
2675 containing data (both when sending and receiving), and not of a variable
2676 being a pointer to the data. Let's say that I like it better ;)
2677 The price for that is constructs like "int msg=*(int*)payload" in
2678 handlers, but it's a fine price, IMHO.
2680 - Let it work in RL (yuhu)
2682 2004-06-21 Martin Quinson
2684 - porting SG plugin and SG select to new standards (works almost).
2685 - plug memleaks and fix bugs around.
2688 - cleanup the prototype of data recv and force users to specify when they
2689 want to handle references to objects. Test case working even for cycles.
2690 - plug memleaks. Valgrind is perfectly ok with this.
2692 2004-06-12 Martin Quinson
2694 - cleanup the separation between plugin and main code in plugin creation
2696 2004-06-11 Martin Quinson
2698 - Reput hook for raw sockets, needed for BW experiments
2699 - kill a few lines of dead code
2700 [Data description] Interface cleanup
2701 - gras_datadesc_by_name returns the searched type or NULL.
2702 That way, no variable is needed to use a type desc once, which makes
2704 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2705 two parameters were strings (field name, type name), leading to
2707 [Dicos] Interface cleanup
2708 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2709 This is consistant with the dynar API.
2711 2004-04-21 Martin Quinson
2713 - Porting to new standards.
2715 - interface cleanup.
2716 There is no bag anymore, no need to take extra provision to mask the
2717 pointers behind "ID".
2718 Better splitup of functions between files create/exchange/convert.
2719 This is still a bit artificial since convert and receive are so
2720 interleaved, but anyway.
2722 - add a queued message list to procdata (the ones not matching criteria
2724 - factorize some more code between SG and RL wrt procdata
2726 - use gras_exit in example to track memleaks
2727 - get rid of gs_example now that GS is properly integrated into gras
2728 - update run_test to integrate the latest tests (datadesc)
2730 - rename WARNINGn macros to WARNn since it prooved error-prone
2732 2004-04-19 Martin Quinson
2734 - register init/exit functions within gras module mechanism
2735 - send/receive function.
2736 Convertion is not implemented, but short-cutted if not needed.
2737 struct/array elements are sent one by one (instead of block-wise), but
2738 nobody really cares (yet). Get a prototype before optimizing.
2739 - tests (using a file socket) for DD send/receive on:
2740 - base types: int, float
2741 - array: fixed size, string (ie ref to dynamic string)
2742 - structure: homogeneous, heterogeneous
2743 - chained list, graph with cycle
2744 Believe it or not, valgrind is not too unhappy with the results. The
2745 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2749 - Bugfix when using a filename explicitely (instead of '-')
2751 2004-04-09 Martin Quinson
2753 - factorize more code between RL and SG in socket creation
2754 - Complete the implementation and tests of:
2756 o file (only in RL, and mainly for debugging)
2758 I lost 3 days to design a portable address resolver, and then decided
2759 that the prototype mainly have to run on my box.
2760 Addressing portability too early may be like optimizing too early :-/
2762 - use gras_init in the Tests instead of the crappy parse_log_opt
2763 (the latter function is removed)
2764 [Conditional execution]
2765 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2766 [Code reorganisation]
2767 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2768 Build examples against the RL library, since there is no way to disable
2769 its creation for now.
2771 For information, the beginning of coding on GRAS was back in june
2772 2003. I guess that every line has been rewritten at least twice since