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 documentation
47 - Increase code coverage in test suites
51 * Protect context stacks against stack overflow. The number of protected memory
52 pages allocated on the top of each stack (1 by default) can be configured
53 with the new command line option --cfg=contexts/guard_size:<value>.
54 * Simcalls are now generated by a python script that
55 - generates files included by SimGrid
56 - checks that all the functions exist, and proposes prototypes
58 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
59 host_set_data, host_get_data
61 - simcall_process_join(smx_process_t process, double timeout)
62 * Fix bug where sleeping processing could not be suspended.
65 * Translate surf models from C to C++
66 - Generic classes for all models: Model, Resource, Action
67 - A generic interface for each kind of model (CPU, Network, Storage
68 Workstation, WorkstationVM)
70 * Translate surf routings from C to C++
71 * Add callbacks using sigc++ or boost::signals2
72 - Add callback functions for resource creation/destruction
73 - Add callback functions for action state change
74 - Handle Energy as a plugin
75 * Replace swag by boost::intrusive
76 * Add new routing model for torus clusters
77 - topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn" parameters
79 - see examples/platforms/torus_cluster.xml
83 * Hostfiles support host:nb_processes construct to deploy several processes on
85 * Collective communication algorithms should not crash if used with
86 improper number of nodes and report the error.
87 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
88 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
89 MPI_Dims_create, MPI_Cart_sub are supported.
90 * New interface to use SMPI programmatically (still depends on MSG for
91 some parts, see examples/smpi/smpi_msg_masterslave) :
92 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
96 * Global variables privatization in MPI executables is now performed at runtime
97 with the option smpi/privatize_global_variables (default:no).
98 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
99 dynamic libraries loaded by the application are not privatized (static
100 linking with these libraries is advised in this case)
103 * Options defined in XML work correctly now.
106 * New cmake option, enable_lib_in_jar, to control whether native libraries are
107 copied into simgrid.jar or not (ON by default). Use this option if you want
108 to reduce the size of the installed simgrid.jar, *and* the native libraries
109 are kept installed elsewhere.
110 * Surf binding with SWIG (code generated in maintainer mode only):
111 - plugin to handle callbacks
112 - CPU model only for the moment
115 * Supernovae build mode is definitively removed. It was used to improve
116 inlining and inter-module optimizations. It is nowadays superseded by
117 link-time optimizations commonly available in compilers.
118 * Update ns-3 find lib. Bindings for ns-3 should work again now.
119 * Add boost dependency for surf++
120 * Add new macro for tests
121 - ADD_TESH(name <tesh_args>)
122 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
126 - xbt_bar_t XBT_barrier_init(unsigned int count)
127 - int XBT_barrier_wait(xbt_bar_t barrier)
128 - void XBT_barrier_destroy(xbt_bar_t barrier)
129 * Make the xbt_os_time module public
131 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
133 SimGrid (3.10) stable; urgency=low
135 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
138 * Reintegrate Java to the main archive as desynchronizing these
139 package is not acceptable anymore (Java is now considered stable)
140 * Add explicit synchronization facilities through semaphores
141 * Bug fix: Task.setDataSize() only changed the C world, not the value
142 cached in the Java world
145 * Dramatically change the way files are handled. API and internals changed, but
146 this part of MSG was not considered as production grade either.
147 * Add explicit synchronization facilities through semaphores
148 * Add a new function MSG_host_get_process_list()
149 * Preliminary DVFS support (see examples/msg/energy/ for details)
152 * SMPI is now included directly in the libsimgrid as the windows
153 linker doesn't force us on splitting it anymore.
154 * Improvements of the SMPI replay tool:
155 - Most of the collective communications are now rooted in the same process as
156 in the original application.
157 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
158 used until now). Multiple data types can now be used in a trace.
159 - The replay tool now supports traces produce either by TAU or a modified
161 - Bug Fix: the compute part of the reduce action is now taken into account.
162 - Gatherv collective is now supported
163 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
164 of smpirun outputs time independent traces for the current run. One file
165 is created per process. If too many processes are simulated, this behavior
166 can be changed to one file for all processes by using the
167 tracing/smpi/format/ti_one_file flag
168 * smpirun generates the host file if needed (with given host count and platform)
169 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
170 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
171 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
172 --cfg=smpi/coll_selector:(mpich/ompi)
173 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
175 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
176 happening inside a collective SMPI call.
177 * Fix the behavior of complex data types handling.
178 * Make MPI_Wtime another synchronization point to take computations into account.
179 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
180 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
181 * Add all missing Fortran bindings, SMPI should work with Fortran 90
182 (no privatization of global variables yet)
183 * Preliminary DVFS support (see examples/smpi/energy/ for details)
186 * Verification of liveness properties is now available for SMPI applications
187 (in addition to MSG applications)
188 * Bugged examples using SMPI in examples/smpi/mc/
189 * Add --cfg=model-check/visited option. Allows the verification of infinite
190 programs. Detection of loops in the execution thanks to the system state
191 comparison and reduction of the state space to explore. Can be combined with
192 DPOR for safety properties.
195 * Allow to change SimGrid configuration (see --help) within the code
196 thanks to SD_config() as it can be done in MSG.
197 * Add a new function SD_task_set_amount() upon user request.
200 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
201 * Remove rule based routing (no more PCRE dependency)
202 * Add a limiter_link option to cluster tag, to specify a maximum reachable
203 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
204 * Add a loopback_bw and loopback_lat options to cluster tag.
205 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
206 routing. Now peers are encapsulated in an AS and have their own private
207 router but this is transparent.
210 * Our own implementation of getline is renamed xbt_getline, and gets
211 used even if the OS provide a getline(). This should reduce the
212 configuration complexity by using the same code on all platforms.
213 * New type: xbt_cfg_elm_boolean.
214 * Allow to use yes/no for boolean configuration options in the command line.
215 * Allow to disable SimGrid cleanups at exit from command line option.
216 There are situations where one may want a simulation to end with an exit.
217 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
218 annoying when scripting around the simulator. Adding a
219 --cfg=clean_atexit:no allows to circumvent this issue.
222 * Lots of memory leaks were corrected in this release.
223 There is no known memory leaks anymore, in all of our 600+ tests.
224 * New command line option --version, to get SimGrid version information.
225 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
227 * Supernovae builds are deprecated, and expected to be removed in the next
230 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
232 SimGrid (3.9) stable; urgency=low
234 The Grasgory release: GRAS is really dead now.
236 * Complete overhaul of the internal host structures scheme.
239 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
240 was considered as experimental and badly maintained since then.
241 * Keeping it was thus a trap to our potential users, that could take
242 it instead of MSG or SMPI by mistake despite is pity state.
243 * GRAS seems to have very few users (if any), and no one volunteered
244 to maintain it further. It also induces a lot of XBT code (for
245 portability sake), that must be maintained too.
246 * For all these reasons, we killed GRAS. If someone wants to revive it
247 in the future, don't cry, our git history still remembers of GRAS.
250 * Major overhaul. Merge our documentation again as time proved that
251 splitting it was really not helping our users.
252 * Further improve the developer documentation to help newcomers
253 hacking on SimGrid itself. The user documentation (and in
254 particular, the beginner documentation) is still in a sorry state.
257 * Now works on Windows too!
258 * Much more extensive test suite, from MPICH
261 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
262 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
263 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
264 * Bug fix: let task be scheduled when the last dependency to be solved is
265 a control dependency.
266 * Remove SD_load_environment_script function.
267 Use the C sg_platf function if you want to declare a platform
271 * New function: MSG_process_get_number()
272 * Old function documented: MSG_config()
273 * Remove MSG_load_platform_script function
274 Use the C sg_platf function if you want to declare a platform
278 * Change the default value of the TCP_gamma constant (maximal size of TCP
279 congestion window) to a more realistic 4MiB value. If you notice changes in
280 your simulation results, you can fall back to the previous 20k tiny window
281 by adding --cfg=network/TCP_gamma:20000 on command line.
282 * (Hopefully) fix a bug wrt periodic availability and state traces
283 * Bug fix: use default values at start when first event in availability/state
284 trace is not at time 0.
287 * remove the "new_" part of function name sg_platf_new_trace_connect
288 (resulting in sg_platf_trace_connect), since it does not create
292 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
293 use case where it's really mandatory, and maintaining it was a pain in
295 * New: xbt_fifo_search(), search an item with a user-provided
296 comparison function instead of dumb pointer comparison.
299 * Fix the lua deployment:
300 Use `simgrid.init_application()` before deployment instead of
301 `simgrid.msg_register_application()` after.
304 * Transfer the tracing files into the corresponding modules.
306 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
308 SimGrid (3.8.1) stable; urgency=low
310 The "we are told that some people want to also *install* the simgrid
313 * Add missing file "tesh.1" to the archive.
315 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
317 SimGrid (3.8) stable; urgency=low
319 The Psssshiiiit release: SimGrid jumps into the Cloud.
322 * Add an experimental interface to manipulate VMs. They are mainly
323 process groups with very few intrinsic semantic, but they should
324 allow you to build the semantic you want easily.
325 * New function: MSG_host_set_property_value()
326 * New function: MSG_process_on_exit(). To clean memory in all cases.
327 * Bug fixes that made the host (and link) failures unusable.
328 * Add a way to auto-restart process when the host in which they are
329 executing comes back (ON_FAILURE="RESTART" on deployment file,
330 MSG_process_auto_restart_set).
331 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
332 please stop using the old ones, they are DEPRECATED.
334 * Deprecate functions MSG_global_init() / MSG_global_init_args()
335 Please use MSG_init() instead. (reducing the amount of entry
336 points in the library helps us).
337 * Make it impossible to link against the wrong version of the lib
338 * Deprecate MSG_clean(). No need to call it anymore.
339 * Function MSG_get_host_number() is not deprecated anymore.
342 * Split the doc into a user guide and a reference guide.
343 * Start a developper guide to help people hacking on SimGrid.
346 * Enable tracing by default. This modules rocks you should use it.
347 * Remove option custom_flags. Now use environment variables CFLAGS
349 * Use default cmake things to detect lua instead of home grown ones.
350 * New option "enable_mallocators" to disable mallocators, for debugging
351 purpose ("on" by default).
354 * Bug fixes around the resource failures: don't let the processes
355 survive the host they are running onto.
356 * Add an interface to auto-restart processes when the host in which they are
357 executing comes back.
358 * Ensures that SIMIX_clean is called automatically. It's not part of
359 the public interface anymore (bindings should be updated).
362 * Bug fix for when SD_Simulate is called with a positive value: be careful
363 when comparing doubles. Sometimes they are different for non significant
365 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
366 parallel task whose initial work is distributed among host according
367 to the Amdahl's law. Such tasks are created with a parameter alpha
368 that corresponds to the non-parallelizable part of the computation.
369 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
370 between two sets of workstations assuming a 1D block distribution (each
371 workstation owns a similar share of data) on both sides.
373 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
374 Data redistribution will be automatically scheduled once parent and child
375 are both scheduled. The filling of computation_amount and
376 communication_amount structures is now done seamlessly thanks to the chosen
378 * New function SD_workstation_dump to display various information
379 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
380 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
381 bandwidth on the route onto which the task is scheduled and the amount of
383 To divide the nominal bandwidth by 2, the rate then has to be :
384 rate = bandwidth/(2*amount)
385 * Compute tasks that have failed can now be rescheduled and executed again
386 (from their beginning)
387 * Increasing source code coverage (src/simdag is now covered at 95.8%
391 * Re-implement time-independent trace replay using SMPI (at the
392 smpi_smp_* level) instead of MSG. This should replace
393 examples/msg/actions/actions.c
394 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
395 hindexed and structs)
396 * Implement the exchange of non-contiguous data.
397 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
398 * Correct behavior of smpi/sender_gap and set its default value to 0
399 * Add option to asynchronously send small messages to allow better
400 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
401 specifies the size in bytes under which messages will be asynchronously sent.
402 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
403 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
406 * New C interface to define a platform: XML is now optional.
407 For more info, please check include/simgrid/platf.h.
408 * New interface to define random platforms from the C:
409 For more info, please check include/simgrid/platf_generator.h and
410 examples/msg/masterslave/masterslave_platfgen.c
411 * Export a sg_cmdline dynar containing all the arguments we got from
415 * Two new tracing options for adding comments to trace file so you
416 can track your experiments (see --help-tracing for details).
417 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
418 * Adding the SimGrid version that generated the trace file as a comment.
419 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
420 * Fix to avoid key clashes on Paje links
421 * Other minor fixes related to the Paje specification
424 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
425 and renamed to xbt_str_hash() and xbt_str_hash_ext().
426 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
427 * Greatly improve the robustness of mmalloc to user errors (such as
428 using an area after freeing it, or freeing it twice)
430 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
432 SimGrid (3.7.1) stable; urgency=low
435 * Restore the prototype of MSG_process_create_with_environment() to
436 the pre-3.7 situation by removing the kill_time argument.
437 * Add a MSG_process_set_kill_time() function instead.
440 * Fix weird behaviors when dealing with parallel tasks.
443 * Simgrid is now built as a dll.
444 * Simgrid-java now works on Windows.
445 * Simgrid-Java is now included into Windows package.
448 * First pre-build package for MacOSX.
451 * Fix compilation when using MSG_USE_DEPRECATED.
452 * Fix some compilation issues on Macs and Windows.
453 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
454 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
456 We discovered that the Lua console is broken, but we are missing the
457 manpower to fix it right now. The problem existed in 3.7 too, so we
458 are not blocking the release for that. Sorry if you depended on this
459 feature, any help would be really welcome.
461 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
463 SimGrid (3.7) stable; urgency=low
465 The "spring cleanups (before the next Big Project kicks in)" release.
468 * We can specify the SMPI latency/bandwidth factor with command line
469 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
470 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
471 You can also use the "config tag" from platform file by adding this line
472 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
473 (see "example/platforms/tag_config.xml" to use "config tag").
474 Note that the command line supersedes the platform file configuration.
475 * Change the correction factors used in LMM model, according to
476 the latest experiments described in INRIA RR-7821.
477 Accuracy should be improved this way.
478 * Use the partial invalidation optimization by default for the
479 network too. Should produce the exact same results, only faster.
480 * Major cleanup in surf to merge models and split some optimization
481 mechanisms from the core of the models. As a result you can now
482 specify which model to use (e.g., --cfg=network/model:LV08
483 --cfg=cpu/model:Cas01) and which optimization mode to use
484 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
485 Incompatible combinations should err at initialization. See
486 --help-models for the list of all models and optimization modes.
487 * The CLM03 workstation model was dropped for simplicity because it
488 used the deprecated CM02 network model. Use default instead.
489 * Rename the TCP_gamma configuration option to network/TCP_gamma
490 * Rename the coordinates configuration option to
491 network/coordinates, and document it
492 * Use now crosstraffic keyword instead of the terribly misleading
493 fullduplex keyword. It is activated by default now in the current
494 default model, use --cfg=network/crosstraffic:0 to turn it off.
495 * Ongoing refactoring the model parsing to make XML files optional
496 See include/simgrid/platf.h for details (still to be completed)
499 * Major overhaul of the documentation. Almost instructive now :/
500 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
501 functions and friends. This interface was considered as
502 deprecated since over 2 years, it's time to inform our users that it is.
503 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
504 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
505 use these (crappy) functions in your code.
506 These functions will be removed soon. Stop using them now.
507 * Deprecate MSG_get_host_{table,number}
508 Implement MSG_hosts_as_dynar() instead.
509 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
510 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
513 * Stabilize the parallel execution mode of user contexts
514 * Introduce configuration variables to control parallel execution:
515 - contexts/synchro: Synchronization mode to use when running
516 contexts in parallel (either futex, posix or busy_wait)
517 - contexts/parallel_threshold: Minimal number of user contexts
518 that must be part of a scheduling round to switch to parallel
519 execution mode (raw contexts only)
520 * Fix bugs that prevented to use suspend/resume along with
521 synchronization structures.
522 * Fix bugs in process termination that lead to invalid memory access
523 in very specific conditions.
526 * Introduce a parallel mode for the models (controlled by surf/nthreads
527 configuration item). In our tests, running the models in parallel
528 never lead to any speedups because they are so fast that the gain
529 of computing each model in parallel does not amortizes the
530 synchronization costs, even when ultra fast futexes are used.
531 This is released anyway because YMMV.
534 * Performance boost by using a swag internally to compute the set of
535 tasks that are finished and should constitute the return value of
539 * Enable it by default now that it is considered rather stable.
542 * Documentation of the tracing functions.
543 * Performance gains when tracing categorized/uncategorized resource
544 utilization by avoiding calls to get route when updating resource
545 variables. LMM constraints are being used instead.
546 * API changed to set task categories. Use MSG_task_set_category instead
547 of TRACE_msg_set_task_category, and SD_task_set_category instead
548 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
549 * Bugfix for graphicator, routes not correctly obtained, memory leaks
550 * Examples for link user variables added (see at examples/msg/tracing/)
551 * Deprecated function TRACE_msg_set_process_category completely removed
552 * Trace header updated according to the latest Paje file format
553 * Tracing network lazy updates, no longer obligate users to use full updates
554 * --cfg=tracing/platform:1 also registers power/bandwidth variables
555 * Experimental: let user code declare/set/push/pop application states for hosts
556 * API changed to allow the manual creation of graph configuration files
557 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
560 * Improve the API of Lua MSG bindings, using the Lua spirit.
561 * Each simulated process now lives in its own Lua world (globals are
562 automatically duplicated). It helps writing simulators. Will allow
563 to run Splay programs within SimGrid in the future.
564 * Add a Chord example in Lua, equivalent to the MSG one.
567 * Start the implementation of a solution to express temporal
568 properties, not only local assertions. This is still an
569 experimental work in progress, stay clear from it to be safe.
573 - Add new runtime parameters --help-logs and --help-log-categories
574 to display informations about supported logging parameters and
576 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
577 don't exists anymore.
578 * Mallocators: allow value NULL for the reset function.
580 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
581 create homogeneous dictionaries, where all the elements share the
582 same free function. Non homogeneous dictionaries will be
583 deprecated in the next release.
584 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
585 - Multi-level dictionaries are deprecated.
587 - new function xbt_dynar_search_or_negative() that is useful when
588 you have less than 2 million elements in your dynar and don't
589 want of the extra complexity of catching exceptions when the
590 element is not found.
592 - Make xbt_os_thread module (for thread portability) public.
593 Documentation is still to come, sorry.
595 - Cleanups and simplifications to make it maintainable again.
596 - Exotic features (such as memalign and valloc) were removed.
597 - The metadata were extended and improved so that the
598 model-checker becomes able to explore and inspect the heaps.
599 - This may induce a performance drop when enable_model-checking is
600 ON in cmake (even if it's not used in the simulation), but it is
601 necessary at this point to get MC working.
603 Turn model-checking OFF if simulation performance matters to you.
604 Not enabling it at runtime is not enough, disable it in cmake.
606 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
608 SimGrid (3.6.2) stable; urgency=low
610 The "Not coding new stuff allows to polish old things" release.
613 * New bindings to the NS3 packet level simulator (experimental)
614 * Use the raw (efficient) execution contexts instead of the sysv
615 (portable) ones when possible.
616 * libpcre is now mandatory in any cases since not using it led to
617 severe performance loss and possibly other issues
618 * Update the XML platforms:
619 - G5K: include the latest machine in Nancy
620 - GridPP and LCG: new platforms
621 * Documentation was partially updated, at least (more to come)
623 Bug fixes, cosmetics and small improvements
624 * Free terminated processes before the end of the simulation to avoid
625 exhausting the memory of users having very dynamic amount of
627 * Bug fix and cosmetics about canceling non-running tasks
628 * Bug fix about the dot loader's issues when using libcgraph
631 * Create an installer for windows with nsis (amd64 and win32)
632 - Add an hello world project to illustrate simgrid project creation.
633 - Embed libpcre into the Simgrid installer to avoid
634 its compilation burden
635 * The raw execution contexts should work on Apple now
636 * Port to Windows 64 bits
637 - Sysv contexts now have an implementation for this arch
638 - GRAS communication features now support this arch
639 * Drop support for borland compiler on windows
640 - this code was not maintained, and we kinda depend on gcc nowadays
641 * Fix portability issues on kfreebsd/gnu: build error about semaphores
642 * Fix portability issue on unstable ubuntu: linker became picky on
645 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
647 SimGrid (3.6.1) stable; urgency=low
649 The "Oops, we broke Macs too" release
652 * Fixed contexts detection so that raw ones are used when possible
653 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
654 leads to a strange error, with user code segfaulting sometimes when
655 the generated code is not perfectly aligned (which is not
656 controllable from the user side, depends on the amount of code)
659 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
661 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
663 SimGrid (3.6) stable; urgency=medium
665 The Summer Release, also known as the "OMG! They Killed Kenny!" version
668 * Bindings now constitute their own package, separated from the main one.
669 Rationale: reduce our maintenance nightmare by reducing the module coupling
670 They will soon be released on their own on gforge.
672 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
673 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
675 GRAS: It is not considered as stable anymore, but experimental. Sorry.
676 * It's not quite deprecated for now because we have no replacement,
677 but it may soon become the case.
680 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
681 * Fortran: New user-level cache variable to store the rank of the running
682 process. This improves performance by an order of magnitude.
683 * C: New coccinelle script to automatically locate and modify global and
684 local static variables.
685 * Improved SMPI network model with a sender-side gap to account for multiple
689 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
690 only say if a communication is finished, no matter whether it succeeded or
691 failed. You can call MSG_comm_get_status() to know the status of a finished
693 * New function MSG_task_dsend() to send a task and detach it. When a
694 communication is detached, you are never notified of its success or failure
695 and the memory is released automatically once it is finished. This function
696 is useful when you don't care about the end nor the success of a
698 * Change the prototypes of action replay. Sorry for inconvenience,
699 but this is really more efficient this way (and to adapt your code,
700 you just have to fix the initialization, that shouldn't be too long)
701 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
702 ever managed to do anything useful with it.
703 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
704 takes a dynar of communications. It returns immediately and gives the
705 index of a finished communication (if any).
706 * New example: a basic implementation of the Chord P2P algorithm.
709 * New model for multi-core CPUs. You can now use the core attribute to
710 precise the number of cores of a host. This is a basic model. Every
711 process running on the host receives at most the power provided in
712 the DTD (throughput<=power). Total throughput of process cannot exceed
714 * New peer tag. This peer tag creates a tiny AS comprising a host and a
715 router linked by an up-link and a down-link (possibly asymmetrical).
716 This kind of pattern allows to easily build last-mile model style platforms.
717 Aggregating such patterns in a rule-based AS is thus the technique of
718 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
719 * New model for Vivaldi routing. We transformed the Vivaldi network model
720 into a Vivaldi routing model (based on the rule-based model). This allows to
721 combine Vivaldi based latencies with last-mile platforms.
724 * Added a check for NaN of IEEE754 infinite in the double entries of
726 * Introduce a new context factory "raw", highly inspirated from the
727 ucontext factory, but using manually crafted functions in assembly to
728 do the work in an efficient manner.
729 * Allow to change the used context factory at run time, not only at
730 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
731 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
732 contexts at runtime (only with raw contexts or ucontexts).
733 * Completely rewrote this module to allow parallel execution of user
734 processes. Use --cfg=contexts/nthreads:N to execute user processes
735 with N parallel threads (the default is 1, meaning no parallelism).
736 * Allow to decide dynamically between sequential and parallel modes.
737 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
738 processes in parallel only when their number is greater than or equal to P
740 * Added a check for NaN of IEEE754 infinite in the double entries of
744 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
745 won't output the state of processes when interrupted with Ctrl-C
746 * Add a new function xbt_dynar_to_array that transforms a dynar into a
747 NULL-terminated array. This may solve backward compatibility issues
748 due to the change to return type of SD_simulate. See also:
749 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
750 * Add new macros with variable number of arguments.
751 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
752 - in xbt/asserts.h: xbt_assert
753 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
754 - in xbt/ex.h: THROWF and RETHROWF.
755 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
757 * Change xbt_die() to accept a format string with arguments, just like printf.
758 * New data structure: xbt_lib_t, like a dict but more general and with better
762 * New configuration options
763 Options triva/categorized and triva/uncategorized can be used to generate
764 graph configuration files for Triva visualization tool.
765 * Configuration option tracing/platform is renamed to tracing/categorized
766 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
767 * New TRACE_user_link_variable interface:
768 User provides the name of the link and the tracing variable to attach to it
769 * the declaration of tracing categories must be done after the environment
771 * simpler tracing interface, just one way to declare categories
772 TRACE_category or TRACE_category_with_color, it is up to you
773 * links in the trace file are again identified by their names
774 * trace contains the full platform hierarchy exactly as declared using the ASes
775 * Options tracing/msg/[task|process]:1 groups the process by hosts
776 for both cases, tasks and processes must have names that are unique during the simulation
777 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
778 * The experimental option tracing/msg/volume is deprecated
779 its functionality may be reincorporated if needed
781 The tracing generates a trace file with unordered timestamped events,
782 because of the way the core simulator (surf) works. A script available
783 at the tools directory (fix-paje-trace.sh) can be used to put the events
784 in order. We have changed the tracing so it can generate ordered timestamped
785 events in the final trace, but depending on the simulator (and how much time
786 is simulated) that can lead to a huge memory utilization. It is deactivated
787 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
790 * Define a SIMGRID_VERSION macro in simgrid_config.h.
791 - We are trying hard to keep the API stable, but it may happen that
792 some things change (we're a research project after all, not a
793 nuclear plant operating system). If such things should happen, you
794 could rely on that macro to adapt.
795 - current value: 30600 for 3.06.00, aka 3.6
796 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
797 a number that can be compared with SIMGRID_VERSION.
798 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
799 assertions and verbose/debug logging events are disabled at compile time.
801 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
803 SimGrid (3.5) stable; urgency=medium
806 * New feature to any SimGrid-based simulator: Model-Checking
807 Check SIN#1 for more details.
810 * New Model SMPI (three-interval linear regression for correction factors)
811 See RR-7426, available at http://hal.inria.fr/inria-00527150
812 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
813 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
814 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
815 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
816 * See also src/smpi/README
821 - Tracing API changes: TRACE_start and TRACE_end should not be called
822 by user-code. They are automatically called by simulators created
823 with SimDAG, MSG and SMPI if the toolkit is compiled with
824 tracing_enabled=ON. Categories declaration and utilization remain the
825 same for MSG and SimDag.
826 - A function was added to the tracing API to declare categories with
828 - TRACE_category_with_color (char *category, char *color)
829 where color must be in the following format
830 "%f %f %f", red, green, blue
831 and red, green, blue are float values in the interval [0, 1]
832 - User can specify NULL as color parameter, or continue calling
834 On that case, the tracing system will define random colors
835 - The following command-line options are supported:
836 --cfg=tracing/filename:msg.trace
837 --cfg=tracing:1 (activate tracing, needed to use others)
838 --cfg=tracing/platform:1 (categorized resource use)
839 --cfg=tracing/uncategorized:1 (uncategorized resource use)
840 --cfg=tracing/msg/task:1 (task creation)
841 --cfg=tracing/msg/process:1 (process creation, migration)
842 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
843 --cfg=tracing/smpi:1 (SMPI interface tracing)
844 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
845 - examples of examples/msg/tracing updated
847 - DAXLoader and DOTLoader functions can generate tasks with categories
848 - A new function to attribute a category to SD tasks:
849 TRACE_sd_set_task_category (SD_task_t task, char *category)
850 * Tracing the MPI interface implemented by SMPI
851 - Collective operations are traced with states
852 - Point-to-Point operations are traced with states/links
853 - Tracing activated by a parameter "-trace filename" passed
854 to smpirun during execution (considering that simgrid
855 is compiled with tracing enabled)
856 - To run the simulation with gdb, the simulator
857 accepts --cfg=tracing/smpi:1 to trace SMPI
858 - tesh files to check if smpi tracing is ok
859 - See examples/smpi/NAS/DT-trace
860 * GTNetS tracing re-worked
861 - adaptation to the tracing system of GTNets to cope
862 with modifications regarding the fullduplex mode
863 - new tesh files to check if gtnets tracing is ok
866 * Asynchronous communications through the functions:
867 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
868 * New function: MSG_load_platform_script()
869 to make possible using a lua script instead of XML files to set up platforms
870 * New function: MSG_set_function
871 to associate functions to processes, used when bypassing the parser
872 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
874 Platforms: Add some more examples in examples/platforms
875 * Grid'5000: see www.grid5000.fr
876 * *_30000_hosts.xml: various huge files [mainly scalability testing]
879 * Change the XML format. This is a very important modification. SimGrid
880 3.5 introduces a new hierarchical format based on the notion of
881 Autonomous Systems. Compatibility with old format is ensured
882 through the perl script provided in the install bin directory
883 bin/simgrid_update_xml.
884 It is now possible to build platforms with specific routing
885 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
886 connect several platforms together. We will try to provide soon
887 set of realistic platforms exploiting these properties (have a
888 look at examples/platforms/ for the moment).
889 * Take the opportunity of the XML format change to be a good XML citizen:
890 rename link:ctn to link_ctn and similar changes (also dealt with by
892 * Add a new routing scheme (rule-based) using regular expressions. It
893 enables to have an extremely low memory footprint when the
894 underlying routing is simple and can be compactly described. You
895 need to have libpcre4-dev (perl regular expressions) installed if
896 you want to use this routing scheme.
897 * Revive the cluster TAG and allow to easily and efficiently (both in
898 term of memory and speed) connect clusters together. Have a look
899 at teshsuite/simdag/platforms/ to see how this can be done. With
900 this tag, you can create clusters with thousands of tasks at no
901 cost (have a look at examples/platforms/).
902 Note: clusters are implemented as ASes, so there is no need for an
903 enclosing AS tag if you have only one cluster in your platform.
904 * Add new generic functions in the public interface that allows the user
905 to call SURF 'create_resource' methods from your code (same
906 functionality as the XML bypass mechanism but with a much lighter
908 * Add a new model (enabled through command line --cfg=network/model:SMPI)
909 that uses a piecewise linear approximation to produce better
910 results when exchanging small messages.
911 * Add a new parameter to handle correctly full duplex link and account
912 for interferences between uplink and downlink communications
913 (activate with --cfg=fullduplex:1).
916 * Rename the SD_READY (all dependencies are satisfied and task is
917 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
918 dependencies are satisfied) state.
919 This prevents a confusion between the notion of "ready to schedule"
920 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
921 simulated" (SD_RUNNABLE) used by the simulation kernel.
922 * Change the way a task is considered as ready. Instead of removing
923 dependencies when a task is done, a counter is decreased. This way,
924 it is always possible to reach ancestors thanks to the
925 SD_taks_get_parents function (even after the end of the simulation.)
926 * Change the return type of SD_Simulate from (SD_task_t*) into
927 xbt_dynar_t. This function was in handling a dynar internally and
928 converted it into a NULL terminated array for historical reasons.
929 * New function SD_dotload(char*) to load a DAG described in dot
930 format. This loader and the corresponding examples require the
931 installation of the graphviz library.
932 * Fix a bug in the management of tasks of size 0 in the surf network
933 models. This problem was only visible with SIMDAG and you should
934 thus disregard results produced with earlier versions if you
935 relied on this feature (some tasks were blocked because of this).
936 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
937 even though your applications were doing only point-to-point
938 communications and sequential computations. Now you can really use any
939 model you want (of course, if you create real parallel tasks, which are
940 not implemented in most models beside ptaskL07, this will abort).
941 * Add an example that schedules a DAX on an heterogeneous platform
942 using a Min-Min strategy.
943 * New function SD_workstation_get_current_task() that returns the kind
944 of task currently running on a workstation in the sequential access
946 * Raise some warnings when unexecuted tasks remains at the end of the
947 simulation. This is usually caused by cycles in the DAG.
950 * New function: SIMIX_process_set_function() called by MSG_set_function
951 * Change the underlying waiting queue in semaphores so that a process
952 can wait on several of them simultaneously (as in waitany).
953 * Fix the way to handle tokens in semaphores so that all access patterns
954 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
955 * kill the dirty pimple SIMIX_message_sizes_output()
956 Please use (proper) visualization instead
959 * New data container: setset (set of sets of elements)
960 * New module: mmalloc (mapped malloc, allowing to have several
961 independent segments of malloc)
962 * New function: xbt_dict_cursor_set_data()
963 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
964 * New function: xbt_dynar_is_empty()
965 * New function: xbt_fifo_get_last_item()
966 * Fix xbt_dynar_shrink(): use the right element size.
967 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
968 uninitialized areas during expand.
969 * Fix semaphores: previous implementation was severely broken.
970 * Use library init/fini functions for our initialization.
971 - you can use logs and other feature as soon as you want in your
972 code (even before the xbt_init / MSG_init)
973 - xbt_exit is now a no-op and produce a warning when used.
976 * Port GRAS to new SIMIX mechanisms. This allows gras users to
977 benefit from the latest improvement to the simulation kernel.
978 * Kill measurement sockets for now. If you rely on them, sorry. This
979 release is not for you. This feature will be reintroduced in the
980 future, but we cannot delay the release any further.
981 * New function: gras_msgtype_get_name().
982 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
983 * Fix (at last) the pmm example: it should not randomly fail anymore.
985 Build chain: bug fixes and overall polishing
986 * Cmake is now stable enough. Hence, we killed the autotools.
987 * Port to windows ( TM :)
988 * Fix the 'make install' target.
989 No need to use 'make install-simgrid' anymore
990 * Introduce a 'make dist' target compiling a *source* archive
991 'make package' compiles a binary archive
992 * Compile java files only on need
993 * Add --cd and --setenv command line options to tesh
994 * Out of source builds are not fully supported yet, but we are close
995 * Enable supernovae and optimization flags by default for our users
998 * Add layer to set up environment directly from lua, without XML.
999 * The effect of gras_stub_generator can be achieved through
1000 lua too (check examples/gras/console/ping_generator.lua)
1002 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1004 SimGrid (3.4.1) stable; urgency=low
1006 The "Polishing easter eggs is probably a good idea" release.
1007 This is a bug fixes release only.
1011 * Fix a bug preventing the tasks from begin garbage collected.
1014 * Fix a bug occuring when a host involved in a communication fails.
1015 This was not detected properly by the other peer involved in the
1016 communication. Now, it's reported as a network error.
1019 * Warn the user about loop dependencies in data flow of DAX files
1020 * Obey the control-flow dependencies of DAX files
1023 * Add option "enable_smpi" allowing to not compile SMPI.
1024 Probably useful for the (Mac) users experiencing a build error here
1025 * Improve the detection of lua5.1 and ruby1.8
1027 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1029 SimGrid (3.4) stable; urgency=low
1031 The "Easter in Cargese" release. Also known as (major changes):
1033 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1034 ~> bindings were greatly improved
1035 ~> new tracing infrastructure for better visualization introduced
1037 * the "Welcome to configury modernity" release.
1038 ~> we switched from autotools to cmake, and improved our cdash
1041 A more detailled list of changes follow (full detail in svn log).
1043 Java Bindings: Various Cleanups
1044 * (install java-gcj-compat-dev on debian-like to use them)
1045 * Remove put/get: no need to export deprecated interface in Java
1046 Use send/receive instead.
1047 * Cleanup the examples and add a README per directory
1048 * Remove example autoDestination (that's the only way to go now)
1049 * Remove example explicitDestination (was a plain copy of basic)
1050 * Make JniException a runtime exception, so that there is no need to
1051 declare the fact that you may encounter such a beast. I guess that
1052 nobody will ever want to survive such error.
1053 * Create specific errors for each MSG case of failure:
1054 host failure, transfer failure, timeout, task cancelled
1055 * Cleanup the exceptions that may get thrown by each function
1056 * Other internal cleanups in Java bindings. Performance still bad :/
1057 Ruby and Lua Bindings: create them
1058 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1059 * That's new and great, you should try them out.
1060 Same functionalities than Java bindings, only even less polished
1062 * Kill the useless "rate" argument of SD_task_get_execution_time()
1063 Everyone used to provide -1 as a value, it was not used, and the
1064 semantic of a possible use wasn't even clear.
1065 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1066 either as comm costs or compute costs to mean that there is no
1067 such thing for that specific task.
1068 * Add a SD_task_set_name() function
1069 * Fix SD_task_unschedule() on typed tasks
1070 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1071 * In DAX loader, accept useless 'level' attributes to <job> since
1072 LIGO DAGs have them (seem to be to ease graphical representation).
1074 * Add an example masterslave_mailbox.c using send/receive and not
1075 the deprecated put/get interface.
1076 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1077 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1079 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1080 (declare MSG_USE_DEPRECATED to still have the old name)
1081 * Add a MSG_task_set_data() function
1082 * About trace replay (see examples/msg/actions):
1084 - Allow to work with splitted trace files for each process
1085 Give the specific trace file as argument of each process,
1086 and call MSG_action_trace_run(NULL)
1087 You can still have one merged file for all processes.
1088 - Fix implementation of collective operations
1089 * Allow task_execute() on 0-sized tasks (closes #10063)
1091 * This is the first release of SimGrid where SMPI is not considered
1092 beta anymore (even if some corners should still be improved)
1093 * Port over the new SIMIX_network submodule (internal refactoring)
1094 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1095 * Implement more missing elements of the standard:
1097 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1099 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1100 MPI_Scan() MPI_Get_processor_name()
1101 - Added implementation of missing case for Alltoall (warning: it's
1102 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1103 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1104 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1105 SMPI_MPI_Allgatherv()
1106 * Bug fixes include:
1107 - MPI_Waitsome() was broken
1108 - Allow relative includes in smpicc
1109 - Command line cfg argument 'reference_speed' was ignored...
1110 - Some functions did not properly lead to auto-benching of user code
1111 - smpicc passes -O2 by default (just like openmpi one)
1113 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1114 * Bug fixes about timeouts during communications
1115 * add SIMIX_message_sizes_output() as a pimple to write to file the
1116 amount of messages per size. Use gnuplot to get histogram.
1117 Pimple because that's the only user-visible function of simix,
1118 defined directly in xbt.h (irk, sorry)
1120 - Add a SIMIX_sem_get_capacity() function
1121 - Fix interactions with processe resume/suspende
1122 - release_forever() was stupidly broken
1123 - Fix SIMIX_display_process_status() for processes in a semaphore
1124 - Make SIMIX_sem_block_onto() user-visible
1125 * Refactoring context stuff:
1126 - Use pseudo-OOP for better modularity
1127 - reimplement SIMIX_process_kill() without process_schedule() so
1128 that the latter can take as invariant that it is called from
1130 - Merge context_start into context_new for sake of simplicity
1132 * Add a Vivaldi network model, coded live during SUD'10 ;)
1133 * Rename configuration variables to start a hierarchy:
1134 o cpu_model -> cpu/model
1135 o network_model -> network/model
1136 o workstation_model -> workstation/model
1137 * New configuration variables:
1138 o network/bandwidth_factor: correction to bandwith
1139 o network/latency_factor: correction to latency
1140 o netwotk/weight_S: correction to the weight of competing streams
1141 * Add a long description to the models, that users can see with such
1142 argument on the command line: --cfg=cpu/model:help
1143 * --help-models display the long description of all known models
1145 * config: add the ability to set a default value after registration
1146 Does not override any previously set value (e.g. from cmd line)
1147 * dict: allow to have integer key and data.
1148 When so, you need to use the following functions
1149 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1150 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1151 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1152 In contrary to regular dicts, the key is not malloced before copy.
1153 Mixing scalar and regular elements in the same dict is not tested
1155 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1156 Tracing for Visualization:
1157 * SimGrid is now instrumented in order to generate a trace file for
1158 visualization analysis: to use it, need to compile SimGrid with the
1159 "tracing" option enabled, and instrument the program using SimGrid with
1160 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1161 (among other functions).
1162 * The instrumentation only traces the platform utilization for now
1163 * Documentation to use the tracing functions and how to analyze the
1164 traces with the Triva tool is written.
1165 * More information about: SimGrid FAQ (in the section Tracing Simulations
1168 * We moved to cmake as default build system. Autotools support will
1169 be dropped soon. Check the FAQ for more info about how to use it.
1170 * Greatly improved our cdash/ctest interactions
1171 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1172 * Added memory checking tests with valgrind; lot of memleak fixing.
1173 This may be the first release of simgrid with so few memory issues
1174 * Added code coverage tests.
1175 Our coverage is still improvable, but at least we see it on cdash.
1177 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1179 SimGrid (3.3.4) stable; urgency=low
1181 The "Desktop Grid needs love too" release (also called Xmas release).
1183 Models improvements:
1184 * Major speedup in the maxmin system solving by using lazy evaluation
1185 Instead of solving completely the maxmin system at each iteration,
1186 only invalidate (and recompute) the modified parts.
1187 This new feature is enabled in default models but you can try to
1188 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1189 * Cas01 IMproved as default CPU model
1190 This CPU model is the same Cas01 model, but it uses the
1191 maxmin-selective-update flag and a heap structure to manage
1192 actions on SURF kernel.
1193 It reduces the complexity to find the next action to finish and,
1194 consequently, it's faster than the old Cas01.
1195 This is the new default CPU model (Cas01).
1196 * Rename the old Cas01 model to Cas01_fullupdate
1197 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1198 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1199 * CpuTI (CPU Trace Integration)
1200 A new CPU model whose objective is simulate faster when using
1201 availability trace files.
1202 Instead of using a full featured, over engineered maxmin system for
1203 CPU modeling, this model does the pre-integration of traces files
1204 to calculate the amount of CPU power available, and so, executes
1205 faster than the old CPU models.
1206 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1207 * Use LV08 as default network model since it gives better accuracy
1208 for small messages and shouldn't change things for big ones.
1209 Use --cfg=network_model:CM02 to get the previous behavior.
1212 ******************************************
1213 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1214 ******************************************
1215 * The new CPU model may changes simulations!
1216 The point is that events occurring at the exact same timestamp
1217 are not scheduled in the same order with the old and new
1218 version. This may be enough to completely change the execution
1219 of simulations in some cases.
1220 * The new network model will change simulations!
1221 This new model is more realistic than the previous one, so you
1222 should consider redoing your old experiments with this model.
1223 Sorry for the inconvenience.
1226 * Introduce the supernovae compilation mode
1227 When compiled that way, the whole SimGrid (or almost) is put in a
1228 single compilation unit and compiled in one shoot.
1229 This is to help gcc which has difficulties to inline stuff from one
1231 The speedup seem to be above 15%, althrough more tests are needed on
1232 amd64 to confirm that gain.
1235 * Port of MSG's mailbox on top of SIMIX network
1236 The put/get mechanism was greatly simplified on the way.
1239 * New SIMIX network module. Provides:
1240 - Mailbox: rendez-vous mecanism to find with who you want to speak
1241 - Synchronous send/recv: easier and hopefully faster since the
1242 logic is handled in the maestro process directly now
1243 - Asynchronous send/recv: you dreamt of it? It's here now
1244 Too bad that nobody cared enough to propagate the change to MSG.
1245 * Add semaphores as SIMIX synchronization mechanism.
1248 * new function SD_daxload(char*) to load a DAX file
1249 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1250 * Introduce typed tasks. Specify its kind and cost at creation.
1251 At scheduling, just give where it should be placed, and the cost
1252 for each involved resource is automatically computed.
1253 Existing constructors so far (more to come of course):
1254 - SD_task_create_comm_e2e() for end-to-end communication
1255 - SD_task_create_comp_seq() for sequential computation
1256 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1257 * new function SD_task_dump() for debuging display
1258 * new function SD_task_dotty(task,FILE*) writing to file the info
1259 about the task in dotty format
1260 * SD_task_dependency_exists() can now cope with having one of its
1261 arguments NULL. If so, it tests whether the other argument has any
1263 * Add getters on list of preceding/following tasks:
1264 SD_task_get_parents(task) and SD_task_get_children(task)
1265 * Add getters on amount of workstations and list:
1266 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1267 * Add getter on task kind: SD_task_get_kind(task)
1268 * Update the start_time and finish_time of tasks on completion/failure
1269 * Bugfix: Remove task from state swags when destroyed
1272 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1275 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1276 dynar with one separate thread per value of the dynar.
1277 * Change the prototype of xbt_thread_create(), sorry.
1278 Added a boolean parameter indicating whether we want to join this
1279 thread (used in SG only for now)
1280 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1283 * GTNetS wrappers should now be usable again (and betterly tested too)
1284 * Fix a major regression from 3.2 where the timeout provided to
1285 MSG_task_put_with_timeout() was used as absolute time before which
1286 the comm should be done.
1287 * Start to fix the <cluster> tag.
1288 - Internal links should be good now (beside of the loopback, which
1289 use the private link instead)
1290 - paths to the external world is still rather broken
1291 - the <route:multi> tag is just broken. Actually that's brain-dead.
1292 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1293 to make it less stupid
1294 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1295 * Fix a source-level compatibility glitch from 3.2: after defining
1296 MSG_USE_DEPRECATED, you can use the old name
1297 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1298 * Allow to compile from the SVN with automake 1.11
1299 * Fix some problems when using the "start_time" tag in deployment XMLs.
1300 * Fix #8569: XBT/synchro.h has redundant declarations
1301 * Fix #8563: MSG return values and exceptions
1302 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1303 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1304 Thanks to Robson Peixoto.
1305 * Fix a vicious bug in dictionaries inducing that some elements were
1306 not freed on xbt_dict_free()
1308 Portability report of this version:
1309 * Main portability targets:
1310 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1311 on (amd64/i386/ia64)
1312 - mac leopard on i386
1313 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1314 but nothing critical.
1315 * Other platforms: windows, AIX and others were not tested for this release
1317 Timing report of this version:
1318 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1319 scenario parameters). From 8h to a few seconds in desktop grid settings.
1320 * Supernovae brings about 25% speedup on i386.
1322 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1324 SimGrid (3.3.3) stable; urgency=low
1326 The "Need for Speed" release.
1328 The timings done to validate the 3.3.2 were faulty.
1329 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1331 The problem was a conversion from a manually handled vector to
1332 xbt_dynar_t on the critical path.
1333 xbt_dynar_foreach calls functions, inducing stack management crap.
1335 We inlined these functions and xbt_dynar_foreach is now breath taking.
1336 We also inlined xbt_swag_belong on the way.
1338 Here are some approximate speedup measurements (on master/slaves
1339 simulations lasting between 10s and 20s each):
1340 3.3.1 -> 3.3.2: about same performance
1341 3.3.2 -> 3.3.3: 40% speedup
1342 3.3.1 -> 3.3.3: 40% speedup
1343 3.3.1 with inline patch -> 3.3.3: 30% speedup
1345 Our reading is that the refactoring which occurred in 3.3.2 made us
1346 suffer much more from the xbt_dynar_foreach low performance, but
1347 once we solved this, this refactoring proved to be very performance
1348 effective. From the 40% speedup, somehow, 10% are due to the
1349 inlining and 30% to the refactoring.
1351 That's a pitty that gcc cannot inline functions placed in other files
1352 alone. We have to choose between:
1353 - break the encapsulation (by putting private data structures and
1354 accessors in headers files to help gcc)
1355 - live with low performance
1356 - switch to a decent compiler such as icc (not quite possible).
1358 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1360 SimGrid (3.3.2) stable; urgency=low
1362 The "Simplicity does not preceed complexity, but follows it" release.
1364 The main contributors of this release were (lexical order):
1365 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1368 * Extract the routing logic into its own object.
1369 (was dupplicated in network.c and workstation_LV07.c;
1370 Allows to implement other ways of storing that info)
1371 => kill now useless network_card concept
1372 - Use dynar to represent routes (instead of void** + int*)
1373 - kill link_set (use surf_network_model->resource_set instead)
1374 - Add a command-line option to choose the routing schema to use
1375 - Add three new models:
1376 * Floyd (shortest path computed at initialization)
1377 * Dijikstra (shortest path recomputed all the time)
1378 * Cached Dijikstra (shortest path computed on need)
1379 All these models where contributed by Silas De Munck, and are
1380 described in his ICCS09 paper.
1382 * Simplify model declaration
1383 (less redirections, less function to write when defining a model)
1384 - Factorize stuff between models:
1387 surf_model_resource_set(model)
1388 surf_model_resource_by_name(model, name)
1389 - Unify the types of models in s_surf_model_t (using an union)
1390 - Embeed fields of common_public directly into s_surf_model_t
1391 - Rename model methods:
1392 action_free ~> action_unref
1393 action_change_state ~> action_state_set
1394 action_get_state ~> action_state_get
1395 - Change model methods into functions :
1396 (model)->common_public->action_use ~> surf_action_ref
1398 * Implement a generic resource; use it as ancestor to specific ones
1399 (allows to kill duplicated code in models)
1400 Drawback: timer command don't need no name nor properties;
1401 workstation_CLM03 don't need no properties
1402 (but I guess we can live with those few bytes wasted)
1404 * Improve the action object model
1405 - implement a constructor avoiding dupplicated code about field
1406 initialization in generic_action part.
1408 * Kill the SDP model: it has an external dependency, is deprecated
1409 in flavor of modern lmm models, and didn't compile since a while
1412 * Relocation of the context module from XBT to SIMIX.
1413 (the context were decoupled from the simix processes, duplicating a lot of code)
1414 => a lot of code was factorized
1415 - less overhead is introduced during scheduling
1416 - simpler API for the context factory
1417 - the logic for process creation,destruction and manipulation was simplified
1418 * Simplification of the s_smx_process_t data structure.
1419 => accesing the simix level data associated to a process is faster now,
1420 and the code is a lot more readable.
1423 * Implement some more MPI primitives:
1424 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1425 -implementation: Bcast: flat or 2-ary tree (default),
1426 Barrier: 4-ary tree,
1428 Allreduce: Reduce then Bcast
1429 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1430 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1431 Alltoallv: flat tree, like ompi
1433 * Add support for optimized collectives (Bcast is now binomial by default)
1434 * Port smpirun and smpicc to OS X
1437 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1438 Shout out if you used it.
1441 * Display the list of still queued messages in SG mode when existing
1445 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1446 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1447 * Add xbt_str_from_file(FILE*)
1448 * Add xbt_dict_get_key achieving a linear reverse search
1449 * Remove the context module
1451 Portability report of this version:
1452 * Main portability targets:
1453 - Linux(debian)/x86/context
1454 - Linux(debian)/x86/pthreads
1455 - Linux(debian)/amd64/context
1456 - Linux(debian)/amd64/pthreads
1457 On these, we still have the eratic breakages of gras/pmm and
1458 amok/saturate_sg reported in previous version. We still think
1459 that the tests are the cause of the fault, not the tested code.
1461 - Mac OSX Leopard/x86/context
1462 Still false negative in tesh autotesting.
1463 Smpi still fails, but this time because readlink does not accept -f
1464 Everything seems to work properly beside of that.
1467 - AIX version 5.3 (only tested contexts this time)
1468 Smpi still fails there because mktemp is not installed.
1469 Everything seems to work properly beside of that.
1471 I managed to compile it for the first time, but several breakages.
1472 Won't delay the release for this exotic platform.
1474 * Windows: it's still lagging behind. If you want to help, please
1477 Timing report of this version:
1478 This version seem to be more than 5% faster than 3.3.1 (on linux
1479 64bits with contextes). The gain is less than expected, we are
1480 investigating this for next release.
1482 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1484 SimGrid (3.3.1) stable; urgency=low
1487 * Implement a --cfg-help to show existing configuration variables
1488 * Build chain do not require doxygen in maintainer mode
1491 * fix a bug on struct sizeof computation, which prevented the
1492 exchange of arrays of structs in some conditions
1493 - added a regression test about this in datadesc_usage
1494 * Allow the exchange of 0-long dynamic vectors.
1495 - for that, use -1 as indicator of dynamic size instead of 0
1496 - This implied to change any size from unsigned long to long,
1497 reducing a bit communication abilities, but I guess that with
1498 64bits being quite common, this is more than enough.
1499 - This also induce a protocol change, thus bumping network protocol
1500 version from 0 to 1 (if we have external users, we have to get
1501 clean on that point too ;)
1502 - added two regression tests about this in datadesc_usage
1503 * Be more verbose when propagating local exceptions
1504 This helps debugging.
1505 * Display the status of simulated processes when receiving SIGINT in
1509 * Allow to control the simulation from a trace file.
1510 New functions MSG_action_register() and MSG_action_trace_run()
1511 The first one allows to associate a function execution to each
1512 kind of action while the second one parses a trace file and
1513 triggers the corresponding actions within the system.
1514 For now, only a toy example is provided in examples/msg/actions
1515 * Add an exemple of process migration in examples/msg/migration
1516 * Fix a bug in task exchange which broke MSG_task_get_sender()
1517 Add a teshsuite regression test for that.
1518 [Bug: if MSG_task_get_sender() is called after sender exit,
1520 * Fix a bug which prevented suspend/resume to work properly
1521 * Display the status of simulated processes when receiving SIGINT
1522 This fixes a regression of v3.3. due to the introduction of SIMIX
1523 * Bug fixing in failure management:
1524 - trace could not start by a failure at time 0
1525 - failure during communications were not working
1528 * Add SIMIX_process_set_name() to change the name of the current
1529 process in the log messages.
1530 * Store smx_hosts in a dict since we only retrieve them by name
1531 * Move the configuration infrastructure to surf
1534 * Move the configuration infrastructure to surf
1537 * Massive internal cleanups:
1538 - Store internal structures on processes instead of hosts (allows
1539 to have more than one process per host, in addition of being more
1541 - Cleanup the initialization/finalization process
1542 - Kill a whole bunch of unneeded synchronization:
1543 processes run in exclusive manner within the simulator
1544 - Move queues from global tables to process data fields
1546 - now accept -platform and -hostfile arguments
1547 - Pass the right rank value to processes according to the hostfile
1548 * Compile the examples by default, and use them as regression tests
1549 * Implement MPI_Wtime()
1550 * Change the reference speed to a command line option
1553 * TCP_gamma can now be specified as command line option using
1554 --cfg=TCP_gamma:10000000.0
1555 * Change the --surf-path cmd line option into --cfg=path:
1558 * Also include strbuff from xbt.h public header
1559 * xbt_ex_display(): do not free the exception after displaying
1560 This allows to do more with the given exception afterward.
1561 Users should call xbt_ex_free() themselves.
1565 Portability report of this version:
1566 * Main portability targets:
1567 - Linux(debian)/x86/context
1568 - Linux(debian)/x86/pthreads
1569 - Linux(debian)/amd64/context
1570 - Linux(debian)/amd64/pthreads
1571 These targets fail about 1/10 of times on gras/pmm, but we believe
1572 that this is because of the test, not because of simgrid.
1573 amok/saturate_sg fails even more rarely, and the test may not be
1576 - Mac OSX Leopard/x86/context
1577 The test suite still spits tons of errors because some obscure
1578 force prevents us from removing the temporary directories
1579 arguing that they still contain some metadata I've never heard of.
1580 Smpi fails because seq is not installed.
1581 Everything seems to work properly beside of that.
1584 - AIX version 5.3 (both contexts and pthread)
1585 Smpi still fails there because mktemp is not installed.
1586 XML inclusions seems rosty on AIX.
1588 * Windows: it's still lagging behind. If you want to help, please
1591 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1593 SimGrid (3.3) stable; urgency=high
1597 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1598 [Malek Cherier & Mt]
1600 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1603 Until now, GRAS were using MSG as an interface to SURF. It was
1604 quite difficult because both interface have several differences
1605 (MSG channels vs GRAS sockets were the most notable point).
1607 This also opens the gate to SMPI (which should occur soon) and speed
1608 up simulations by to 40% (even if it were not the main goal).
1610 **************************************
1611 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1612 **************************************
1613 The point is that events occuring at the exact same timestamp are
1614 not scheduled in the same order with the old and new version. This
1615 may be enough to completely change the execution of simulations in
1616 some cases. Sorry for the inconvenience.
1618 * Cleanup and upgrade the XML format to push further scalability
1619 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1621 * Improve the testing infrastructure with tesh. Now a very large part of
1622 the code is tested not only by being run but also by checking that the
1623 output match an expected output [Mt].
1625 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1626 is really less memory-demanding, which should allow you to use
1627 larger files in SimGrid [AL].
1629 * Inform valgrind about our contextes, so that it becomes usable
1630 with the default (and more effecient) version of SimGrid
1631 [contributed by Sékou Diakite, many thanks]
1634 * Introduce a listener thread in charge of receiving incoming
1635 messages from the network. It allows to overlap communication and
1636 computation but most notably, it removes some stupid deadlocks due
1637 to the fact that so far, a process could not send and receive at
1638 the same time. This made most non trivial communication schema
1640 * Convert the PIDs from long int to int to match the MSG ones (and
1641 linux ones too) [Mt]
1642 * New function: gras_agent_spawn() to launch a new process on
1643 current host. Only working in simulation for now. [Mt]
1644 * New function: gras_os_hostport() returning a constant form (ie,
1645 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1648 * Make the backtrace of exceptions more human readable [Mt]
1649 * New module: xbt/str [Mt]
1650 a ton of string utility functions (split, join, printf to a newly
1651 allocated buffer, trim, etc)
1652 * New module: xbt/hash [Mt]
1653 SHA1 hashing algorithm (more to come if needed)
1654 * New module: xbt/synchro [Mt]
1655 synchronization tools (mutex and conditions) working the same way
1656 in simulation and in real life (mainly useful for GRAS, but not
1658 * New module: xbt/queue [Mt]
1659 classical producer/consumer synchronization scheme
1660 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1661 (using the classical functions will get serialized) [Mt]
1662 * Make dictionary internal table dynamic. No need to specify its size
1663 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1665 * Make sure the log channels are organized as a tree under windows
1666 (because of ANSI C compatibility issue, any channel were child of
1667 root directly) [Mt].
1670 * Cleaned many thing in surf and fixed a few bugs [AL].
1671 * Add a nice command line configuration mechanism to compose models [AL].
1672 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1673 the previous one (KCCFLN05). It relies on something that looks like
1674 a max-min sharing mechanism but cannot be written as such. A new solver
1675 was thus designed [AL].
1676 * Add a new solver to lmm. Based on Lagrange optimization and
1677 gradient-based descent, it enables to efficiently maximise systems s.a
1679 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1681 This solver enables to propose two new network models for TCP Reno and
1682 TCP Vegas based on Low's work. These models still need to be fully
1683 tested though [Pedro Velho].
1686 * Bug fix in SD_simulate. Now the time bound given as argument is
1688 * Use the new parallel task model (ptask_L07) as default.
1689 * Use the SURF command line configuration mechanism.
1690 * 0-size tasks (for synchronization) should now work.
1692 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1694 SimGrid (3.2) stable; urgency=high
1698 We still experience issues on this platform, but we believe that at
1699 least MSG is usable.
1701 GRAS API BREAKAGE (for simplification purpose, sorry):
1702 * the gras_msgtype_by_name is not used anymore. Instead of
1703 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1704 you can write (and must)
1705 gras_msg_send(toserver, "request", &request);
1706 - If you still want to pass a gras_msgtype_t to the function (to cache
1707 the type and avoid the lookup time), use the gras_msg_send_() variant.
1708 - Impacted functions:
1709 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1710 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1711 * The callbacks are now expected to return 0 when everything went well
1712 (just like the main() function)
1714 GRAS new features and improvements:
1715 * New module mecanism where user code can use per process globals [Mt]
1716 This is similar to gras_userdata_*() functions, but for libraries. It
1717 factorize some code developped over and over in the examples and AMOK.
1718 It has still to be documented and used (only amok/peermanagement is
1720 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1721 loss when they were small enough to fit into the buffer and sent quickly
1722 enough so that they can all get received in one shoot.
1723 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1724 if not found. Use the *_or_null() variant for the old semantic.
1725 * In gras_msg_handle, do not discard messages without callback.
1726 They are probably messages to be explicitly awaited later (ie, proofs of
1727 mis-synchronization in userland since they are sent before being awaited)
1728 No big deal usually.
1729 * gras_socket_meas_send/recv: semantic changed!
1730 The numerical arguments used to be (1) the total amount of data to send
1731 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1732 messages. This was need for the fool willing to send more than MAXINT
1733 bytes on quite fat pipes.
1736 * Do really rename the hostmanagement module to peermanagement. [Mt]
1737 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1738 but this is rather new and this was documented in the module
1739 documentation (poor excuses, I admit)
1740 * Bandwidth measurement semantic changed! This follows the changes to
1741 gras_socket_meas_send/recv explained above.
1744 * A sequential mode has been added to the workstations. When a workstation
1745 is in sequential mode, it can execute only one task, and the other tasks
1746 are waiting in a FIFO. [Christophe Thiery]
1749 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1750 model for SIMDAG. [Christophe Thiery]
1751 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1752 equal to 0, which caused some bad behaviors in
1753 saturated_constraint_set_update. I now use a threshold mechanism like in
1757 * When running manually src/testall, you select specific units [Mt]
1758 testall is the result of our cunit mecanism, and should replace all
1759 the scripty thingy around since bash don't run easily on billware.
1761 * A mallocator system has been added. [Christophe Thiery]
1762 Mallocators allow you to recycle your unused objects instead of freeing them
1763 and allocating new ones.
1765 Documentation update:
1766 * FAQ reworking + New FAQs:
1767 - "Valgrind spits tons of errors!" [Mt]
1768 - "How to repport bugs" [Mt]
1769 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1770 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1771 - Communication time measurement within MSG [AL]
1772 - I experience weird communication times when I change the latency [AL]
1773 * GRAS tutorial [Mt]
1775 - an introduction to the framework and to the used communication model
1776 - an initiatic tour introducing the most proheminent features:
1778 . Lesson 0: Installing GRAS
1779 . Lesson 1: Setting up your own project
1780 o Part 2: Message passing
1781 . Lesson 2: Exchanging simple messages
1782 . Lesson 3: Passing arguments to the processes (in SG)
1783 . Lesson 4: Attaching callbacks to messages
1784 . Lesson 5: Using globals in processes
1785 . Lesson 6: Logging informations properly
1786 . Lesson 7: Using internal timers
1787 . Lesson 8: Handling errors through exceptions
1788 . Lesson 9: Exchanging simple data
1789 . Lesson 10: Remote Procedure Calling (RPC)
1790 . Lesson 11: Explicitely waiting for messages
1791 . Recapping of message passing features in GRAS
1792 - A HOWTO section containing:
1793 o HOWTO design a GRAS application
1794 More are due, of course. They will come latter. In the meanwhile, you can
1795 check the examples which are still here.
1797 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1799 SimGrid (3.1) stable; urgency=high
1803 There was a stack corruption somewhere, visible only when optimizing
1804 with these versions. [Vince]
1807 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1808 module that enabled to play with Directed Acyclic Graphs. It is built
1809 directly on top of SURF and provides an API rather close to the old
1810 SG. Some old codes using SG are currently under rewrite to check that
1811 all needful functions are provided. [Christophe Thiery]
1814 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1815 extension of the classical CLM03 model that gracefully handles
1816 failures. This is now the default model for MSG and GRAS. It doesn't
1817 handle parallel tasks yet though. [AL]
1818 * Bug fix: Weights were not correctly set in the network part.
1819 WARNING: This may have resulted in incorrect results with simulations
1820 where there are more than one flow on a given link. [AL]
1823 * After a (long ?) discussion on simgrid-devel, we have decided that the
1824 convention we had on units was stupid. That is why it has been decided
1825 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1826 WARNING : This means that all previous platform files will not work as
1827 such with this version! A warning is issued to ask users to update
1829 A conversion script can be found in the contrib module of the CVS, under
1830 the name contrib/platform_generation/surfxml_update.pl [MQ]
1833 * Bug fix: Processes were started in reverse order, wrt deployment file.
1834 WARNING: if your code relies on this bug, please fix it. [AL]
1835 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1836 being executed on two different locations. [AL]
1837 * Bug fix: Failures are now better supported thanks to Derrick's tests
1838 (there was many failure situations I hadn't thought of and that weren't
1839 correctly handled). [AL]
1840 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1844 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1845 Exception occurring on server-side are propagated back to client (!).
1847 API CHANGE: the callback changed their prototype. Change:
1848 int my_handler(gras_socket_t expeditor, void *payload_data) {
1850 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1851 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1853 * New! function: gras_msg_handleall to deal with all messages arriving
1854 within a given period.
1855 * New! function: gras_socket_server_range to get a server socket in a
1856 range of port numbers (ease to avoid port number conflicts) [MQ]
1857 * New! gras processes display their backtrace when they get a SIGUSR1
1858 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1859 Sweet to debug RL processes [MQ]
1863 - Do not force experiment sizes to be expressed in kb, or it becomes
1864 impossible to measure the latency this way (needs one byte-long tests)
1865 WARNING: this changes the amok_bw_* function semantic. [MQ]
1866 - Implements the link saturation stuff. [MQ]
1867 * Peer management module:
1868 New! module factorizing code that we wrote over and over [MQ].
1871 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1872 - Test units are placed directly into the library code, they get extracted
1873 automatically and placed into the src/testall binary.
1874 - Convert most of the XBT tests to this system.
1875 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1876 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1877 live in the same C code. This required to change a little bit the API
1878 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1879 * New module: sparse graph structure with basic algorithms (this is work
1880 in progress and the API is not considered to be frozen yet). [AL]
1881 * Display more information on backtraces: source line & function names are
1882 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1883 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1884 * Reworked a little bit some #include statements to load only required
1885 headers. Some user code that relied on SimGrid to include stdlib or
1886 stdio may need to include it by themselves. [AL]
1887 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1889 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1890 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1891 but I guess that almost nobody used those parts. [MQ]
1893 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1895 SimGrid (3.0.1) stable; urgency=low
1898 * Unfortunately, I had missed 5 misnamed functions:
1899 xbt_fifo_item_t xbt_fifo_newitem(void);
1900 void xbt_fifo_freeitem(xbt_fifo_item_t);
1901 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1902 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1903 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1904 They're now deprecated. Please use their new versions:
1905 xbt_fifo_item_t xbt_fifo_new_item(void);
1906 void xbt_fifo_free_item(xbt_fifo_item_t);
1907 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1908 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1909 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1911 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1912 * Documentation: xbt_log module unmercifully reworked [MQ]
1913 * Bugfix: there was a problem with the ending of contexts with
1914 the pthread backend. It caused some weird deadlock or behavior
1915 depending on the pthread implementation. [AL]
1916 * Bugfix: get the exceptions raised in the simulator repport where
1917 and why they come from when they are not catched in time [AL, MQ]
1920 * Bugfix: Do repport the error when two non-connected hosts try to
1921 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1924 * Add additionnal checkings on communications. Assert that two
1925 communicating hosts are connected by a set of links... [AL]
1928 * Add additionnal checkings on channel values in communication [AL]
1929 * New: MSG_task_get_source to see on which host a task was generated [HC]
1930 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1931 the number of tasks waiting to be received on channel and sent
1933 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1934 waits for the first task coming from a given host.. [AL]
1936 GRAS new functionnalities: [MQ]
1937 * Enhance the parsing macro to allow the size of multidimentional objects
1938 to be given thru annotations.
1939 * New example (and documentation): Matrix Multiplication a la RPC
1940 (as when I was young!) and fix a bunch of bugs found on the way.
1942 GRAS performance improvements: [MQ]
1944 * Reduce the amount of cbps creation/destruction by making it static to
1945 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1948 * Change libdata to a set so that we can search for stuff by ID (and thus
1949 reduce the insane amount of dict lookups)
1952 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1953 RL communication on top of SG since it's so uneffective.
1954 It may also be used for inter-thread communication in RL, one day.
1955 * Use gras_datadesc_copy() to exchange messages on top of SG
1957 - improve message exchange performance on top of SG
1958 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1959 * Don't exchange on the network the size of the used part of buffer,
1960 instead, specify the possible buffer size to read().
1962 - reduces the amount of read/write calls (one pair per exchange)
1963 - reduces the amount of exchanged data (the size)
1964 - allows to retrieve all arrived data on receiver side, if we don't need
1965 it right now (subsequent read will peek the buffer)
1966 - allows the receiver to proceed with the begining of the stream before
1967 everything is arrived
1968 - make it possible to build an iov transport (using readv/writev)
1970 - take care of the data with non-stable storage (like stacked data),
1972 * If possible, TCP send uses vector I/O (when writev() is here)
1973 - Don't use it for receive since we send data sizes and data on the
1974 same stream, so we wouldn't be able to chain large amount of chunks
1975 before having to flush the stuff to read the size.
1976 * Rework the transport plugin mecanism to simplify it and reduce the
1977 amount of pointer dereferencement when searching for the right function
1980 * I guess that now, we do almost as few system calls as possible while
1981 doing as few data copy as possible.
1983 To improve it further, we could try to send all the sizes first and then
1984 all the data (to use iov on receiving size), but it's only a partial
1985 solution: when you have 2 dimensional data, the sizes of the second
1986 dimension is data of the first dimension, so you need 3 streams.
1988 I'm not sure the potential performance gains justify the coding burden.
1990 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1992 SimGrid (3.00) stable; urgency=high
1995 * New! Give the possibility to hijack the surf parser and thus bypass
1996 MSG_create_environment and MSG_launch_application. Have a look at
1997 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1999 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2001 SimGrid (2.96) unstable; urgency=low
2006 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2007 This deprecates the xbt_error_t mecanisms.
2008 It modifies (simplifies) all XBT and GRAS API.
2009 MSG API keeps unchanged (exceptions raised by XBT are catched from
2010 within MSG and masked with existing error handling facilities)
2013 * New! Add a FATPIPE model. [AL]
2014 * New! Add a parallel task model. [AL]
2015 * New! Add automatically a loopback interface (in the default
2016 network model) if none was precised.
2019 * Bugfix: MSG_process_resume now works with the current running process.
2021 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2022 * Modification of MSG_task_get_compute_duration. Once a task has been
2023 processed, the value returned by this function is now equal to 0. [AL]
2024 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2025 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2026 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2027 of a m_task. For now, MSG_task_cancel only works with computation
2029 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2030 of the processor (in Mflop/s) regardless of the current load on the
2032 * API Change: use proper naming convention for MSG_getClock and
2033 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2035 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2036 This function changes the priority of a computation task. This priority
2037 doesn't affect the transfer rate. A priority of 2 will make a task
2038 receive two times more cpu power than the other ones. This function
2039 has been added to suit the needs of Nguyen The Loc and hasn't been that
2040 much tested yet. So if it fails, please report it and send me your code.
2042 * API Change: removed all functions and types that were marked "deprecated"
2043 since many months. Renamed MSG_global_init_args to MSG_global_init.
2045 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2047 SimGrid (2.95) unstable; urgency=low
2050 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2051 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2052 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2055 * Bugfix: Allow absolute paths for platform description files [MQ]
2056 * Bugfix: do free the variables after use. Leads to drastic performance
2058 * Implement max_duration (ie, timeouts) on resources [AL]
2061 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2063 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2064 the needed support (timeouts) [AL]
2067 * Implement measurement sockets. You can now get the bandwidth between two
2068 hosts thanks to AMOK (see below). [MQ]
2069 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2070 dynar over the network (yeah) [MQ]
2071 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2072 * simple/static token-ring example. [Alexandre Colucci and MQ]
2073 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2074 in place before [MQ]
2077 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2078 type parsing. "short" and "long" modifiers were also ignored; other
2079 modifier (such as reference level) are still ignored. [MQ]
2080 * Embeed the buffer size within the buffer itself on SG. [MQ]
2081 That way, send() are atomic and cannot get intermixed anymore (at least
2082 the ones which are less than 100k; bigger messages still have the issue)
2083 * Array size pushed by the field, not by the field type (or each
2084 and every long int will push stuff to the cbps) [MQ]
2085 * use select() to sleep since it allows to portably sleep less than one
2088 GRAS (minor cleanups)
2089 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2092 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2093 * type category 'ignored' killed as it was never used and were difficult
2095 * whether a type can cycle or not is now a flag, leaving room for more
2096 flags (as "ignored", if we feel the need one day ;)
2097 * Rename raw sockets to measurement sockets since "raw" has another
2098 meaning in networking community.
2101 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2102 and offers features not belonging to GRAS but that most applications
2103 need. One day, it may be a set of plugins loadable at runtime.
2104 * New module: bandwidth
2105 bandwidth measurement between arbitrary nodes running this module. [MQ]
2107 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2109 SimGrid (2.94) unstable; urgency=low
2111 The first beta release of SimGrid 3 !
2115 * Update the main page and the FAQ. Adding references to gforge.
2118 * Add a gras_os_getpid function.
2121 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2122 * Extend the logs so that they also print PID, hostname, date, ... if
2124 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2125 and kill the old version which were in testsuite/
2126 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2130 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2131 for users. Instead, each programming environment (GRAS, MSG,...) use it
2132 when needed to provide such a feature to users.
2133 Don't shortcut the mecanism or you will also shortcut the virtualization
2134 you need on the simulator.
2138 * Cleanups in configury with regard to compile optimization/warning flags.
2139 Also add -fno-loop-optimize to any powerpc since it's the optimization
2140 killing gcc (< 3.4.0).
2141 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2142 needed by MSG examples complications
2143 * Borrow configury beautifications from PHP
2146 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2147 warning (thanks loris for stumbling into this one).
2148 * Bugfix: stop loading private headers (gras_config.h) from the public
2152 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2153 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2154 RL would imply a DNS resolver, which is *hard* to do in a portable way
2155 (and therefore delayed).
2156 * Implement a real timer mecanism and use it in timing macros. This allows
2157 to avoid rounding errors and get a 0.000005 sec precision in timing
2158 macros. While I was at it, various cleanups:
2159 - allow to declare more than one timed section per file (fix a stupid bug)
2160 - move some private declaration to the right place
2161 - merge conditional execution and timing macros into emulation module
2162 - document the module
2163 - make sure the module cleanups its mess on gras_exit
2164 * Documentation improvements:
2165 - (new) how to compile applications using GRAS
2166 - (new) emulation support (timing macros)
2168 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2170 SimGrid (2.93) unstable; urgency=low
2172 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2175 - Use Paje properly where used. Still to be sanitized properly.
2176 - Portability fix: Add an implementation of the contexts using pthread
2180 - Add xbt_procname(): returns the name of the current process.
2181 Use it to show the current process's name in all logging.
2183 - fix detection of older flex version and the reaction, since we do
2184 depend on modern ones (we use lex_destroy)
2185 - Better separation of SG and RL in the libs: remove all simulation code
2186 from libgras. As a result, this lib is now only 200k when stripped.
2187 Some of the xbt modules may also be duplicated (two sets and such) and
2188 should be cleaned/killed before SG3.
2189 - Insist on using xlC on AIX because of weird problems involving gcc there.
2190 - Cleanup the make remote stuff. This is now done by scripts
2191 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2192 mainly for our private use, but we're working on changing them to user
2195 - Bugfix: flush the socket on close only if there is some *output*.
2196 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2198 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2201 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2203 SimGrid (2.92) unstable; urgency=low
2205 Alpha 3 on the path to SimGrid 3
2209 - New! First try of benchmarking macros.
2210 - New! First try so that gras_stub_generator generate deployment and
2211 remote compilation helpers.
2213 - Bugfix: Initialization fix in msg_test.
2217 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2219 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2221 - API Change: gras_os_sleep to take the amount of seconds as a double.
2222 Accepting an int was error prone since it was the only location where
2223 seconds were coded as such. It leaded to damn rounding errors.
2224 - Bugfix: Hard to belive that timers ever worked before this.
2226 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2228 SimGrid (2.91) unstable; urgency=low
2230 Alpha 2 on the path to SimGrid 3
2234 - Bug fix in the lmm_solver.
2236 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2237 through the function MSG_paje_output.
2238 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2239 - It is possible to bound the rate of a communication in MSG with
2240 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2241 to write it in the changelog).
2242 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2247 - Various cleanups to the autotools stuff
2248 - Begin to move Gras examples to examples/gras/
2249 - Let make distcheck work again (yeah!)
2251 - documentation overhauled using doxygen.
2252 gtk-doc-tools is dead in SimGrid now.
2253 - Automatically extract all existing logging categories, and add the list
2254 to the documentation (long standing one, to say the less)
2256 - Cleanup the known architecture table. Reorder the entries to group what
2257 should be, and use a more consistent naming scheme.
2258 (some of the test dataset are still to be regenerated)
2259 - New! Allow library to register globals on each process just as userdata
2261 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2262 lookup time (for now).
2263 Use it in msg and trp.
2264 This cleans a lot the internals and helps enforcing privacy of the
2265 headers between the gras components.
2266 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2267 - Bugfix: gras_os_time was delirious in RL.
2268 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2270 - Reenable GRAS now that it works.
2272 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2274 SimGrid (2.90) unstable; urgency=low
2276 Alpha 1 on the path to SimGrid 3
2278 * It is a long time since the last release of SimGrid. I'm sorry about
2279 that but as I had told you, I was rewriting a lot of things. I apologize
2280 to those who had been reporting bugs to me and that I had not answered.
2281 If your bug is still in the new version, please tell me. Here is a
2282 summary of the main changes.
2284 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2285 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2286 portability, speed, and a lot more but you'll figure it out later.
2287 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2288 I wanted to release the new MSG as soon as possible and I have
2289 broken GRAS, which is the reason why, for now, only MSG is fully
2290 functional. A laconic description of these projects is available
2291 in the documentation.
2293 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2294 kernel optimized for our needs (called SURF but only the developers
2295 should use it). Hence, MSG is now roughly 30 times faster and I think
2296 that by rewriting a little bit MSG, I could event speed it up a little
2297 bit more. Beside the gain in speed, it is also much easier to encode a
2298 new platform model with SURF than it was with SG. More to come...
2300 * REVOLUTION 3: I have tried to change a little as possible the API of
2301 MSG but a few things really had to disappear. The main differences
2302 with the previous version are :
2303 1) no more m_links_t and the corresponding functions. Platforms are
2304 directly read from a XML description and cannot be hard-coded
2305 anymore. The same format is used for application deployment
2306 description. The new format is described in the documentation.
2307 Have a look in tools/platform_generation. There is a tiny script
2308 that converts from the old platform format to the new one. Concerning
2309 the application deployment format, parsing the old one is tricky.
2310 I think most of you should however be able to convert your files. If
2311 it is really an issue, I can write a C code that does the conversion.
2313 2) the toolbox tbx does not exist anymore. We now have a library
2314 with much more data-structures but without the hash-tables (we have
2315 dictionaries that are much faster).
2317 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2319 *****************************************************************************
2320 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2321 * don't feel like converting the oldies. (Mt) *
2322 *****************************************************************************
2325 Version 2.90: "the long awaited one"
2326 - Finished rewriting and debugging MSG. Rewrote the documentation.
2327 - disable GRAS for now since it needs to be ported to the newest SG
2330 - Finish the port to windows (using mingw32 for cross-compile)
2333 - Main loop and datastructures of SURF. A cpu resource object is
2334 functional. Surf can thus be used to create cpu's with variable
2335 performance on which you can execute some actions.
2337 2004-11-15 Martin Quinson
2338 - Port to ARM. Simply added the alignment and size descriptions. Should
2339 work, but the ARM machines are so slow that I didn't had the opportunity
2340 to 'make check' over there yet.
2342 2004-11-15 Arnaud Legrand
2343 - Trace manager now written. It uses a heap structure and is therefore
2344 expected to be efficient. It may however be speeded up (particularly
2345 when many events occur at the same date) by using red and black
2346 trees. One day maybe...
2347 - Max-min linear system solver written. It uses a sparse matrix
2348 structure taking advantage of its expected use. Most operations are
2349 O(1) and free/calloc are called as few as possible. The computation of
2350 the minimum could however be improved by using a red and black tree
2353 2004-11-03 Arnaud Legrand
2354 - Rename every gras_* function that was in xbt/ to its xbt_
2356 - Add a heap and a doubly-linked list to xbt
2357 - Added a dichotomy to the dictionaries. make check works as well before
2358 so I assume that the patch is correct. I do not know however if things
2359 run effectively faster than before now. :)
2361 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2362 SimGrid, and the version number is bumped to 2.x
2364 2004-10-29 Martin Quinson
2365 - Introduction of the remote errors.
2366 They are the result of a RMI/RPC on the remote machine.
2367 ErrCodes being scalar values, you can't get the host on which those
2368 errors did happen. Extending the error mechanism as in Gnome is possible.
2369 No idea yet whether it is a good idea.
2371 2004-10-28 Martin Quinson
2372 - Interface revolution: the Starred Structure Eradication.
2373 I used to do typedef struct {} toto_t; and then handle *toto_t.
2374 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2375 - ???_t is a valid type (builded with typedef)
2376 - s_toto_t is a structure (access to fields with .)
2377 - s_toto is a structure needing 'struct' keyword to be used
2378 - e_toto_t is an enum
2379 - toto_t is an 'object' (struct*)
2381 typedef struct s_toto {} s_toto_t, *toto_t;
2382 typedef enum {} e_toto_t;
2383 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2384 s_toto_t) is private.
2386 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2387 so that it can set the variable to NULL. It was so for dicts and sets,
2388 it changed for dynars.
2390 - Fix a bunch of memleaks in dict_remove
2391 - Fix a bug in sg/server_socket opening: it failed all the time.
2393 2004-10-07 Martin Quinson
2394 - Speed up dynar lookup operation a bit.
2396 gras_dynar_get is dead.
2398 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2399 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2400 (which gives you the address of the stored data).
2402 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2403 copy of the data using an affectation to do the job and not a memcpy.
2405 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2406 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2408 It does not really speedup the dynar test because they are
2409 setting elements all the time (and look them seldom). But the dict does
2410 far more lookup than setting.
2412 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2414 2004-10-05 Martin Quinson
2415 - Allow to (en/dis)able the cycle detection at run time.
2417 Whether we should check for cycle or not is now a property of each
2418 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2419 datadesc_cycle_unset allow to remove this property when previously set.
2421 Note that the cycle detection is off by default since it impacts the
2422 performance. Watch the data you feed GRAS with ;)
2424 This property is hereditary. Any element embedded in a structure having it
2425 set have it set for the time of this data exchange.
2427 You should set it both on sender and receiver side. If you don't set it on
2428 sender side, it will enter an endless loop. If you forget on receiver
2429 side, the cycles won't be recreated after communication.
2431 - Header reorganization.
2432 Kill gras_private.h, each submodule must load the headers it needs.
2434 2004-10-04 Martin Quinson
2435 - Interface revolution: do not try to survive to malloc failure.
2437 Now, gras_malloc and friends call gras_abort() on failure.
2438 As a conclusion, malloc_error is not a valid error anymore, and all
2439 functions for which it was the only gras_error_t return value are
2440 changed. They now return void, or there result directly.
2441 This simplify the API a lot.
2443 2004-09-29 Martin Quinson
2444 - Re-enable raw sockets.
2445 Created by gras_socket_{client,server}_ext;
2446 Used with gras_raw_{send,recv}
2449 It should allow to kill the last bits of gras first version soon.
2451 This is not completely satisfactory yet (duplicate code with
2452 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2455 - Simplify transport plugin (internal) interface by not passing any
2456 argument to _server and _client, but embedding them in the socket
2459 2004-09-28 Martin Quinson
2460 - Finish the port to AIX.
2461 autoconf was my problem (segfault within the malloc replacement
2462 function. No idea why)
2464 2004-09-16 Martin Quinson
2465 - Fix some size_t madness on 64bit architectures.
2467 2004-09-08 Martin Quinson
2468 - Reduce the number of system headers loaded, overload some more system
2469 calls (such as malloc to cast the result of the system one, and work
2471 - Fix and reintroduce the config support
2473 2004-09-07 Martin Quinson
2474 - Source code reorganization to allow Arnaud to surf all over there.
2475 - Allow to document the logging categories.
2476 - Remove all uppercase from logging categories and useless cleanup in names.
2478 2004-08-18 Martin Quinson
2479 Version 0.6.2 (protocol not changed; API changed)
2480 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2481 gras_error_t), and NULL when not found. Functions expecting a msgtype
2482 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2483 hopefully usefull message.
2484 - Portability to prehistoric sparcs again
2486 2004-08-17 Martin Quinson
2487 Version 0.6.1 (protocol not changed; ABI not changed)
2488 - prealloc some buffers to speed things up
2490 2004-08-11 Martin Quinson
2491 Version 0.6 (protocol not changed; ABI expended)
2492 - The parsing macro can deal with the references, provided that you add
2493 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2495 2004-08-09 Martin Quinson
2496 Version 0.5 (protocol not changed; ABI changed)
2497 - Allow to off turn the cycle detection code in data exchange at
2498 compilation time. It should be at run time, but I'm short of time (and
2499 the config stuff is still broken). That way, we keep dict out of the
2500 critical path, which is good because the performance is poor:
2501 - search not dichotomial yet
2502 - dynar give no way to access their content and memcpy everytime
2503 - In composed data description (struct, ref and so on), stop foolness of
2504 keeping the subtype's ID, but store the type itself. This keeps sets out
2505 of the critical path, which is good since they rely on dynar and
2506 dictionnaries. The only loose of that is that we cannot detect the
2507 redeclaration of a structure/union with another content (but I'm not sure
2508 the code detected well this error before anyway). We still can detect
2509 the redefinition discrepancy for the other types.
2510 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2511 it breaks the code because of type-punning used all over the place).
2512 This breaks on all non-gcc architectures (for now).
2514 All those changes (plus the buffer of last time) allow me to gain 2 order
2515 of magnitude on cruel tests consisting of 800000 array of integers on two
2516 level of a hierarchical structure (200 secondes -> 4 secondes)
2519 - the selector of reference must now return the type it points to, not
2520 the ID of this type.
2522 2004-08-06 Martin Quinson
2523 Version 0.4 (protocol changed; ABI not changed)
2524 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2525 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2526 - Implement a buffer transport, and use it by default (it relies on tcp in
2527 real life and on sg in simulation).
2528 That's a bit hackish since I had a new field to the structure to store
2529 its data without interfering with the subtype ones. Inheritance
2530 is tricky in C. And that's a kind of reverse inheritance with one class
2531 derivating two classes. Or maybe a game with java interfaces. Anyway,
2532 that's damn hard in C (at least).
2533 Moreover, I got tired while trying to ensure plugin separation and
2534 genericity in SG mode. MSG wants me to do weird things, so let's go for
2535 cruel hacks (temporarily of course ;).
2536 See comment in transport_private.h:71
2537 - do not include all the _interface headers in private but in the files
2538 which really need them (to cut the compilation time when they are
2541 2004-07-26 Martin Quinson
2542 Version 0.3 (protocol not changed; ABI changed)
2543 - Major overhault of the datadesc interface to simplify it:
2544 - shorted the function names:
2545 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2546 - add a trivial way to push/pop integers into the cbps without malloc.
2547 This allows to make really generic sub_type description, which simply
2548 pop their size of the stack.
2549 - add a function gras_datadesc_ref_pop_arr() which does what users want
2550 most of the time: Declare a dynamic array (which pops its size of the
2551 stack) and declare a reference to it. Poor name, but anyway.
2552 - kill the post-send callback, add a post-receive one
2554 2004-07-23 Martin Quinson
2555 Version 0.2 (protocol changed; ABI changed)
2556 - add some testing for cpbs in the test cases, and fix some more bugs.
2557 This invalidate again the little64 data file, since I cannot regenerate
2559 - remove an awfull optimization in the logging stuff, allowing me to:
2560 - understand it again
2561 - learn gcc how to check that the argument match the provided format
2562 - fix all errors revealed by gcc after that
2563 - internal keys of dict are not \0 terminated. Deal with it properly in
2564 loggings instead of segfaulting when the user want to see the logs :-/
2566 2004-07-22 Martin Quinson
2567 - Fix some stupid bug preventing cbps (callback postit) from working
2569 2004-07-21 Martin Quinson
2570 - Some documentation cleanups
2571 - remove the useless last argument of msgtype_declare
2572 - rename the Virtu functions to fit into the 'os' namespace
2573 - move headers src/include -> src/include/gras/ and stop fooling with
2574 gras -> . symbolic link
2575 - make distcheck is now successful
2577 2004-07-19 Martin Quinson
2579 - Build shared library also
2580 - Install html doc to the right location
2581 - stop removing maintainer files in make clean
2582 - build tests only on make check
2584 2004-07-13 Martin Quinson
2586 - No major issue in previous version => change versionning schema
2587 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2588 corresponding dataset.
2590 2004-07-11 Martin Quinson
2592 - Get it working with any kind of structure (we can compute the padding
2593 bytes remotely for all the architectures I have access to)
2594 - Implement the structure parsing macro (still not quite robust/complete)
2595 - Improvement to the remote testing toysuite
2597 2004-07-10 Martin Quinson
2598 [autoconf mechanism]
2599 - get ride of a bunch of deprecated macros
2600 - actually run the test for two-compliment, not only compile it :-/
2601 - test whether the structures get packed (and bail out if yes. Damn.
2602 Alignment is a serious matter)
2603 - test whether the structures get compacted (but respecting the alignment
2604 constraints of each types)
2605 - test whether the array fields of structures can straddle alignment boundaries
2607 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2608 (took me 2 hours to find that bug, looking at the wrong place)
2609 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2610 before sending/receiving and are used to compute the offsets of fields
2611 - Given that padding size depend even on compiler options, keep track of
2612 alignment and aligned_size only for the current architecture. Not a big
2613 deal since we send structure fields one after the other (seems
2615 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2616 program, let it work on linux/gcc/little32. portability todo.
2618 2004-07-08 Martin Quinson
2619 - import and improve remote compilation support from FAST
2620 - make sure make check works on half a dozen of machines out there
2622 2004-07-07 Martin Quinson
2623 Let's say it's version 0.0.3 ;)
2624 - Implement conversions (yuhu!)
2625 - Let it work on solaris (beside conversion, of course)
2626 - Stupid me, using rand() to generate the conversion datatests in not wise.
2628 2004-07-06 Martin Quinson
2629 - Let make dist work, since I'm gonna need it to compile on remote hosts
2630 - Let Tests/datadesc_usage write the architecture on which the file was
2631 generated as first byte.
2632 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2633 architecture descriptions.
2634 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2635 on those architectures.
2636 - Optimization: send/recv array of scalar in one shoot
2638 2004-07-05 Martin Quinson
2639 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2641 - Plug a whole bunch of memleaks
2642 - each process now have to call gras_{init,exit}. One day, their log
2643 settings will be separated
2644 - Continue the code factorisation between SG, RL and common in Transport.
2646 2004-07-04 Martin Quinson
2648 - Redistribution between SG and RL.
2649 We wanna have to accept in SG, so move accepted related parts of RL in
2650 the common part. (more precisely, the dynar of all known sockets is no
2651 more a static in transport.c, but part of the process_data)
2653 [gras_stub_generator]
2654 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2656 2004-07-03 Martin Quinson
2657 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2658 them all up in one shot)
2660 - Fix the ugly case of reference to dynamic array.
2661 - New (semi-public) function gras_datadesc_size to allow the messaging
2662 layer to malloc the needed space for the buffer.
2664 - gras_socket_close now expect the socket to close (and not its address to
2665 put NULL in it after it). This is because the socket passed to handlers
2666 is one of their argument (=> not writable).
2668 - propagate the interface cleanup from last week in datadesc, ie remove a
2669 superfluous level of indirection. User pass adress of variable
2670 containing data (both when sending and receiving), and not of a variable
2671 being a pointer to the data. Let's say that I like it better ;)
2672 The price for that is constructs like "int msg=*(int*)payload" in
2673 handlers, but it's a fine price, IMHO.
2675 - Let it work in RL (yuhu)
2677 2004-06-21 Martin Quinson
2679 - porting SG plugin and SG select to new standards (works almost).
2680 - plug memleaks and fix bugs around.
2683 - cleanup the prototype of data recv and force users to specify when they
2684 want to handle references to objects. Test case working even for cycles.
2685 - plug memleaks. Valgrind is perfectly ok with this.
2687 2004-06-12 Martin Quinson
2689 - cleanup the separation between plugin and main code in plugin creation
2691 2004-06-11 Martin Quinson
2693 - Reput hook for raw sockets, needed for BW experiments
2694 - kill a few lines of dead code
2695 [Data description] Interface cleanup
2696 - gras_datadesc_by_name returns the searched type or NULL.
2697 That way, no variable is needed to use a type desc once, which makes
2699 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2700 two parameters were strings (field name, type name), leading to
2702 [Dicos] Interface cleanup
2703 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2704 This is consistant with the dynar API.
2706 2004-04-21 Martin Quinson
2708 - Porting to new standards.
2710 - interface cleanup.
2711 There is no bag anymore, no need to take extra provision to mask the
2712 pointers behind "ID".
2713 Better splitup of functions between files create/exchange/convert.
2714 This is still a bit artificial since convert and receive are so
2715 interleaved, but anyway.
2717 - add a queued message list to procdata (the ones not matching criteria
2719 - factorize some more code between SG and RL wrt procdata
2721 - use gras_exit in example to track memleaks
2722 - get rid of gs_example now that GS is properly integrated into gras
2723 - update run_test to integrate the latest tests (datadesc)
2725 - rename WARNINGn macros to WARNn since it prooved error-prone
2727 2004-04-19 Martin Quinson
2729 - register init/exit functions within gras module mechanism
2730 - send/receive function.
2731 Convertion is not implemented, but short-cutted if not needed.
2732 struct/array elements are sent one by one (instead of block-wise), but
2733 nobody really cares (yet). Get a prototype before optimizing.
2734 - tests (using a file socket) for DD send/receive on:
2735 - base types: int, float
2736 - array: fixed size, string (ie ref to dynamic string)
2737 - structure: homogeneous, heterogeneous
2738 - chained list, graph with cycle
2739 Believe it or not, valgrind is not too unhappy with the results. The
2740 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2744 - Bugfix when using a filename explicitely (instead of '-')
2746 2004-04-09 Martin Quinson
2748 - factorize more code between RL and SG in socket creation
2749 - Complete the implementation and tests of:
2751 o file (only in RL, and mainly for debugging)
2753 I lost 3 days to design a portable address resolver, and then decided
2754 that the prototype mainly have to run on my box.
2755 Addressing portability too early may be like optimizing too early :-/
2757 - use gras_init in the Tests instead of the crappy parse_log_opt
2758 (the latter function is removed)
2759 [Conditional execution]
2760 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2761 [Code reorganisation]
2762 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2763 Build examples against the RL library, since there is no way to disable
2764 its creation for now.
2766 For information, the beginning of coding on GRAS was back in june
2767 2003. I guess that every line has been rewritten at least twice since