1 SimGrid (3.11) NOT RELEASED; urgency=low
3 TENTATIVE RELEASE GOALS for 3.11:
4 * Consider the removal of Supernovae mode if no user has manifested since the
5 release of version 3.10.
6 * Switch to tesh.pl, and kill the now unused parts of xbt that seem fragile
7 * Switch to surf++, and reintegrate the hypervisor branch on top of it
8 * Kill support for 32bits host architecture.
9 - Integers are too short to store a date or a size
10 - Plus, less platforms ~> simpler codebase ~> easier to maintain
11 * Clean up CMake files (may need a full rewrite).
12 Non exhaustive list of subgoals:
13 - Use genuine cmake mechanisms and variables when available,
14 instead of reinventing the wheel.
15 - Correctly determine system and architecture (e.g. x32).
16 - Correctly determine compiler type and version (e.g. clang).
17 - Correctly set compiler flags according to compiler type and version.
18 - Correctly set compiler flags for C++, Java, and Fortran compilers too.
19 - Use git to generate the dist archive. Either use git-archive to
20 generate the tarball, or keep using cmake -E tar, but use git-ls-files
21 to get the list of files to pack.
23 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
25 SimGrid (3.10) stable; urgency=low
27 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
30 * Reintegrate Java to the main archive as desynchronizing these
31 package is not acceptable anymore (Java is now considered stable)
32 * Add explicit synchronization facilities through semaphores
33 * Bug fix: Task.setDataSize() only changed the C world, not the value
34 cached in the Java world
37 * Dramatically change the way files are handled. API and internals changed, but
38 this part of MSG was not considered as production grade either.
39 * Add explicit synchronization facilities through semaphores
40 * Add a new function MSG_host_get_process_list()
41 * Preliminary DVFS support (see examples/msg/energy/ for details)
44 * SMPI is now included directly in the libsimgrid as the windows
45 linker doesn't force us on splitting it anymore.
46 * Improvements of the SMPI replay tool:
47 - Most of the collective communications are now rooted in the same process as
48 in the original application.
49 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
50 used until now). Multiple data types can now be used in a trace.
51 - The replay tool now supports traces produce either by TAU or a modified
53 - Bug Fix: the compute part of the reduce action is now taken into account.
54 - Gatherv collective is now supported
55 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
56 of smpirun outputs time independent traces for the current run. One file
57 is created per process. If too many processes are simulated, this behavior
58 can be changed to one file for all processes by using the
59 tracing/smpi/format/ti_one_file flag
60 * smpirun generates the host file if needed (with given host count and platform)
61 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
62 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
63 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
64 --cfg=smpi/coll_selector:(mpich/ompi)
65 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
67 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
68 happening inside a collective SMPI call.
69 * Fix the behavior of complex data types handling.
70 * Make MPI_Wtime another synchronization point to take computations into account.
71 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
72 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
73 * Add all missing Fortran bindings, SMPI should work with Fortran 90
74 (no privatization of global variables yet)
75 * Preliminary DVFS support (see examples/smpi/energy/ for details)
78 * Verification of liveness properties is now available for SMPI applications
79 (in addition to MSG applications)
80 * Bugged examples using SMPI in examples/smpi/mc/
81 * Add --cfg=model-check/visited option. Allows the verification of infinite
82 programs. Detection of loops in the execution thanks to the system state
83 comparison and reduction of the state space to explore. Can be combined with
84 DPOR for safety properties.
87 * Allow to change SimGrid configuration (see --help) within the code
88 thanks to SD_config() as it can be done in MSG.
89 * Add a new function SD_task_set_amount() upon user request.
92 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
93 * Remove rule based routing (no more PCRE dependency)
94 * Add a limiter_link option to cluster tag, to specify a maximum reachable
95 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
96 * Add a loopback_bw and loopback_lat options to cluster tag.
97 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
98 routing. Now peers are encapsulated in an AS and have their own private
99 router but this is transparent.
102 * Our own implementation of getline is renamed xbt_getline, and gets
103 used even if the OS provide a getline(). This should reduce the
104 configuration complexity by using the same code on all platforms.
105 * New type: xbt_cfg_elm_boolean.
106 * Allow to use yes/no for boolean configuration options in the command line.
107 * Allow to disable SimGrid cleanups at exit from command line option.
108 There are situations where one may want a simulation to end with an exit.
109 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
110 annoying when scripting around the simulator. Adding a
111 --cfg=clean_atexit:no allows to circumvent this issue.
114 * Lots of memory leaks were corrected in this release.
115 There is no known memory leaks anymore, in all of our 600+ tests.
116 * New command line option --version, to get SimGrid version information.
117 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
119 * Supernovae builds are deprecated, and expected to be removed in the next
122 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
124 SimGrid (3.9) stable; urgency=low
126 The Grasgory release: GRAS is really dead now.
128 * Complete overhaul of the internal host structures scheme.
131 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
132 was considered as experimental and badly maintained since then.
133 * Keeping it was thus a trap to our potential users, that could take
134 it instead of MSG or SMPI by mistake despite is pity state.
135 * GRAS seems to have very few users (if any), and no one volunteered
136 to maintain it further. It also induces a lot of XBT code (for
137 portability sake), that must be maintained too.
138 * For all these reasons, we killed GRAS. If someone wants to revive it
139 in the future, don't cry, our git history still remembers of GRAS.
142 * Major overhaul. Merge our documentation again as time proved that
143 splitting it was really not helping our users.
144 * Further improve the developer documentation to help newcomers
145 hacking on SimGrid itself. The user documentation (and in
146 particular, the beginner documentation) is still in a sorry state.
149 * Now works on Windows too!
150 * Much more extensive test suite, from MPICH
153 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
154 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
155 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
156 * Bug fix: let task be scheduled when the last dependency to be solved is
157 a control dependency.
158 * Remove SD_load_environment_script function.
159 Use the C sg_platf function if you want to declare a platform
163 * New function: MSG_process_get_number()
164 * Old function documented: MSG_config()
165 * Remove MSG_load_platform_script function
166 Use the C sg_platf function if you want to declare a platform
170 * Change the default value of the TCP_gamma constant (maximal size of TCP
171 congestion window) to a more realistic 4MiB value. If you notice changes in
172 your simulation results, you can fall back to the previous 20k tiny window
173 by adding --cfg=network/TCP_gamma:20000 on command line.
174 * (Hopefully) fix a bug wrt periodic availability and state traces
175 * Bug fix: use default values at start when first event in availability/state
176 trace is not at time 0.
179 * remove the "new_" part of function name sg_platf_new_trace_connect
180 (resulting in sg_platf_trace_connect), since it does not create
184 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
185 use case where it's really mandatory, and maintaining it was a pain in
187 * New: xbt_fifo_search(), search an item with a user-provided
188 comparison function instead of dumb pointer comparison.
191 * Fix the lua deployment:
192 Use `simgrid.init_application()` before deployment instead of
193 `simgrid.msg_register_application()` after.
196 * Transfer the tracing files into the corresponding modules.
198 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
200 SimGrid (3.8.1) stable; urgency=low
202 The "we are told that some people want to also *install* the simgrid
205 * Add missing file "tesh.1" to the archive.
207 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
209 SimGrid (3.8) stable; urgency=low
211 The Psssshiiiit release: SimGrid jumps into the Cloud.
214 * Add an experimental interface to manipulate VMs. They are mainly
215 process groups with very few intrinsic semantic, but they should
216 allow you to build the semantic you want easily.
217 * New function: MSG_host_set_property_value()
218 * New function: MSG_process_on_exit(). To clean memory in all cases.
219 * Bug fixes that made the host (and link) failures unusable.
220 * Add a way to auto-restart process when the host in which they are
221 executing comes back (ON_FAILURE="RESTART" on deployment file,
222 MSG_process_auto_restart_set).
223 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
224 please stop using the old ones, they are DEPRECATED.
226 * Deprecate functions MSG_global_init() / MSG_global_init_args()
227 Please use MSG_init() instead. (reducing the amount of entry
228 points in the library helps us).
229 * Make it impossible to link against the wrong version of the lib
230 * Deprecate MSG_clean(). No need to call it anymore.
231 * Function MSG_get_host_number() is not deprecated anymore.
234 * Split the doc into a user guide and a reference guide.
235 * Start a developper guide to help people hacking on SimGrid.
238 * Enable tracing by default. This modules rocks you should use it.
239 * Remove option custom_flags. Now use environment variables CFLAGS
241 * Use default cmake things to detect lua instead of home grown ones.
242 * New option "enable_mallocators" to disable mallocators, for debugging
243 purpose ("on" by default).
246 * Bug fixes around the resource failures: don't let the processes
247 survive the host they are running onto.
248 * Add an interface to auto-restart processes when the host in which they are
249 executing comes back.
250 * Ensures that SIMIX_clean is called automatically. It's not part of
251 the public interface anymore (bindings should be updated).
254 * Bug fix for when SD_Simulate is called with a positive value: be careful
255 when comparing doubles. Sometimes they are different for non significant
257 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
258 parallel task whose initial work is distributed among host according
259 to the Amdahl's law. Such tasks are created with a parameter alpha
260 that corresponds to the non-parallelizable part of the computation.
261 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
262 between two sets of workstations assuming a 1D block distribution (each
263 workstation owns a similar share of data) on both sides.
265 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
266 Data redistribution will be automatically scheduled once parent and child
267 are both scheduled. The filling of computation_amount and
268 communication_amount structures is now done seamlessly thanks to the chosen
270 * New function SD_workstation_dump to display various information
271 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
272 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
273 bandwidth on the route onto which the task is scheduled and the amount of
275 To divide the nominal bandwidth by 2, the rate then has to be :
276 rate = bandwidth/(2*amount)
277 * Compute tasks that have failed can now be rescheduled and executed again
278 (from their beginning)
279 * Increasing source code coverage (src/simdag is now covered at 95.8%
283 * Re-implement time-independent trace replay using SMPI (at the
284 smpi_smp_* level) instead of MSG. This should replace
285 examples/msg/actions/actions.c
286 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
287 hindexed and structs)
288 * Implement the exchange of non-contiguous data.
289 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
290 * Correct behavior of smpi/sender_gap and set its default value to 0
291 * Add option to asynchronously send small messages to allow better
292 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
293 specifies the size in bytes under which messages will be asynchronously sent.
294 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
295 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
298 * New C interface to define a platform: XML is now optional.
299 For more info, please check include/simgrid/platf.h.
300 * New interface to define random platforms from the C:
301 For more info, please check include/simgrid/platf_generator.h and
302 examples/msg/masterslave/masterslave_platfgen.c
303 * Export a sg_cmdline dynar containing all the arguments we got from
307 * Two new tracing options for adding comments to trace file so you
308 can track your experiments (see --help-tracing for details).
309 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
310 * Adding the SimGrid version that generated the trace file as a comment.
311 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
312 * Fix to avoid key clashes on Paje links
313 * Other minor fixes related to the Paje specification
316 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
317 and renamed to xbt_str_hash() and xbt_str_hash_ext().
318 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
319 * Greatly improve the robustness of mmalloc to user errors (such as
320 using an area after freeing it, or freeing it twice)
322 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
324 SimGrid (3.7.1) stable; urgency=low
327 * Restore the prototype of MSG_process_create_with_environment() to
328 the pre-3.7 situation by removing the kill_time argument.
329 * Add a MSG_process_set_kill_time() function instead.
332 * Fix weird behaviors when dealing with parallel tasks.
335 * Simgrid is now built as a dll.
336 * Simgrid-java now works on Windows.
337 * Simgrid-Java is now included into Windows package.
340 * First pre-build package for MacOSX.
343 * Fix compilation when using MSG_USE_DEPRECATED.
344 * Fix some compilation issues on Macs and Windows.
345 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
346 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
348 We discovered that the Lua console is broken, but we are missing the
349 manpower to fix it right now. The problem existed in 3.7 too, so we
350 are not blocking the release for that. Sorry if you depended on this
351 feature, any help would be really welcome.
353 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
355 SimGrid (3.7) stable; urgency=low
357 The "spring cleanups (before the next Big Project kicks in)" release.
360 * We can specify the SMPI latency/bandwidth factor with command line
361 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
362 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
363 You can also use the "config tag" from platform file by adding this line
364 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
365 (see "example/platforms/tag_config.xml" to use "config tag").
366 Note that the command line supersedes the platform file configuration.
367 * Change the correction factors used in LMM model, according to
368 the latest experiments described in INRIA RR-7821.
369 Accuracy should be improved this way.
370 * Use the partial invalidation optimization by default for the
371 network too. Should produce the exact same results, only faster.
372 * Major cleanup in surf to merge models and split some optimization
373 mechanisms from the core of the models. As a result you can now
374 specify which model to use (e.g., --cfg=network/model:LV08
375 --cfg=cpu/model:Cas01) and which optimization mode to use
376 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
377 Incompatible combinations should err at initialization. See
378 --help-models for the list of all models and optimization modes.
379 * The CLM03 workstation model was dropped for simplicity because it
380 used the deprecated CM02 network model. Use default instead.
381 * Rename the TCP_gamma configuration option to network/TCP_gamma
382 * Rename the coordinates configuration option to
383 network/coordinates, and document it
384 * Use now crosstraffic keyword instead of the terribly misleading
385 fullduplex keyword. It is activated by default now in the current
386 default model, use --cfg=network/crosstraffic:0 to turn it off.
387 * Ongoing refactoring the model parsing to make XML files optional
388 See include/simgrid/platf.h for details (still to be completed)
391 * Major overhaul of the documentation. Almost instructive now :/
392 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
393 functions and friends. This interface was considered as
394 deprecated since over 2 years, it's time to inform our users that it is.
395 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
396 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
397 use these (crappy) functions in your code.
398 These functions will be removed soon. Stop using them now.
399 * Deprecate MSG_get_host_{table,number}
400 Implement MSG_hosts_as_dynar() instead.
401 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
402 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
405 * Stabilize the parallel execution mode of user contexts
406 * Introduce configuration variables to control parallel execution:
407 - contexts/synchro: Synchronization mode to use when running
408 contexts in parallel (either futex, posix or busy_wait)
409 - contexts/parallel_threshold: Minimal number of user contexts
410 that must be part of a scheduling round to switch to parallel
411 execution mode (raw contexts only)
412 * Fix bugs that prevented to use suspend/resume along with
413 synchronization structures.
414 * Fix bugs in process termination that lead to invalid memory access
415 in very specific conditions.
418 * Introduce a parallel mode for the models (controlled by surf/nthreads
419 configuration item). In our tests, running the models in parallel
420 never lead to any speedups because they are so fast that the gain
421 of computing each model in parallel does not amortizes the
422 synchronization costs, even when ultra fast futexes are used.
423 This is released anyway because YMMV.
426 * Performance boost by using a swag internally to compute the set of
427 tasks that are finished and should constitute the return value of
431 * Enable it by default now that it is considered rather stable.
434 * Documentation of the tracing functions.
435 * Performance gains when tracing categorized/uncategorized resource
436 utilization by avoiding calls to get route when updating resource
437 variables. LMM constraints are being used instead.
438 * API changed to set task categories. Use MSG_task_set_category instead
439 of TRACE_msg_set_task_category, and SD_task_set_category instead
440 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
441 * Bugfix for graphicator, routes not correctly obtained, memory leaks
442 * Examples for link user variables added (see at examples/msg/tracing/)
443 * Deprecated function TRACE_msg_set_process_category completely removed
444 * Trace header updated according to the latest Paje file format
445 * Tracing network lazy updates, no longer obligate users to use full updates
446 * --cfg=tracing/platform:1 also registers power/bandwidth variables
447 * Experimental: let user code declare/set/push/pop application states for hosts
448 * API changed to allow the manual creation of graph configuration files
449 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
452 * Improve the API of Lua MSG bindings, using the Lua spirit.
453 * Each simulated process now lives in its own Lua world (globals are
454 automatically duplicated). It helps writing simulators. Will allow
455 to run Splay programs within SimGrid in the future.
456 * Add a Chord example in Lua, equivalent to the MSG one.
459 * Start the implementation of a solution to express temporal
460 properties, not only local assertions. This is still an
461 experimental work in progress, stay clear from it to be safe.
465 - Add new runtime parameters --help-logs and --help-log-categories
466 to display informations about supported logging parameters and
468 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
469 don't exists anymore.
470 * Mallocators: allow value NULL for the reset function.
472 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
473 create homogeneous dictionaries, where all the elements share the
474 same free function. Non homogeneous dictionaries will be
475 deprecated in the next release.
476 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
477 - Multi-level dictionaries are deprecated.
479 - new function xbt_dynar_search_or_negative() that is useful when
480 you have less than 2 million elements in your dynar and don't
481 want of the extra complexity of catching exceptions when the
482 element is not found.
484 - Make xbt_os_thread module (for thread portability) public.
485 Documentation is still to come, sorry.
487 - Cleanups and simplifications to make it maintainable again.
488 - Exotic features (such as memalign and valloc) were removed.
489 - The metadata were extended and improved so that the
490 model-checker becomes able to explore and inspect the heaps.
491 - This may induce a performance drop when enable_model-checking is
492 ON in cmake (even if it's not used in the simulation), but it is
493 necessary at this point to get MC working.
495 Turn model-checking OFF if simulation performance matters to you.
496 Not enabling it at runtime is not enough, disable it in cmake.
498 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
500 SimGrid (3.6.2) stable; urgency=low
502 The "Not coding new stuff allows to polish old things" release.
505 * New bindings to the NS3 packet level simulator (experimental)
506 * Use the raw (efficient) execution contexts instead of the sysv
507 (portable) ones when possible.
508 * libpcre is now mandatory in any cases since not using it led to
509 severe performance loss and possibly other issues
510 * Update the XML platforms:
511 - G5K: include the latest machine in Nancy
512 - GridPP and LCG: new platforms
513 * Documentation was partially updated, at least (more to come)
515 Bug fixes, cosmetics and small improvements
516 * Free terminated processes before the end of the simulation to avoid
517 exhausting the memory of users having very dynamic amount of
519 * Bug fix and cosmetics about canceling non-running tasks
520 * Bug fix about the dot loader's issues when using libcgraph
523 * Create an installer for windows with nsis (amd64 and win32)
524 - Add an hello world project to illustrate simgrid project creation.
525 - Embed libpcre into the Simgrid installer to avoid
526 its compilation burden
527 * The raw execution contexts should work on Apple now
528 * Port to Windows 64 bits
529 - Sysv contexts now have an implementation for this arch
530 - GRAS communication features now support this arch
531 * Drop support for borland compiler on windows
532 - this code was not maintained, and we kinda depend on gcc nowadays
533 * Fix portability issues on kfreebsd/gnu: build error about semaphores
534 * Fix portability issue on unstable ubuntu: linker became picky on
537 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
539 SimGrid (3.6.1) stable; urgency=low
541 The "Oops, we broke Macs too" release
544 * Fixed contexts detection so that raw ones are used when possible
545 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
546 leads to a strange error, with user code segfaulting sometimes when
547 the generated code is not perfectly aligned (which is not
548 controllable from the user side, depends on the amount of code)
551 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
553 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
555 SimGrid (3.6) stable; urgency=medium
557 The Summer Release, also known as the "OMG! They Killed Kenny!" version
560 * Bindings now constitute their own package, separated from the main one.
561 Rationale: reduce our maintenance nightmare by reducing the module coupling
562 They will soon be released on their own on gforge.
564 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
565 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
567 GRAS: It is not considered as stable anymore, but experimental. Sorry.
568 * It's not quite deprecated for now because we have no replacement,
569 but it may soon become the case.
572 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
573 * Fortran: New user-level cache variable to store the rank of the running
574 process. This improves performance by an order of magnitude.
575 * C: New coccinelle script to automatically locate and modify global and
576 local static variables.
577 * Improved SMPI network model with a sender-side gap to account for multiple
581 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
582 only say if a communication is finished, no matter whether it succeeded or
583 failed. You can call MSG_comm_get_status() to know the status of a finished
585 * New function MSG_task_dsend() to send a task and detach it. When a
586 communication is detached, you are never notified of its success or failure
587 and the memory is released automatically once it is finished. This function
588 is useful when you don't care about the end nor the success of a
590 * Change the prototypes of action replay. Sorry for inconvenience,
591 but this is really more efficient this way (and to adapt your code,
592 you just have to fix the initialization, that shouldn't be too long)
593 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
594 ever managed to do anything useful with it.
595 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
596 takes a dynar of communications. It returns immediately and gives the
597 index of a finished communication (if any).
598 * New example: a basic implementation of the Chord P2P algorithm.
601 * New model for multi-core CPUs. You can now use the core attribute to
602 precise the number of cores of a host. This is a basic model. Every
603 process running on the host receives at most the power provided in
604 the DTD (throughput<=power). Total throughput of process cannot exceed
606 * New peer tag. This peer tag creates a tiny AS comprising a host and a
607 router linked by an up-link and a down-link (possibly asymmetrical).
608 This kind of pattern allows to easily build last-mile model style platforms.
609 Aggregating such patterns in a rule-based AS is thus the technique of
610 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
611 * New model for Vivaldi routing. We transformed the Vivaldi network model
612 into a Vivaldi routing model (based on the rule-based model). This allows to
613 combine Vivaldi based latencies with last-mile platforms.
616 * Added a check for NaN of IEEE754 infinite in the double entries of
618 * Introduce a new context factory "raw", highly inspirated from the
619 ucontext factory, but using manually crafted functions in assembly to
620 do the work in an efficient manner.
621 * Allow to change the used context factory at run time, not only at
622 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
623 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
624 contexts at runtime (only with raw contexts or ucontexts).
625 * Completely rewrote this module to allow parallel execution of user
626 processes. Use --cfg=contexts/nthreads:N to execute user processes
627 with N parallel threads (the default is 1, meaning no parallelism).
628 * Allow to decide dynamically between sequential and parallel modes.
629 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
630 processes in parallel only when their number is greater than or equal to P
632 * Added a check for NaN of IEEE754 infinite in the double entries of
636 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
637 won't output the state of processes when interrupted with Ctrl-C
638 * Add a new function xbt_dynar_to_array that transforms a dynar into a
639 NULL-terminated array. This may solve backward compatibility issues
640 due to the change to return type of SD_simulate. See also:
641 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
642 * Add new macros with variable number of arguments.
643 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
644 - in xbt/asserts.h: xbt_assert
645 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
646 - in xbt/ex.h: THROWF and RETHROWF.
647 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
649 * Change xbt_die() to accept a format string with arguments, just like printf.
650 * New data structure: xbt_lib_t, like a dict but more general and with better
654 * New configuration options
655 Options triva/categorized and triva/uncategorized can be used to generate
656 graph configuration files for Triva visualization tool.
657 * Configuration option tracing/platform is renamed to tracing/categorized
658 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
659 * New TRACE_user_link_variable interface:
660 User provides the name of the link and the tracing variable to attach to it
661 * the declaration of tracing categories must be done after the environment
663 * simpler tracing interface, just one way to declare categories
664 TRACE_category or TRACE_category_with_color, it is up to you
665 * links in the trace file are again identified by their names
666 * trace contains the full platform hierarchy exactly as declared using the ASes
667 * Options tracing/msg/[task|process]:1 groups the process by hosts
668 for both cases, tasks and processes must have names that are unique during the simulation
669 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
670 * The experimental option tracing/msg/volume is deprecated
671 its functionality may be reincorporated if needed
673 The tracing generates a trace file with unordered timestamped events,
674 because of the way the core simulator (surf) works. A script available
675 at the tools directory (fix-paje-trace.sh) can be used to put the events
676 in order. We have changed the tracing so it can generate ordered timestamped
677 events in the final trace, but depending on the simulator (and how much time
678 is simulated) that can lead to a huge memory utilization. It is deactivated
679 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
682 * Define a SIMGRID_VERSION macro in simgrid_config.h.
683 - We are trying hard to keep the API stable, but it may happen that
684 some things change (we're a research project after all, not a
685 nuclear plant operating system). If such things should happen, you
686 could rely on that macro to adapt.
687 - current value: 30600 for 3.06.00, aka 3.6
688 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
689 a number that can be compared with SIMGRID_VERSION.
690 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
691 assertions and verbose/debug logging events are disabled at compile time.
693 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
695 SimGrid (3.5) stable; urgency=medium
698 * New feature to any SimGrid-based simulator: Model-Checking
699 Check SIN#1 for more details.
702 * New Model SMPI (three-interval linear regression for correction factors)
703 See RR-7426, available at http://hal.inria.fr/inria-00527150
704 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
705 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
706 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
707 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
708 * See also src/smpi/README
713 - Tracing API changes: TRACE_start and TRACE_end should not be called
714 by user-code. They are automatically called by simulators created
715 with SimDAG, MSG and SMPI if the toolkit is compiled with
716 tracing_enabled=ON. Categories declaration and utilization remain the
717 same for MSG and SimDag.
718 - A function was added to the tracing API to declare categories with
720 - TRACE_category_with_color (char *category, char *color)
721 where color must be in the following format
722 "%f %f %f", red, green, blue
723 and red, green, blue are float values in the interval [0, 1]
724 - User can specify NULL as color parameter, or continue calling
726 On that case, the tracing system will define random colors
727 - The following command-line options are supported:
728 --cfg=tracing/filename:msg.trace
729 --cfg=tracing:1 (activate tracing, needed to use others)
730 --cfg=tracing/platform:1 (categorized resource use)
731 --cfg=tracing/uncategorized:1 (uncategorized resource use)
732 --cfg=tracing/msg/task:1 (task creation)
733 --cfg=tracing/msg/process:1 (process creation, migration)
734 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
735 --cfg=tracing/smpi:1 (SMPI interface tracing)
736 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
737 - examples of examples/msg/tracing updated
739 - DAXLoader and DOTLoader functions can generate tasks with categories
740 - A new function to attribute a category to SD tasks:
741 TRACE_sd_set_task_category (SD_task_t task, char *category)
742 * Tracing the MPI interface implemented by SMPI
743 - Collective operations are traced with states
744 - Point-to-Point operations are traced with states/links
745 - Tracing activated by a parameter "-trace filename" passed
746 to smpirun during execution (considering that simgrid
747 is compiled with tracing enabled)
748 - To run the simulation with gdb, the simulator
749 accepts --cfg=tracing/smpi:1 to trace SMPI
750 - tesh files to check if smpi tracing is ok
751 - See examples/smpi/NAS/DT-trace
752 * GTNetS tracing re-worked
753 - adaptation to the tracing system of GTNets to cope
754 with modifications regarding the fullduplex mode
755 - new tesh files to check if gtnets tracing is ok
758 * Asynchronous communications through the functions:
759 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
760 * New function: MSG_load_platform_script()
761 to make possible using a lua script instead of XML files to set up platforms
762 * New function: MSG_set_function
763 to associate functions to processes, used when bypassing the parser
764 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
766 Platforms: Add some more examples in examples/platforms
767 * Grid'5000: see www.grid5000.fr
768 * *_30000_hosts.xml: various huge files [mainly scalability testing]
771 * Change the XML format. This is a very important modification. SimGrid
772 3.5 introduces a new hierarchical format based on the notion of
773 Autonomous Systems. Compatibility with old format is ensured
774 through the perl script provided in the install bin directory
775 bin/simgrid_update_xml.
776 It is now possible to build platforms with specific routing
777 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
778 connect several platforms together. We will try to provide soon
779 set of realistic platforms exploiting these properties (have a
780 look at examples/platforms/ for the moment).
781 * Take the opportunity of the XML format change to be a good XML citizen:
782 rename link:ctn to link_ctn and similar changes (also dealt with by
784 * Add a new routing scheme (rule-based) using regular expressions. It
785 enables to have an extremely low memory footprint when the
786 underlying routing is simple and can be compactly described. You
787 need to have libpcre4-dev (perl regular expressions) installed if
788 you want to use this routing scheme.
789 * Revive the cluster TAG and allow to easily and efficiently (both in
790 term of memory and speed) connect clusters together. Have a look
791 at teshsuite/simdag/platforms/ to see how this can be done. With
792 this tag, you can create clusters with thousands of tasks at no
793 cost (have a look at examples/platforms/).
794 Note: clusters are implemented as ASes, so there is no need for an
795 enclosing AS tag if you have only one cluster in your platform.
796 * Add new generic functions in the public interface that allows the user
797 to call SURF 'create_resource' methods from your code (same
798 functionality as the XML bypass mechanism but with a much lighter
800 * Add a new model (enabled through command line --cfg=network/model:SMPI)
801 that uses a piecewise linear approximation to produce better
802 results when exchanging small messages.
803 * Add a new parameter to handle correctly full duplex link and account
804 for interferences between uplink and downlink communications
805 (activate with --cfg=fullduplex:1).
808 * Rename the SD_READY (all dependencies are satisfied and task is
809 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
810 dependencies are satisfied) state.
811 This prevents a confusion between the notion of "ready to schedule"
812 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
813 simulated" (SD_RUNNABLE) used by the simulation kernel.
814 * Change the way a task is considered as ready. Instead of removing
815 dependencies when a task is done, a counter is decreased. This way,
816 it is always possible to reach ancestors thanks to the
817 SD_taks_get_parents function (even after the end of the simulation.)
818 * Change the return type of SD_Simulate from (SD_task_t*) into
819 xbt_dynar_t. This function was in handling a dynar internally and
820 converted it into a NULL terminated array for historical reasons.
821 * New function SD_dotload(char*) to load a DAG described in dot
822 format. This loader and the corresponding examples require the
823 installation of the graphviz library.
824 * Fix a bug in the management of tasks of size 0 in the surf network
825 models. This problem was only visible with SIMDAG and you should
826 thus disregard results produced with earlier versions if you
827 relied on this feature (some tasks were blocked because of this).
828 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
829 even though your applications were doing only point-to-point
830 communications and sequential computations. Now you can really use any
831 model you want (of course, if you create real parallel tasks, which are
832 not implemented in most models beside ptaskL07, this will abort).
833 * Add an example that schedules a DAX on an heterogeneous platform
834 using a Min-Min strategy.
835 * New function SD_workstation_get_current_task() that returns the kind
836 of task currently running on a workstation in the sequential access
838 * Raise some warnings when unexecuted tasks remains at the end of the
839 simulation. This is usually caused by cycles in the DAG.
842 * New function: SIMIX_process_set_function() called by MSG_set_function
843 * Change the underlying waiting queue in semaphores so that a process
844 can wait on several of them simultaneously (as in waitany).
845 * Fix the way to handle tokens in semaphores so that all access patterns
846 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
847 * kill the dirty pimple SIMIX_message_sizes_output()
848 Please use (proper) visualization instead
851 * New data container: setset (set of sets of elements)
852 * New module: mmalloc (mapped malloc, allowing to have several
853 independent segments of malloc)
854 * New function: xbt_dict_cursor_set_data()
855 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
856 * New function: xbt_dynar_is_empty()
857 * New function: xbt_fifo_get_last_item()
858 * Fix xbt_dynar_shrink(): use the right element size.
859 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
860 uninitialized areas during expand.
861 * Fix semaphores: previous implementation was severely broken.
862 * Use library init/fini functions for our initialization.
863 - you can use logs and other feature as soon as you want in your
864 code (even before the xbt_init / MSG_init)
865 - xbt_exit is now a no-op and produce a warning when used.
868 * Port GRAS to new SIMIX mechanisms. This allows gras users to
869 benefit from the latest improvement to the simulation kernel.
870 * Kill measurement sockets for now. If you rely on them, sorry. This
871 release is not for you. This feature will be reintroduced in the
872 future, but we cannot delay the release any further.
873 * New function: gras_msgtype_get_name().
874 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
875 * Fix (at last) the pmm example: it should not randomly fail anymore.
877 Build chain: bug fixes and overall polishing
878 * Cmake is now stable enough. Hence, we killed the autotools.
879 * Port to windows ( TM :)
880 * Fix the 'make install' target.
881 No need to use 'make install-simgrid' anymore
882 * Introduce a 'make dist' target compiling a *source* archive
883 'make package' compiles a binary archive
884 * Compile java files only on need
885 * Add --cd and --setenv command line options to tesh
886 * Out of source builds are not fully supported yet, but we are close
887 * Enable supernovae and optimization flags by default for our users
890 * Add layer to set up environment directly from lua, without XML.
891 * The effect of gras_stub_generator can be achieved through
892 lua too (check examples/gras/console/ping_generator.lua)
894 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
896 SimGrid (3.4.1) stable; urgency=low
898 The "Polishing easter eggs is probably a good idea" release.
899 This is a bug fixes release only.
903 * Fix a bug preventing the tasks from begin garbage collected.
906 * Fix a bug occuring when a host involved in a communication fails.
907 This was not detected properly by the other peer involved in the
908 communication. Now, it's reported as a network error.
911 * Warn the user about loop dependencies in data flow of DAX files
912 * Obey the control-flow dependencies of DAX files
915 * Add option "enable_smpi" allowing to not compile SMPI.
916 Probably useful for the (Mac) users experiencing a build error here
917 * Improve the detection of lua5.1 and ruby1.8
919 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
921 SimGrid (3.4) stable; urgency=low
923 The "Easter in Cargese" release. Also known as (major changes):
925 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
926 ~> bindings were greatly improved
927 ~> new tracing infrastructure for better visualization introduced
929 * the "Welcome to configury modernity" release.
930 ~> we switched from autotools to cmake, and improved our cdash
933 A more detailled list of changes follow (full detail in svn log).
935 Java Bindings: Various Cleanups
936 * (install java-gcj-compat-dev on debian-like to use them)
937 * Remove put/get: no need to export deprecated interface in Java
938 Use send/receive instead.
939 * Cleanup the examples and add a README per directory
940 * Remove example autoDestination (that's the only way to go now)
941 * Remove example explicitDestination (was a plain copy of basic)
942 * Make JniException a runtime exception, so that there is no need to
943 declare the fact that you may encounter such a beast. I guess that
944 nobody will ever want to survive such error.
945 * Create specific errors for each MSG case of failure:
946 host failure, transfer failure, timeout, task cancelled
947 * Cleanup the exceptions that may get thrown by each function
948 * Other internal cleanups in Java bindings. Performance still bad :/
949 Ruby and Lua Bindings: create them
950 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
951 * That's new and great, you should try them out.
952 Same functionalities than Java bindings, only even less polished
954 * Kill the useless "rate" argument of SD_task_get_execution_time()
955 Everyone used to provide -1 as a value, it was not used, and the
956 semantic of a possible use wasn't even clear.
957 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
958 either as comm costs or compute costs to mean that there is no
959 such thing for that specific task.
960 * Add a SD_task_set_name() function
961 * Fix SD_task_unschedule() on typed tasks
962 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
963 * In DAX loader, accept useless 'level' attributes to <job> since
964 LIGO DAGs have them (seem to be to ease graphical representation).
966 * Add an example masterslave_mailbox.c using send/receive and not
967 the deprecated put/get interface.
968 * Kill the MSG_paje_output() function. It's a noop since 2 years.
969 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
971 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
972 (declare MSG_USE_DEPRECATED to still have the old name)
973 * Add a MSG_task_set_data() function
974 * About trace replay (see examples/msg/actions):
976 - Allow to work with splitted trace files for each process
977 Give the specific trace file as argument of each process,
978 and call MSG_action_trace_run(NULL)
979 You can still have one merged file for all processes.
980 - Fix implementation of collective operations
981 * Allow task_execute() on 0-sized tasks (closes #10063)
983 * This is the first release of SimGrid where SMPI is not considered
984 beta anymore (even if some corners should still be improved)
985 * Port over the new SIMIX_network submodule (internal refactoring)
986 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
987 * Implement more missing elements of the standard:
989 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
991 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
992 MPI_Scan() MPI_Get_processor_name()
993 - Added implementation of missing case for Alltoall (warning: it's
994 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
995 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
996 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
997 SMPI_MPI_Allgatherv()
999 - MPI_Waitsome() was broken
1000 - Allow relative includes in smpicc
1001 - Command line cfg argument 'reference_speed' was ignored...
1002 - Some functions did not properly lead to auto-benching of user code
1003 - smpicc passes -O2 by default (just like openmpi one)
1005 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1006 * Bug fixes about timeouts during communications
1007 * add SIMIX_message_sizes_output() as a pimple to write to file the
1008 amount of messages per size. Use gnuplot to get histogram.
1009 Pimple because that's the only user-visible function of simix,
1010 defined directly in xbt.h (irk, sorry)
1012 - Add a SIMIX_sem_get_capacity() function
1013 - Fix interactions with processe resume/suspende
1014 - release_forever() was stupidly broken
1015 - Fix SIMIX_display_process_status() for processes in a semaphore
1016 - Make SIMIX_sem_block_onto() user-visible
1017 * Refactoring context stuff:
1018 - Use pseudo-OOP for better modularity
1019 - reimplement SIMIX_process_kill() without process_schedule() so
1020 that the latter can take as invariant that it is called from
1022 - Merge context_start into context_new for sake of simplicity
1024 * Add a Vivaldi network model, coded live during SUD'10 ;)
1025 * Rename configuration variables to start a hierarchy:
1026 o cpu_model -> cpu/model
1027 o network_model -> network/model
1028 o workstation_model -> workstation/model
1029 * New configuration variables:
1030 o network/bandwidth_factor: correction to bandwith
1031 o network/latency_factor: correction to latency
1032 o netwotk/weight_S: correction to the weight of competing streams
1033 * Add a long description to the models, that users can see with such
1034 argument on the command line: --cfg=cpu/model:help
1035 * --help-models display the long description of all known models
1037 * config: add the ability to set a default value after registration
1038 Does not override any previously set value (e.g. from cmd line)
1039 * dict: allow to have integer key and data.
1040 When so, you need to use the following functions
1041 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1042 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1043 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1044 In contrary to regular dicts, the key is not malloced before copy.
1045 Mixing scalar and regular elements in the same dict is not tested
1047 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1048 Tracing for Visualization:
1049 * SimGrid is now instrumented in order to generate a trace file for
1050 visualization analysis: to use it, need to compile SimGrid with the
1051 "tracing" option enabled, and instrument the program using SimGrid with
1052 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1053 (among other functions).
1054 * The instrumentation only traces the platform utilization for now
1055 * Documentation to use the tracing functions and how to analyze the
1056 traces with the Triva tool is written.
1057 * More information about: SimGrid FAQ (in the section Tracing Simulations
1060 * We moved to cmake as default build system. Autotools support will
1061 be dropped soon. Check the FAQ for more info about how to use it.
1062 * Greatly improved our cdash/ctest interactions
1063 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1064 * Added memory checking tests with valgrind; lot of memleak fixing.
1065 This may be the first release of simgrid with so few memory issues
1066 * Added code coverage tests.
1067 Our coverage is still improvable, but at least we see it on cdash.
1069 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1071 SimGrid (3.3.4) stable; urgency=low
1073 The "Desktop Grid needs love too" release (also called Xmas release).
1075 Models improvements:
1076 * Major speedup in the maxmin system solving by using lazy evaluation
1077 Instead of solving completely the maxmin system at each iteration,
1078 only invalidate (and recompute) the modified parts.
1079 This new feature is enabled in default models but you can try to
1080 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1081 * Cas01 IMproved as default CPU model
1082 This CPU model is the same Cas01 model, but it uses the
1083 maxmin-selective-update flag and a heap structure to manage
1084 actions on SURF kernel.
1085 It reduces the complexity to find the next action to finish and,
1086 consequently, it's faster than the old Cas01.
1087 This is the new default CPU model (Cas01).
1088 * Rename the old Cas01 model to Cas01_fullupdate
1089 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1090 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1091 * CpuTI (CPU Trace Integration)
1092 A new CPU model whose objective is simulate faster when using
1093 availability trace files.
1094 Instead of using a full featured, over engineered maxmin system for
1095 CPU modeling, this model does the pre-integration of traces files
1096 to calculate the amount of CPU power available, and so, executes
1097 faster than the old CPU models.
1098 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1099 * Use LV08 as default network model since it gives better accuracy
1100 for small messages and shouldn't change things for big ones.
1101 Use --cfg=network_model:CM02 to get the previous behavior.
1104 ******************************************
1105 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1106 ******************************************
1107 * The new CPU model may changes simulations!
1108 The point is that events occurring at the exact same timestamp
1109 are not scheduled in the same order with the old and new
1110 version. This may be enough to completely change the execution
1111 of simulations in some cases.
1112 * The new network model will change simulations!
1113 This new model is more realistic than the previous one, so you
1114 should consider redoing your old experiments with this model.
1115 Sorry for the inconvenience.
1118 * Introduce the supernovae compilation mode
1119 When compiled that way, the whole SimGrid (or almost) is put in a
1120 single compilation unit and compiled in one shoot.
1121 This is to help gcc which has difficulties to inline stuff from one
1123 The speedup seem to be above 15%, althrough more tests are needed on
1124 amd64 to confirm that gain.
1127 * Port of MSG's mailbox on top of SIMIX network
1128 The put/get mechanism was greatly simplified on the way.
1131 * New SIMIX network module. Provides:
1132 - Mailbox: rendez-vous mecanism to find with who you want to speak
1133 - Synchronous send/recv: easier and hopefully faster since the
1134 logic is handled in the maestro process directly now
1135 - Asynchronous send/recv: you dreamt of it? It's here now
1136 Too bad that nobody cared enough to propagate the change to MSG.
1137 * Add semaphores as SIMIX synchronization mechanism.
1140 * new function SD_daxload(char*) to load a DAX file
1141 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1142 * Introduce typed tasks. Specify its kind and cost at creation.
1143 At scheduling, just give where it should be placed, and the cost
1144 for each involved resource is automatically computed.
1145 Existing constructors so far (more to come of course):
1146 - SD_task_create_comm_e2e() for end-to-end communication
1147 - SD_task_create_comp_seq() for sequential computation
1148 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1149 * new function SD_task_dump() for debuging display
1150 * new function SD_task_dotty(task,FILE*) writing to file the info
1151 about the task in dotty format
1152 * SD_task_dependency_exists() can now cope with having one of its
1153 arguments NULL. If so, it tests whether the other argument has any
1155 * Add getters on list of preceding/following tasks:
1156 SD_task_get_parents(task) and SD_task_get_children(task)
1157 * Add getters on amount of workstations and list:
1158 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1159 * Add getter on task kind: SD_task_get_kind(task)
1160 * Update the start_time and finish_time of tasks on completion/failure
1161 * Bugfix: Remove task from state swags when destroyed
1164 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1167 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1168 dynar with one separate thread per value of the dynar.
1169 * Change the prototype of xbt_thread_create(), sorry.
1170 Added a boolean parameter indicating whether we want to join this
1171 thread (used in SG only for now)
1172 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1175 * GTNetS wrappers should now be usable again (and betterly tested too)
1176 * Fix a major regression from 3.2 where the timeout provided to
1177 MSG_task_put_with_timeout() was used as absolute time before which
1178 the comm should be done.
1179 * Start to fix the <cluster> tag.
1180 - Internal links should be good now (beside of the loopback, which
1181 use the private link instead)
1182 - paths to the external world is still rather broken
1183 - the <route:multi> tag is just broken. Actually that's brain-dead.
1184 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1185 to make it less stupid
1186 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1187 * Fix a source-level compatibility glitch from 3.2: after defining
1188 MSG_USE_DEPRECATED, you can use the old name
1189 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1190 * Allow to compile from the SVN with automake 1.11
1191 * Fix some problems when using the "start_time" tag in deployment XMLs.
1192 * Fix #8569: XBT/synchro.h has redundant declarations
1193 * Fix #8563: MSG return values and exceptions
1194 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1195 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1196 Thanks to Robson Peixoto.
1197 * Fix a vicious bug in dictionaries inducing that some elements were
1198 not freed on xbt_dict_free()
1200 Portability report of this version:
1201 * Main portability targets:
1202 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1203 on (amd64/i386/ia64)
1204 - mac leopard on i386
1205 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1206 but nothing critical.
1207 * Other platforms: windows, AIX and others were not tested for this release
1209 Timing report of this version:
1210 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1211 scenario parameters). From 8h to a few seconds in desktop grid settings.
1212 * Supernovae brings about 25% speedup on i386.
1214 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1216 SimGrid (3.3.3) stable; urgency=low
1218 The "Need for Speed" release.
1220 The timings done to validate the 3.3.2 were faulty.
1221 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1223 The problem was a conversion from a manually handled vector to
1224 xbt_dynar_t on the critical path.
1225 xbt_dynar_foreach calls functions, inducing stack management crap.
1227 We inlined these functions and xbt_dynar_foreach is now breath taking.
1228 We also inlined xbt_swag_belong on the way.
1230 Here are some approximate speedup measurements (on master/slaves
1231 simulations lasting between 10s and 20s each):
1232 3.3.1 -> 3.3.2: about same performance
1233 3.3.2 -> 3.3.3: 40% speedup
1234 3.3.1 -> 3.3.3: 40% speedup
1235 3.3.1 with inline patch -> 3.3.3: 30% speedup
1237 Our reading is that the refactoring which occurred in 3.3.2 made us
1238 suffer much more from the xbt_dynar_foreach low performance, but
1239 once we solved this, this refactoring proved to be very performance
1240 effective. From the 40% speedup, somehow, 10% are due to the
1241 inlining and 30% to the refactoring.
1243 That's a pitty that gcc cannot inline functions placed in other files
1244 alone. We have to choose between:
1245 - break the encapsulation (by putting private data structures and
1246 accessors in headers files to help gcc)
1247 - live with low performance
1248 - switch to a decent compiler such as icc (not quite possible).
1250 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1252 SimGrid (3.3.2) stable; urgency=low
1254 The "Simplicity does not preceed complexity, but follows it" release.
1256 The main contributors of this release were (lexical order):
1257 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1260 * Extract the routing logic into its own object.
1261 (was dupplicated in network.c and workstation_LV07.c;
1262 Allows to implement other ways of storing that info)
1263 => kill now useless network_card concept
1264 - Use dynar to represent routes (instead of void** + int*)
1265 - kill link_set (use surf_network_model->resource_set instead)
1266 - Add a command-line option to choose the routing schema to use
1267 - Add three new models:
1268 * Floyd (shortest path computed at initialization)
1269 * Dijikstra (shortest path recomputed all the time)
1270 * Cached Dijikstra (shortest path computed on need)
1271 All these models where contributed by Silas De Munck, and are
1272 described in his ICCS09 paper.
1274 * Simplify model declaration
1275 (less redirections, less function to write when defining a model)
1276 - Factorize stuff between models:
1279 surf_model_resource_set(model)
1280 surf_model_resource_by_name(model, name)
1281 - Unify the types of models in s_surf_model_t (using an union)
1282 - Embeed fields of common_public directly into s_surf_model_t
1283 - Rename model methods:
1284 action_free ~> action_unref
1285 action_change_state ~> action_state_set
1286 action_get_state ~> action_state_get
1287 - Change model methods into functions :
1288 (model)->common_public->action_use ~> surf_action_ref
1290 * Implement a generic resource; use it as ancestor to specific ones
1291 (allows to kill duplicated code in models)
1292 Drawback: timer command don't need no name nor properties;
1293 workstation_CLM03 don't need no properties
1294 (but I guess we can live with those few bytes wasted)
1296 * Improve the action object model
1297 - implement a constructor avoiding dupplicated code about field
1298 initialization in generic_action part.
1300 * Kill the SDP model: it has an external dependency, is deprecated
1301 in flavor of modern lmm models, and didn't compile since a while
1304 * Relocation of the context module from XBT to SIMIX.
1305 (the context were decoupled from the simix processes, duplicating a lot of code)
1306 => a lot of code was factorized
1307 - less overhead is introduced during scheduling
1308 - simpler API for the context factory
1309 - the logic for process creation,destruction and manipulation was simplified
1310 * Simplification of the s_smx_process_t data structure.
1311 => accesing the simix level data associated to a process is faster now,
1312 and the code is a lot more readable.
1315 * Implement some more MPI primitives:
1316 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1317 -implementation: Bcast: flat or 2-ary tree (default),
1318 Barrier: 4-ary tree,
1320 Allreduce: Reduce then Bcast
1321 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1322 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1323 Alltoallv: flat tree, like ompi
1325 * Add support for optimized collectives (Bcast is now binomial by default)
1326 * Port smpirun and smpicc to OS X
1329 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1330 Shout out if you used it.
1333 * Display the list of still queued messages in SG mode when existing
1337 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1338 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1339 * Add xbt_str_from_file(FILE*)
1340 * Add xbt_dict_get_key achieving a linear reverse search
1341 * Remove the context module
1343 Portability report of this version:
1344 * Main portability targets:
1345 - Linux(debian)/x86/context
1346 - Linux(debian)/x86/pthreads
1347 - Linux(debian)/amd64/context
1348 - Linux(debian)/amd64/pthreads
1349 On these, we still have the eratic breakages of gras/pmm and
1350 amok/saturate_sg reported in previous version. We still think
1351 that the tests are the cause of the fault, not the tested code.
1353 - Mac OSX Leopard/x86/context
1354 Still false negative in tesh autotesting.
1355 Smpi still fails, but this time because readlink does not accept -f
1356 Everything seems to work properly beside of that.
1359 - AIX version 5.3 (only tested contexts this time)
1360 Smpi still fails there because mktemp is not installed.
1361 Everything seems to work properly beside of that.
1363 I managed to compile it for the first time, but several breakages.
1364 Won't delay the release for this exotic platform.
1366 * Windows: it's still lagging behind. If you want to help, please
1369 Timing report of this version:
1370 This version seem to be more than 5% faster than 3.3.1 (on linux
1371 64bits with contextes). The gain is less than expected, we are
1372 investigating this for next release.
1374 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1376 SimGrid (3.3.1) stable; urgency=low
1379 * Implement a --cfg-help to show existing configuration variables
1380 * Build chain do not require doxygen in maintainer mode
1383 * fix a bug on struct sizeof computation, which prevented the
1384 exchange of arrays of structs in some conditions
1385 - added a regression test about this in datadesc_usage
1386 * Allow the exchange of 0-long dynamic vectors.
1387 - for that, use -1 as indicator of dynamic size instead of 0
1388 - This implied to change any size from unsigned long to long,
1389 reducing a bit communication abilities, but I guess that with
1390 64bits being quite common, this is more than enough.
1391 - This also induce a protocol change, thus bumping network protocol
1392 version from 0 to 1 (if we have external users, we have to get
1393 clean on that point too ;)
1394 - added two regression tests about this in datadesc_usage
1395 * Be more verbose when propagating local exceptions
1396 This helps debugging.
1397 * Display the status of simulated processes when receiving SIGINT in
1401 * Allow to control the simulation from a trace file.
1402 New functions MSG_action_register() and MSG_action_trace_run()
1403 The first one allows to associate a function execution to each
1404 kind of action while the second one parses a trace file and
1405 triggers the corresponding actions within the system.
1406 For now, only a toy example is provided in examples/msg/actions
1407 * Add an exemple of process migration in examples/msg/migration
1408 * Fix a bug in task exchange which broke MSG_task_get_sender()
1409 Add a teshsuite regression test for that.
1410 [Bug: if MSG_task_get_sender() is called after sender exit,
1412 * Fix a bug which prevented suspend/resume to work properly
1413 * Display the status of simulated processes when receiving SIGINT
1414 This fixes a regression of v3.3. due to the introduction of SIMIX
1415 * Bug fixing in failure management:
1416 - trace could not start by a failure at time 0
1417 - failure during communications were not working
1420 * Add SIMIX_process_set_name() to change the name of the current
1421 process in the log messages.
1422 * Store smx_hosts in a dict since we only retrieve them by name
1423 * Move the configuration infrastructure to surf
1426 * Move the configuration infrastructure to surf
1429 * Massive internal cleanups:
1430 - Store internal structures on processes instead of hosts (allows
1431 to have more than one process per host, in addition of being more
1433 - Cleanup the initialization/finalization process
1434 - Kill a whole bunch of unneeded synchronization:
1435 processes run in exclusive manner within the simulator
1436 - Move queues from global tables to process data fields
1438 - now accept -platform and -hostfile arguments
1439 - Pass the right rank value to processes according to the hostfile
1440 * Compile the examples by default, and use them as regression tests
1441 * Implement MPI_Wtime()
1442 * Change the reference speed to a command line option
1445 * TCP_gamma can now be specified as command line option using
1446 --cfg=TCP_gamma:10000000.0
1447 * Change the --surf-path cmd line option into --cfg=path:
1450 * Also include strbuff from xbt.h public header
1451 * xbt_ex_display(): do not free the exception after displaying
1452 This allows to do more with the given exception afterward.
1453 Users should call xbt_ex_free() themselves.
1457 Portability report of this version:
1458 * Main portability targets:
1459 - Linux(debian)/x86/context
1460 - Linux(debian)/x86/pthreads
1461 - Linux(debian)/amd64/context
1462 - Linux(debian)/amd64/pthreads
1463 These targets fail about 1/10 of times on gras/pmm, but we believe
1464 that this is because of the test, not because of simgrid.
1465 amok/saturate_sg fails even more rarely, and the test may not be
1468 - Mac OSX Leopard/x86/context
1469 The test suite still spits tons of errors because some obscure
1470 force prevents us from removing the temporary directories
1471 arguing that they still contain some metadata I've never heard of.
1472 Smpi fails because seq is not installed.
1473 Everything seems to work properly beside of that.
1476 - AIX version 5.3 (both contexts and pthread)
1477 Smpi still fails there because mktemp is not installed.
1478 XML inclusions seems rosty on AIX.
1480 * Windows: it's still lagging behind. If you want to help, please
1483 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1485 SimGrid (3.3) stable; urgency=high
1489 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1490 [Malek Cherier & Mt]
1492 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1495 Until now, GRAS were using MSG as an interface to SURF. It was
1496 quite difficult because both interface have several differences
1497 (MSG channels vs GRAS sockets were the most notable point).
1499 This also opens the gate to SMPI (which should occur soon) and speed
1500 up simulations by to 40% (even if it were not the main goal).
1502 **************************************
1503 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1504 **************************************
1505 The point is that events occuring at the exact same timestamp are
1506 not scheduled in the same order with the old and new version. This
1507 may be enough to completely change the execution of simulations in
1508 some cases. Sorry for the inconvenience.
1510 * Cleanup and upgrade the XML format to push further scalability
1511 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1513 * Improve the testing infrastructure with tesh. Now a very large part of
1514 the code is tested not only by being run but also by checking that the
1515 output match an expected output [Mt].
1517 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1518 is really less memory-demanding, which should allow you to use
1519 larger files in SimGrid [AL].
1521 * Inform valgrind about our contextes, so that it becomes usable
1522 with the default (and more effecient) version of SimGrid
1523 [contributed by Sékou Diakite, many thanks]
1526 * Introduce a listener thread in charge of receiving incoming
1527 messages from the network. It allows to overlap communication and
1528 computation but most notably, it removes some stupid deadlocks due
1529 to the fact that so far, a process could not send and receive at
1530 the same time. This made most non trivial communication schema
1532 * Convert the PIDs from long int to int to match the MSG ones (and
1533 linux ones too) [Mt]
1534 * New function: gras_agent_spawn() to launch a new process on
1535 current host. Only working in simulation for now. [Mt]
1536 * New function: gras_os_hostport() returning a constant form (ie,
1537 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1540 * Make the backtrace of exceptions more human readable [Mt]
1541 * New module: xbt/str [Mt]
1542 a ton of string utility functions (split, join, printf to a newly
1543 allocated buffer, trim, etc)
1544 * New module: xbt/hash [Mt]
1545 SHA1 hashing algorithm (more to come if needed)
1546 * New module: xbt/synchro [Mt]
1547 synchronization tools (mutex and conditions) working the same way
1548 in simulation and in real life (mainly useful for GRAS, but not
1550 * New module: xbt/queue [Mt]
1551 classical producer/consumer synchronization scheme
1552 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1553 (using the classical functions will get serialized) [Mt]
1554 * Make dictionary internal table dynamic. No need to specify its size
1555 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1557 * Make sure the log channels are organized as a tree under windows
1558 (because of ANSI C compatibility issue, any channel were child of
1559 root directly) [Mt].
1562 * Cleaned many thing in surf and fixed a few bugs [AL].
1563 * Add a nice command line configuration mechanism to compose models [AL].
1564 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1565 the previous one (KCCFLN05). It relies on something that looks like
1566 a max-min sharing mechanism but cannot be written as such. A new solver
1567 was thus designed [AL].
1568 * Add a new solver to lmm. Based on Lagrange optimization and
1569 gradient-based descent, it enables to efficiently maximise systems s.a
1571 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1573 This solver enables to propose two new network models for TCP Reno and
1574 TCP Vegas based on Low's work. These models still need to be fully
1575 tested though [Pedro Velho].
1578 * Bug fix in SD_simulate. Now the time bound given as argument is
1580 * Use the new parallel task model (ptask_L07) as default.
1581 * Use the SURF command line configuration mechanism.
1582 * 0-size tasks (for synchronization) should now work.
1584 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1586 SimGrid (3.2) stable; urgency=high
1590 We still experience issues on this platform, but we believe that at
1591 least MSG is usable.
1593 GRAS API BREAKAGE (for simplification purpose, sorry):
1594 * the gras_msgtype_by_name is not used anymore. Instead of
1595 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1596 you can write (and must)
1597 gras_msg_send(toserver, "request", &request);
1598 - If you still want to pass a gras_msgtype_t to the function (to cache
1599 the type and avoid the lookup time), use the gras_msg_send_() variant.
1600 - Impacted functions:
1601 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1602 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1603 * The callbacks are now expected to return 0 when everything went well
1604 (just like the main() function)
1606 GRAS new features and improvements:
1607 * New module mecanism where user code can use per process globals [Mt]
1608 This is similar to gras_userdata_*() functions, but for libraries. It
1609 factorize some code developped over and over in the examples and AMOK.
1610 It has still to be documented and used (only amok/peermanagement is
1612 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1613 loss when they were small enough to fit into the buffer and sent quickly
1614 enough so that they can all get received in one shoot.
1615 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1616 if not found. Use the *_or_null() variant for the old semantic.
1617 * In gras_msg_handle, do not discard messages without callback.
1618 They are probably messages to be explicitly awaited later (ie, proofs of
1619 mis-synchronization in userland since they are sent before being awaited)
1620 No big deal usually.
1621 * gras_socket_meas_send/recv: semantic changed!
1622 The numerical arguments used to be (1) the total amount of data to send
1623 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1624 messages. This was need for the fool willing to send more than MAXINT
1625 bytes on quite fat pipes.
1628 * Do really rename the hostmanagement module to peermanagement. [Mt]
1629 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1630 but this is rather new and this was documented in the module
1631 documentation (poor excuses, I admit)
1632 * Bandwidth measurement semantic changed! This follows the changes to
1633 gras_socket_meas_send/recv explained above.
1636 * A sequential mode has been added to the workstations. When a workstation
1637 is in sequential mode, it can execute only one task, and the other tasks
1638 are waiting in a FIFO. [Christophe Thiery]
1641 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1642 model for SIMDAG. [Christophe Thiery]
1643 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1644 equal to 0, which caused some bad behaviors in
1645 saturated_constraint_set_update. I now use a threshold mechanism like in
1649 * When running manually src/testall, you select specific units [Mt]
1650 testall is the result of our cunit mecanism, and should replace all
1651 the scripty thingy around since bash don't run easily on billware.
1653 * A mallocator system has been added. [Christophe Thiery]
1654 Mallocators allow you to recycle your unused objects instead of freeing them
1655 and allocating new ones.
1657 Documentation update:
1658 * FAQ reworking + New FAQs:
1659 - "Valgrind spits tons of errors!" [Mt]
1660 - "How to repport bugs" [Mt]
1661 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1662 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1663 - Communication time measurement within MSG [AL]
1664 - I experience weird communication times when I change the latency [AL]
1665 * GRAS tutorial [Mt]
1667 - an introduction to the framework and to the used communication model
1668 - an initiatic tour introducing the most proheminent features:
1670 . Lesson 0: Installing GRAS
1671 . Lesson 1: Setting up your own project
1672 o Part 2: Message passing
1673 . Lesson 2: Exchanging simple messages
1674 . Lesson 3: Passing arguments to the processes (in SG)
1675 . Lesson 4: Attaching callbacks to messages
1676 . Lesson 5: Using globals in processes
1677 . Lesson 6: Logging informations properly
1678 . Lesson 7: Using internal timers
1679 . Lesson 8: Handling errors through exceptions
1680 . Lesson 9: Exchanging simple data
1681 . Lesson 10: Remote Procedure Calling (RPC)
1682 . Lesson 11: Explicitely waiting for messages
1683 . Recapping of message passing features in GRAS
1684 - A HOWTO section containing:
1685 o HOWTO design a GRAS application
1686 More are due, of course. They will come latter. In the meanwhile, you can
1687 check the examples which are still here.
1689 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1691 SimGrid (3.1) stable; urgency=high
1695 There was a stack corruption somewhere, visible only when optimizing
1696 with these versions. [Vince]
1699 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1700 module that enabled to play with Directed Acyclic Graphs. It is built
1701 directly on top of SURF and provides an API rather close to the old
1702 SG. Some old codes using SG are currently under rewrite to check that
1703 all needful functions are provided. [Christophe Thiery]
1706 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1707 extension of the classical CLM03 model that gracefully handles
1708 failures. This is now the default model for MSG and GRAS. It doesn't
1709 handle parallel tasks yet though. [AL]
1710 * Bug fix: Weights were not correctly set in the network part.
1711 WARNING: This may have resulted in incorrect results with simulations
1712 where there are more than one flow on a given link. [AL]
1715 * After a (long ?) discussion on simgrid-devel, we have decided that the
1716 convention we had on units was stupid. That is why it has been decided
1717 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1718 WARNING : This means that all previous platform files will not work as
1719 such with this version! A warning is issued to ask users to update
1721 A conversion script can be found in the contrib module of the CVS, under
1722 the name contrib/platform_generation/surfxml_update.pl [MQ]
1725 * Bug fix: Processes were started in reverse order, wrt deployment file.
1726 WARNING: if your code relies on this bug, please fix it. [AL]
1727 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1728 being executed on two different locations. [AL]
1729 * Bug fix: Failures are now better supported thanks to Derrick's tests
1730 (there was many failure situations I hadn't thought of and that weren't
1731 correctly handled). [AL]
1732 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1736 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1737 Exception occurring on server-side are propagated back to client (!).
1739 API CHANGE: the callback changed their prototype. Change:
1740 int my_handler(gras_socket_t expeditor, void *payload_data) {
1742 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1743 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1745 * New! function: gras_msg_handleall to deal with all messages arriving
1746 within a given period.
1747 * New! function: gras_socket_server_range to get a server socket in a
1748 range of port numbers (ease to avoid port number conflicts) [MQ]
1749 * New! gras processes display their backtrace when they get a SIGUSR1
1750 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1751 Sweet to debug RL processes [MQ]
1755 - Do not force experiment sizes to be expressed in kb, or it becomes
1756 impossible to measure the latency this way (needs one byte-long tests)
1757 WARNING: this changes the amok_bw_* function semantic. [MQ]
1758 - Implements the link saturation stuff. [MQ]
1759 * Peer management module:
1760 New! module factorizing code that we wrote over and over [MQ].
1763 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1764 - Test units are placed directly into the library code, they get extracted
1765 automatically and placed into the src/testall binary.
1766 - Convert most of the XBT tests to this system.
1767 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1768 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1769 live in the same C code. This required to change a little bit the API
1770 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1771 * New module: sparse graph structure with basic algorithms (this is work
1772 in progress and the API is not considered to be frozen yet). [AL]
1773 * Display more information on backtraces: source line & function names are
1774 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1775 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1776 * Reworked a little bit some #include statements to load only required
1777 headers. Some user code that relied on SimGrid to include stdlib or
1778 stdio may need to include it by themselves. [AL]
1779 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1781 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1782 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1783 but I guess that almost nobody used those parts. [MQ]
1785 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1787 SimGrid (3.0.1) stable; urgency=low
1790 * Unfortunately, I had missed 5 misnamed functions:
1791 xbt_fifo_item_t xbt_fifo_newitem(void);
1792 void xbt_fifo_freeitem(xbt_fifo_item_t);
1793 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1794 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1795 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1796 They're now deprecated. Please use their new versions:
1797 xbt_fifo_item_t xbt_fifo_new_item(void);
1798 void xbt_fifo_free_item(xbt_fifo_item_t);
1799 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1800 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1801 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1803 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1804 * Documentation: xbt_log module unmercifully reworked [MQ]
1805 * Bugfix: there was a problem with the ending of contexts with
1806 the pthread backend. It caused some weird deadlock or behavior
1807 depending on the pthread implementation. [AL]
1808 * Bugfix: get the exceptions raised in the simulator repport where
1809 and why they come from when they are not catched in time [AL, MQ]
1812 * Bugfix: Do repport the error when two non-connected hosts try to
1813 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1816 * Add additionnal checkings on communications. Assert that two
1817 communicating hosts are connected by a set of links... [AL]
1820 * Add additionnal checkings on channel values in communication [AL]
1821 * New: MSG_task_get_source to see on which host a task was generated [HC]
1822 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1823 the number of tasks waiting to be received on channel and sent
1825 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1826 waits for the first task coming from a given host.. [AL]
1828 GRAS new functionnalities: [MQ]
1829 * Enhance the parsing macro to allow the size of multidimentional objects
1830 to be given thru annotations.
1831 * New example (and documentation): Matrix Multiplication a la RPC
1832 (as when I was young!) and fix a bunch of bugs found on the way.
1834 GRAS performance improvements: [MQ]
1836 * Reduce the amount of cbps creation/destruction by making it static to
1837 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1840 * Change libdata to a set so that we can search for stuff by ID (and thus
1841 reduce the insane amount of dict lookups)
1844 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1845 RL communication on top of SG since it's so uneffective.
1846 It may also be used for inter-thread communication in RL, one day.
1847 * Use gras_datadesc_copy() to exchange messages on top of SG
1849 - improve message exchange performance on top of SG
1850 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1851 * Don't exchange on the network the size of the used part of buffer,
1852 instead, specify the possible buffer size to read().
1854 - reduces the amount of read/write calls (one pair per exchange)
1855 - reduces the amount of exchanged data (the size)
1856 - allows to retrieve all arrived data on receiver side, if we don't need
1857 it right now (subsequent read will peek the buffer)
1858 - allows the receiver to proceed with the begining of the stream before
1859 everything is arrived
1860 - make it possible to build an iov transport (using readv/writev)
1862 - take care of the data with non-stable storage (like stacked data),
1864 * If possible, TCP send uses vector I/O (when writev() is here)
1865 - Don't use it for receive since we send data sizes and data on the
1866 same stream, so we wouldn't be able to chain large amount of chunks
1867 before having to flush the stuff to read the size.
1868 * Rework the transport plugin mecanism to simplify it and reduce the
1869 amount of pointer dereferencement when searching for the right function
1872 * I guess that now, we do almost as few system calls as possible while
1873 doing as few data copy as possible.
1875 To improve it further, we could try to send all the sizes first and then
1876 all the data (to use iov on receiving size), but it's only a partial
1877 solution: when you have 2 dimensional data, the sizes of the second
1878 dimension is data of the first dimension, so you need 3 streams.
1880 I'm not sure the potential performance gains justify the coding burden.
1882 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1884 SimGrid (3.00) stable; urgency=high
1887 * New! Give the possibility to hijack the surf parser and thus bypass
1888 MSG_create_environment and MSG_launch_application. Have a look at
1889 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1891 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1893 SimGrid (2.96) unstable; urgency=low
1898 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1899 This deprecates the xbt_error_t mecanisms.
1900 It modifies (simplifies) all XBT and GRAS API.
1901 MSG API keeps unchanged (exceptions raised by XBT are catched from
1902 within MSG and masked with existing error handling facilities)
1905 * New! Add a FATPIPE model. [AL]
1906 * New! Add a parallel task model. [AL]
1907 * New! Add automatically a loopback interface (in the default
1908 network model) if none was precised.
1911 * Bugfix: MSG_process_resume now works with the current running process.
1913 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1914 * Modification of MSG_task_get_compute_duration. Once a task has been
1915 processed, the value returned by this function is now equal to 0. [AL]
1916 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1917 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1918 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1919 of a m_task. For now, MSG_task_cancel only works with computation
1921 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1922 of the processor (in Mflop/s) regardless of the current load on the
1924 * API Change: use proper naming convention for MSG_getClock and
1925 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1927 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1928 This function changes the priority of a computation task. This priority
1929 doesn't affect the transfer rate. A priority of 2 will make a task
1930 receive two times more cpu power than the other ones. This function
1931 has been added to suit the needs of Nguyen The Loc and hasn't been that
1932 much tested yet. So if it fails, please report it and send me your code.
1934 * API Change: removed all functions and types that were marked "deprecated"
1935 since many months. Renamed MSG_global_init_args to MSG_global_init.
1937 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1939 SimGrid (2.95) unstable; urgency=low
1942 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1943 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1944 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1947 * Bugfix: Allow absolute paths for platform description files [MQ]
1948 * Bugfix: do free the variables after use. Leads to drastic performance
1950 * Implement max_duration (ie, timeouts) on resources [AL]
1953 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1955 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1956 the needed support (timeouts) [AL]
1959 * Implement measurement sockets. You can now get the bandwidth between two
1960 hosts thanks to AMOK (see below). [MQ]
1961 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1962 dynar over the network (yeah) [MQ]
1963 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1964 * simple/static token-ring example. [Alexandre Colucci and MQ]
1965 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1966 in place before [MQ]
1969 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1970 type parsing. "short" and "long" modifiers were also ignored; other
1971 modifier (such as reference level) are still ignored. [MQ]
1972 * Embeed the buffer size within the buffer itself on SG. [MQ]
1973 That way, send() are atomic and cannot get intermixed anymore (at least
1974 the ones which are less than 100k; bigger messages still have the issue)
1975 * Array size pushed by the field, not by the field type (or each
1976 and every long int will push stuff to the cbps) [MQ]
1977 * use select() to sleep since it allows to portably sleep less than one
1980 GRAS (minor cleanups)
1981 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1984 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1985 * type category 'ignored' killed as it was never used and were difficult
1987 * whether a type can cycle or not is now a flag, leaving room for more
1988 flags (as "ignored", if we feel the need one day ;)
1989 * Rename raw sockets to measurement sockets since "raw" has another
1990 meaning in networking community.
1993 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1994 and offers features not belonging to GRAS but that most applications
1995 need. One day, it may be a set of plugins loadable at runtime.
1996 * New module: bandwidth
1997 bandwidth measurement between arbitrary nodes running this module. [MQ]
1999 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2001 SimGrid (2.94) unstable; urgency=low
2003 The first beta release of SimGrid 3 !
2007 * Update the main page and the FAQ. Adding references to gforge.
2010 * Add a gras_os_getpid function.
2013 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2014 * Extend the logs so that they also print PID, hostname, date, ... if
2016 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2017 and kill the old version which were in testsuite/
2018 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2022 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2023 for users. Instead, each programming environment (GRAS, MSG,...) use it
2024 when needed to provide such a feature to users.
2025 Don't shortcut the mecanism or you will also shortcut the virtualization
2026 you need on the simulator.
2030 * Cleanups in configury with regard to compile optimization/warning flags.
2031 Also add -fno-loop-optimize to any powerpc since it's the optimization
2032 killing gcc (< 3.4.0).
2033 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2034 needed by MSG examples complications
2035 * Borrow configury beautifications from PHP
2038 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2039 warning (thanks loris for stumbling into this one).
2040 * Bugfix: stop loading private headers (gras_config.h) from the public
2044 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2045 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2046 RL would imply a DNS resolver, which is *hard* to do in a portable way
2047 (and therefore delayed).
2048 * Implement a real timer mecanism and use it in timing macros. This allows
2049 to avoid rounding errors and get a 0.000005 sec precision in timing
2050 macros. While I was at it, various cleanups:
2051 - allow to declare more than one timed section per file (fix a stupid bug)
2052 - move some private declaration to the right place
2053 - merge conditional execution and timing macros into emulation module
2054 - document the module
2055 - make sure the module cleanups its mess on gras_exit
2056 * Documentation improvements:
2057 - (new) how to compile applications using GRAS
2058 - (new) emulation support (timing macros)
2060 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2062 SimGrid (2.93) unstable; urgency=low
2064 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2067 - Use Paje properly where used. Still to be sanitized properly.
2068 - Portability fix: Add an implementation of the contexts using pthread
2072 - Add xbt_procname(): returns the name of the current process.
2073 Use it to show the current process's name in all logging.
2075 - fix detection of older flex version and the reaction, since we do
2076 depend on modern ones (we use lex_destroy)
2077 - Better separation of SG and RL in the libs: remove all simulation code
2078 from libgras. As a result, this lib is now only 200k when stripped.
2079 Some of the xbt modules may also be duplicated (two sets and such) and
2080 should be cleaned/killed before SG3.
2081 - Insist on using xlC on AIX because of weird problems involving gcc there.
2082 - Cleanup the make remote stuff. This is now done by scripts
2083 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2084 mainly for our private use, but we're working on changing them to user
2087 - Bugfix: flush the socket on close only if there is some *output*.
2088 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2090 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2093 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2095 SimGrid (2.92) unstable; urgency=low
2097 Alpha 3 on the path to SimGrid 3
2101 - New! First try of benchmarking macros.
2102 - New! First try so that gras_stub_generator generate deployment and
2103 remote compilation helpers.
2105 - Bugfix: Initialization fix in msg_test.
2109 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2111 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2113 - API Change: gras_os_sleep to take the amount of seconds as a double.
2114 Accepting an int was error prone since it was the only location where
2115 seconds were coded as such. It leaded to damn rounding errors.
2116 - Bugfix: Hard to belive that timers ever worked before this.
2118 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2120 SimGrid (2.91) unstable; urgency=low
2122 Alpha 2 on the path to SimGrid 3
2126 - Bug fix in the lmm_solver.
2128 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2129 through the function MSG_paje_output.
2130 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2131 - It is possible to bound the rate of a communication in MSG with
2132 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2133 to write it in the changelog).
2134 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2139 - Various cleanups to the autotools stuff
2140 - Begin to move Gras examples to examples/gras/
2141 - Let make distcheck work again (yeah!)
2143 - documentation overhauled using doxygen.
2144 gtk-doc-tools is dead in SimGrid now.
2145 - Automatically extract all existing logging categories, and add the list
2146 to the documentation (long standing one, to say the less)
2148 - Cleanup the known architecture table. Reorder the entries to group what
2149 should be, and use a more consistent naming scheme.
2150 (some of the test dataset are still to be regenerated)
2151 - New! Allow library to register globals on each process just as userdata
2153 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2154 lookup time (for now).
2155 Use it in msg and trp.
2156 This cleans a lot the internals and helps enforcing privacy of the
2157 headers between the gras components.
2158 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2159 - Bugfix: gras_os_time was delirious in RL.
2160 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2162 - Reenable GRAS now that it works.
2164 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2166 SimGrid (2.90) unstable; urgency=low
2168 Alpha 1 on the path to SimGrid 3
2170 * It is a long time since the last release of SimGrid. I'm sorry about
2171 that but as I had told you, I was rewriting a lot of things. I apologize
2172 to those who had been reporting bugs to me and that I had not answered.
2173 If your bug is still in the new version, please tell me. Here is a
2174 summary of the main changes.
2176 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2177 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2178 portability, speed, and a lot more but you'll figure it out later.
2179 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2180 I wanted to release the new MSG as soon as possible and I have
2181 broken GRAS, which is the reason why, for now, only MSG is fully
2182 functional. A laconic description of these projects is available
2183 in the documentation.
2185 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2186 kernel optimized for our needs (called SURF but only the developers
2187 should use it). Hence, MSG is now roughly 30 times faster and I think
2188 that by rewriting a little bit MSG, I could event speed it up a little
2189 bit more. Beside the gain in speed, it is also much easier to encode a
2190 new platform model with SURF than it was with SG. More to come...
2192 * REVOLUTION 3: I have tried to change a little as possible the API of
2193 MSG but a few things really had to disappear. The main differences
2194 with the previous version are :
2195 1) no more m_links_t and the corresponding functions. Platforms are
2196 directly read from a XML description and cannot be hard-coded
2197 anymore. The same format is used for application deployment
2198 description. The new format is described in the documentation.
2199 Have a look in tools/platform_generation. There is a tiny script
2200 that converts from the old platform format to the new one. Concerning
2201 the application deployment format, parsing the old one is tricky.
2202 I think most of you should however be able to convert your files. If
2203 it is really an issue, I can write a C code that does the conversion.
2205 2) the toolbox tbx does not exist anymore. We now have a library
2206 with much more data-structures but without the hash-tables (we have
2207 dictionaries that are much faster).
2209 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2211 *****************************************************************************
2212 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2213 * don't feel like converting the oldies. (Mt) *
2214 *****************************************************************************
2217 Version 2.90: "the long awaited one"
2218 - Finished rewriting and debugging MSG. Rewrote the documentation.
2219 - disable GRAS for now since it needs to be ported to the newest SG
2222 - Finish the port to windows (using mingw32 for cross-compile)
2225 - Main loop and datastructures of SURF. A cpu resource object is
2226 functional. Surf can thus be used to create cpu's with variable
2227 performance on which you can execute some actions.
2229 2004-11-15 Martin Quinson
2230 - Port to ARM. Simply added the alignment and size descriptions. Should
2231 work, but the ARM machines are so slow that I didn't had the opportunity
2232 to 'make check' over there yet.
2234 2004-11-15 Arnaud Legrand
2235 - Trace manager now written. It uses a heap structure and is therefore
2236 expected to be efficient. It may however be speeded up (particularly
2237 when many events occur at the same date) by using red and black
2238 trees. One day maybe...
2239 - Max-min linear system solver written. It uses a sparse matrix
2240 structure taking advantage of its expected use. Most operations are
2241 O(1) and free/calloc are called as few as possible. The computation of
2242 the minimum could however be improved by using a red and black tree
2245 2004-11-03 Arnaud Legrand
2246 - Rename every gras_* function that was in xbt/ to its xbt_
2248 - Add a heap and a doubly-linked list to xbt
2249 - Added a dichotomy to the dictionaries. make check works as well before
2250 so I assume that the patch is correct. I do not know however if things
2251 run effectively faster than before now. :)
2253 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2254 SimGrid, and the version number is bumped to 2.x
2256 2004-10-29 Martin Quinson
2257 - Introduction of the remote errors.
2258 They are the result of a RMI/RPC on the remote machine.
2259 ErrCodes being scalar values, you can't get the host on which those
2260 errors did happen. Extending the error mechanism as in Gnome is possible.
2261 No idea yet whether it is a good idea.
2263 2004-10-28 Martin Quinson
2264 - Interface revolution: the Starred Structure Eradication.
2265 I used to do typedef struct {} toto_t; and then handle *toto_t.
2266 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2267 - ???_t is a valid type (builded with typedef)
2268 - s_toto_t is a structure (access to fields with .)
2269 - s_toto is a structure needing 'struct' keyword to be used
2270 - e_toto_t is an enum
2271 - toto_t is an 'object' (struct*)
2273 typedef struct s_toto {} s_toto_t, *toto_t;
2274 typedef enum {} e_toto_t;
2275 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2276 s_toto_t) is private.
2278 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2279 so that it can set the variable to NULL. It was so for dicts and sets,
2280 it changed for dynars.
2282 - Fix a bunch of memleaks in dict_remove
2283 - Fix a bug in sg/server_socket opening: it failed all the time.
2285 2004-10-07 Martin Quinson
2286 - Speed up dynar lookup operation a bit.
2288 gras_dynar_get is dead.
2290 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2291 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2292 (which gives you the address of the stored data).
2294 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2295 copy of the data using an affectation to do the job and not a memcpy.
2297 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2298 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2300 It does not really speedup the dynar test because they are
2301 setting elements all the time (and look them seldom). But the dict does
2302 far more lookup than setting.
2304 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2306 2004-10-05 Martin Quinson
2307 - Allow to (en/dis)able the cycle detection at run time.
2309 Whether we should check for cycle or not is now a property of each
2310 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2311 datadesc_cycle_unset allow to remove this property when previously set.
2313 Note that the cycle detection is off by default since it impacts the
2314 performance. Watch the data you feed GRAS with ;)
2316 This property is hereditary. Any element embedded in a structure having it
2317 set have it set for the time of this data exchange.
2319 You should set it both on sender and receiver side. If you don't set it on
2320 sender side, it will enter an endless loop. If you forget on receiver
2321 side, the cycles won't be recreated after communication.
2323 - Header reorganization.
2324 Kill gras_private.h, each submodule must load the headers it needs.
2326 2004-10-04 Martin Quinson
2327 - Interface revolution: do not try to survive to malloc failure.
2329 Now, gras_malloc and friends call gras_abort() on failure.
2330 As a conclusion, malloc_error is not a valid error anymore, and all
2331 functions for which it was the only gras_error_t return value are
2332 changed. They now return void, or there result directly.
2333 This simplify the API a lot.
2335 2004-09-29 Martin Quinson
2336 - Re-enable raw sockets.
2337 Created by gras_socket_{client,server}_ext;
2338 Used with gras_raw_{send,recv}
2341 It should allow to kill the last bits of gras first version soon.
2343 This is not completely satisfactory yet (duplicate code with
2344 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2347 - Simplify transport plugin (internal) interface by not passing any
2348 argument to _server and _client, but embedding them in the socket
2351 2004-09-28 Martin Quinson
2352 - Finish the port to AIX.
2353 autoconf was my problem (segfault within the malloc replacement
2354 function. No idea why)
2356 2004-09-16 Martin Quinson
2357 - Fix some size_t madness on 64bit architectures.
2359 2004-09-08 Martin Quinson
2360 - Reduce the number of system headers loaded, overload some more system
2361 calls (such as malloc to cast the result of the system one, and work
2363 - Fix and reintroduce the config support
2365 2004-09-07 Martin Quinson
2366 - Source code reorganization to allow Arnaud to surf all over there.
2367 - Allow to document the logging categories.
2368 - Remove all uppercase from logging categories and useless cleanup in names.
2370 2004-08-18 Martin Quinson
2371 Version 0.6.2 (protocol not changed; API changed)
2372 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2373 gras_error_t), and NULL when not found. Functions expecting a msgtype
2374 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2375 hopefully usefull message.
2376 - Portability to prehistoric sparcs again
2378 2004-08-17 Martin Quinson
2379 Version 0.6.1 (protocol not changed; ABI not changed)
2380 - prealloc some buffers to speed things up
2382 2004-08-11 Martin Quinson
2383 Version 0.6 (protocol not changed; ABI expended)
2384 - The parsing macro can deal with the references, provided that you add
2385 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2387 2004-08-09 Martin Quinson
2388 Version 0.5 (protocol not changed; ABI changed)
2389 - Allow to off turn the cycle detection code in data exchange at
2390 compilation time. It should be at run time, but I'm short of time (and
2391 the config stuff is still broken). That way, we keep dict out of the
2392 critical path, which is good because the performance is poor:
2393 - search not dichotomial yet
2394 - dynar give no way to access their content and memcpy everytime
2395 - In composed data description (struct, ref and so on), stop foolness of
2396 keeping the subtype's ID, but store the type itself. This keeps sets out
2397 of the critical path, which is good since they rely on dynar and
2398 dictionnaries. The only loose of that is that we cannot detect the
2399 redeclaration of a structure/union with another content (but I'm not sure
2400 the code detected well this error before anyway). We still can detect
2401 the redefinition discrepancy for the other types.
2402 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2403 it breaks the code because of type-punning used all over the place).
2404 This breaks on all non-gcc architectures (for now).
2406 All those changes (plus the buffer of last time) allow me to gain 2 order
2407 of magnitude on cruel tests consisting of 800000 array of integers on two
2408 level of a hierarchical structure (200 secondes -> 4 secondes)
2411 - the selector of reference must now return the type it points to, not
2412 the ID of this type.
2414 2004-08-06 Martin Quinson
2415 Version 0.4 (protocol changed; ABI not changed)
2416 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2417 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2418 - Implement a buffer transport, and use it by default (it relies on tcp in
2419 real life and on sg in simulation).
2420 That's a bit hackish since I had a new field to the structure to store
2421 its data without interfering with the subtype ones. Inheritance
2422 is tricky in C. And that's a kind of reverse inheritance with one class
2423 derivating two classes. Or maybe a game with java interfaces. Anyway,
2424 that's damn hard in C (at least).
2425 Moreover, I got tired while trying to ensure plugin separation and
2426 genericity in SG mode. MSG wants me to do weird things, so let's go for
2427 cruel hacks (temporarily of course ;).
2428 See comment in transport_private.h:71
2429 - do not include all the _interface headers in private but in the files
2430 which really need them (to cut the compilation time when they are
2433 2004-07-26 Martin Quinson
2434 Version 0.3 (protocol not changed; ABI changed)
2435 - Major overhault of the datadesc interface to simplify it:
2436 - shorted the function names:
2437 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2438 - add a trivial way to push/pop integers into the cbps without malloc.
2439 This allows to make really generic sub_type description, which simply
2440 pop their size of the stack.
2441 - add a function gras_datadesc_ref_pop_arr() which does what users want
2442 most of the time: Declare a dynamic array (which pops its size of the
2443 stack) and declare a reference to it. Poor name, but anyway.
2444 - kill the post-send callback, add a post-receive one
2446 2004-07-23 Martin Quinson
2447 Version 0.2 (protocol changed; ABI changed)
2448 - add some testing for cpbs in the test cases, and fix some more bugs.
2449 This invalidate again the little64 data file, since I cannot regenerate
2451 - remove an awfull optimization in the logging stuff, allowing me to:
2452 - understand it again
2453 - learn gcc how to check that the argument match the provided format
2454 - fix all errors revealed by gcc after that
2455 - internal keys of dict are not \0 terminated. Deal with it properly in
2456 loggings instead of segfaulting when the user want to see the logs :-/
2458 2004-07-22 Martin Quinson
2459 - Fix some stupid bug preventing cbps (callback postit) from working
2461 2004-07-21 Martin Quinson
2462 - Some documentation cleanups
2463 - remove the useless last argument of msgtype_declare
2464 - rename the Virtu functions to fit into the 'os' namespace
2465 - move headers src/include -> src/include/gras/ and stop fooling with
2466 gras -> . symbolic link
2467 - make distcheck is now successful
2469 2004-07-19 Martin Quinson
2471 - Build shared library also
2472 - Install html doc to the right location
2473 - stop removing maintainer files in make clean
2474 - build tests only on make check
2476 2004-07-13 Martin Quinson
2478 - No major issue in previous version => change versionning schema
2479 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2480 corresponding dataset.
2482 2004-07-11 Martin Quinson
2484 - Get it working with any kind of structure (we can compute the padding
2485 bytes remotely for all the architectures I have access to)
2486 - Implement the structure parsing macro (still not quite robust/complete)
2487 - Improvement to the remote testing toysuite
2489 2004-07-10 Martin Quinson
2490 [autoconf mechanism]
2491 - get ride of a bunch of deprecated macros
2492 - actually run the test for two-compliment, not only compile it :-/
2493 - test whether the structures get packed (and bail out if yes. Damn.
2494 Alignment is a serious matter)
2495 - test whether the structures get compacted (but respecting the alignment
2496 constraints of each types)
2497 - test whether the array fields of structures can straddle alignment boundaries
2499 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2500 (took me 2 hours to find that bug, looking at the wrong place)
2501 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2502 before sending/receiving and are used to compute the offsets of fields
2503 - Given that padding size depend even on compiler options, keep track of
2504 alignment and aligned_size only for the current architecture. Not a big
2505 deal since we send structure fields one after the other (seems
2507 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2508 program, let it work on linux/gcc/little32. portability todo.
2510 2004-07-08 Martin Quinson
2511 - import and improve remote compilation support from FAST
2512 - make sure make check works on half a dozen of machines out there
2514 2004-07-07 Martin Quinson
2515 Let's say it's version 0.0.3 ;)
2516 - Implement conversions (yuhu!)
2517 - Let it work on solaris (beside conversion, of course)
2518 - Stupid me, using rand() to generate the conversion datatests in not wise.
2520 2004-07-06 Martin Quinson
2521 - Let make dist work, since I'm gonna need it to compile on remote hosts
2522 - Let Tests/datadesc_usage write the architecture on which the file was
2523 generated as first byte.
2524 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2525 architecture descriptions.
2526 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2527 on those architectures.
2528 - Optimization: send/recv array of scalar in one shoot
2530 2004-07-05 Martin Quinson
2531 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2533 - Plug a whole bunch of memleaks
2534 - each process now have to call gras_{init,exit}. One day, their log
2535 settings will be separated
2536 - Continue the code factorisation between SG, RL and common in Transport.
2538 2004-07-04 Martin Quinson
2540 - Redistribution between SG and RL.
2541 We wanna have to accept in SG, so move accepted related parts of RL in
2542 the common part. (more precisely, the dynar of all known sockets is no
2543 more a static in transport.c, but part of the process_data)
2545 [gras_stub_generator]
2546 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2548 2004-07-03 Martin Quinson
2549 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2550 them all up in one shot)
2552 - Fix the ugly case of reference to dynamic array.
2553 - New (semi-public) function gras_datadesc_size to allow the messaging
2554 layer to malloc the needed space for the buffer.
2556 - gras_socket_close now expect the socket to close (and not its address to
2557 put NULL in it after it). This is because the socket passed to handlers
2558 is one of their argument (=> not writable).
2560 - propagate the interface cleanup from last week in datadesc, ie remove a
2561 superfluous level of indirection. User pass adress of variable
2562 containing data (both when sending and receiving), and not of a variable
2563 being a pointer to the data. Let's say that I like it better ;)
2564 The price for that is constructs like "int msg=*(int*)payload" in
2565 handlers, but it's a fine price, IMHO.
2567 - Let it work in RL (yuhu)
2569 2004-06-21 Martin Quinson
2571 - porting SG plugin and SG select to new standards (works almost).
2572 - plug memleaks and fix bugs around.
2575 - cleanup the prototype of data recv and force users to specify when they
2576 want to handle references to objects. Test case working even for cycles.
2577 - plug memleaks. Valgrind is perfectly ok with this.
2579 2004-06-12 Martin Quinson
2581 - cleanup the separation between plugin and main code in plugin creation
2583 2004-06-11 Martin Quinson
2585 - Reput hook for raw sockets, needed for BW experiments
2586 - kill a few lines of dead code
2587 [Data description] Interface cleanup
2588 - gras_datadesc_by_name returns the searched type or NULL.
2589 That way, no variable is needed to use a type desc once, which makes
2591 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2592 two parameters were strings (field name, type name), leading to
2594 [Dicos] Interface cleanup
2595 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2596 This is consistant with the dynar API.
2598 2004-04-21 Martin Quinson
2600 - Porting to new standards.
2602 - interface cleanup.
2603 There is no bag anymore, no need to take extra provision to mask the
2604 pointers behind "ID".
2605 Better splitup of functions between files create/exchange/convert.
2606 This is still a bit artificial since convert and receive are so
2607 interleaved, but anyway.
2609 - add a queued message list to procdata (the ones not matching criteria
2611 - factorize some more code between SG and RL wrt procdata
2613 - use gras_exit in example to track memleaks
2614 - get rid of gs_example now that GS is properly integrated into gras
2615 - update run_test to integrate the latest tests (datadesc)
2617 - rename WARNINGn macros to WARNn since it prooved error-prone
2619 2004-04-19 Martin Quinson
2621 - register init/exit functions within gras module mechanism
2622 - send/receive function.
2623 Convertion is not implemented, but short-cutted if not needed.
2624 struct/array elements are sent one by one (instead of block-wise), but
2625 nobody really cares (yet). Get a prototype before optimizing.
2626 - tests (using a file socket) for DD send/receive on:
2627 - base types: int, float
2628 - array: fixed size, string (ie ref to dynamic string)
2629 - structure: homogeneous, heterogeneous
2630 - chained list, graph with cycle
2631 Believe it or not, valgrind is not too unhappy with the results. The
2632 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2636 - Bugfix when using a filename explicitely (instead of '-')
2638 2004-04-09 Martin Quinson
2640 - factorize more code between RL and SG in socket creation
2641 - Complete the implementation and tests of:
2643 o file (only in RL, and mainly for debugging)
2645 I lost 3 days to design a portable address resolver, and then decided
2646 that the prototype mainly have to run on my box.
2647 Addressing portability too early may be like optimizing too early :-/
2649 - use gras_init in the Tests instead of the crappy parse_log_opt
2650 (the latter function is removed)
2651 [Conditional execution]
2652 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2653 [Code reorganisation]
2654 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2655 Build examples against the RL library, since there is no way to disable
2656 its creation for now.
2658 For information, the beginning of coding on GRAS was back in june
2659 2003. I guess that every line has been rewritten at least twice since