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
9 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
11 SimGrid (3.10-rc1) unstable; urgency=low
14 * Reintegrate Java to the main archive as desynchronizing these
15 package is not acceptable anymore (Java is now considered stable)
16 * Add explicit synchronization facilities through semaphores
17 * Bug fix: Task.setDataSize() only changed the C world, not the value
18 cached in the Java world
21 * Dramatically change the way files are handled. API and internals changed, but
22 this part of MSG was not considered as production grade either.
23 * Add explicit synchronization facilities through semaphores
24 * Add a new function MSG_host_get_process_list()
25 * Preliminary DVFS support (see examples/msg/energy/ for details)
28 * SMPI is now included directly in the libsimgrid as the windows
29 linker doesn't force us on splitting it anymore.
30 * Improvements of the SMPI replay tool:
31 - Most of the collective communications are now rooted in the same process as
32 in the original application.
33 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
34 used until now). Multiple data types can now be used in a trace.
35 - The replay tool now supports traces produce either by TAU or a modified
37 - Bug Fix: the compute part of the reduce action is now taken into account.
38 - Gatherv collective is now supported
39 - Simgrid (SMPI for now) can generate replay traces as well. Option -trace-ti
40 of smpirun outputs time independent traces for the current run. One file
41 is created per process. If too many processes are simulated, this behavior
42 can be changed to one file for all processes by using the
43 tracing/smpi/format/ti_one_file flag
44 * smpirun generates the host file if needed (with given host count and platform)
45 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
46 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
47 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
48 --cfg=smpi/coll_selector:(mpich/ompi)
49 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
51 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
52 happening inside a collective SMPI call.
53 * Fix the behavior of complex data types handling.
54 * Make MPI_Wtime another synchronization point to take computations into account.
55 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
56 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
57 * Add all missing Fortran bindings, SMPI should work with Fortran 90
58 (no privatization of global variables yet)
61 * Verification of liveness properties is now available for SMPI applications
62 (in addition to MSG applications)
63 * Bugged examples using SMPI in examples/smpi/mc/
64 * Add --cfg=model-check/visited option. Allows the verification of infinite
65 programs. Detection of loops in the execution thanks to the system state
66 comparison and reduction of the state space to explore. Can be combined with
67 DPOR for safety properties.
70 * Allow to change SimGrid configuration (see --help) within the code
71 thanks to SD_config() as it can be done in MSG.
72 * Add a new function SD_task_set_amount() upon user request.
75 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
76 * Remove rule based routing (no more PCRE dependency)
77 * Add a limiter_link option to cluster tag, to specify a maximum reachable
78 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
79 * Add a loopback_bw and loopback_lat options to cluster tag.
80 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
81 routing. Now peers are encapsulated in an AS and have their own private
82 router but this is transparent.
85 * Our own implementation of getline is renamed xbt_getline, and gets
86 used even if the OS provide a getline(). This should reduce the
87 configuration complexity by using the same code on all platforms.
88 * New type: xbt_cfg_elm_boolean.
89 * Allow to use yes/no for boolean configuration options in the command line.
90 * Allow to disable SimGrid cleanups at exit from command line option.
91 There are situations where one may want a simulation to end with an exit.
92 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
93 annoying when scripting around the simulator. Adding a
94 --cfg=clean_atexit:no allows to circumvent this issue.
97 * New command line option --version, to get SimGrid version information.
98 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
100 * Supernovae builds are deprecated, and expected to be removed in the next
103 -- Mon Oct 7 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
105 SimGrid (3.9) stable; urgency=low
107 The Grasgory release: GRAS is really dead now.
109 * Complete overhaul of the internal host structures scheme.
112 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
113 was considered as experimental and badly maintained since then.
114 * Keeping it was thus a trap to our potential users, that could take
115 it instead of MSG or SMPI by mistake despite is pity state.
116 * GRAS seems to have very few users (if any), and no one volunteered
117 to maintain it further. It also induces a lot of XBT code (for
118 portability sake), that must be maintained too.
119 * For all these reasons, we killed GRAS. If someone wants to revive it
120 in the future, don't cry, our git history still remembers of GRAS.
123 * Major overhaul. Merge our documentation again as time proved that
124 splitting it was really not helping our users.
125 * Further improve the developer documentation to help newcomers
126 hacking on SimGrid itself. The user documentation (and in
127 particular, the beginner documentation) is still in a sorry state.
130 * Now works on Windows too!
131 * Much more extensive test suite, from MPICH
134 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
135 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
136 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
137 * Bug fix: let task be scheduled when the last dependency to be solved is
138 a control dependency.
139 * Remove SD_load_environment_script function.
140 Use the C sg_platf function if you want to declare a platform
144 * New function: MSG_process_get_number()
145 * Old function documented: MSG_config()
146 * Remove MSG_load_platform_script function
147 Use the C sg_platf function if you want to declare a platform
151 * Change the default value of the TCP_gamma constant (maximal size of TCP
152 congestion window) to a more realistic 4MiB value. If you notice changes in
153 your simulation results, you can fall back to the previous 20k tiny window
154 by adding --cfg=network/TCP_gamma:20000 on command line.
155 * (Hopefully) fix a bug wrt periodic availability and state traces
156 * Bug fix: use default values at start when first event in availability/state
157 trace is not at time 0.
160 * remove the "new_" part of function name sg_platf_new_trace_connect
161 (resulting in sg_platf_trace_connect), since it does not create
165 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
166 use case where it's really mandatory, and maintaining it was a pain in
168 * New: xbt_fifo_search(), search an item with a user-provided
169 comparison function instead of dumb pointer comparison.
172 * Fix the lua deployment:
173 Use `simgrid.init_application()` before deployment instead of
174 `simgrid.msg_register_application()` after.
177 * Transfer the tracing files into the corresponding modules.
179 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
181 SimGrid (3.8.1) stable; urgency=low
183 The "we are told that some people want to also *install* the simgrid
186 * Add missing file "tesh.1" to the archive.
188 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
190 SimGrid (3.8) stable; urgency=low
192 The Psssshiiiit release: SimGrid jumps into the Cloud.
195 * Add an experimental interface to manipulate VMs. They are mainly
196 process groups with very few intrinsic semantic, but they should
197 allow you to build the semantic you want easily.
198 * New function: MSG_host_set_property_value()
199 * New function: MSG_process_on_exit(). To clean memory in all cases.
200 * Bug fixes that made the host (and link) failures unusable.
201 * Add a way to auto-restart process when the host in which they are
202 executing comes back (ON_FAILURE="RESTART" on deployment file,
203 MSG_process_auto_restart_set).
204 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
205 please stop using the old ones, they are DEPRECATED.
207 * Deprecate functions MSG_global_init() / MSG_global_init_args()
208 Please use MSG_init() instead. (reducing the amount of entry
209 points in the library helps us).
210 * Make it impossible to link against the wrong version of the lib
211 * Deprecate MSG_clean(). No need to call it anymore.
212 * Function MSG_get_host_number() is not deprecated anymore.
215 * Split the doc into a user guide and a reference guide.
216 * Start a developper guide to help people hacking on SimGrid.
219 * Enable tracing by default. This modules rocks you should use it.
220 * Remove option custom_flags. Now use environment variables CFLAGS
222 * Use default cmake things to detect lua instead of home grown ones.
223 * New option "enable_mallocators" to disable mallocators, for debugging
224 purpose ("on" by default).
227 * Bug fixes around the resource failures: don't let the processes
228 survive the host they are running onto.
229 * Add an interface to auto-restart processes when the host in which they are
230 executing comes back.
231 * Ensures that SIMIX_clean is called automatically. It's not part of
232 the public interface anymore (bindings should be updated).
235 * Bug fix for when SD_Simulate is called with a positive value: be careful
236 when comparing doubles. Sometimes they are different for non significant
238 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
239 parallel task whose initial work is distributed among host according
240 to the Amdahl's law. Such tasks are created with a parameter alpha
241 that corresponds to the non-parallelizable part of the computation.
242 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
243 between two sets of workstations assuming a 1D block distribution (each
244 workstation owns a similar share of data) on both sides.
246 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
247 Data redistribution will be automatically scheduled once parent and child
248 are both scheduled. The filling of computation_amount and
249 communication_amount structures is now done seamlessly thanks to the chosen
251 * New function SD_workstation_dump to display various information
252 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
253 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
254 bandwidth on the route onto which the task is scheduled and the amount of
256 To divide the nominal bandwidth by 2, the rate then has to be :
257 rate = bandwidth/(2*amount)
258 * Compute tasks that have failed can now be rescheduled and executed again
259 (from their beginning)
260 * Increasing source code coverage (src/simdag is now covered at 95.8%
264 * Re-implement time-independent trace replay using SMPI (at the
265 smpi_smp_* level) instead of MSG. This should replace
266 examples/msg/actions/actions.c
267 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
268 hindexed and structs)
269 * Implement the exchange of non-contiguous data.
270 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
271 * Correct behavior of smpi/sender_gap and set its default value to 0
272 * Add option to asynchronously send small messages to allow better
273 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
274 specifies the size in bytes under which messages will be asynchronously sent.
275 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
276 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
279 * New C interface to define a platform: XML is now optional.
280 For more info, please check include/simgrid/platf.h.
281 * New interface to define random platforms from the C:
282 For more info, please check include/simgrid/platf_generator.h and
283 examples/msg/masterslave/masterslave_platfgen.c
284 * Export a sg_cmdline dynar containing all the arguments we got from
288 * Two new tracing options for adding comments to trace file so you
289 can track your experiments (see --help-tracing for details).
290 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
291 * Adding the SimGrid version that generated the trace file as a comment.
292 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
293 * Fix to avoid key clashes on Paje links
294 * Other minor fixes related to the Paje specification
297 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
298 and renamed to xbt_str_hash() and xbt_str_hash_ext().
299 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
300 * Greatly improve the robustness of mmalloc to user errors (such as
301 using an area after freeing it, or freeing it twice)
303 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
305 SimGrid (3.7.1) stable; urgency=low
308 * Restore the prototype of MSG_process_create_with_environment() to
309 the pre-3.7 situation by removing the kill_time argument.
310 * Add a MSG_process_set_kill_time() function instead.
313 * Fix weird behaviors when dealing with parallel tasks.
316 * Simgrid is now built as a dll.
317 * Simgrid-java now works on Windows.
318 * Simgrid-Java is now included into Windows package.
321 * First pre-build package for MacOSX.
324 * Fix compilation when using MSG_USE_DEPRECATED.
325 * Fix some compilation issues on Macs and Windows.
326 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
327 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
329 We discovered that the Lua console is broken, but we are missing the
330 manpower to fix it right now. The problem existed in 3.7 too, so we
331 are not blocking the release for that. Sorry if you depended on this
332 feature, any help would be really welcome.
334 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
336 SimGrid (3.7) stable; urgency=low
338 The "spring cleanups (before the next Big Project kicks in)" release.
341 * We can specify the SMPI latency/bandwidth factor with command line
342 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
343 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
344 You can also use the "config tag" from platform file by adding this line
345 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
346 (see "example/platforms/tag_config.xml" to use "config tag").
347 Note that the command line supersedes the platform file configuration.
348 * Change the correction factors used in LMM model, according to
349 the latest experiments described in INRIA RR-7821.
350 Accuracy should be improved this way.
351 * Use the partial invalidation optimization by default for the
352 network too. Should produce the exact same results, only faster.
353 * Major cleanup in surf to merge models and split some optimization
354 mechanisms from the core of the models. As a result you can now
355 specify which model to use (e.g., --cfg=network/model:LV08
356 --cfg=cpu/model:Cas01) and which optimization mode to use
357 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
358 Incompatible combinations should err at initialization. See
359 --help-models for the list of all models and optimization modes.
360 * The CLM03 workstation model was dropped for simplicity because it
361 used the deprecated CM02 network model. Use default instead.
362 * Rename the TCP_gamma configuration option to network/TCP_gamma
363 * Rename the coordinates configuration option to
364 network/coordinates, and document it
365 * Use now crosstraffic keyword instead of the terribly misleading
366 fullduplex keyword. It is activated by default now in the current
367 default model, use --cfg=network/crosstraffic:0 to turn it off.
368 * Ongoing refactoring the model parsing to make XML files optional
369 See include/simgrid/platf.h for details (still to be completed)
372 * Major overhaul of the documentation. Almost instructive now :/
373 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
374 functions and friends. This interface was considered as
375 deprecated since over 2 years, it's time to inform our users that it is.
376 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
377 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
378 use these (crappy) functions in your code.
379 These functions will be removed soon. Stop using them now.
380 * Deprecate MSG_get_host_{table,number}
381 Implement MSG_hosts_as_dynar() instead.
382 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
383 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
386 * Stabilize the parallel execution mode of user contexts
387 * Introduce configuration variables to control parallel execution:
388 - contexts/synchro: Synchronization mode to use when running
389 contexts in parallel (either futex, posix or busy_wait)
390 - contexts/parallel_threshold: Minimal number of user contexts
391 that must be part of a scheduling round to switch to parallel
392 execution mode (raw contexts only)
393 * Fix bugs that prevented to use suspend/resume along with
394 synchronization structures.
395 * Fix bugs in process termination that lead to invalid memory access
396 in very specific conditions.
399 * Introduce a parallel mode for the models (controlled by surf/nthreads
400 configuration item). In our tests, running the models in parallel
401 never lead to any speedups because they are so fast that the gain
402 of computing each model in parallel does not amortizes the
403 synchronization costs, even when ultra fast futexes are used.
404 This is released anyway because YMMV.
407 * Performance boost by using a swag internally to compute the set of
408 tasks that are finished and should constitute the return value of
412 * Enable it by default now that it is considered rather stable.
415 * Documentation of the tracing functions.
416 * Performance gains when tracing categorized/uncategorized resource
417 utilization by avoiding calls to get route when updating resource
418 variables. LMM constraints are being used instead.
419 * API changed to set task categories. Use MSG_task_set_category instead
420 of TRACE_msg_set_task_category, and SD_task_set_category instead
421 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
422 * Bugfix for graphicator, routes not correctly obtained, memory leaks
423 * Examples for link user variables added (see at examples/msg/tracing/)
424 * Deprecated function TRACE_msg_set_process_category completely removed
425 * Trace header updated according to the latest Paje file format
426 * Tracing network lazy updates, no longer obligate users to use full updates
427 * --cfg=tracing/platform:1 also registers power/bandwidth variables
428 * Experimental: let user code declare/set/push/pop application states for hosts
429 * API changed to allow the manual creation of graph configuration files
430 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
433 * Improve the API of Lua MSG bindings, using the Lua spirit.
434 * Each simulated process now lives in its own Lua world (globals are
435 automatically duplicated). It helps writing simulators. Will allow
436 to run Splay programs within SimGrid in the future.
437 * Add a Chord example in Lua, equivalent to the MSG one.
440 * Start the implementation of a solution to express temporal
441 properties, not only local assertions. This is still an
442 experimental work in progress, stay clear from it to be safe.
446 - Add new runtime parameters --help-logs and --help-log-categories
447 to display informations about supported logging parameters and
449 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
450 don't exists anymore.
451 * Mallocators: allow value NULL for the reset function.
453 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
454 create homogeneous dictionaries, where all the elements share the
455 same free function. Non homogeneous dictionaries will be
456 deprecated in the next release.
457 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
458 - Multi-level dictionaries are deprecated.
460 - new function xbt_dynar_search_or_negative() that is useful when
461 you have less than 2 million elements in your dynar and don't
462 want of the extra complexity of catching exceptions when the
463 element is not found.
465 - Make xbt_os_thread module (for thread portability) public.
466 Documentation is still to come, sorry.
468 - Cleanups and simplifications to make it maintainable again.
469 - Exotic features (such as memalign and valloc) were removed.
470 - The metadata were extended and improved so that the
471 model-checker becomes able to explore and inspect the heaps.
472 - This may induce a performance drop when enable_model-checking is
473 ON in cmake (even if it's not used in the simulation), but it is
474 necessary at this point to get MC working.
476 Turn model-checking OFF if simulation performance matters to you.
477 Not enabling it at runtime is not enough, disable it in cmake.
479 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
481 SimGrid (3.6.2) stable; urgency=low
483 The "Not coding new stuff allows to polish old things" release.
486 * New bindings to the NS3 packet level simulator (experimental)
487 * Use the raw (efficient) execution contexts instead of the sysv
488 (portable) ones when possible.
489 * libpcre is now mandatory in any cases since not using it led to
490 severe performance loss and possibly other issues
491 * Update the XML platforms:
492 - G5K: include the latest machine in Nancy
493 - GridPP and LCG: new platforms
494 * Documentation was partially updated, at least (more to come)
496 Bug fixes, cosmetics and small improvements
497 * Free terminated processes before the end of the simulation to avoid
498 exhausting the memory of users having very dynamic amount of
500 * Bug fix and cosmetics about canceling non-running tasks
501 * Bug fix about the dot loader's issues when using libcgraph
504 * Create an installer for windows with nsis (amd64 and win32)
505 - Add an hello world project to illustrate simgrid project creation.
506 - Embed libpcre into the Simgrid installer to avoid
507 its compilation burden
508 * The raw execution contexts should work on Apple now
509 * Port to Windows 64 bits
510 - Sysv contexts now have an implementation for this arch
511 - GRAS communication features now support this arch
512 * Drop support for borland compiler on windows
513 - this code was not maintained, and we kinda depend on gcc nowadays
514 * Fix portability issues on kfreebsd/gnu: build error about semaphores
515 * Fix portability issue on unstable ubuntu: linker became picky on
518 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
520 SimGrid (3.6.1) stable; urgency=low
522 The "Oops, we broke Macs too" release
525 * Fixed contexts detection so that raw ones are used when possible
526 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
527 leads to a strange error, with user code segfaulting sometimes when
528 the generated code is not perfectly aligned (which is not
529 controllable from the user side, depends on the amount of code)
532 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
534 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
536 SimGrid (3.6) stable; urgency=medium
538 The Summer Release, also known as the "OMG! They Killed Kenny!" version
541 * Bindings now constitute their own package, separated from the main one.
542 Rationale: reduce our maintenance nightmare by reducing the module coupling
543 They will soon be released on their own on gforge.
545 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
546 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
548 GRAS: It is not considered as stable anymore, but experimental. Sorry.
549 * It's not quite deprecated for now because we have no replacement,
550 but it may soon become the case.
553 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
554 * Fortran: New user-level cache variable to store the rank of the running
555 process. This improves performance by an order of magnitude.
556 * C: New coccinelle script to automatically locate and modify global and
557 local static variables.
558 * Improved SMPI network model with a sender-side gap to account for multiple
562 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
563 only say if a communication is finished, no matter whether it succeeded or
564 failed. You can call MSG_comm_get_status() to know the status of a finished
566 * New function MSG_task_dsend() to send a task and detach it. When a
567 communication is detached, you are never notified of its success or failure
568 and the memory is released automatically once it is finished. This function
569 is useful when you don't care about the end nor the success of a
571 * Change the prototypes of action replay. Sorry for inconvenience,
572 but this is really more efficient this way (and to adapt your code,
573 you just have to fix the initialization, that shouldn't be too long)
574 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
575 ever managed to do anything useful with it.
576 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
577 takes a dynar of communications. It returns immediately and gives the
578 index of a finished communication (if any).
579 * New example: a basic implementation of the Chord P2P algorithm.
582 * New model for multi-core CPUs. You can now use the core attribute to
583 precise the number of cores of a host. This is a basic model. Every
584 process running on the host receives at most the power provided in
585 the DTD (throughput<=power). Total throughput of process cannot exceed
587 * New peer tag. This peer tag creates a tiny AS comprising a host and a
588 router linked by an up-link and a down-link (possibly asymmetrical).
589 This kind of pattern allows to easily build last-mile model style platforms.
590 Aggregating such patterns in a rule-based AS is thus the technique of
591 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
592 * New model for Vivaldi routing. We transformed the Vivaldi network model
593 into a Vivaldi routing model (based on the rule-based model). This allows to
594 combine Vivaldi based latencies with last-mile platforms.
597 * Added a check for NaN of IEEE754 infinite in the double entries of
599 * Introduce a new context factory "raw", highly inspirated from the
600 ucontext factory, but using manually crafted functions in assembly to
601 do the work in an efficient manner.
602 * Allow to change the used context factory at run time, not only at
603 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
604 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
605 contexts at runtime (only with raw contexts or ucontexts).
606 * Completely rewrote this module to allow parallel execution of user
607 processes. Use --cfg=contexts/nthreads:N to execute user processes
608 with N parallel threads (the default is 1, meaning no parallelism).
609 * Allow to decide dynamically between sequential and parallel modes.
610 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
611 processes in parallel only when their number is greater than or equal to P
613 * Added a check for NaN of IEEE754 infinite in the double entries of
617 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
618 won't output the state of processes when interrupted with Ctrl-C
619 * Add a new function xbt_dynar_to_array that transforms a dynar into a
620 NULL-terminated array. This may solve backward compatibility issues
621 due to the change to return type of SD_simulate. See also:
622 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
623 * Add new macros with variable number of arguments.
624 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
625 - in xbt/asserts.h: xbt_assert
626 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
627 - in xbt/ex.h: THROWF and RETHROWF.
628 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
630 * Change xbt_die() to accept a format string with arguments, just like printf.
631 * New data structure: xbt_lib_t, like a dict but more general and with better
635 * New configuration options
636 Options triva/categorized and triva/uncategorized can be used to generate
637 graph configuration files for Triva visualization tool.
638 * Configuration option tracing/platform is renamed to tracing/categorized
639 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
640 * New TRACE_user_link_variable interface:
641 User provides the name of the link and the tracing variable to attach to it
642 * the declaration of tracing categories must be done after the environment
644 * simpler tracing interface, just one way to declare categories
645 TRACE_category or TRACE_category_with_color, it is up to you
646 * links in the trace file are again identified by their names
647 * trace contains the full platform hierarchy exactly as declared using the ASes
648 * Options tracing/msg/[task|process]:1 groups the process by hosts
649 for both cases, tasks and processes must have names that are unique during the simulation
650 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
651 * The experimental option tracing/msg/volume is deprecated
652 its functionality may be reincorporated if needed
654 The tracing generates a trace file with unordered timestamped events,
655 because of the way the core simulator (surf) works. A script available
656 at the tools directory (fix-paje-trace.sh) can be used to put the events
657 in order. We have changed the tracing so it can generate ordered timestamped
658 events in the final trace, but depending on the simulator (and how much time
659 is simulated) that can lead to a huge memory utilization. It is deactivated
660 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
663 * Define a SIMGRID_VERSION macro in simgrid_config.h.
664 - We are trying hard to keep the API stable, but it may happen that
665 some things change (we're a research project after all, not a
666 nuclear plant operating system). If such things should happen, you
667 could rely on that macro to adapt.
668 - current value: 30600 for 3.06.00, aka 3.6
669 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
670 a number that can be compared with SIMGRID_VERSION.
671 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
672 assertions and verbose/debug logging events are disabled at compile time.
674 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
676 SimGrid (3.5) stable; urgency=medium
679 * New feature to any SimGrid-based simulator: Model-Checking
680 Check SIN#1 for more details.
683 * New Model SMPI (three-interval linear regression for correction factors)
684 See RR-7426, available at http://hal.inria.fr/inria-00527150
685 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
686 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
687 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
688 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
689 * See also src/smpi/README
694 - Tracing API changes: TRACE_start and TRACE_end should not be called
695 by user-code. They are automatically called by simulators created
696 with SimDAG, MSG and SMPI if the toolkit is compiled with
697 tracing_enabled=ON. Categories declaration and utilization remain the
698 same for MSG and SimDag.
699 - A function was added to the tracing API to declare categories with
701 - TRACE_category_with_color (char *category, char *color)
702 where color must be in the following format
703 "%f %f %f", red, green, blue
704 and red, green, blue are float values in the interval [0, 1]
705 - User can specify NULL as color parameter, or continue calling
707 On that case, the tracing system will define random colors
708 - The following command-line options are supported:
709 --cfg=tracing/filename:msg.trace
710 --cfg=tracing:1 (activate tracing, needed to use others)
711 --cfg=tracing/platform:1 (categorized resource use)
712 --cfg=tracing/uncategorized:1 (uncategorized resource use)
713 --cfg=tracing/msg/task:1 (task creation)
714 --cfg=tracing/msg/process:1 (process creation, migration)
715 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
716 --cfg=tracing/smpi:1 (SMPI interface tracing)
717 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
718 - examples of examples/msg/tracing updated
720 - DAXLoader and DOTLoader functions can generate tasks with categories
721 - A new function to attribute a category to SD tasks:
722 TRACE_sd_set_task_category (SD_task_t task, char *category)
723 * Tracing the MPI interface implemented by SMPI
724 - Collective operations are traced with states
725 - Point-to-Point operations are traced with states/links
726 - Tracing activated by a parameter "-trace filename" passed
727 to smpirun during execution (considering that simgrid
728 is compiled with tracing enabled)
729 - To run the simulation with gdb, the simulator
730 accepts --cfg=tracing/smpi:1 to trace SMPI
731 - tesh files to check if smpi tracing is ok
732 - See examples/smpi/NAS/DT-trace
733 * GTNetS tracing re-worked
734 - adaptation to the tracing system of GTNets to cope
735 with modifications regarding the fullduplex mode
736 - new tesh files to check if gtnets tracing is ok
739 * Asynchronous communications through the functions:
740 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
741 * New function: MSG_load_platform_script()
742 to make possible using a lua script instead of XML files to set up platforms
743 * New function: MSG_set_function
744 to associate functions to processes, used when bypassing the parser
745 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
747 Platforms: Add some more examples in examples/platforms
748 * Grid'5000: see www.grid5000.fr
749 * *_30000_hosts.xml: various huge files [mainly scalability testing]
752 * Change the XML format. This is a very important modification. SimGrid
753 3.5 introduces a new hierarchical format based on the notion of
754 Autonomous Systems. Compatibility with old format is ensured
755 through the perl script provided in the install bin directory
756 bin/simgrid_update_xml.
757 It is now possible to build platforms with specific routing
758 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
759 connect several platforms together. We will try to provide soon
760 set of realistic platforms exploiting these properties (have a
761 look at examples/platforms/ for the moment).
762 * Take the opportunity of the XML format change to be a good XML citizen:
763 rename link:ctn to link_ctn and similar changes (also dealt with by
765 * Add a new routing scheme (rule-based) using regular expressions. It
766 enables to have an extremely low memory footprint when the
767 underlying routing is simple and can be compactly described. You
768 need to have libpcre4-dev (perl regular expressions) installed if
769 you want to use this routing scheme.
770 * Revive the cluster TAG and allow to easily and efficiently (both in
771 term of memory and speed) connect clusters together. Have a look
772 at teshsuite/simdag/platforms/ to see how this can be done. With
773 this tag, you can create clusters with thousands of tasks at no
774 cost (have a look at examples/platforms/).
775 Note: clusters are implemented as ASes, so there is no need for an
776 enclosing AS tag if you have only one cluster in your platform.
777 * Add new generic functions in the public interface that allows the user
778 to call SURF 'create_resource' methods from your code (same
779 functionality as the XML bypass mechanism but with a much lighter
781 * Add a new model (enabled through command line --cfg=network/model:SMPI)
782 that uses a piecewise linear approximation to produce better
783 results when exchanging small messages.
784 * Add a new parameter to handle correctly full duplex link and account
785 for interferences between uplink and downlink communications
786 (activate with --cfg=fullduplex:1).
789 * Rename the SD_READY (all dependencies are satisfied and task is
790 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
791 dependencies are satisfied) state.
792 This prevents a confusion between the notion of "ready to schedule"
793 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
794 simulated" (SD_RUNNABLE) used by the simulation kernel.
795 * Change the way a task is considered as ready. Instead of removing
796 dependencies when a task is done, a counter is decreased. This way,
797 it is always possible to reach ancestors thanks to the
798 SD_taks_get_parents function (even after the end of the simulation.)
799 * Change the return type of SD_Simulate from (SD_task_t*) into
800 xbt_dynar_t. This function was in handling a dynar internally and
801 converted it into a NULL terminated array for historical reasons.
802 * New function SD_dotload(char*) to load a DAG described in dot
803 format. This loader and the corresponding examples require the
804 installation of the graphviz library.
805 * Fix a bug in the management of tasks of size 0 in the surf network
806 models. This problem was only visible with SIMDAG and you should
807 thus disregard results produced with earlier versions if you
808 relied on this feature (some tasks were blocked because of this).
809 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
810 even though your applications were doing only point-to-point
811 communications and sequential computations. Now you can really use any
812 model you want (of course, if you create real parallel tasks, which are
813 not implemented in most models beside ptaskL07, this will abort).
814 * Add an example that schedules a DAX on an heterogeneous platform
815 using a Min-Min strategy.
816 * New function SD_workstation_get_current_task() that returns the kind
817 of task currently running on a workstation in the sequential access
819 * Raise some warnings when unexecuted tasks remains at the end of the
820 simulation. This is usually caused by cycles in the DAG.
823 * New function: SIMIX_process_set_function() called by MSG_set_function
824 * Change the underlying waiting queue in semaphores so that a process
825 can wait on several of them simultaneously (as in waitany).
826 * Fix the way to handle tokens in semaphores so that all access patterns
827 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
828 * kill the dirty pimple SIMIX_message_sizes_output()
829 Please use (proper) visualization instead
832 * New data container: setset (set of sets of elements)
833 * New module: mmalloc (mapped malloc, allowing to have several
834 independent segments of malloc)
835 * New function: xbt_dict_cursor_set_data()
836 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
837 * New function: xbt_dynar_is_empty()
838 * New function: xbt_fifo_get_last_item()
839 * Fix xbt_dynar_shrink(): use the right element size.
840 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
841 uninitialized areas during expand.
842 * Fix semaphores: previous implementation was severely broken.
843 * Use library init/fini functions for our initialization.
844 - you can use logs and other feature as soon as you want in your
845 code (even before the xbt_init / MSG_init)
846 - xbt_exit is now a no-op and produce a warning when used.
849 * Port GRAS to new SIMIX mechanisms. This allows gras users to
850 benefit from the latest improvement to the simulation kernel.
851 * Kill measurement sockets for now. If you rely on them, sorry. This
852 release is not for you. This feature will be reintroduced in the
853 future, but we cannot delay the release any further.
854 * New function: gras_msgtype_get_name().
855 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
856 * Fix (at last) the pmm example: it should not randomly fail anymore.
858 Build chain: bug fixes and overall polishing
859 * Cmake is now stable enough. Hence, we killed the autotools.
860 * Port to windows ( TM :)
861 * Fix the 'make install' target.
862 No need to use 'make install-simgrid' anymore
863 * Introduce a 'make dist' target compiling a *source* archive
864 'make package' compiles a binary archive
865 * Compile java files only on need
866 * Add --cd and --setenv command line options to tesh
867 * Out of source builds are not fully supported yet, but we are close
868 * Enable supernovae and optimization flags by default for our users
871 * Add layer to set up environment directly from lua, without XML.
872 * The effect of gras_stub_generator can be achieved through
873 lua too (check examples/gras/console/ping_generator.lua)
875 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
877 SimGrid (3.4.1) stable; urgency=low
879 The "Polishing easter eggs is probably a good idea" release.
880 This is a bug fixes release only.
884 * Fix a bug preventing the tasks from begin garbage collected.
887 * Fix a bug occuring when a host involved in a communication fails.
888 This was not detected properly by the other peer involved in the
889 communication. Now, it's reported as a network error.
892 * Warn the user about loop dependencies in data flow of DAX files
893 * Obey the control-flow dependencies of DAX files
896 * Add option "enable_smpi" allowing to not compile SMPI.
897 Probably useful for the (Mac) users experiencing a build error here
898 * Improve the detection of lua5.1 and ruby1.8
900 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
902 SimGrid (3.4) stable; urgency=low
904 The "Easter in Cargese" release. Also known as (major changes):
906 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
907 ~> bindings were greatly improved
908 ~> new tracing infrastructure for better visualization introduced
910 * the "Welcome to configury modernity" release.
911 ~> we switched from autotools to cmake, and improved our cdash
914 A more detailled list of changes follow (full detail in svn log).
916 Java Bindings: Various Cleanups
917 * (install java-gcj-compat-dev on debian-like to use them)
918 * Remove put/get: no need to export deprecated interface in Java
919 Use send/receive instead.
920 * Cleanup the examples and add a README per directory
921 * Remove example autoDestination (that's the only way to go now)
922 * Remove example explicitDestination (was a plain copy of basic)
923 * Make JniException a runtime exception, so that there is no need to
924 declare the fact that you may encounter such a beast. I guess that
925 nobody will ever want to survive such error.
926 * Create specific errors for each MSG case of failure:
927 host failure, transfer failure, timeout, task cancelled
928 * Cleanup the exceptions that may get thrown by each function
929 * Other internal cleanups in Java bindings. Performance still bad :/
930 Ruby and Lua Bindings: create them
931 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
932 * That's new and great, you should try them out.
933 Same functionalities than Java bindings, only even less polished
935 * Kill the useless "rate" argument of SD_task_get_execution_time()
936 Everyone used to provide -1 as a value, it was not used, and the
937 semantic of a possible use wasn't even clear.
938 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
939 either as comm costs or compute costs to mean that there is no
940 such thing for that specific task.
941 * Add a SD_task_set_name() function
942 * Fix SD_task_unschedule() on typed tasks
943 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
944 * In DAX loader, accept useless 'level' attributes to <job> since
945 LIGO DAGs have them (seem to be to ease graphical representation).
947 * Add an example masterslave_mailbox.c using send/receive and not
948 the deprecated put/get interface.
949 * Kill the MSG_paje_output() function. It's a noop since 2 years.
950 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
952 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
953 (declare MSG_USE_DEPRECATED to still have the old name)
954 * Add a MSG_task_set_data() function
955 * About trace replay (see examples/msg/actions):
957 - Allow to work with splitted trace files for each process
958 Give the specific trace file as argument of each process,
959 and call MSG_action_trace_run(NULL)
960 You can still have one merged file for all processes.
961 - Fix implementation of collective operations
962 * Allow task_execute() on 0-sized tasks (closes #10063)
964 * This is the first release of SimGrid where SMPI is not considered
965 beta anymore (even if some corners should still be improved)
966 * Port over the new SIMIX_network submodule (internal refactoring)
967 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
968 * Implement more missing elements of the standard:
970 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
972 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
973 MPI_Scan() MPI_Get_processor_name()
974 - Added implementation of missing case for Alltoall (warning: it's
975 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
976 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
977 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
978 SMPI_MPI_Allgatherv()
980 - MPI_Waitsome() was broken
981 - Allow relative includes in smpicc
982 - Command line cfg argument 'reference_speed' was ignored...
983 - Some functions did not properly lead to auto-benching of user code
984 - smpicc passes -O2 by default (just like openmpi one)
986 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
987 * Bug fixes about timeouts during communications
988 * add SIMIX_message_sizes_output() as a pimple to write to file the
989 amount of messages per size. Use gnuplot to get histogram.
990 Pimple because that's the only user-visible function of simix,
991 defined directly in xbt.h (irk, sorry)
993 - Add a SIMIX_sem_get_capacity() function
994 - Fix interactions with processe resume/suspende
995 - release_forever() was stupidly broken
996 - Fix SIMIX_display_process_status() for processes in a semaphore
997 - Make SIMIX_sem_block_onto() user-visible
998 * Refactoring context stuff:
999 - Use pseudo-OOP for better modularity
1000 - reimplement SIMIX_process_kill() without process_schedule() so
1001 that the latter can take as invariant that it is called from
1003 - Merge context_start into context_new for sake of simplicity
1005 * Add a Vivaldi network model, coded live during SUD'10 ;)
1006 * Rename configuration variables to start a hierarchy:
1007 o cpu_model -> cpu/model
1008 o network_model -> network/model
1009 o workstation_model -> workstation/model
1010 * New configuration variables:
1011 o network/bandwidth_factor: correction to bandwith
1012 o network/latency_factor: correction to latency
1013 o netwotk/weight_S: correction to the weight of competing streams
1014 * Add a long description to the models, that users can see with such
1015 argument on the command line: --cfg=cpu/model:help
1016 * --help-models display the long description of all known models
1018 * config: add the ability to set a default value after registration
1019 Does not override any previously set value (e.g. from cmd line)
1020 * dict: allow to have integer key and data.
1021 When so, you need to use the following functions
1022 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1023 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1024 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1025 In contrary to regular dicts, the key is not malloced before copy.
1026 Mixing scalar and regular elements in the same dict is not tested
1028 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1029 Tracing for Visualization:
1030 * SimGrid is now instrumented in order to generate a trace file for
1031 visualization analysis: to use it, need to compile SimGrid with the
1032 "tracing" option enabled, and instrument the program using SimGrid with
1033 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1034 (among other functions).
1035 * The instrumentation only traces the platform utilization for now
1036 * Documentation to use the tracing functions and how to analyze the
1037 traces with the Triva tool is written.
1038 * More information about: SimGrid FAQ (in the section Tracing Simulations
1041 * We moved to cmake as default build system. Autotools support will
1042 be dropped soon. Check the FAQ for more info about how to use it.
1043 * Greatly improved our cdash/ctest interactions
1044 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1045 * Added memory checking tests with valgrind; lot of memleak fixing.
1046 This may be the first release of simgrid with so few memory issues
1047 * Added code coverage tests.
1048 Our coverage is still improvable, but at least we see it on cdash.
1050 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1052 SimGrid (3.3.4) stable; urgency=low
1054 The "Desktop Grid needs love too" release (also called Xmas release).
1056 Models improvements:
1057 * Major speedup in the maxmin system solving by using lazy evaluation
1058 Instead of solving completely the maxmin system at each iteration,
1059 only invalidate (and recompute) the modified parts.
1060 This new feature is enabled in default models but you can try to
1061 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1062 * Cas01 IMproved as default CPU model
1063 This CPU model is the same Cas01 model, but it uses the
1064 maxmin-selective-update flag and a heap structure to manage
1065 actions on SURF kernel.
1066 It reduces the complexity to find the next action to finish and,
1067 consequently, it's faster than the old Cas01.
1068 This is the new default CPU model (Cas01).
1069 * Rename the old Cas01 model to Cas01_fullupdate
1070 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1071 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1072 * CpuTI (CPU Trace Integration)
1073 A new CPU model whose objective is simulate faster when using
1074 availability trace files.
1075 Instead of using a full featured, over engineered maxmin system for
1076 CPU modeling, this model does the pre-integration of traces files
1077 to calculate the amount of CPU power available, and so, executes
1078 faster than the old CPU models.
1079 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1080 * Use LV08 as default network model since it gives better accuracy
1081 for small messages and shouldn't change things for big ones.
1082 Use --cfg=network_model:CM02 to get the previous behavior.
1085 ******************************************
1086 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1087 ******************************************
1088 * The new CPU model may changes simulations!
1089 The point is that events occurring at the exact same timestamp
1090 are not scheduled in the same order with the old and new
1091 version. This may be enough to completely change the execution
1092 of simulations in some cases.
1093 * The new network model will change simulations!
1094 This new model is more realistic than the previous one, so you
1095 should consider redoing your old experiments with this model.
1096 Sorry for the inconvenience.
1099 * Introduce the supernovae compilation mode
1100 When compiled that way, the whole SimGrid (or almost) is put in a
1101 single compilation unit and compiled in one shoot.
1102 This is to help gcc which has difficulties to inline stuff from one
1104 The speedup seem to be above 15%, althrough more tests are needed on
1105 amd64 to confirm that gain.
1108 * Port of MSG's mailbox on top of SIMIX network
1109 The put/get mechanism was greatly simplified on the way.
1112 * New SIMIX network module. Provides:
1113 - Mailbox: rendez-vous mecanism to find with who you want to speak
1114 - Synchronous send/recv: easier and hopefully faster since the
1115 logic is handled in the maestro process directly now
1116 - Asynchronous send/recv: you dreamt of it? It's here now
1117 Too bad that nobody cared enough to propagate the change to MSG.
1118 * Add semaphores as SIMIX synchronization mechanism.
1121 * new function SD_daxload(char*) to load a DAX file
1122 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1123 * Introduce typed tasks. Specify its kind and cost at creation.
1124 At scheduling, just give where it should be placed, and the cost
1125 for each involved resource is automatically computed.
1126 Existing constructors so far (more to come of course):
1127 - SD_task_create_comm_e2e() for end-to-end communication
1128 - SD_task_create_comp_seq() for sequential computation
1129 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1130 * new function SD_task_dump() for debuging display
1131 * new function SD_task_dotty(task,FILE*) writing to file the info
1132 about the task in dotty format
1133 * SD_task_dependency_exists() can now cope with having one of its
1134 arguments NULL. If so, it tests whether the other argument has any
1136 * Add getters on list of preceding/following tasks:
1137 SD_task_get_parents(task) and SD_task_get_children(task)
1138 * Add getters on amount of workstations and list:
1139 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1140 * Add getter on task kind: SD_task_get_kind(task)
1141 * Update the start_time and finish_time of tasks on completion/failure
1142 * Bugfix: Remove task from state swags when destroyed
1145 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1148 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1149 dynar with one separate thread per value of the dynar.
1150 * Change the prototype of xbt_thread_create(), sorry.
1151 Added a boolean parameter indicating whether we want to join this
1152 thread (used in SG only for now)
1153 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1156 * GTNetS wrappers should now be usable again (and betterly tested too)
1157 * Fix a major regression from 3.2 where the timeout provided to
1158 MSG_task_put_with_timeout() was used as absolute time before which
1159 the comm should be done.
1160 * Start to fix the <cluster> tag.
1161 - Internal links should be good now (beside of the loopback, which
1162 use the private link instead)
1163 - paths to the external world is still rather broken
1164 - the <route:multi> tag is just broken. Actually that's brain-dead.
1165 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1166 to make it less stupid
1167 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1168 * Fix a source-level compatibility glitch from 3.2: after defining
1169 MSG_USE_DEPRECATED, you can use the old name
1170 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1171 * Allow to compile from the SVN with automake 1.11
1172 * Fix some problems when using the "start_time" tag in deployment XMLs.
1173 * Fix #8569: XBT/synchro.h has redundant declarations
1174 * Fix #8563: MSG return values and exceptions
1175 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1176 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1177 Thanks to Robson Peixoto.
1178 * Fix a vicious bug in dictionaries inducing that some elements were
1179 not freed on xbt_dict_free()
1181 Portability report of this version:
1182 * Main portability targets:
1183 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1184 on (amd64/i386/ia64)
1185 - mac leopard on i386
1186 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1187 but nothing critical.
1188 * Other platforms: windows, AIX and others were not tested for this release
1190 Timing report of this version:
1191 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1192 scenario parameters). From 8h to a few seconds in desktop grid settings.
1193 * Supernovae brings about 25% speedup on i386.
1195 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1197 SimGrid (3.3.3) stable; urgency=low
1199 The "Need for Speed" release.
1201 The timings done to validate the 3.3.2 were faulty.
1202 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1204 The problem was a conversion from a manually handled vector to
1205 xbt_dynar_t on the critical path.
1206 xbt_dynar_foreach calls functions, inducing stack management crap.
1208 We inlined these functions and xbt_dynar_foreach is now breath taking.
1209 We also inlined xbt_swag_belong on the way.
1211 Here are some approximate speedup measurements (on master/slaves
1212 simulations lasting between 10s and 20s each):
1213 3.3.1 -> 3.3.2: about same performance
1214 3.3.2 -> 3.3.3: 40% speedup
1215 3.3.1 -> 3.3.3: 40% speedup
1216 3.3.1 with inline patch -> 3.3.3: 30% speedup
1218 Our reading is that the refactoring which occurred in 3.3.2 made us
1219 suffer much more from the xbt_dynar_foreach low performance, but
1220 once we solved this, this refactoring proved to be very performance
1221 effective. From the 40% speedup, somehow, 10% are due to the
1222 inlining and 30% to the refactoring.
1224 That's a pitty that gcc cannot inline functions placed in other files
1225 alone. We have to choose between:
1226 - break the encapsulation (by putting private data structures and
1227 accessors in headers files to help gcc)
1228 - live with low performance
1229 - switch to a decent compiler such as icc (not quite possible).
1231 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1233 SimGrid (3.3.2) stable; urgency=low
1235 The "Simplicity does not preceed complexity, but follows it" release.
1237 The main contributors of this release were (lexical order):
1238 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1241 * Extract the routing logic into its own object.
1242 (was dupplicated in network.c and workstation_LV07.c;
1243 Allows to implement other ways of storing that info)
1244 => kill now useless network_card concept
1245 - Use dynar to represent routes (instead of void** + int*)
1246 - kill link_set (use surf_network_model->resource_set instead)
1247 - Add a command-line option to choose the routing schema to use
1248 - Add three new models:
1249 * Floyd (shortest path computed at initialization)
1250 * Dijikstra (shortest path recomputed all the time)
1251 * Cached Dijikstra (shortest path computed on need)
1252 All these models where contributed by Silas De Munck, and are
1253 described in his ICCS09 paper.
1255 * Simplify model declaration
1256 (less redirections, less function to write when defining a model)
1257 - Factorize stuff between models:
1260 surf_model_resource_set(model)
1261 surf_model_resource_by_name(model, name)
1262 - Unify the types of models in s_surf_model_t (using an union)
1263 - Embeed fields of common_public directly into s_surf_model_t
1264 - Rename model methods:
1265 action_free ~> action_unref
1266 action_change_state ~> action_state_set
1267 action_get_state ~> action_state_get
1268 - Change model methods into functions :
1269 (model)->common_public->action_use ~> surf_action_ref
1271 * Implement a generic resource; use it as ancestor to specific ones
1272 (allows to kill duplicated code in models)
1273 Drawback: timer command don't need no name nor properties;
1274 workstation_CLM03 don't need no properties
1275 (but I guess we can live with those few bytes wasted)
1277 * Improve the action object model
1278 - implement a constructor avoiding dupplicated code about field
1279 initialization in generic_action part.
1281 * Kill the SDP model: it has an external dependency, is deprecated
1282 in flavor of modern lmm models, and didn't compile since a while
1285 * Relocation of the context module from XBT to SIMIX.
1286 (the context were decoupled from the simix processes, duplicating a lot of code)
1287 => a lot of code was factorized
1288 - less overhead is introduced during scheduling
1289 - simpler API for the context factory
1290 - the logic for process creation,destruction and manipulation was simplified
1291 * Simplification of the s_smx_process_t data structure.
1292 => accesing the simix level data associated to a process is faster now,
1293 and the code is a lot more readable.
1296 * Implement some more MPI primitives:
1297 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1298 -implementation: Bcast: flat or 2-ary tree (default),
1299 Barrier: 4-ary tree,
1301 Allreduce: Reduce then Bcast
1302 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1303 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1304 Alltoallv: flat tree, like ompi
1306 * Add support for optimized collectives (Bcast is now binomial by default)
1307 * Port smpirun and smpicc to OS X
1310 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1311 Shout out if you used it.
1314 * Display the list of still queued messages in SG mode when existing
1318 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1319 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1320 * Add xbt_str_from_file(FILE*)
1321 * Add xbt_dict_get_key achieving a linear reverse search
1322 * Remove the context module
1324 Portability report of this version:
1325 * Main portability targets:
1326 - Linux(debian)/x86/context
1327 - Linux(debian)/x86/pthreads
1328 - Linux(debian)/amd64/context
1329 - Linux(debian)/amd64/pthreads
1330 On these, we still have the eratic breakages of gras/pmm and
1331 amok/saturate_sg reported in previous version. We still think
1332 that the tests are the cause of the fault, not the tested code.
1334 - Mac OSX Leopard/x86/context
1335 Still false negative in tesh autotesting.
1336 Smpi still fails, but this time because readlink does not accept -f
1337 Everything seems to work properly beside of that.
1340 - AIX version 5.3 (only tested contexts this time)
1341 Smpi still fails there because mktemp is not installed.
1342 Everything seems to work properly beside of that.
1344 I managed to compile it for the first time, but several breakages.
1345 Won't delay the release for this exotic platform.
1347 * Windows: it's still lagging behind. If you want to help, please
1350 Timing report of this version:
1351 This version seem to be more than 5% faster than 3.3.1 (on linux
1352 64bits with contextes). The gain is less than expected, we are
1353 investigating this for next release.
1355 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1357 SimGrid (3.3.1) stable; urgency=low
1360 * Implement a --cfg-help to show existing configuration variables
1361 * Build chain do not require doxygen in maintainer mode
1364 * fix a bug on struct sizeof computation, which prevented the
1365 exchange of arrays of structs in some conditions
1366 - added a regression test about this in datadesc_usage
1367 * Allow the exchange of 0-long dynamic vectors.
1368 - for that, use -1 as indicator of dynamic size instead of 0
1369 - This implied to change any size from unsigned long to long,
1370 reducing a bit communication abilities, but I guess that with
1371 64bits being quite common, this is more than enough.
1372 - This also induce a protocol change, thus bumping network protocol
1373 version from 0 to 1 (if we have external users, we have to get
1374 clean on that point too ;)
1375 - added two regression tests about this in datadesc_usage
1376 * Be more verbose when propagating local exceptions
1377 This helps debugging.
1378 * Display the status of simulated processes when receiving SIGINT in
1382 * Allow to control the simulation from a trace file.
1383 New functions MSG_action_register() and MSG_action_trace_run()
1384 The first one allows to associate a function execution to each
1385 kind of action while the second one parses a trace file and
1386 triggers the corresponding actions within the system.
1387 For now, only a toy example is provided in examples/msg/actions
1388 * Add an exemple of process migration in examples/msg/migration
1389 * Fix a bug in task exchange which broke MSG_task_get_sender()
1390 Add a teshsuite regression test for that.
1391 [Bug: if MSG_task_get_sender() is called after sender exit,
1393 * Fix a bug which prevented suspend/resume to work properly
1394 * Display the status of simulated processes when receiving SIGINT
1395 This fixes a regression of v3.3. due to the introduction of SIMIX
1396 * Bug fixing in failure management:
1397 - trace could not start by a failure at time 0
1398 - failure during communications were not working
1401 * Add SIMIX_process_set_name() to change the name of the current
1402 process in the log messages.
1403 * Store smx_hosts in a dict since we only retrieve them by name
1404 * Move the configuration infrastructure to surf
1407 * Move the configuration infrastructure to surf
1410 * Massive internal cleanups:
1411 - Store internal structures on processes instead of hosts (allows
1412 to have more than one process per host, in addition of being more
1414 - Cleanup the initialization/finalization process
1415 - Kill a whole bunch of unneeded synchronization:
1416 processes run in exclusive manner within the simulator
1417 - Move queues from global tables to process data fields
1419 - now accept -platform and -hostfile arguments
1420 - Pass the right rank value to processes according to the hostfile
1421 * Compile the examples by default, and use them as regression tests
1422 * Implement MPI_Wtime()
1423 * Change the reference speed to a command line option
1426 * TCP_gamma can now be specified as command line option using
1427 --cfg=TCP_gamma:10000000.0
1428 * Change the --surf-path cmd line option into --cfg=path:
1431 * Also include strbuff from xbt.h public header
1432 * xbt_ex_display(): do not free the exception after displaying
1433 This allows to do more with the given exception afterward.
1434 Users should call xbt_ex_free() themselves.
1438 Portability report of this version:
1439 * Main portability targets:
1440 - Linux(debian)/x86/context
1441 - Linux(debian)/x86/pthreads
1442 - Linux(debian)/amd64/context
1443 - Linux(debian)/amd64/pthreads
1444 These targets fail about 1/10 of times on gras/pmm, but we believe
1445 that this is because of the test, not because of simgrid.
1446 amok/saturate_sg fails even more rarely, and the test may not be
1449 - Mac OSX Leopard/x86/context
1450 The test suite still spits tons of errors because some obscure
1451 force prevents us from removing the temporary directories
1452 arguing that they still contain some metadata I've never heard of.
1453 Smpi fails because seq is not installed.
1454 Everything seems to work properly beside of that.
1457 - AIX version 5.3 (both contexts and pthread)
1458 Smpi still fails there because mktemp is not installed.
1459 XML inclusions seems rosty on AIX.
1461 * Windows: it's still lagging behind. If you want to help, please
1464 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1466 SimGrid (3.3) stable; urgency=high
1470 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1471 [Malek Cherier & Mt]
1473 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1476 Until now, GRAS were using MSG as an interface to SURF. It was
1477 quite difficult because both interface have several differences
1478 (MSG channels vs GRAS sockets were the most notable point).
1480 This also opens the gate to SMPI (which should occur soon) and speed
1481 up simulations by to 40% (even if it were not the main goal).
1483 **************************************
1484 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1485 **************************************
1486 The point is that events occuring at the exact same timestamp are
1487 not scheduled in the same order with the old and new version. This
1488 may be enough to completely change the execution of simulations in
1489 some cases. Sorry for the inconvenience.
1491 * Cleanup and upgrade the XML format to push further scalability
1492 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1494 * Improve the testing infrastructure with tesh. Now a very large part of
1495 the code is tested not only by being run but also by checking that the
1496 output match an expected output [Mt].
1498 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1499 is really less memory-demanding, which should allow you to use
1500 larger files in SimGrid [AL].
1502 * Inform valgrind about our contextes, so that it becomes usable
1503 with the default (and more effecient) version of SimGrid
1504 [contributed by Sékou Diakite, many thanks]
1507 * Introduce a listener thread in charge of receiving incoming
1508 messages from the network. It allows to overlap communication and
1509 computation but most notably, it removes some stupid deadlocks due
1510 to the fact that so far, a process could not send and receive at
1511 the same time. This made most non trivial communication schema
1513 * Convert the PIDs from long int to int to match the MSG ones (and
1514 linux ones too) [Mt]
1515 * New function: gras_agent_spawn() to launch a new process on
1516 current host. Only working in simulation for now. [Mt]
1517 * New function: gras_os_hostport() returning a constant form (ie,
1518 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1521 * Make the backtrace of exceptions more human readable [Mt]
1522 * New module: xbt/str [Mt]
1523 a ton of string utility functions (split, join, printf to a newly
1524 allocated buffer, trim, etc)
1525 * New module: xbt/hash [Mt]
1526 SHA1 hashing algorithm (more to come if needed)
1527 * New module: xbt/synchro [Mt]
1528 synchronization tools (mutex and conditions) working the same way
1529 in simulation and in real life (mainly useful for GRAS, but not
1531 * New module: xbt/queue [Mt]
1532 classical producer/consumer synchronization scheme
1533 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1534 (using the classical functions will get serialized) [Mt]
1535 * Make dictionary internal table dynamic. No need to specify its size
1536 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1538 * Make sure the log channels are organized as a tree under windows
1539 (because of ANSI C compatibility issue, any channel were child of
1540 root directly) [Mt].
1543 * Cleaned many thing in surf and fixed a few bugs [AL].
1544 * Add a nice command line configuration mechanism to compose models [AL].
1545 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1546 the previous one (KCCFLN05). It relies on something that looks like
1547 a max-min sharing mechanism but cannot be written as such. A new solver
1548 was thus designed [AL].
1549 * Add a new solver to lmm. Based on Lagrange optimization and
1550 gradient-based descent, it enables to efficiently maximise systems s.a
1552 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1554 This solver enables to propose two new network models for TCP Reno and
1555 TCP Vegas based on Low's work. These models still need to be fully
1556 tested though [Pedro Velho].
1559 * Bug fix in SD_simulate. Now the time bound given as argument is
1561 * Use the new parallel task model (ptask_L07) as default.
1562 * Use the SURF command line configuration mechanism.
1563 * 0-size tasks (for synchronization) should now work.
1565 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1567 SimGrid (3.2) stable; urgency=high
1571 We still experience issues on this platform, but we believe that at
1572 least MSG is usable.
1574 GRAS API BREAKAGE (for simplification purpose, sorry):
1575 * the gras_msgtype_by_name is not used anymore. Instead of
1576 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1577 you can write (and must)
1578 gras_msg_send(toserver, "request", &request);
1579 - If you still want to pass a gras_msgtype_t to the function (to cache
1580 the type and avoid the lookup time), use the gras_msg_send_() variant.
1581 - Impacted functions:
1582 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1583 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1584 * The callbacks are now expected to return 0 when everything went well
1585 (just like the main() function)
1587 GRAS new features and improvements:
1588 * New module mecanism where user code can use per process globals [Mt]
1589 This is similar to gras_userdata_*() functions, but for libraries. It
1590 factorize some code developped over and over in the examples and AMOK.
1591 It has still to be documented and used (only amok/peermanagement is
1593 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1594 loss when they were small enough to fit into the buffer and sent quickly
1595 enough so that they can all get received in one shoot.
1596 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1597 if not found. Use the *_or_null() variant for the old semantic.
1598 * In gras_msg_handle, do not discard messages without callback.
1599 They are probably messages to be explicitly awaited later (ie, proofs of
1600 mis-synchronization in userland since they are sent before being awaited)
1601 No big deal usually.
1602 * gras_socket_meas_send/recv: semantic changed!
1603 The numerical arguments used to be (1) the total amount of data to send
1604 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1605 messages. This was need for the fool willing to send more than MAXINT
1606 bytes on quite fat pipes.
1609 * Do really rename the hostmanagement module to peermanagement. [Mt]
1610 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1611 but this is rather new and this was documented in the module
1612 documentation (poor excuses, I admit)
1613 * Bandwidth measurement semantic changed! This follows the changes to
1614 gras_socket_meas_send/recv explained above.
1617 * A sequential mode has been added to the workstations. When a workstation
1618 is in sequential mode, it can execute only one task, and the other tasks
1619 are waiting in a FIFO. [Christophe Thiery]
1622 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1623 model for SIMDAG. [Christophe Thiery]
1624 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1625 equal to 0, which caused some bad behaviors in
1626 saturated_constraint_set_update. I now use a threshold mechanism like in
1630 * When running manually src/testall, you select specific units [Mt]
1631 testall is the result of our cunit mecanism, and should replace all
1632 the scripty thingy around since bash don't run easily on billware.
1634 * A mallocator system has been added. [Christophe Thiery]
1635 Mallocators allow you to recycle your unused objects instead of freeing them
1636 and allocating new ones.
1638 Documentation update:
1639 * FAQ reworking + New FAQs:
1640 - "Valgrind spits tons of errors!" [Mt]
1641 - "How to repport bugs" [Mt]
1642 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1643 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1644 - Communication time measurement within MSG [AL]
1645 - I experience weird communication times when I change the latency [AL]
1646 * GRAS tutorial [Mt]
1648 - an introduction to the framework and to the used communication model
1649 - an initiatic tour introducing the most proheminent features:
1651 . Lesson 0: Installing GRAS
1652 . Lesson 1: Setting up your own project
1653 o Part 2: Message passing
1654 . Lesson 2: Exchanging simple messages
1655 . Lesson 3: Passing arguments to the processes (in SG)
1656 . Lesson 4: Attaching callbacks to messages
1657 . Lesson 5: Using globals in processes
1658 . Lesson 6: Logging informations properly
1659 . Lesson 7: Using internal timers
1660 . Lesson 8: Handling errors through exceptions
1661 . Lesson 9: Exchanging simple data
1662 . Lesson 10: Remote Procedure Calling (RPC)
1663 . Lesson 11: Explicitely waiting for messages
1664 . Recapping of message passing features in GRAS
1665 - A HOWTO section containing:
1666 o HOWTO design a GRAS application
1667 More are due, of course. They will come latter. In the meanwhile, you can
1668 check the examples which are still here.
1670 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1672 SimGrid (3.1) stable; urgency=high
1676 There was a stack corruption somewhere, visible only when optimizing
1677 with these versions. [Vince]
1680 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1681 module that enabled to play with Directed Acyclic Graphs. It is built
1682 directly on top of SURF and provides an API rather close to the old
1683 SG. Some old codes using SG are currently under rewrite to check that
1684 all needful functions are provided. [Christophe Thiery]
1687 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1688 extension of the classical CLM03 model that gracefully handles
1689 failures. This is now the default model for MSG and GRAS. It doesn't
1690 handle parallel tasks yet though. [AL]
1691 * Bug fix: Weights were not correctly set in the network part.
1692 WARNING: This may have resulted in incorrect results with simulations
1693 where there are more than one flow on a given link. [AL]
1696 * After a (long ?) discussion on simgrid-devel, we have decided that the
1697 convention we had on units was stupid. That is why it has been decided
1698 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1699 WARNING : This means that all previous platform files will not work as
1700 such with this version! A warning is issued to ask users to update
1702 A conversion script can be found in the contrib module of the CVS, under
1703 the name contrib/platform_generation/surfxml_update.pl [MQ]
1706 * Bug fix: Processes were started in reverse order, wrt deployment file.
1707 WARNING: if your code relies on this bug, please fix it. [AL]
1708 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1709 being executed on two different locations. [AL]
1710 * Bug fix: Failures are now better supported thanks to Derrick's tests
1711 (there was many failure situations I hadn't thought of and that weren't
1712 correctly handled). [AL]
1713 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1717 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1718 Exception occurring on server-side are propagated back to client (!).
1720 API CHANGE: the callback changed their prototype. Change:
1721 int my_handler(gras_socket_t expeditor, void *payload_data) {
1723 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1724 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1726 * New! function: gras_msg_handleall to deal with all messages arriving
1727 within a given period.
1728 * New! function: gras_socket_server_range to get a server socket in a
1729 range of port numbers (ease to avoid port number conflicts) [MQ]
1730 * New! gras processes display their backtrace when they get a SIGUSR1
1731 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1732 Sweet to debug RL processes [MQ]
1736 - Do not force experiment sizes to be expressed in kb, or it becomes
1737 impossible to measure the latency this way (needs one byte-long tests)
1738 WARNING: this changes the amok_bw_* function semantic. [MQ]
1739 - Implements the link saturation stuff. [MQ]
1740 * Peer management module:
1741 New! module factorizing code that we wrote over and over [MQ].
1744 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1745 - Test units are placed directly into the library code, they get extracted
1746 automatically and placed into the src/testall binary.
1747 - Convert most of the XBT tests to this system.
1748 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1749 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1750 live in the same C code. This required to change a little bit the API
1751 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1752 * New module: sparse graph structure with basic algorithms (this is work
1753 in progress and the API is not considered to be frozen yet). [AL]
1754 * Display more information on backtraces: source line & function names are
1755 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1756 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1757 * Reworked a little bit some #include statements to load only required
1758 headers. Some user code that relied on SimGrid to include stdlib or
1759 stdio may need to include it by themselves. [AL]
1760 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1762 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1763 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1764 but I guess that almost nobody used those parts. [MQ]
1766 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1768 SimGrid (3.0.1) stable; urgency=low
1771 * Unfortunately, I had missed 5 misnamed functions:
1772 xbt_fifo_item_t xbt_fifo_newitem(void);
1773 void xbt_fifo_freeitem(xbt_fifo_item_t);
1774 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1775 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1776 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1777 They're now deprecated. Please use their new versions:
1778 xbt_fifo_item_t xbt_fifo_new_item(void);
1779 void xbt_fifo_free_item(xbt_fifo_item_t);
1780 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1781 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1782 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1784 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1785 * Documentation: xbt_log module unmercifully reworked [MQ]
1786 * Bugfix: there was a problem with the ending of contexts with
1787 the pthread backend. It caused some weird deadlock or behavior
1788 depending on the pthread implementation. [AL]
1789 * Bugfix: get the exceptions raised in the simulator repport where
1790 and why they come from when they are not catched in time [AL, MQ]
1793 * Bugfix: Do repport the error when two non-connected hosts try to
1794 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1797 * Add additionnal checkings on communications. Assert that two
1798 communicating hosts are connected by a set of links... [AL]
1801 * Add additionnal checkings on channel values in communication [AL]
1802 * New: MSG_task_get_source to see on which host a task was generated [HC]
1803 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1804 the number of tasks waiting to be received on channel and sent
1806 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1807 waits for the first task coming from a given host.. [AL]
1809 GRAS new functionnalities: [MQ]
1810 * Enhance the parsing macro to allow the size of multidimentional objects
1811 to be given thru annotations.
1812 * New example (and documentation): Matrix Multiplication a la RPC
1813 (as when I was young!) and fix a bunch of bugs found on the way.
1815 GRAS performance improvements: [MQ]
1817 * Reduce the amount of cbps creation/destruction by making it static to
1818 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1821 * Change libdata to a set so that we can search for stuff by ID (and thus
1822 reduce the insane amount of dict lookups)
1825 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1826 RL communication on top of SG since it's so uneffective.
1827 It may also be used for inter-thread communication in RL, one day.
1828 * Use gras_datadesc_copy() to exchange messages on top of SG
1830 - improve message exchange performance on top of SG
1831 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1832 * Don't exchange on the network the size of the used part of buffer,
1833 instead, specify the possible buffer size to read().
1835 - reduces the amount of read/write calls (one pair per exchange)
1836 - reduces the amount of exchanged data (the size)
1837 - allows to retrieve all arrived data on receiver side, if we don't need
1838 it right now (subsequent read will peek the buffer)
1839 - allows the receiver to proceed with the begining of the stream before
1840 everything is arrived
1841 - make it possible to build an iov transport (using readv/writev)
1843 - take care of the data with non-stable storage (like stacked data),
1845 * If possible, TCP send uses vector I/O (when writev() is here)
1846 - Don't use it for receive since we send data sizes and data on the
1847 same stream, so we wouldn't be able to chain large amount of chunks
1848 before having to flush the stuff to read the size.
1849 * Rework the transport plugin mecanism to simplify it and reduce the
1850 amount of pointer dereferencement when searching for the right function
1853 * I guess that now, we do almost as few system calls as possible while
1854 doing as few data copy as possible.
1856 To improve it further, we could try to send all the sizes first and then
1857 all the data (to use iov on receiving size), but it's only a partial
1858 solution: when you have 2 dimensional data, the sizes of the second
1859 dimension is data of the first dimension, so you need 3 streams.
1861 I'm not sure the potential performance gains justify the coding burden.
1863 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1865 SimGrid (3.00) stable; urgency=high
1868 * New! Give the possibility to hijack the surf parser and thus bypass
1869 MSG_create_environment and MSG_launch_application. Have a look at
1870 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1872 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1874 SimGrid (2.96) unstable; urgency=low
1879 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1880 This deprecates the xbt_error_t mecanisms.
1881 It modifies (simplifies) all XBT and GRAS API.
1882 MSG API keeps unchanged (exceptions raised by XBT are catched from
1883 within MSG and masked with existing error handling facilities)
1886 * New! Add a FATPIPE model. [AL]
1887 * New! Add a parallel task model. [AL]
1888 * New! Add automatically a loopback interface (in the default
1889 network model) if none was precised.
1892 * Bugfix: MSG_process_resume now works with the current running process.
1894 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1895 * Modification of MSG_task_get_compute_duration. Once a task has been
1896 processed, the value returned by this function is now equal to 0. [AL]
1897 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1898 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1899 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1900 of a m_task. For now, MSG_task_cancel only works with computation
1902 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1903 of the processor (in Mflop/s) regardless of the current load on the
1905 * API Change: use proper naming convention for MSG_getClock and
1906 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1908 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1909 This function changes the priority of a computation task. This priority
1910 doesn't affect the transfer rate. A priority of 2 will make a task
1911 receive two times more cpu power than the other ones. This function
1912 has been added to suit the needs of Nguyen The Loc and hasn't been that
1913 much tested yet. So if it fails, please report it and send me your code.
1915 * API Change: removed all functions and types that were marked "deprecated"
1916 since many months. Renamed MSG_global_init_args to MSG_global_init.
1918 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1920 SimGrid (2.95) unstable; urgency=low
1923 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1924 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1925 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1928 * Bugfix: Allow absolute paths for platform description files [MQ]
1929 * Bugfix: do free the variables after use. Leads to drastic performance
1931 * Implement max_duration (ie, timeouts) on resources [AL]
1934 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1936 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1937 the needed support (timeouts) [AL]
1940 * Implement measurement sockets. You can now get the bandwidth between two
1941 hosts thanks to AMOK (see below). [MQ]
1942 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1943 dynar over the network (yeah) [MQ]
1944 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1945 * simple/static token-ring example. [Alexandre Colucci and MQ]
1946 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1947 in place before [MQ]
1950 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1951 type parsing. "short" and "long" modifiers were also ignored; other
1952 modifier (such as reference level) are still ignored. [MQ]
1953 * Embeed the buffer size within the buffer itself on SG. [MQ]
1954 That way, send() are atomic and cannot get intermixed anymore (at least
1955 the ones which are less than 100k; bigger messages still have the issue)
1956 * Array size pushed by the field, not by the field type (or each
1957 and every long int will push stuff to the cbps) [MQ]
1958 * use select() to sleep since it allows to portably sleep less than one
1961 GRAS (minor cleanups)
1962 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1965 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1966 * type category 'ignored' killed as it was never used and were difficult
1968 * whether a type can cycle or not is now a flag, leaving room for more
1969 flags (as "ignored", if we feel the need one day ;)
1970 * Rename raw sockets to measurement sockets since "raw" has another
1971 meaning in networking community.
1974 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1975 and offers features not belonging to GRAS but that most applications
1976 need. One day, it may be a set of plugins loadable at runtime.
1977 * New module: bandwidth
1978 bandwidth measurement between arbitrary nodes running this module. [MQ]
1980 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1982 SimGrid (2.94) unstable; urgency=low
1984 The first beta release of SimGrid 3 !
1988 * Update the main page and the FAQ. Adding references to gforge.
1991 * Add a gras_os_getpid function.
1994 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1995 * Extend the logs so that they also print PID, hostname, date, ... if
1997 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1998 and kill the old version which were in testsuite/
1999 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2003 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2004 for users. Instead, each programming environment (GRAS, MSG,...) use it
2005 when needed to provide such a feature to users.
2006 Don't shortcut the mecanism or you will also shortcut the virtualization
2007 you need on the simulator.
2011 * Cleanups in configury with regard to compile optimization/warning flags.
2012 Also add -fno-loop-optimize to any powerpc since it's the optimization
2013 killing gcc (< 3.4.0).
2014 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2015 needed by MSG examples complications
2016 * Borrow configury beautifications from PHP
2019 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2020 warning (thanks loris for stumbling into this one).
2021 * Bugfix: stop loading private headers (gras_config.h) from the public
2025 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2026 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2027 RL would imply a DNS resolver, which is *hard* to do in a portable way
2028 (and therefore delayed).
2029 * Implement a real timer mecanism and use it in timing macros. This allows
2030 to avoid rounding errors and get a 0.000005 sec precision in timing
2031 macros. While I was at it, various cleanups:
2032 - allow to declare more than one timed section per file (fix a stupid bug)
2033 - move some private declaration to the right place
2034 - merge conditional execution and timing macros into emulation module
2035 - document the module
2036 - make sure the module cleanups its mess on gras_exit
2037 * Documentation improvements:
2038 - (new) how to compile applications using GRAS
2039 - (new) emulation support (timing macros)
2041 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2043 SimGrid (2.93) unstable; urgency=low
2045 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2048 - Use Paje properly where used. Still to be sanitized properly.
2049 - Portability fix: Add an implementation of the contexts using pthread
2053 - Add xbt_procname(): returns the name of the current process.
2054 Use it to show the current process's name in all logging.
2056 - fix detection of older flex version and the reaction, since we do
2057 depend on modern ones (we use lex_destroy)
2058 - Better separation of SG and RL in the libs: remove all simulation code
2059 from libgras. As a result, this lib is now only 200k when stripped.
2060 Some of the xbt modules may also be duplicated (two sets and such) and
2061 should be cleaned/killed before SG3.
2062 - Insist on using xlC on AIX because of weird problems involving gcc there.
2063 - Cleanup the make remote stuff. This is now done by scripts
2064 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2065 mainly for our private use, but we're working on changing them to user
2068 - Bugfix: flush the socket on close only if there is some *output*.
2069 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2071 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2074 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2076 SimGrid (2.92) unstable; urgency=low
2078 Alpha 3 on the path to SimGrid 3
2082 - New! First try of benchmarking macros.
2083 - New! First try so that gras_stub_generator generate deployment and
2084 remote compilation helpers.
2086 - Bugfix: Initialization fix in msg_test.
2090 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2092 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2094 - API Change: gras_os_sleep to take the amount of seconds as a double.
2095 Accepting an int was error prone since it was the only location where
2096 seconds were coded as such. It leaded to damn rounding errors.
2097 - Bugfix: Hard to belive that timers ever worked before this.
2099 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2101 SimGrid (2.91) unstable; urgency=low
2103 Alpha 2 on the path to SimGrid 3
2107 - Bug fix in the lmm_solver.
2109 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2110 through the function MSG_paje_output.
2111 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2112 - It is possible to bound the rate of a communication in MSG with
2113 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2114 to write it in the changelog).
2115 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2120 - Various cleanups to the autotools stuff
2121 - Begin to move Gras examples to examples/gras/
2122 - Let make distcheck work again (yeah!)
2124 - documentation overhauled using doxygen.
2125 gtk-doc-tools is dead in SimGrid now.
2126 - Automatically extract all existing logging categories, and add the list
2127 to the documentation (long standing one, to say the less)
2129 - Cleanup the known architecture table. Reorder the entries to group what
2130 should be, and use a more consistent naming scheme.
2131 (some of the test dataset are still to be regenerated)
2132 - New! Allow library to register globals on each process just as userdata
2134 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2135 lookup time (for now).
2136 Use it in msg and trp.
2137 This cleans a lot the internals and helps enforcing privacy of the
2138 headers between the gras components.
2139 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2140 - Bugfix: gras_os_time was delirious in RL.
2141 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2143 - Reenable GRAS now that it works.
2145 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2147 SimGrid (2.90) unstable; urgency=low
2149 Alpha 1 on the path to SimGrid 3
2151 * It is a long time since the last release of SimGrid. I'm sorry about
2152 that but as I had told you, I was rewriting a lot of things. I apologize
2153 to those who had been reporting bugs to me and that I had not answered.
2154 If your bug is still in the new version, please tell me. Here is a
2155 summary of the main changes.
2157 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2158 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2159 portability, speed, and a lot more but you'll figure it out later.
2160 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2161 I wanted to release the new MSG as soon as possible and I have
2162 broken GRAS, which is the reason why, for now, only MSG is fully
2163 functional. A laconic description of these projects is available
2164 in the documentation.
2166 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2167 kernel optimized for our needs (called SURF but only the developers
2168 should use it). Hence, MSG is now roughly 30 times faster and I think
2169 that by rewriting a little bit MSG, I could event speed it up a little
2170 bit more. Beside the gain in speed, it is also much easier to encode a
2171 new platform model with SURF than it was with SG. More to come...
2173 * REVOLUTION 3: I have tried to change a little as possible the API of
2174 MSG but a few things really had to disappear. The main differences
2175 with the previous version are :
2176 1) no more m_links_t and the corresponding functions. Platforms are
2177 directly read from a XML description and cannot be hard-coded
2178 anymore. The same format is used for application deployment
2179 description. The new format is described in the documentation.
2180 Have a look in tools/platform_generation. There is a tiny script
2181 that converts from the old platform format to the new one. Concerning
2182 the application deployment format, parsing the old one is tricky.
2183 I think most of you should however be able to convert your files. If
2184 it is really an issue, I can write a C code that does the conversion.
2186 2) the toolbox tbx does not exist anymore. We now have a library
2187 with much more data-structures but without the hash-tables (we have
2188 dictionaries that are much faster).
2190 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2192 *****************************************************************************
2193 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2194 * don't feel like converting the oldies. (Mt) *
2195 *****************************************************************************
2198 Version 2.90: "the long awaited one"
2199 - Finished rewriting and debugging MSG. Rewrote the documentation.
2200 - disable GRAS for now since it needs to be ported to the newest SG
2203 - Finish the port to windows (using mingw32 for cross-compile)
2206 - Main loop and datastructures of SURF. A cpu resource object is
2207 functional. Surf can thus be used to create cpu's with variable
2208 performance on which you can execute some actions.
2210 2004-11-15 Martin Quinson
2211 - Port to ARM. Simply added the alignment and size descriptions. Should
2212 work, but the ARM machines are so slow that I didn't had the opportunity
2213 to 'make check' over there yet.
2215 2004-11-15 Arnaud Legrand
2216 - Trace manager now written. It uses a heap structure and is therefore
2217 expected to be efficient. It may however be speeded up (particularly
2218 when many events occur at the same date) by using red and black
2219 trees. One day maybe...
2220 - Max-min linear system solver written. It uses a sparse matrix
2221 structure taking advantage of its expected use. Most operations are
2222 O(1) and free/calloc are called as few as possible. The computation of
2223 the minimum could however be improved by using a red and black tree
2226 2004-11-03 Arnaud Legrand
2227 - Rename every gras_* function that was in xbt/ to its xbt_
2229 - Add a heap and a doubly-linked list to xbt
2230 - Added a dichotomy to the dictionaries. make check works as well before
2231 so I assume that the patch is correct. I do not know however if things
2232 run effectively faster than before now. :)
2234 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2235 SimGrid, and the version number is bumped to 2.x
2237 2004-10-29 Martin Quinson
2238 - Introduction of the remote errors.
2239 They are the result of a RMI/RPC on the remote machine.
2240 ErrCodes being scalar values, you can't get the host on which those
2241 errors did happen. Extending the error mechanism as in Gnome is possible.
2242 No idea yet whether it is a good idea.
2244 2004-10-28 Martin Quinson
2245 - Interface revolution: the Starred Structure Eradication.
2246 I used to do typedef struct {} toto_t; and then handle *toto_t.
2247 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2248 - ???_t is a valid type (builded with typedef)
2249 - s_toto_t is a structure (access to fields with .)
2250 - s_toto is a structure needing 'struct' keyword to be used
2251 - e_toto_t is an enum
2252 - toto_t is an 'object' (struct*)
2254 typedef struct s_toto {} s_toto_t, *toto_t;
2255 typedef enum {} e_toto_t;
2256 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2257 s_toto_t) is private.
2259 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2260 so that it can set the variable to NULL. It was so for dicts and sets,
2261 it changed for dynars.
2263 - Fix a bunch of memleaks in dict_remove
2264 - Fix a bug in sg/server_socket opening: it failed all the time.
2266 2004-10-07 Martin Quinson
2267 - Speed up dynar lookup operation a bit.
2269 gras_dynar_get is dead.
2271 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2272 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2273 (which gives you the address of the stored data).
2275 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2276 copy of the data using an affectation to do the job and not a memcpy.
2278 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2279 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2281 It does not really speedup the dynar test because they are
2282 setting elements all the time (and look them seldom). But the dict does
2283 far more lookup than setting.
2285 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2287 2004-10-05 Martin Quinson
2288 - Allow to (en/dis)able the cycle detection at run time.
2290 Whether we should check for cycle or not is now a property of each
2291 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2292 datadesc_cycle_unset allow to remove this property when previously set.
2294 Note that the cycle detection is off by default since it impacts the
2295 performance. Watch the data you feed GRAS with ;)
2297 This property is hereditary. Any element embedded in a structure having it
2298 set have it set for the time of this data exchange.
2300 You should set it both on sender and receiver side. If you don't set it on
2301 sender side, it will enter an endless loop. If you forget on receiver
2302 side, the cycles won't be recreated after communication.
2304 - Header reorganization.
2305 Kill gras_private.h, each submodule must load the headers it needs.
2307 2004-10-04 Martin Quinson
2308 - Interface revolution: do not try to survive to malloc failure.
2310 Now, gras_malloc and friends call gras_abort() on failure.
2311 As a conclusion, malloc_error is not a valid error anymore, and all
2312 functions for which it was the only gras_error_t return value are
2313 changed. They now return void, or there result directly.
2314 This simplify the API a lot.
2316 2004-09-29 Martin Quinson
2317 - Re-enable raw sockets.
2318 Created by gras_socket_{client,server}_ext;
2319 Used with gras_raw_{send,recv}
2322 It should allow to kill the last bits of gras first version soon.
2324 This is not completely satisfactory yet (duplicate code with
2325 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2328 - Simplify transport plugin (internal) interface by not passing any
2329 argument to _server and _client, but embedding them in the socket
2332 2004-09-28 Martin Quinson
2333 - Finish the port to AIX.
2334 autoconf was my problem (segfault within the malloc replacement
2335 function. No idea why)
2337 2004-09-16 Martin Quinson
2338 - Fix some size_t madness on 64bit architectures.
2340 2004-09-08 Martin Quinson
2341 - Reduce the number of system headers loaded, overload some more system
2342 calls (such as malloc to cast the result of the system one, and work
2344 - Fix and reintroduce the config support
2346 2004-09-07 Martin Quinson
2347 - Source code reorganization to allow Arnaud to surf all over there.
2348 - Allow to document the logging categories.
2349 - Remove all uppercase from logging categories and useless cleanup in names.
2351 2004-08-18 Martin Quinson
2352 Version 0.6.2 (protocol not changed; API changed)
2353 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2354 gras_error_t), and NULL when not found. Functions expecting a msgtype
2355 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2356 hopefully usefull message.
2357 - Portability to prehistoric sparcs again
2359 2004-08-17 Martin Quinson
2360 Version 0.6.1 (protocol not changed; ABI not changed)
2361 - prealloc some buffers to speed things up
2363 2004-08-11 Martin Quinson
2364 Version 0.6 (protocol not changed; ABI expended)
2365 - The parsing macro can deal with the references, provided that you add
2366 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2368 2004-08-09 Martin Quinson
2369 Version 0.5 (protocol not changed; ABI changed)
2370 - Allow to off turn the cycle detection code in data exchange at
2371 compilation time. It should be at run time, but I'm short of time (and
2372 the config stuff is still broken). That way, we keep dict out of the
2373 critical path, which is good because the performance is poor:
2374 - search not dichotomial yet
2375 - dynar give no way to access their content and memcpy everytime
2376 - In composed data description (struct, ref and so on), stop foolness of
2377 keeping the subtype's ID, but store the type itself. This keeps sets out
2378 of the critical path, which is good since they rely on dynar and
2379 dictionnaries. The only loose of that is that we cannot detect the
2380 redeclaration of a structure/union with another content (but I'm not sure
2381 the code detected well this error before anyway). We still can detect
2382 the redefinition discrepancy for the other types.
2383 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2384 it breaks the code because of type-punning used all over the place).
2385 This breaks on all non-gcc architectures (for now).
2387 All those changes (plus the buffer of last time) allow me to gain 2 order
2388 of magnitude on cruel tests consisting of 800000 array of integers on two
2389 level of a hierarchical structure (200 secondes -> 4 secondes)
2392 - the selector of reference must now return the type it points to, not
2393 the ID of this type.
2395 2004-08-06 Martin Quinson
2396 Version 0.4 (protocol changed; ABI not changed)
2397 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2398 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2399 - Implement a buffer transport, and use it by default (it relies on tcp in
2400 real life and on sg in simulation).
2401 That's a bit hackish since I had a new field to the structure to store
2402 its data without interfering with the subtype ones. Inheritance
2403 is tricky in C. And that's a kind of reverse inheritance with one class
2404 derivating two classes. Or maybe a game with java interfaces. Anyway,
2405 that's damn hard in C (at least).
2406 Moreover, I got tired while trying to ensure plugin separation and
2407 genericity in SG mode. MSG wants me to do weird things, so let's go for
2408 cruel hacks (temporarily of course ;).
2409 See comment in transport_private.h:71
2410 - do not include all the _interface headers in private but in the files
2411 which really need them (to cut the compilation time when they are
2414 2004-07-26 Martin Quinson
2415 Version 0.3 (protocol not changed; ABI changed)
2416 - Major overhault of the datadesc interface to simplify it:
2417 - shorted the function names:
2418 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2419 - add a trivial way to push/pop integers into the cbps without malloc.
2420 This allows to make really generic sub_type description, which simply
2421 pop their size of the stack.
2422 - add a function gras_datadesc_ref_pop_arr() which does what users want
2423 most of the time: Declare a dynamic array (which pops its size of the
2424 stack) and declare a reference to it. Poor name, but anyway.
2425 - kill the post-send callback, add a post-receive one
2427 2004-07-23 Martin Quinson
2428 Version 0.2 (protocol changed; ABI changed)
2429 - add some testing for cpbs in the test cases, and fix some more bugs.
2430 This invalidate again the little64 data file, since I cannot regenerate
2432 - remove an awfull optimization in the logging stuff, allowing me to:
2433 - understand it again
2434 - learn gcc how to check that the argument match the provided format
2435 - fix all errors revealed by gcc after that
2436 - internal keys of dict are not \0 terminated. Deal with it properly in
2437 loggings instead of segfaulting when the user want to see the logs :-/
2439 2004-07-22 Martin Quinson
2440 - Fix some stupid bug preventing cbps (callback postit) from working
2442 2004-07-21 Martin Quinson
2443 - Some documentation cleanups
2444 - remove the useless last argument of msgtype_declare
2445 - rename the Virtu functions to fit into the 'os' namespace
2446 - move headers src/include -> src/include/gras/ and stop fooling with
2447 gras -> . symbolic link
2448 - make distcheck is now successful
2450 2004-07-19 Martin Quinson
2452 - Build shared library also
2453 - Install html doc to the right location
2454 - stop removing maintainer files in make clean
2455 - build tests only on make check
2457 2004-07-13 Martin Quinson
2459 - No major issue in previous version => change versionning schema
2460 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2461 corresponding dataset.
2463 2004-07-11 Martin Quinson
2465 - Get it working with any kind of structure (we can compute the padding
2466 bytes remotely for all the architectures I have access to)
2467 - Implement the structure parsing macro (still not quite robust/complete)
2468 - Improvement to the remote testing toysuite
2470 2004-07-10 Martin Quinson
2471 [autoconf mechanism]
2472 - get ride of a bunch of deprecated macros
2473 - actually run the test for two-compliment, not only compile it :-/
2474 - test whether the structures get packed (and bail out if yes. Damn.
2475 Alignment is a serious matter)
2476 - test whether the structures get compacted (but respecting the alignment
2477 constraints of each types)
2478 - test whether the array fields of structures can straddle alignment boundaries
2480 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2481 (took me 2 hours to find that bug, looking at the wrong place)
2482 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2483 before sending/receiving and are used to compute the offsets of fields
2484 - Given that padding size depend even on compiler options, keep track of
2485 alignment and aligned_size only for the current architecture. Not a big
2486 deal since we send structure fields one after the other (seems
2488 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2489 program, let it work on linux/gcc/little32. portability todo.
2491 2004-07-08 Martin Quinson
2492 - import and improve remote compilation support from FAST
2493 - make sure make check works on half a dozen of machines out there
2495 2004-07-07 Martin Quinson
2496 Let's say it's version 0.0.3 ;)
2497 - Implement conversions (yuhu!)
2498 - Let it work on solaris (beside conversion, of course)
2499 - Stupid me, using rand() to generate the conversion datatests in not wise.
2501 2004-07-06 Martin Quinson
2502 - Let make dist work, since I'm gonna need it to compile on remote hosts
2503 - Let Tests/datadesc_usage write the architecture on which the file was
2504 generated as first byte.
2505 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2506 architecture descriptions.
2507 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2508 on those architectures.
2509 - Optimization: send/recv array of scalar in one shoot
2511 2004-07-05 Martin Quinson
2512 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2514 - Plug a whole bunch of memleaks
2515 - each process now have to call gras_{init,exit}. One day, their log
2516 settings will be separated
2517 - Continue the code factorisation between SG, RL and common in Transport.
2519 2004-07-04 Martin Quinson
2521 - Redistribution between SG and RL.
2522 We wanna have to accept in SG, so move accepted related parts of RL in
2523 the common part. (more precisely, the dynar of all known sockets is no
2524 more a static in transport.c, but part of the process_data)
2526 [gras_stub_generator]
2527 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2529 2004-07-03 Martin Quinson
2530 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2531 them all up in one shot)
2533 - Fix the ugly case of reference to dynamic array.
2534 - New (semi-public) function gras_datadesc_size to allow the messaging
2535 layer to malloc the needed space for the buffer.
2537 - gras_socket_close now expect the socket to close (and not its address to
2538 put NULL in it after it). This is because the socket passed to handlers
2539 is one of their argument (=> not writable).
2541 - propagate the interface cleanup from last week in datadesc, ie remove a
2542 superfluous level of indirection. User pass adress of variable
2543 containing data (both when sending and receiving), and not of a variable
2544 being a pointer to the data. Let's say that I like it better ;)
2545 The price for that is constructs like "int msg=*(int*)payload" in
2546 handlers, but it's a fine price, IMHO.
2548 - Let it work in RL (yuhu)
2550 2004-06-21 Martin Quinson
2552 - porting SG plugin and SG select to new standards (works almost).
2553 - plug memleaks and fix bugs around.
2556 - cleanup the prototype of data recv and force users to specify when they
2557 want to handle references to objects. Test case working even for cycles.
2558 - plug memleaks. Valgrind is perfectly ok with this.
2560 2004-06-12 Martin Quinson
2562 - cleanup the separation between plugin and main code in plugin creation
2564 2004-06-11 Martin Quinson
2566 - Reput hook for raw sockets, needed for BW experiments
2567 - kill a few lines of dead code
2568 [Data description] Interface cleanup
2569 - gras_datadesc_by_name returns the searched type or NULL.
2570 That way, no variable is needed to use a type desc once, which makes
2572 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2573 two parameters were strings (field name, type name), leading to
2575 [Dicos] Interface cleanup
2576 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2577 This is consistant with the dynar API.
2579 2004-04-21 Martin Quinson
2581 - Porting to new standards.
2583 - interface cleanup.
2584 There is no bag anymore, no need to take extra provision to mask the
2585 pointers behind "ID".
2586 Better splitup of functions between files create/exchange/convert.
2587 This is still a bit artificial since convert and receive are so
2588 interleaved, but anyway.
2590 - add a queued message list to procdata (the ones not matching criteria
2592 - factorize some more code between SG and RL wrt procdata
2594 - use gras_exit in example to track memleaks
2595 - get rid of gs_example now that GS is properly integrated into gras
2596 - update run_test to integrate the latest tests (datadesc)
2598 - rename WARNINGn macros to WARNn since it prooved error-prone
2600 2004-04-19 Martin Quinson
2602 - register init/exit functions within gras module mechanism
2603 - send/receive function.
2604 Convertion is not implemented, but short-cutted if not needed.
2605 struct/array elements are sent one by one (instead of block-wise), but
2606 nobody really cares (yet). Get a prototype before optimizing.
2607 - tests (using a file socket) for DD send/receive on:
2608 - base types: int, float
2609 - array: fixed size, string (ie ref to dynamic string)
2610 - structure: homogeneous, heterogeneous
2611 - chained list, graph with cycle
2612 Believe it or not, valgrind is not too unhappy with the results. The
2613 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2617 - Bugfix when using a filename explicitely (instead of '-')
2619 2004-04-09 Martin Quinson
2621 - factorize more code between RL and SG in socket creation
2622 - Complete the implementation and tests of:
2624 o file (only in RL, and mainly for debugging)
2626 I lost 3 days to design a portable address resolver, and then decided
2627 that the prototype mainly have to run on my box.
2628 Addressing portability too early may be like optimizing too early :-/
2630 - use gras_init in the Tests instead of the crappy parse_log_opt
2631 (the latter function is removed)
2632 [Conditional execution]
2633 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2634 [Code reorganisation]
2635 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2636 Build examples against the RL library, since there is no way to disable
2637 its creation for now.
2639 For information, the beginning of coding on GRAS was back in june
2640 2003. I guess that every line has been rewritten at least twice since