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) stable; urgency=low
13 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
16 * Reintegrate Java to the main archive as desynchronizing these
17 package is not acceptable anymore (Java is now considered stable)
18 * Add explicit synchronization facilities through semaphores
19 * Bug fix: Task.setDataSize() only changed the C world, not the value
20 cached in the Java world
23 * Dramatically change the way files are handled. API and internals changed, but
24 this part of MSG was not considered as production grade either.
25 * Add explicit synchronization facilities through semaphores
26 * Add a new function MSG_host_get_process_list()
27 * Preliminary DVFS support (see examples/msg/energy/ for details)
30 * SMPI is now included directly in the libsimgrid as the windows
31 linker doesn't force us on splitting it anymore.
32 * Improvements of the SMPI replay tool:
33 - Most of the collective communications are now rooted in the same process as
34 in the original application.
35 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
36 used until now). Multiple data types can now be used in a trace.
37 - The replay tool now supports traces produce either by TAU or a modified
39 - Bug Fix: the compute part of the reduce action is now taken into account.
40 - Gatherv collective is now supported
41 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
42 of smpirun outputs time independent traces for the current run. One file
43 is created per process. If too many processes are simulated, this behavior
44 can be changed to one file for all processes by using the
45 tracing/smpi/format/ti_one_file flag
46 * smpirun generates the host file if needed (with given host count and platform)
47 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
48 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
49 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
50 --cfg=smpi/coll_selector:(mpich/ompi)
51 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
53 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
54 happening inside a collective SMPI call.
55 * Fix the behavior of complex data types handling.
56 * Make MPI_Wtime another synchronization point to take computations into account.
57 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
58 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
59 * Add all missing Fortran bindings, SMPI should work with Fortran 90
60 (no privatization of global variables yet)
61 * Preliminary DVFS support (see examples/smpi/energy/ for details)
64 * Verification of liveness properties is now available for SMPI applications
65 (in addition to MSG applications)
66 * Bugged examples using SMPI in examples/smpi/mc/
67 * Add --cfg=model-check/visited option. Allows the verification of infinite
68 programs. Detection of loops in the execution thanks to the system state
69 comparison and reduction of the state space to explore. Can be combined with
70 DPOR for safety properties.
73 * Allow to change SimGrid configuration (see --help) within the code
74 thanks to SD_config() as it can be done in MSG.
75 * Add a new function SD_task_set_amount() upon user request.
78 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
79 * Remove rule based routing (no more PCRE dependency)
80 * Add a limiter_link option to cluster tag, to specify a maximum reachable
81 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
82 * Add a loopback_bw and loopback_lat options to cluster tag.
83 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
84 routing. Now peers are encapsulated in an AS and have their own private
85 router but this is transparent.
88 * Our own implementation of getline is renamed xbt_getline, and gets
89 used even if the OS provide a getline(). This should reduce the
90 configuration complexity by using the same code on all platforms.
91 * New type: xbt_cfg_elm_boolean.
92 * Allow to use yes/no for boolean configuration options in the command line.
93 * Allow to disable SimGrid cleanups at exit from command line option.
94 There are situations where one may want a simulation to end with an exit.
95 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
96 annoying when scripting around the simulator. Adding a
97 --cfg=clean_atexit:no allows to circumvent this issue.
100 * Lots of memory leaks were corrected in this release.
101 There is no known memory leaks anymore, in all of our 600+ tests.
102 * New command line option --version, to get SimGrid version information.
103 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
105 * Supernovae builds are deprecated, and expected to be removed in the next
108 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
110 SimGrid (3.9) stable; urgency=low
112 The Grasgory release: GRAS is really dead now.
114 * Complete overhaul of the internal host structures scheme.
117 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
118 was considered as experimental and badly maintained since then.
119 * Keeping it was thus a trap to our potential users, that could take
120 it instead of MSG or SMPI by mistake despite is pity state.
121 * GRAS seems to have very few users (if any), and no one volunteered
122 to maintain it further. It also induces a lot of XBT code (for
123 portability sake), that must be maintained too.
124 * For all these reasons, we killed GRAS. If someone wants to revive it
125 in the future, don't cry, our git history still remembers of GRAS.
128 * Major overhaul. Merge our documentation again as time proved that
129 splitting it was really not helping our users.
130 * Further improve the developer documentation to help newcomers
131 hacking on SimGrid itself. The user documentation (and in
132 particular, the beginner documentation) is still in a sorry state.
135 * Now works on Windows too!
136 * Much more extensive test suite, from MPICH
139 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
140 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
141 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
142 * Bug fix: let task be scheduled when the last dependency to be solved is
143 a control dependency.
144 * Remove SD_load_environment_script function.
145 Use the C sg_platf function if you want to declare a platform
149 * New function: MSG_process_get_number()
150 * Old function documented: MSG_config()
151 * Remove MSG_load_platform_script function
152 Use the C sg_platf function if you want to declare a platform
156 * Change the default value of the TCP_gamma constant (maximal size of TCP
157 congestion window) to a more realistic 4MiB value. If you notice changes in
158 your simulation results, you can fall back to the previous 20k tiny window
159 by adding --cfg=network/TCP_gamma:20000 on command line.
160 * (Hopefully) fix a bug wrt periodic availability and state traces
161 * Bug fix: use default values at start when first event in availability/state
162 trace is not at time 0.
165 * remove the "new_" part of function name sg_platf_new_trace_connect
166 (resulting in sg_platf_trace_connect), since it does not create
170 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
171 use case where it's really mandatory, and maintaining it was a pain in
173 * New: xbt_fifo_search(), search an item with a user-provided
174 comparison function instead of dumb pointer comparison.
177 * Fix the lua deployment:
178 Use `simgrid.init_application()` before deployment instead of
179 `simgrid.msg_register_application()` after.
182 * Transfer the tracing files into the corresponding modules.
184 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
186 SimGrid (3.8.1) stable; urgency=low
188 The "we are told that some people want to also *install* the simgrid
191 * Add missing file "tesh.1" to the archive.
193 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
195 SimGrid (3.8) stable; urgency=low
197 The Psssshiiiit release: SimGrid jumps into the Cloud.
200 * Add an experimental interface to manipulate VMs. They are mainly
201 process groups with very few intrinsic semantic, but they should
202 allow you to build the semantic you want easily.
203 * New function: MSG_host_set_property_value()
204 * New function: MSG_process_on_exit(). To clean memory in all cases.
205 * Bug fixes that made the host (and link) failures unusable.
206 * Add a way to auto-restart process when the host in which they are
207 executing comes back (ON_FAILURE="RESTART" on deployment file,
208 MSG_process_auto_restart_set).
209 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
210 please stop using the old ones, they are DEPRECATED.
212 * Deprecate functions MSG_global_init() / MSG_global_init_args()
213 Please use MSG_init() instead. (reducing the amount of entry
214 points in the library helps us).
215 * Make it impossible to link against the wrong version of the lib
216 * Deprecate MSG_clean(). No need to call it anymore.
217 * Function MSG_get_host_number() is not deprecated anymore.
220 * Split the doc into a user guide and a reference guide.
221 * Start a developper guide to help people hacking on SimGrid.
224 * Enable tracing by default. This modules rocks you should use it.
225 * Remove option custom_flags. Now use environment variables CFLAGS
227 * Use default cmake things to detect lua instead of home grown ones.
228 * New option "enable_mallocators" to disable mallocators, for debugging
229 purpose ("on" by default).
232 * Bug fixes around the resource failures: don't let the processes
233 survive the host they are running onto.
234 * Add an interface to auto-restart processes when the host in which they are
235 executing comes back.
236 * Ensures that SIMIX_clean is called automatically. It's not part of
237 the public interface anymore (bindings should be updated).
240 * Bug fix for when SD_Simulate is called with a positive value: be careful
241 when comparing doubles. Sometimes they are different for non significant
243 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
244 parallel task whose initial work is distributed among host according
245 to the Amdahl's law. Such tasks are created with a parameter alpha
246 that corresponds to the non-parallelizable part of the computation.
247 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
248 between two sets of workstations assuming a 1D block distribution (each
249 workstation owns a similar share of data) on both sides.
251 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
252 Data redistribution will be automatically scheduled once parent and child
253 are both scheduled. The filling of computation_amount and
254 communication_amount structures is now done seamlessly thanks to the chosen
256 * New function SD_workstation_dump to display various information
257 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
258 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
259 bandwidth on the route onto which the task is scheduled and the amount of
261 To divide the nominal bandwidth by 2, the rate then has to be :
262 rate = bandwidth/(2*amount)
263 * Compute tasks that have failed can now be rescheduled and executed again
264 (from their beginning)
265 * Increasing source code coverage (src/simdag is now covered at 95.8%
269 * Re-implement time-independent trace replay using SMPI (at the
270 smpi_smp_* level) instead of MSG. This should replace
271 examples/msg/actions/actions.c
272 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
273 hindexed and structs)
274 * Implement the exchange of non-contiguous data.
275 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
276 * Correct behavior of smpi/sender_gap and set its default value to 0
277 * Add option to asynchronously send small messages to allow better
278 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
279 specifies the size in bytes under which messages will be asynchronously sent.
280 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
281 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
284 * New C interface to define a platform: XML is now optional.
285 For more info, please check include/simgrid/platf.h.
286 * New interface to define random platforms from the C:
287 For more info, please check include/simgrid/platf_generator.h and
288 examples/msg/masterslave/masterslave_platfgen.c
289 * Export a sg_cmdline dynar containing all the arguments we got from
293 * Two new tracing options for adding comments to trace file so you
294 can track your experiments (see --help-tracing for details).
295 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
296 * Adding the SimGrid version that generated the trace file as a comment.
297 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
298 * Fix to avoid key clashes on Paje links
299 * Other minor fixes related to the Paje specification
302 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
303 and renamed to xbt_str_hash() and xbt_str_hash_ext().
304 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
305 * Greatly improve the robustness of mmalloc to user errors (such as
306 using an area after freeing it, or freeing it twice)
308 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
310 SimGrid (3.7.1) stable; urgency=low
313 * Restore the prototype of MSG_process_create_with_environment() to
314 the pre-3.7 situation by removing the kill_time argument.
315 * Add a MSG_process_set_kill_time() function instead.
318 * Fix weird behaviors when dealing with parallel tasks.
321 * Simgrid is now built as a dll.
322 * Simgrid-java now works on Windows.
323 * Simgrid-Java is now included into Windows package.
326 * First pre-build package for MacOSX.
329 * Fix compilation when using MSG_USE_DEPRECATED.
330 * Fix some compilation issues on Macs and Windows.
331 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
332 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
334 We discovered that the Lua console is broken, but we are missing the
335 manpower to fix it right now. The problem existed in 3.7 too, so we
336 are not blocking the release for that. Sorry if you depended on this
337 feature, any help would be really welcome.
339 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
341 SimGrid (3.7) stable; urgency=low
343 The "spring cleanups (before the next Big Project kicks in)" release.
346 * We can specify the SMPI latency/bandwidth factor with command line
347 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
348 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
349 You can also use the "config tag" from platform file by adding this line
350 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
351 (see "example/platforms/tag_config.xml" to use "config tag").
352 Note that the command line supersedes the platform file configuration.
353 * Change the correction factors used in LMM model, according to
354 the latest experiments described in INRIA RR-7821.
355 Accuracy should be improved this way.
356 * Use the partial invalidation optimization by default for the
357 network too. Should produce the exact same results, only faster.
358 * Major cleanup in surf to merge models and split some optimization
359 mechanisms from the core of the models. As a result you can now
360 specify which model to use (e.g., --cfg=network/model:LV08
361 --cfg=cpu/model:Cas01) and which optimization mode to use
362 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
363 Incompatible combinations should err at initialization. See
364 --help-models for the list of all models and optimization modes.
365 * The CLM03 workstation model was dropped for simplicity because it
366 used the deprecated CM02 network model. Use default instead.
367 * Rename the TCP_gamma configuration option to network/TCP_gamma
368 * Rename the coordinates configuration option to
369 network/coordinates, and document it
370 * Use now crosstraffic keyword instead of the terribly misleading
371 fullduplex keyword. It is activated by default now in the current
372 default model, use --cfg=network/crosstraffic:0 to turn it off.
373 * Ongoing refactoring the model parsing to make XML files optional
374 See include/simgrid/platf.h for details (still to be completed)
377 * Major overhaul of the documentation. Almost instructive now :/
378 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
379 functions and friends. This interface was considered as
380 deprecated since over 2 years, it's time to inform our users that it is.
381 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
382 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
383 use these (crappy) functions in your code.
384 These functions will be removed soon. Stop using them now.
385 * Deprecate MSG_get_host_{table,number}
386 Implement MSG_hosts_as_dynar() instead.
387 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
388 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
391 * Stabilize the parallel execution mode of user contexts
392 * Introduce configuration variables to control parallel execution:
393 - contexts/synchro: Synchronization mode to use when running
394 contexts in parallel (either futex, posix or busy_wait)
395 - contexts/parallel_threshold: Minimal number of user contexts
396 that must be part of a scheduling round to switch to parallel
397 execution mode (raw contexts only)
398 * Fix bugs that prevented to use suspend/resume along with
399 synchronization structures.
400 * Fix bugs in process termination that lead to invalid memory access
401 in very specific conditions.
404 * Introduce a parallel mode for the models (controlled by surf/nthreads
405 configuration item). In our tests, running the models in parallel
406 never lead to any speedups because they are so fast that the gain
407 of computing each model in parallel does not amortizes the
408 synchronization costs, even when ultra fast futexes are used.
409 This is released anyway because YMMV.
412 * Performance boost by using a swag internally to compute the set of
413 tasks that are finished and should constitute the return value of
417 * Enable it by default now that it is considered rather stable.
420 * Documentation of the tracing functions.
421 * Performance gains when tracing categorized/uncategorized resource
422 utilization by avoiding calls to get route when updating resource
423 variables. LMM constraints are being used instead.
424 * API changed to set task categories. Use MSG_task_set_category instead
425 of TRACE_msg_set_task_category, and SD_task_set_category instead
426 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
427 * Bugfix for graphicator, routes not correctly obtained, memory leaks
428 * Examples for link user variables added (see at examples/msg/tracing/)
429 * Deprecated function TRACE_msg_set_process_category completely removed
430 * Trace header updated according to the latest Paje file format
431 * Tracing network lazy updates, no longer obligate users to use full updates
432 * --cfg=tracing/platform:1 also registers power/bandwidth variables
433 * Experimental: let user code declare/set/push/pop application states for hosts
434 * API changed to allow the manual creation of graph configuration files
435 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
438 * Improve the API of Lua MSG bindings, using the Lua spirit.
439 * Each simulated process now lives in its own Lua world (globals are
440 automatically duplicated). It helps writing simulators. Will allow
441 to run Splay programs within SimGrid in the future.
442 * Add a Chord example in Lua, equivalent to the MSG one.
445 * Start the implementation of a solution to express temporal
446 properties, not only local assertions. This is still an
447 experimental work in progress, stay clear from it to be safe.
451 - Add new runtime parameters --help-logs and --help-log-categories
452 to display informations about supported logging parameters and
454 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
455 don't exists anymore.
456 * Mallocators: allow value NULL for the reset function.
458 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
459 create homogeneous dictionaries, where all the elements share the
460 same free function. Non homogeneous dictionaries will be
461 deprecated in the next release.
462 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
463 - Multi-level dictionaries are deprecated.
465 - new function xbt_dynar_search_or_negative() that is useful when
466 you have less than 2 million elements in your dynar and don't
467 want of the extra complexity of catching exceptions when the
468 element is not found.
470 - Make xbt_os_thread module (for thread portability) public.
471 Documentation is still to come, sorry.
473 - Cleanups and simplifications to make it maintainable again.
474 - Exotic features (such as memalign and valloc) were removed.
475 - The metadata were extended and improved so that the
476 model-checker becomes able to explore and inspect the heaps.
477 - This may induce a performance drop when enable_model-checking is
478 ON in cmake (even if it's not used in the simulation), but it is
479 necessary at this point to get MC working.
481 Turn model-checking OFF if simulation performance matters to you.
482 Not enabling it at runtime is not enough, disable it in cmake.
484 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
486 SimGrid (3.6.2) stable; urgency=low
488 The "Not coding new stuff allows to polish old things" release.
491 * New bindings to the NS3 packet level simulator (experimental)
492 * Use the raw (efficient) execution contexts instead of the sysv
493 (portable) ones when possible.
494 * libpcre is now mandatory in any cases since not using it led to
495 severe performance loss and possibly other issues
496 * Update the XML platforms:
497 - G5K: include the latest machine in Nancy
498 - GridPP and LCG: new platforms
499 * Documentation was partially updated, at least (more to come)
501 Bug fixes, cosmetics and small improvements
502 * Free terminated processes before the end of the simulation to avoid
503 exhausting the memory of users having very dynamic amount of
505 * Bug fix and cosmetics about canceling non-running tasks
506 * Bug fix about the dot loader's issues when using libcgraph
509 * Create an installer for windows with nsis (amd64 and win32)
510 - Add an hello world project to illustrate simgrid project creation.
511 - Embed libpcre into the Simgrid installer to avoid
512 its compilation burden
513 * The raw execution contexts should work on Apple now
514 * Port to Windows 64 bits
515 - Sysv contexts now have an implementation for this arch
516 - GRAS communication features now support this arch
517 * Drop support for borland compiler on windows
518 - this code was not maintained, and we kinda depend on gcc nowadays
519 * Fix portability issues on kfreebsd/gnu: build error about semaphores
520 * Fix portability issue on unstable ubuntu: linker became picky on
523 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
525 SimGrid (3.6.1) stable; urgency=low
527 The "Oops, we broke Macs too" release
530 * Fixed contexts detection so that raw ones are used when possible
531 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
532 leads to a strange error, with user code segfaulting sometimes when
533 the generated code is not perfectly aligned (which is not
534 controllable from the user side, depends on the amount of code)
537 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
539 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
541 SimGrid (3.6) stable; urgency=medium
543 The Summer Release, also known as the "OMG! They Killed Kenny!" version
546 * Bindings now constitute their own package, separated from the main one.
547 Rationale: reduce our maintenance nightmare by reducing the module coupling
548 They will soon be released on their own on gforge.
550 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
551 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
553 GRAS: It is not considered as stable anymore, but experimental. Sorry.
554 * It's not quite deprecated for now because we have no replacement,
555 but it may soon become the case.
558 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
559 * Fortran: New user-level cache variable to store the rank of the running
560 process. This improves performance by an order of magnitude.
561 * C: New coccinelle script to automatically locate and modify global and
562 local static variables.
563 * Improved SMPI network model with a sender-side gap to account for multiple
567 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
568 only say if a communication is finished, no matter whether it succeeded or
569 failed. You can call MSG_comm_get_status() to know the status of a finished
571 * New function MSG_task_dsend() to send a task and detach it. When a
572 communication is detached, you are never notified of its success or failure
573 and the memory is released automatically once it is finished. This function
574 is useful when you don't care about the end nor the success of a
576 * Change the prototypes of action replay. Sorry for inconvenience,
577 but this is really more efficient this way (and to adapt your code,
578 you just have to fix the initialization, that shouldn't be too long)
579 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
580 ever managed to do anything useful with it.
581 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
582 takes a dynar of communications. It returns immediately and gives the
583 index of a finished communication (if any).
584 * New example: a basic implementation of the Chord P2P algorithm.
587 * New model for multi-core CPUs. You can now use the core attribute to
588 precise the number of cores of a host. This is a basic model. Every
589 process running on the host receives at most the power provided in
590 the DTD (throughput<=power). Total throughput of process cannot exceed
592 * New peer tag. This peer tag creates a tiny AS comprising a host and a
593 router linked by an up-link and a down-link (possibly asymmetrical).
594 This kind of pattern allows to easily build last-mile model style platforms.
595 Aggregating such patterns in a rule-based AS is thus the technique of
596 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
597 * New model for Vivaldi routing. We transformed the Vivaldi network model
598 into a Vivaldi routing model (based on the rule-based model). This allows to
599 combine Vivaldi based latencies with last-mile platforms.
602 * Added a check for NaN of IEEE754 infinite in the double entries of
604 * Introduce a new context factory "raw", highly inspirated from the
605 ucontext factory, but using manually crafted functions in assembly to
606 do the work in an efficient manner.
607 * Allow to change the used context factory at run time, not only at
608 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
609 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
610 contexts at runtime (only with raw contexts or ucontexts).
611 * Completely rewrote this module to allow parallel execution of user
612 processes. Use --cfg=contexts/nthreads:N to execute user processes
613 with N parallel threads (the default is 1, meaning no parallelism).
614 * Allow to decide dynamically between sequential and parallel modes.
615 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
616 processes in parallel only when their number is greater than or equal to P
618 * Added a check for NaN of IEEE754 infinite in the double entries of
622 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
623 won't output the state of processes when interrupted with Ctrl-C
624 * Add a new function xbt_dynar_to_array that transforms a dynar into a
625 NULL-terminated array. This may solve backward compatibility issues
626 due to the change to return type of SD_simulate. See also:
627 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
628 * Add new macros with variable number of arguments.
629 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
630 - in xbt/asserts.h: xbt_assert
631 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
632 - in xbt/ex.h: THROWF and RETHROWF.
633 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
635 * Change xbt_die() to accept a format string with arguments, just like printf.
636 * New data structure: xbt_lib_t, like a dict but more general and with better
640 * New configuration options
641 Options triva/categorized and triva/uncategorized can be used to generate
642 graph configuration files for Triva visualization tool.
643 * Configuration option tracing/platform is renamed to tracing/categorized
644 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
645 * New TRACE_user_link_variable interface:
646 User provides the name of the link and the tracing variable to attach to it
647 * the declaration of tracing categories must be done after the environment
649 * simpler tracing interface, just one way to declare categories
650 TRACE_category or TRACE_category_with_color, it is up to you
651 * links in the trace file are again identified by their names
652 * trace contains the full platform hierarchy exactly as declared using the ASes
653 * Options tracing/msg/[task|process]:1 groups the process by hosts
654 for both cases, tasks and processes must have names that are unique during the simulation
655 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
656 * The experimental option tracing/msg/volume is deprecated
657 its functionality may be reincorporated if needed
659 The tracing generates a trace file with unordered timestamped events,
660 because of the way the core simulator (surf) works. A script available
661 at the tools directory (fix-paje-trace.sh) can be used to put the events
662 in order. We have changed the tracing so it can generate ordered timestamped
663 events in the final trace, but depending on the simulator (and how much time
664 is simulated) that can lead to a huge memory utilization. It is deactivated
665 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
668 * Define a SIMGRID_VERSION macro in simgrid_config.h.
669 - We are trying hard to keep the API stable, but it may happen that
670 some things change (we're a research project after all, not a
671 nuclear plant operating system). If such things should happen, you
672 could rely on that macro to adapt.
673 - current value: 30600 for 3.06.00, aka 3.6
674 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
675 a number that can be compared with SIMGRID_VERSION.
676 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
677 assertions and verbose/debug logging events are disabled at compile time.
679 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
681 SimGrid (3.5) stable; urgency=medium
684 * New feature to any SimGrid-based simulator: Model-Checking
685 Check SIN#1 for more details.
688 * New Model SMPI (three-interval linear regression for correction factors)
689 See RR-7426, available at http://hal.inria.fr/inria-00527150
690 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
691 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
692 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
693 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
694 * See also src/smpi/README
699 - Tracing API changes: TRACE_start and TRACE_end should not be called
700 by user-code. They are automatically called by simulators created
701 with SimDAG, MSG and SMPI if the toolkit is compiled with
702 tracing_enabled=ON. Categories declaration and utilization remain the
703 same for MSG and SimDag.
704 - A function was added to the tracing API to declare categories with
706 - TRACE_category_with_color (char *category, char *color)
707 where color must be in the following format
708 "%f %f %f", red, green, blue
709 and red, green, blue are float values in the interval [0, 1]
710 - User can specify NULL as color parameter, or continue calling
712 On that case, the tracing system will define random colors
713 - The following command-line options are supported:
714 --cfg=tracing/filename:msg.trace
715 --cfg=tracing:1 (activate tracing, needed to use others)
716 --cfg=tracing/platform:1 (categorized resource use)
717 --cfg=tracing/uncategorized:1 (uncategorized resource use)
718 --cfg=tracing/msg/task:1 (task creation)
719 --cfg=tracing/msg/process:1 (process creation, migration)
720 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
721 --cfg=tracing/smpi:1 (SMPI interface tracing)
722 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
723 - examples of examples/msg/tracing updated
725 - DAXLoader and DOTLoader functions can generate tasks with categories
726 - A new function to attribute a category to SD tasks:
727 TRACE_sd_set_task_category (SD_task_t task, char *category)
728 * Tracing the MPI interface implemented by SMPI
729 - Collective operations are traced with states
730 - Point-to-Point operations are traced with states/links
731 - Tracing activated by a parameter "-trace filename" passed
732 to smpirun during execution (considering that simgrid
733 is compiled with tracing enabled)
734 - To run the simulation with gdb, the simulator
735 accepts --cfg=tracing/smpi:1 to trace SMPI
736 - tesh files to check if smpi tracing is ok
737 - See examples/smpi/NAS/DT-trace
738 * GTNetS tracing re-worked
739 - adaptation to the tracing system of GTNets to cope
740 with modifications regarding the fullduplex mode
741 - new tesh files to check if gtnets tracing is ok
744 * Asynchronous communications through the functions:
745 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
746 * New function: MSG_load_platform_script()
747 to make possible using a lua script instead of XML files to set up platforms
748 * New function: MSG_set_function
749 to associate functions to processes, used when bypassing the parser
750 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
752 Platforms: Add some more examples in examples/platforms
753 * Grid'5000: see www.grid5000.fr
754 * *_30000_hosts.xml: various huge files [mainly scalability testing]
757 * Change the XML format. This is a very important modification. SimGrid
758 3.5 introduces a new hierarchical format based on the notion of
759 Autonomous Systems. Compatibility with old format is ensured
760 through the perl script provided in the install bin directory
761 bin/simgrid_update_xml.
762 It is now possible to build platforms with specific routing
763 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
764 connect several platforms together. We will try to provide soon
765 set of realistic platforms exploiting these properties (have a
766 look at examples/platforms/ for the moment).
767 * Take the opportunity of the XML format change to be a good XML citizen:
768 rename link:ctn to link_ctn and similar changes (also dealt with by
770 * Add a new routing scheme (rule-based) using regular expressions. It
771 enables to have an extremely low memory footprint when the
772 underlying routing is simple and can be compactly described. You
773 need to have libpcre4-dev (perl regular expressions) installed if
774 you want to use this routing scheme.
775 * Revive the cluster TAG and allow to easily and efficiently (both in
776 term of memory and speed) connect clusters together. Have a look
777 at teshsuite/simdag/platforms/ to see how this can be done. With
778 this tag, you can create clusters with thousands of tasks at no
779 cost (have a look at examples/platforms/).
780 Note: clusters are implemented as ASes, so there is no need for an
781 enclosing AS tag if you have only one cluster in your platform.
782 * Add new generic functions in the public interface that allows the user
783 to call SURF 'create_resource' methods from your code (same
784 functionality as the XML bypass mechanism but with a much lighter
786 * Add a new model (enabled through command line --cfg=network/model:SMPI)
787 that uses a piecewise linear approximation to produce better
788 results when exchanging small messages.
789 * Add a new parameter to handle correctly full duplex link and account
790 for interferences between uplink and downlink communications
791 (activate with --cfg=fullduplex:1).
794 * Rename the SD_READY (all dependencies are satisfied and task is
795 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
796 dependencies are satisfied) state.
797 This prevents a confusion between the notion of "ready to schedule"
798 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
799 simulated" (SD_RUNNABLE) used by the simulation kernel.
800 * Change the way a task is considered as ready. Instead of removing
801 dependencies when a task is done, a counter is decreased. This way,
802 it is always possible to reach ancestors thanks to the
803 SD_taks_get_parents function (even after the end of the simulation.)
804 * Change the return type of SD_Simulate from (SD_task_t*) into
805 xbt_dynar_t. This function was in handling a dynar internally and
806 converted it into a NULL terminated array for historical reasons.
807 * New function SD_dotload(char*) to load a DAG described in dot
808 format. This loader and the corresponding examples require the
809 installation of the graphviz library.
810 * Fix a bug in the management of tasks of size 0 in the surf network
811 models. This problem was only visible with SIMDAG and you should
812 thus disregard results produced with earlier versions if you
813 relied on this feature (some tasks were blocked because of this).
814 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
815 even though your applications were doing only point-to-point
816 communications and sequential computations. Now you can really use any
817 model you want (of course, if you create real parallel tasks, which are
818 not implemented in most models beside ptaskL07, this will abort).
819 * Add an example that schedules a DAX on an heterogeneous platform
820 using a Min-Min strategy.
821 * New function SD_workstation_get_current_task() that returns the kind
822 of task currently running on a workstation in the sequential access
824 * Raise some warnings when unexecuted tasks remains at the end of the
825 simulation. This is usually caused by cycles in the DAG.
828 * New function: SIMIX_process_set_function() called by MSG_set_function
829 * Change the underlying waiting queue in semaphores so that a process
830 can wait on several of them simultaneously (as in waitany).
831 * Fix the way to handle tokens in semaphores so that all access patterns
832 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
833 * kill the dirty pimple SIMIX_message_sizes_output()
834 Please use (proper) visualization instead
837 * New data container: setset (set of sets of elements)
838 * New module: mmalloc (mapped malloc, allowing to have several
839 independent segments of malloc)
840 * New function: xbt_dict_cursor_set_data()
841 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
842 * New function: xbt_dynar_is_empty()
843 * New function: xbt_fifo_get_last_item()
844 * Fix xbt_dynar_shrink(): use the right element size.
845 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
846 uninitialized areas during expand.
847 * Fix semaphores: previous implementation was severely broken.
848 * Use library init/fini functions for our initialization.
849 - you can use logs and other feature as soon as you want in your
850 code (even before the xbt_init / MSG_init)
851 - xbt_exit is now a no-op and produce a warning when used.
854 * Port GRAS to new SIMIX mechanisms. This allows gras users to
855 benefit from the latest improvement to the simulation kernel.
856 * Kill measurement sockets for now. If you rely on them, sorry. This
857 release is not for you. This feature will be reintroduced in the
858 future, but we cannot delay the release any further.
859 * New function: gras_msgtype_get_name().
860 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
861 * Fix (at last) the pmm example: it should not randomly fail anymore.
863 Build chain: bug fixes and overall polishing
864 * Cmake is now stable enough. Hence, we killed the autotools.
865 * Port to windows ( TM :)
866 * Fix the 'make install' target.
867 No need to use 'make install-simgrid' anymore
868 * Introduce a 'make dist' target compiling a *source* archive
869 'make package' compiles a binary archive
870 * Compile java files only on need
871 * Add --cd and --setenv command line options to tesh
872 * Out of source builds are not fully supported yet, but we are close
873 * Enable supernovae and optimization flags by default for our users
876 * Add layer to set up environment directly from lua, without XML.
877 * The effect of gras_stub_generator can be achieved through
878 lua too (check examples/gras/console/ping_generator.lua)
880 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
882 SimGrid (3.4.1) stable; urgency=low
884 The "Polishing easter eggs is probably a good idea" release.
885 This is a bug fixes release only.
889 * Fix a bug preventing the tasks from begin garbage collected.
892 * Fix a bug occuring when a host involved in a communication fails.
893 This was not detected properly by the other peer involved in the
894 communication. Now, it's reported as a network error.
897 * Warn the user about loop dependencies in data flow of DAX files
898 * Obey the control-flow dependencies of DAX files
901 * Add option "enable_smpi" allowing to not compile SMPI.
902 Probably useful for the (Mac) users experiencing a build error here
903 * Improve the detection of lua5.1 and ruby1.8
905 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
907 SimGrid (3.4) stable; urgency=low
909 The "Easter in Cargese" release. Also known as (major changes):
911 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
912 ~> bindings were greatly improved
913 ~> new tracing infrastructure for better visualization introduced
915 * the "Welcome to configury modernity" release.
916 ~> we switched from autotools to cmake, and improved our cdash
919 A more detailled list of changes follow (full detail in svn log).
921 Java Bindings: Various Cleanups
922 * (install java-gcj-compat-dev on debian-like to use them)
923 * Remove put/get: no need to export deprecated interface in Java
924 Use send/receive instead.
925 * Cleanup the examples and add a README per directory
926 * Remove example autoDestination (that's the only way to go now)
927 * Remove example explicitDestination (was a plain copy of basic)
928 * Make JniException a runtime exception, so that there is no need to
929 declare the fact that you may encounter such a beast. I guess that
930 nobody will ever want to survive such error.
931 * Create specific errors for each MSG case of failure:
932 host failure, transfer failure, timeout, task cancelled
933 * Cleanup the exceptions that may get thrown by each function
934 * Other internal cleanups in Java bindings. Performance still bad :/
935 Ruby and Lua Bindings: create them
936 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
937 * That's new and great, you should try them out.
938 Same functionalities than Java bindings, only even less polished
940 * Kill the useless "rate" argument of SD_task_get_execution_time()
941 Everyone used to provide -1 as a value, it was not used, and the
942 semantic of a possible use wasn't even clear.
943 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
944 either as comm costs or compute costs to mean that there is no
945 such thing for that specific task.
946 * Add a SD_task_set_name() function
947 * Fix SD_task_unschedule() on typed tasks
948 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
949 * In DAX loader, accept useless 'level' attributes to <job> since
950 LIGO DAGs have them (seem to be to ease graphical representation).
952 * Add an example masterslave_mailbox.c using send/receive and not
953 the deprecated put/get interface.
954 * Kill the MSG_paje_output() function. It's a noop since 2 years.
955 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
957 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
958 (declare MSG_USE_DEPRECATED to still have the old name)
959 * Add a MSG_task_set_data() function
960 * About trace replay (see examples/msg/actions):
962 - Allow to work with splitted trace files for each process
963 Give the specific trace file as argument of each process,
964 and call MSG_action_trace_run(NULL)
965 You can still have one merged file for all processes.
966 - Fix implementation of collective operations
967 * Allow task_execute() on 0-sized tasks (closes #10063)
969 * This is the first release of SimGrid where SMPI is not considered
970 beta anymore (even if some corners should still be improved)
971 * Port over the new SIMIX_network submodule (internal refactoring)
972 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
973 * Implement more missing elements of the standard:
975 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
977 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
978 MPI_Scan() MPI_Get_processor_name()
979 - Added implementation of missing case for Alltoall (warning: it's
980 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
981 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
982 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
983 SMPI_MPI_Allgatherv()
985 - MPI_Waitsome() was broken
986 - Allow relative includes in smpicc
987 - Command line cfg argument 'reference_speed' was ignored...
988 - Some functions did not properly lead to auto-benching of user code
989 - smpicc passes -O2 by default (just like openmpi one)
991 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
992 * Bug fixes about timeouts during communications
993 * add SIMIX_message_sizes_output() as a pimple to write to file the
994 amount of messages per size. Use gnuplot to get histogram.
995 Pimple because that's the only user-visible function of simix,
996 defined directly in xbt.h (irk, sorry)
998 - Add a SIMIX_sem_get_capacity() function
999 - Fix interactions with processe resume/suspende
1000 - release_forever() was stupidly broken
1001 - Fix SIMIX_display_process_status() for processes in a semaphore
1002 - Make SIMIX_sem_block_onto() user-visible
1003 * Refactoring context stuff:
1004 - Use pseudo-OOP for better modularity
1005 - reimplement SIMIX_process_kill() without process_schedule() so
1006 that the latter can take as invariant that it is called from
1008 - Merge context_start into context_new for sake of simplicity
1010 * Add a Vivaldi network model, coded live during SUD'10 ;)
1011 * Rename configuration variables to start a hierarchy:
1012 o cpu_model -> cpu/model
1013 o network_model -> network/model
1014 o workstation_model -> workstation/model
1015 * New configuration variables:
1016 o network/bandwidth_factor: correction to bandwith
1017 o network/latency_factor: correction to latency
1018 o netwotk/weight_S: correction to the weight of competing streams
1019 * Add a long description to the models, that users can see with such
1020 argument on the command line: --cfg=cpu/model:help
1021 * --help-models display the long description of all known models
1023 * config: add the ability to set a default value after registration
1024 Does not override any previously set value (e.g. from cmd line)
1025 * dict: allow to have integer key and data.
1026 When so, you need to use the following functions
1027 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1028 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1029 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1030 In contrary to regular dicts, the key is not malloced before copy.
1031 Mixing scalar and regular elements in the same dict is not tested
1033 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1034 Tracing for Visualization:
1035 * SimGrid is now instrumented in order to generate a trace file for
1036 visualization analysis: to use it, need to compile SimGrid with the
1037 "tracing" option enabled, and instrument the program using SimGrid with
1038 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1039 (among other functions).
1040 * The instrumentation only traces the platform utilization for now
1041 * Documentation to use the tracing functions and how to analyze the
1042 traces with the Triva tool is written.
1043 * More information about: SimGrid FAQ (in the section Tracing Simulations
1046 * We moved to cmake as default build system. Autotools support will
1047 be dropped soon. Check the FAQ for more info about how to use it.
1048 * Greatly improved our cdash/ctest interactions
1049 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1050 * Added memory checking tests with valgrind; lot of memleak fixing.
1051 This may be the first release of simgrid with so few memory issues
1052 * Added code coverage tests.
1053 Our coverage is still improvable, but at least we see it on cdash.
1055 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1057 SimGrid (3.3.4) stable; urgency=low
1059 The "Desktop Grid needs love too" release (also called Xmas release).
1061 Models improvements:
1062 * Major speedup in the maxmin system solving by using lazy evaluation
1063 Instead of solving completely the maxmin system at each iteration,
1064 only invalidate (and recompute) the modified parts.
1065 This new feature is enabled in default models but you can try to
1066 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1067 * Cas01 IMproved as default CPU model
1068 This CPU model is the same Cas01 model, but it uses the
1069 maxmin-selective-update flag and a heap structure to manage
1070 actions on SURF kernel.
1071 It reduces the complexity to find the next action to finish and,
1072 consequently, it's faster than the old Cas01.
1073 This is the new default CPU model (Cas01).
1074 * Rename the old Cas01 model to Cas01_fullupdate
1075 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1076 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1077 * CpuTI (CPU Trace Integration)
1078 A new CPU model whose objective is simulate faster when using
1079 availability trace files.
1080 Instead of using a full featured, over engineered maxmin system for
1081 CPU modeling, this model does the pre-integration of traces files
1082 to calculate the amount of CPU power available, and so, executes
1083 faster than the old CPU models.
1084 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1085 * Use LV08 as default network model since it gives better accuracy
1086 for small messages and shouldn't change things for big ones.
1087 Use --cfg=network_model:CM02 to get the previous behavior.
1090 ******************************************
1091 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1092 ******************************************
1093 * The new CPU model may changes simulations!
1094 The point is that events occurring at the exact same timestamp
1095 are not scheduled in the same order with the old and new
1096 version. This may be enough to completely change the execution
1097 of simulations in some cases.
1098 * The new network model will change simulations!
1099 This new model is more realistic than the previous one, so you
1100 should consider redoing your old experiments with this model.
1101 Sorry for the inconvenience.
1104 * Introduce the supernovae compilation mode
1105 When compiled that way, the whole SimGrid (or almost) is put in a
1106 single compilation unit and compiled in one shoot.
1107 This is to help gcc which has difficulties to inline stuff from one
1109 The speedup seem to be above 15%, althrough more tests are needed on
1110 amd64 to confirm that gain.
1113 * Port of MSG's mailbox on top of SIMIX network
1114 The put/get mechanism was greatly simplified on the way.
1117 * New SIMIX network module. Provides:
1118 - Mailbox: rendez-vous mecanism to find with who you want to speak
1119 - Synchronous send/recv: easier and hopefully faster since the
1120 logic is handled in the maestro process directly now
1121 - Asynchronous send/recv: you dreamt of it? It's here now
1122 Too bad that nobody cared enough to propagate the change to MSG.
1123 * Add semaphores as SIMIX synchronization mechanism.
1126 * new function SD_daxload(char*) to load a DAX file
1127 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1128 * Introduce typed tasks. Specify its kind and cost at creation.
1129 At scheduling, just give where it should be placed, and the cost
1130 for each involved resource is automatically computed.
1131 Existing constructors so far (more to come of course):
1132 - SD_task_create_comm_e2e() for end-to-end communication
1133 - SD_task_create_comp_seq() for sequential computation
1134 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1135 * new function SD_task_dump() for debuging display
1136 * new function SD_task_dotty(task,FILE*) writing to file the info
1137 about the task in dotty format
1138 * SD_task_dependency_exists() can now cope with having one of its
1139 arguments NULL. If so, it tests whether the other argument has any
1141 * Add getters on list of preceding/following tasks:
1142 SD_task_get_parents(task) and SD_task_get_children(task)
1143 * Add getters on amount of workstations and list:
1144 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1145 * Add getter on task kind: SD_task_get_kind(task)
1146 * Update the start_time and finish_time of tasks on completion/failure
1147 * Bugfix: Remove task from state swags when destroyed
1150 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1153 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1154 dynar with one separate thread per value of the dynar.
1155 * Change the prototype of xbt_thread_create(), sorry.
1156 Added a boolean parameter indicating whether we want to join this
1157 thread (used in SG only for now)
1158 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1161 * GTNetS wrappers should now be usable again (and betterly tested too)
1162 * Fix a major regression from 3.2 where the timeout provided to
1163 MSG_task_put_with_timeout() was used as absolute time before which
1164 the comm should be done.
1165 * Start to fix the <cluster> tag.
1166 - Internal links should be good now (beside of the loopback, which
1167 use the private link instead)
1168 - paths to the external world is still rather broken
1169 - the <route:multi> tag is just broken. Actually that's brain-dead.
1170 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1171 to make it less stupid
1172 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1173 * Fix a source-level compatibility glitch from 3.2: after defining
1174 MSG_USE_DEPRECATED, you can use the old name
1175 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1176 * Allow to compile from the SVN with automake 1.11
1177 * Fix some problems when using the "start_time" tag in deployment XMLs.
1178 * Fix #8569: XBT/synchro.h has redundant declarations
1179 * Fix #8563: MSG return values and exceptions
1180 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1181 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1182 Thanks to Robson Peixoto.
1183 * Fix a vicious bug in dictionaries inducing that some elements were
1184 not freed on xbt_dict_free()
1186 Portability report of this version:
1187 * Main portability targets:
1188 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1189 on (amd64/i386/ia64)
1190 - mac leopard on i386
1191 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1192 but nothing critical.
1193 * Other platforms: windows, AIX and others were not tested for this release
1195 Timing report of this version:
1196 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1197 scenario parameters). From 8h to a few seconds in desktop grid settings.
1198 * Supernovae brings about 25% speedup on i386.
1200 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1202 SimGrid (3.3.3) stable; urgency=low
1204 The "Need for Speed" release.
1206 The timings done to validate the 3.3.2 were faulty.
1207 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1209 The problem was a conversion from a manually handled vector to
1210 xbt_dynar_t on the critical path.
1211 xbt_dynar_foreach calls functions, inducing stack management crap.
1213 We inlined these functions and xbt_dynar_foreach is now breath taking.
1214 We also inlined xbt_swag_belong on the way.
1216 Here are some approximate speedup measurements (on master/slaves
1217 simulations lasting between 10s and 20s each):
1218 3.3.1 -> 3.3.2: about same performance
1219 3.3.2 -> 3.3.3: 40% speedup
1220 3.3.1 -> 3.3.3: 40% speedup
1221 3.3.1 with inline patch -> 3.3.3: 30% speedup
1223 Our reading is that the refactoring which occurred in 3.3.2 made us
1224 suffer much more from the xbt_dynar_foreach low performance, but
1225 once we solved this, this refactoring proved to be very performance
1226 effective. From the 40% speedup, somehow, 10% are due to the
1227 inlining and 30% to the refactoring.
1229 That's a pitty that gcc cannot inline functions placed in other files
1230 alone. We have to choose between:
1231 - break the encapsulation (by putting private data structures and
1232 accessors in headers files to help gcc)
1233 - live with low performance
1234 - switch to a decent compiler such as icc (not quite possible).
1236 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1238 SimGrid (3.3.2) stable; urgency=low
1240 The "Simplicity does not preceed complexity, but follows it" release.
1242 The main contributors of this release were (lexical order):
1243 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1246 * Extract the routing logic into its own object.
1247 (was dupplicated in network.c and workstation_LV07.c;
1248 Allows to implement other ways of storing that info)
1249 => kill now useless network_card concept
1250 - Use dynar to represent routes (instead of void** + int*)
1251 - kill link_set (use surf_network_model->resource_set instead)
1252 - Add a command-line option to choose the routing schema to use
1253 - Add three new models:
1254 * Floyd (shortest path computed at initialization)
1255 * Dijikstra (shortest path recomputed all the time)
1256 * Cached Dijikstra (shortest path computed on need)
1257 All these models where contributed by Silas De Munck, and are
1258 described in his ICCS09 paper.
1260 * Simplify model declaration
1261 (less redirections, less function to write when defining a model)
1262 - Factorize stuff between models:
1265 surf_model_resource_set(model)
1266 surf_model_resource_by_name(model, name)
1267 - Unify the types of models in s_surf_model_t (using an union)
1268 - Embeed fields of common_public directly into s_surf_model_t
1269 - Rename model methods:
1270 action_free ~> action_unref
1271 action_change_state ~> action_state_set
1272 action_get_state ~> action_state_get
1273 - Change model methods into functions :
1274 (model)->common_public->action_use ~> surf_action_ref
1276 * Implement a generic resource; use it as ancestor to specific ones
1277 (allows to kill duplicated code in models)
1278 Drawback: timer command don't need no name nor properties;
1279 workstation_CLM03 don't need no properties
1280 (but I guess we can live with those few bytes wasted)
1282 * Improve the action object model
1283 - implement a constructor avoiding dupplicated code about field
1284 initialization in generic_action part.
1286 * Kill the SDP model: it has an external dependency, is deprecated
1287 in flavor of modern lmm models, and didn't compile since a while
1290 * Relocation of the context module from XBT to SIMIX.
1291 (the context were decoupled from the simix processes, duplicating a lot of code)
1292 => a lot of code was factorized
1293 - less overhead is introduced during scheduling
1294 - simpler API for the context factory
1295 - the logic for process creation,destruction and manipulation was simplified
1296 * Simplification of the s_smx_process_t data structure.
1297 => accesing the simix level data associated to a process is faster now,
1298 and the code is a lot more readable.
1301 * Implement some more MPI primitives:
1302 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1303 -implementation: Bcast: flat or 2-ary tree (default),
1304 Barrier: 4-ary tree,
1306 Allreduce: Reduce then Bcast
1307 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1308 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1309 Alltoallv: flat tree, like ompi
1311 * Add support for optimized collectives (Bcast is now binomial by default)
1312 * Port smpirun and smpicc to OS X
1315 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1316 Shout out if you used it.
1319 * Display the list of still queued messages in SG mode when existing
1323 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1324 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1325 * Add xbt_str_from_file(FILE*)
1326 * Add xbt_dict_get_key achieving a linear reverse search
1327 * Remove the context module
1329 Portability report of this version:
1330 * Main portability targets:
1331 - Linux(debian)/x86/context
1332 - Linux(debian)/x86/pthreads
1333 - Linux(debian)/amd64/context
1334 - Linux(debian)/amd64/pthreads
1335 On these, we still have the eratic breakages of gras/pmm and
1336 amok/saturate_sg reported in previous version. We still think
1337 that the tests are the cause of the fault, not the tested code.
1339 - Mac OSX Leopard/x86/context
1340 Still false negative in tesh autotesting.
1341 Smpi still fails, but this time because readlink does not accept -f
1342 Everything seems to work properly beside of that.
1345 - AIX version 5.3 (only tested contexts this time)
1346 Smpi still fails there because mktemp is not installed.
1347 Everything seems to work properly beside of that.
1349 I managed to compile it for the first time, but several breakages.
1350 Won't delay the release for this exotic platform.
1352 * Windows: it's still lagging behind. If you want to help, please
1355 Timing report of this version:
1356 This version seem to be more than 5% faster than 3.3.1 (on linux
1357 64bits with contextes). The gain is less than expected, we are
1358 investigating this for next release.
1360 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1362 SimGrid (3.3.1) stable; urgency=low
1365 * Implement a --cfg-help to show existing configuration variables
1366 * Build chain do not require doxygen in maintainer mode
1369 * fix a bug on struct sizeof computation, which prevented the
1370 exchange of arrays of structs in some conditions
1371 - added a regression test about this in datadesc_usage
1372 * Allow the exchange of 0-long dynamic vectors.
1373 - for that, use -1 as indicator of dynamic size instead of 0
1374 - This implied to change any size from unsigned long to long,
1375 reducing a bit communication abilities, but I guess that with
1376 64bits being quite common, this is more than enough.
1377 - This also induce a protocol change, thus bumping network protocol
1378 version from 0 to 1 (if we have external users, we have to get
1379 clean on that point too ;)
1380 - added two regression tests about this in datadesc_usage
1381 * Be more verbose when propagating local exceptions
1382 This helps debugging.
1383 * Display the status of simulated processes when receiving SIGINT in
1387 * Allow to control the simulation from a trace file.
1388 New functions MSG_action_register() and MSG_action_trace_run()
1389 The first one allows to associate a function execution to each
1390 kind of action while the second one parses a trace file and
1391 triggers the corresponding actions within the system.
1392 For now, only a toy example is provided in examples/msg/actions
1393 * Add an exemple of process migration in examples/msg/migration
1394 * Fix a bug in task exchange which broke MSG_task_get_sender()
1395 Add a teshsuite regression test for that.
1396 [Bug: if MSG_task_get_sender() is called after sender exit,
1398 * Fix a bug which prevented suspend/resume to work properly
1399 * Display the status of simulated processes when receiving SIGINT
1400 This fixes a regression of v3.3. due to the introduction of SIMIX
1401 * Bug fixing in failure management:
1402 - trace could not start by a failure at time 0
1403 - failure during communications were not working
1406 * Add SIMIX_process_set_name() to change the name of the current
1407 process in the log messages.
1408 * Store smx_hosts in a dict since we only retrieve them by name
1409 * Move the configuration infrastructure to surf
1412 * Move the configuration infrastructure to surf
1415 * Massive internal cleanups:
1416 - Store internal structures on processes instead of hosts (allows
1417 to have more than one process per host, in addition of being more
1419 - Cleanup the initialization/finalization process
1420 - Kill a whole bunch of unneeded synchronization:
1421 processes run in exclusive manner within the simulator
1422 - Move queues from global tables to process data fields
1424 - now accept -platform and -hostfile arguments
1425 - Pass the right rank value to processes according to the hostfile
1426 * Compile the examples by default, and use them as regression tests
1427 * Implement MPI_Wtime()
1428 * Change the reference speed to a command line option
1431 * TCP_gamma can now be specified as command line option using
1432 --cfg=TCP_gamma:10000000.0
1433 * Change the --surf-path cmd line option into --cfg=path:
1436 * Also include strbuff from xbt.h public header
1437 * xbt_ex_display(): do not free the exception after displaying
1438 This allows to do more with the given exception afterward.
1439 Users should call xbt_ex_free() themselves.
1443 Portability report of this version:
1444 * Main portability targets:
1445 - Linux(debian)/x86/context
1446 - Linux(debian)/x86/pthreads
1447 - Linux(debian)/amd64/context
1448 - Linux(debian)/amd64/pthreads
1449 These targets fail about 1/10 of times on gras/pmm, but we believe
1450 that this is because of the test, not because of simgrid.
1451 amok/saturate_sg fails even more rarely, and the test may not be
1454 - Mac OSX Leopard/x86/context
1455 The test suite still spits tons of errors because some obscure
1456 force prevents us from removing the temporary directories
1457 arguing that they still contain some metadata I've never heard of.
1458 Smpi fails because seq is not installed.
1459 Everything seems to work properly beside of that.
1462 - AIX version 5.3 (both contexts and pthread)
1463 Smpi still fails there because mktemp is not installed.
1464 XML inclusions seems rosty on AIX.
1466 * Windows: it's still lagging behind. If you want to help, please
1469 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1471 SimGrid (3.3) stable; urgency=high
1475 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1476 [Malek Cherier & Mt]
1478 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1481 Until now, GRAS were using MSG as an interface to SURF. It was
1482 quite difficult because both interface have several differences
1483 (MSG channels vs GRAS sockets were the most notable point).
1485 This also opens the gate to SMPI (which should occur soon) and speed
1486 up simulations by to 40% (even if it were not the main goal).
1488 **************************************
1489 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1490 **************************************
1491 The point is that events occuring at the exact same timestamp are
1492 not scheduled in the same order with the old and new version. This
1493 may be enough to completely change the execution of simulations in
1494 some cases. Sorry for the inconvenience.
1496 * Cleanup and upgrade the XML format to push further scalability
1497 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1499 * Improve the testing infrastructure with tesh. Now a very large part of
1500 the code is tested not only by being run but also by checking that the
1501 output match an expected output [Mt].
1503 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1504 is really less memory-demanding, which should allow you to use
1505 larger files in SimGrid [AL].
1507 * Inform valgrind about our contextes, so that it becomes usable
1508 with the default (and more effecient) version of SimGrid
1509 [contributed by Sékou Diakite, many thanks]
1512 * Introduce a listener thread in charge of receiving incoming
1513 messages from the network. It allows to overlap communication and
1514 computation but most notably, it removes some stupid deadlocks due
1515 to the fact that so far, a process could not send and receive at
1516 the same time. This made most non trivial communication schema
1518 * Convert the PIDs from long int to int to match the MSG ones (and
1519 linux ones too) [Mt]
1520 * New function: gras_agent_spawn() to launch a new process on
1521 current host. Only working in simulation for now. [Mt]
1522 * New function: gras_os_hostport() returning a constant form (ie,
1523 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1526 * Make the backtrace of exceptions more human readable [Mt]
1527 * New module: xbt/str [Mt]
1528 a ton of string utility functions (split, join, printf to a newly
1529 allocated buffer, trim, etc)
1530 * New module: xbt/hash [Mt]
1531 SHA1 hashing algorithm (more to come if needed)
1532 * New module: xbt/synchro [Mt]
1533 synchronization tools (mutex and conditions) working the same way
1534 in simulation and in real life (mainly useful for GRAS, but not
1536 * New module: xbt/queue [Mt]
1537 classical producer/consumer synchronization scheme
1538 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1539 (using the classical functions will get serialized) [Mt]
1540 * Make dictionary internal table dynamic. No need to specify its size
1541 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1543 * Make sure the log channels are organized as a tree under windows
1544 (because of ANSI C compatibility issue, any channel were child of
1545 root directly) [Mt].
1548 * Cleaned many thing in surf and fixed a few bugs [AL].
1549 * Add a nice command line configuration mechanism to compose models [AL].
1550 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1551 the previous one (KCCFLN05). It relies on something that looks like
1552 a max-min sharing mechanism but cannot be written as such. A new solver
1553 was thus designed [AL].
1554 * Add a new solver to lmm. Based on Lagrange optimization and
1555 gradient-based descent, it enables to efficiently maximise systems s.a
1557 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1559 This solver enables to propose two new network models for TCP Reno and
1560 TCP Vegas based on Low's work. These models still need to be fully
1561 tested though [Pedro Velho].
1564 * Bug fix in SD_simulate. Now the time bound given as argument is
1566 * Use the new parallel task model (ptask_L07) as default.
1567 * Use the SURF command line configuration mechanism.
1568 * 0-size tasks (for synchronization) should now work.
1570 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1572 SimGrid (3.2) stable; urgency=high
1576 We still experience issues on this platform, but we believe that at
1577 least MSG is usable.
1579 GRAS API BREAKAGE (for simplification purpose, sorry):
1580 * the gras_msgtype_by_name is not used anymore. Instead of
1581 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1582 you can write (and must)
1583 gras_msg_send(toserver, "request", &request);
1584 - If you still want to pass a gras_msgtype_t to the function (to cache
1585 the type and avoid the lookup time), use the gras_msg_send_() variant.
1586 - Impacted functions:
1587 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1588 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1589 * The callbacks are now expected to return 0 when everything went well
1590 (just like the main() function)
1592 GRAS new features and improvements:
1593 * New module mecanism where user code can use per process globals [Mt]
1594 This is similar to gras_userdata_*() functions, but for libraries. It
1595 factorize some code developped over and over in the examples and AMOK.
1596 It has still to be documented and used (only amok/peermanagement is
1598 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1599 loss when they were small enough to fit into the buffer and sent quickly
1600 enough so that they can all get received in one shoot.
1601 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1602 if not found. Use the *_or_null() variant for the old semantic.
1603 * In gras_msg_handle, do not discard messages without callback.
1604 They are probably messages to be explicitly awaited later (ie, proofs of
1605 mis-synchronization in userland since they are sent before being awaited)
1606 No big deal usually.
1607 * gras_socket_meas_send/recv: semantic changed!
1608 The numerical arguments used to be (1) the total amount of data to send
1609 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1610 messages. This was need for the fool willing to send more than MAXINT
1611 bytes on quite fat pipes.
1614 * Do really rename the hostmanagement module to peermanagement. [Mt]
1615 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1616 but this is rather new and this was documented in the module
1617 documentation (poor excuses, I admit)
1618 * Bandwidth measurement semantic changed! This follows the changes to
1619 gras_socket_meas_send/recv explained above.
1622 * A sequential mode has been added to the workstations. When a workstation
1623 is in sequential mode, it can execute only one task, and the other tasks
1624 are waiting in a FIFO. [Christophe Thiery]
1627 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1628 model for SIMDAG. [Christophe Thiery]
1629 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1630 equal to 0, which caused some bad behaviors in
1631 saturated_constraint_set_update. I now use a threshold mechanism like in
1635 * When running manually src/testall, you select specific units [Mt]
1636 testall is the result of our cunit mecanism, and should replace all
1637 the scripty thingy around since bash don't run easily on billware.
1639 * A mallocator system has been added. [Christophe Thiery]
1640 Mallocators allow you to recycle your unused objects instead of freeing them
1641 and allocating new ones.
1643 Documentation update:
1644 * FAQ reworking + New FAQs:
1645 - "Valgrind spits tons of errors!" [Mt]
1646 - "How to repport bugs" [Mt]
1647 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1648 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1649 - Communication time measurement within MSG [AL]
1650 - I experience weird communication times when I change the latency [AL]
1651 * GRAS tutorial [Mt]
1653 - an introduction to the framework and to the used communication model
1654 - an initiatic tour introducing the most proheminent features:
1656 . Lesson 0: Installing GRAS
1657 . Lesson 1: Setting up your own project
1658 o Part 2: Message passing
1659 . Lesson 2: Exchanging simple messages
1660 . Lesson 3: Passing arguments to the processes (in SG)
1661 . Lesson 4: Attaching callbacks to messages
1662 . Lesson 5: Using globals in processes
1663 . Lesson 6: Logging informations properly
1664 . Lesson 7: Using internal timers
1665 . Lesson 8: Handling errors through exceptions
1666 . Lesson 9: Exchanging simple data
1667 . Lesson 10: Remote Procedure Calling (RPC)
1668 . Lesson 11: Explicitely waiting for messages
1669 . Recapping of message passing features in GRAS
1670 - A HOWTO section containing:
1671 o HOWTO design a GRAS application
1672 More are due, of course. They will come latter. In the meanwhile, you can
1673 check the examples which are still here.
1675 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1677 SimGrid (3.1) stable; urgency=high
1681 There was a stack corruption somewhere, visible only when optimizing
1682 with these versions. [Vince]
1685 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1686 module that enabled to play with Directed Acyclic Graphs. It is built
1687 directly on top of SURF and provides an API rather close to the old
1688 SG. Some old codes using SG are currently under rewrite to check that
1689 all needful functions are provided. [Christophe Thiery]
1692 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1693 extension of the classical CLM03 model that gracefully handles
1694 failures. This is now the default model for MSG and GRAS. It doesn't
1695 handle parallel tasks yet though. [AL]
1696 * Bug fix: Weights were not correctly set in the network part.
1697 WARNING: This may have resulted in incorrect results with simulations
1698 where there are more than one flow on a given link. [AL]
1701 * After a (long ?) discussion on simgrid-devel, we have decided that the
1702 convention we had on units was stupid. That is why it has been decided
1703 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1704 WARNING : This means that all previous platform files will not work as
1705 such with this version! A warning is issued to ask users to update
1707 A conversion script can be found in the contrib module of the CVS, under
1708 the name contrib/platform_generation/surfxml_update.pl [MQ]
1711 * Bug fix: Processes were started in reverse order, wrt deployment file.
1712 WARNING: if your code relies on this bug, please fix it. [AL]
1713 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1714 being executed on two different locations. [AL]
1715 * Bug fix: Failures are now better supported thanks to Derrick's tests
1716 (there was many failure situations I hadn't thought of and that weren't
1717 correctly handled). [AL]
1718 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1722 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1723 Exception occurring on server-side are propagated back to client (!).
1725 API CHANGE: the callback changed their prototype. Change:
1726 int my_handler(gras_socket_t expeditor, void *payload_data) {
1728 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1729 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1731 * New! function: gras_msg_handleall to deal with all messages arriving
1732 within a given period.
1733 * New! function: gras_socket_server_range to get a server socket in a
1734 range of port numbers (ease to avoid port number conflicts) [MQ]
1735 * New! gras processes display their backtrace when they get a SIGUSR1
1736 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1737 Sweet to debug RL processes [MQ]
1741 - Do not force experiment sizes to be expressed in kb, or it becomes
1742 impossible to measure the latency this way (needs one byte-long tests)
1743 WARNING: this changes the amok_bw_* function semantic. [MQ]
1744 - Implements the link saturation stuff. [MQ]
1745 * Peer management module:
1746 New! module factorizing code that we wrote over and over [MQ].
1749 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1750 - Test units are placed directly into the library code, they get extracted
1751 automatically and placed into the src/testall binary.
1752 - Convert most of the XBT tests to this system.
1753 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1754 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1755 live in the same C code. This required to change a little bit the API
1756 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1757 * New module: sparse graph structure with basic algorithms (this is work
1758 in progress and the API is not considered to be frozen yet). [AL]
1759 * Display more information on backtraces: source line & function names are
1760 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1761 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1762 * Reworked a little bit some #include statements to load only required
1763 headers. Some user code that relied on SimGrid to include stdlib or
1764 stdio may need to include it by themselves. [AL]
1765 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1767 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1768 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1769 but I guess that almost nobody used those parts. [MQ]
1771 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1773 SimGrid (3.0.1) stable; urgency=low
1776 * Unfortunately, I had missed 5 misnamed functions:
1777 xbt_fifo_item_t xbt_fifo_newitem(void);
1778 void xbt_fifo_freeitem(xbt_fifo_item_t);
1779 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1780 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1781 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1782 They're now deprecated. Please use their new versions:
1783 xbt_fifo_item_t xbt_fifo_new_item(void);
1784 void xbt_fifo_free_item(xbt_fifo_item_t);
1785 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1786 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1787 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1789 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1790 * Documentation: xbt_log module unmercifully reworked [MQ]
1791 * Bugfix: there was a problem with the ending of contexts with
1792 the pthread backend. It caused some weird deadlock or behavior
1793 depending on the pthread implementation. [AL]
1794 * Bugfix: get the exceptions raised in the simulator repport where
1795 and why they come from when they are not catched in time [AL, MQ]
1798 * Bugfix: Do repport the error when two non-connected hosts try to
1799 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1802 * Add additionnal checkings on communications. Assert that two
1803 communicating hosts are connected by a set of links... [AL]
1806 * Add additionnal checkings on channel values in communication [AL]
1807 * New: MSG_task_get_source to see on which host a task was generated [HC]
1808 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1809 the number of tasks waiting to be received on channel and sent
1811 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1812 waits for the first task coming from a given host.. [AL]
1814 GRAS new functionnalities: [MQ]
1815 * Enhance the parsing macro to allow the size of multidimentional objects
1816 to be given thru annotations.
1817 * New example (and documentation): Matrix Multiplication a la RPC
1818 (as when I was young!) and fix a bunch of bugs found on the way.
1820 GRAS performance improvements: [MQ]
1822 * Reduce the amount of cbps creation/destruction by making it static to
1823 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1826 * Change libdata to a set so that we can search for stuff by ID (and thus
1827 reduce the insane amount of dict lookups)
1830 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1831 RL communication on top of SG since it's so uneffective.
1832 It may also be used for inter-thread communication in RL, one day.
1833 * Use gras_datadesc_copy() to exchange messages on top of SG
1835 - improve message exchange performance on top of SG
1836 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1837 * Don't exchange on the network the size of the used part of buffer,
1838 instead, specify the possible buffer size to read().
1840 - reduces the amount of read/write calls (one pair per exchange)
1841 - reduces the amount of exchanged data (the size)
1842 - allows to retrieve all arrived data on receiver side, if we don't need
1843 it right now (subsequent read will peek the buffer)
1844 - allows the receiver to proceed with the begining of the stream before
1845 everything is arrived
1846 - make it possible to build an iov transport (using readv/writev)
1848 - take care of the data with non-stable storage (like stacked data),
1850 * If possible, TCP send uses vector I/O (when writev() is here)
1851 - Don't use it for receive since we send data sizes and data on the
1852 same stream, so we wouldn't be able to chain large amount of chunks
1853 before having to flush the stuff to read the size.
1854 * Rework the transport plugin mecanism to simplify it and reduce the
1855 amount of pointer dereferencement when searching for the right function
1858 * I guess that now, we do almost as few system calls as possible while
1859 doing as few data copy as possible.
1861 To improve it further, we could try to send all the sizes first and then
1862 all the data (to use iov on receiving size), but it's only a partial
1863 solution: when you have 2 dimensional data, the sizes of the second
1864 dimension is data of the first dimension, so you need 3 streams.
1866 I'm not sure the potential performance gains justify the coding burden.
1868 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1870 SimGrid (3.00) stable; urgency=high
1873 * New! Give the possibility to hijack the surf parser and thus bypass
1874 MSG_create_environment and MSG_launch_application. Have a look at
1875 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1877 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1879 SimGrid (2.96) unstable; urgency=low
1884 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1885 This deprecates the xbt_error_t mecanisms.
1886 It modifies (simplifies) all XBT and GRAS API.
1887 MSG API keeps unchanged (exceptions raised by XBT are catched from
1888 within MSG and masked with existing error handling facilities)
1891 * New! Add a FATPIPE model. [AL]
1892 * New! Add a parallel task model. [AL]
1893 * New! Add automatically a loopback interface (in the default
1894 network model) if none was precised.
1897 * Bugfix: MSG_process_resume now works with the current running process.
1899 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1900 * Modification of MSG_task_get_compute_duration. Once a task has been
1901 processed, the value returned by this function is now equal to 0. [AL]
1902 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1903 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1904 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1905 of a m_task. For now, MSG_task_cancel only works with computation
1907 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1908 of the processor (in Mflop/s) regardless of the current load on the
1910 * API Change: use proper naming convention for MSG_getClock and
1911 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1913 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1914 This function changes the priority of a computation task. This priority
1915 doesn't affect the transfer rate. A priority of 2 will make a task
1916 receive two times more cpu power than the other ones. This function
1917 has been added to suit the needs of Nguyen The Loc and hasn't been that
1918 much tested yet. So if it fails, please report it and send me your code.
1920 * API Change: removed all functions and types that were marked "deprecated"
1921 since many months. Renamed MSG_global_init_args to MSG_global_init.
1923 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1925 SimGrid (2.95) unstable; urgency=low
1928 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1929 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1930 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1933 * Bugfix: Allow absolute paths for platform description files [MQ]
1934 * Bugfix: do free the variables after use. Leads to drastic performance
1936 * Implement max_duration (ie, timeouts) on resources [AL]
1939 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1941 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1942 the needed support (timeouts) [AL]
1945 * Implement measurement sockets. You can now get the bandwidth between two
1946 hosts thanks to AMOK (see below). [MQ]
1947 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1948 dynar over the network (yeah) [MQ]
1949 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1950 * simple/static token-ring example. [Alexandre Colucci and MQ]
1951 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1952 in place before [MQ]
1955 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1956 type parsing. "short" and "long" modifiers were also ignored; other
1957 modifier (such as reference level) are still ignored. [MQ]
1958 * Embeed the buffer size within the buffer itself on SG. [MQ]
1959 That way, send() are atomic and cannot get intermixed anymore (at least
1960 the ones which are less than 100k; bigger messages still have the issue)
1961 * Array size pushed by the field, not by the field type (or each
1962 and every long int will push stuff to the cbps) [MQ]
1963 * use select() to sleep since it allows to portably sleep less than one
1966 GRAS (minor cleanups)
1967 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1970 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1971 * type category 'ignored' killed as it was never used and were difficult
1973 * whether a type can cycle or not is now a flag, leaving room for more
1974 flags (as "ignored", if we feel the need one day ;)
1975 * Rename raw sockets to measurement sockets since "raw" has another
1976 meaning in networking community.
1979 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1980 and offers features not belonging to GRAS but that most applications
1981 need. One day, it may be a set of plugins loadable at runtime.
1982 * New module: bandwidth
1983 bandwidth measurement between arbitrary nodes running this module. [MQ]
1985 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1987 SimGrid (2.94) unstable; urgency=low
1989 The first beta release of SimGrid 3 !
1993 * Update the main page and the FAQ. Adding references to gforge.
1996 * Add a gras_os_getpid function.
1999 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2000 * Extend the logs so that they also print PID, hostname, date, ... if
2002 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2003 and kill the old version which were in testsuite/
2004 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2008 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2009 for users. Instead, each programming environment (GRAS, MSG,...) use it
2010 when needed to provide such a feature to users.
2011 Don't shortcut the mecanism or you will also shortcut the virtualization
2012 you need on the simulator.
2016 * Cleanups in configury with regard to compile optimization/warning flags.
2017 Also add -fno-loop-optimize to any powerpc since it's the optimization
2018 killing gcc (< 3.4.0).
2019 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2020 needed by MSG examples complications
2021 * Borrow configury beautifications from PHP
2024 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2025 warning (thanks loris for stumbling into this one).
2026 * Bugfix: stop loading private headers (gras_config.h) from the public
2030 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2031 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2032 RL would imply a DNS resolver, which is *hard* to do in a portable way
2033 (and therefore delayed).
2034 * Implement a real timer mecanism and use it in timing macros. This allows
2035 to avoid rounding errors and get a 0.000005 sec precision in timing
2036 macros. While I was at it, various cleanups:
2037 - allow to declare more than one timed section per file (fix a stupid bug)
2038 - move some private declaration to the right place
2039 - merge conditional execution and timing macros into emulation module
2040 - document the module
2041 - make sure the module cleanups its mess on gras_exit
2042 * Documentation improvements:
2043 - (new) how to compile applications using GRAS
2044 - (new) emulation support (timing macros)
2046 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2048 SimGrid (2.93) unstable; urgency=low
2050 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2053 - Use Paje properly where used. Still to be sanitized properly.
2054 - Portability fix: Add an implementation of the contexts using pthread
2058 - Add xbt_procname(): returns the name of the current process.
2059 Use it to show the current process's name in all logging.
2061 - fix detection of older flex version and the reaction, since we do
2062 depend on modern ones (we use lex_destroy)
2063 - Better separation of SG and RL in the libs: remove all simulation code
2064 from libgras. As a result, this lib is now only 200k when stripped.
2065 Some of the xbt modules may also be duplicated (two sets and such) and
2066 should be cleaned/killed before SG3.
2067 - Insist on using xlC on AIX because of weird problems involving gcc there.
2068 - Cleanup the make remote stuff. This is now done by scripts
2069 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2070 mainly for our private use, but we're working on changing them to user
2073 - Bugfix: flush the socket on close only if there is some *output*.
2074 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2076 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2079 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2081 SimGrid (2.92) unstable; urgency=low
2083 Alpha 3 on the path to SimGrid 3
2087 - New! First try of benchmarking macros.
2088 - New! First try so that gras_stub_generator generate deployment and
2089 remote compilation helpers.
2091 - Bugfix: Initialization fix in msg_test.
2095 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2097 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2099 - API Change: gras_os_sleep to take the amount of seconds as a double.
2100 Accepting an int was error prone since it was the only location where
2101 seconds were coded as such. It leaded to damn rounding errors.
2102 - Bugfix: Hard to belive that timers ever worked before this.
2104 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2106 SimGrid (2.91) unstable; urgency=low
2108 Alpha 2 on the path to SimGrid 3
2112 - Bug fix in the lmm_solver.
2114 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2115 through the function MSG_paje_output.
2116 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2117 - It is possible to bound the rate of a communication in MSG with
2118 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2119 to write it in the changelog).
2120 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2125 - Various cleanups to the autotools stuff
2126 - Begin to move Gras examples to examples/gras/
2127 - Let make distcheck work again (yeah!)
2129 - documentation overhauled using doxygen.
2130 gtk-doc-tools is dead in SimGrid now.
2131 - Automatically extract all existing logging categories, and add the list
2132 to the documentation (long standing one, to say the less)
2134 - Cleanup the known architecture table. Reorder the entries to group what
2135 should be, and use a more consistent naming scheme.
2136 (some of the test dataset are still to be regenerated)
2137 - New! Allow library to register globals on each process just as userdata
2139 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2140 lookup time (for now).
2141 Use it in msg and trp.
2142 This cleans a lot the internals and helps enforcing privacy of the
2143 headers between the gras components.
2144 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2145 - Bugfix: gras_os_time was delirious in RL.
2146 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2148 - Reenable GRAS now that it works.
2150 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2152 SimGrid (2.90) unstable; urgency=low
2154 Alpha 1 on the path to SimGrid 3
2156 * It is a long time since the last release of SimGrid. I'm sorry about
2157 that but as I had told you, I was rewriting a lot of things. I apologize
2158 to those who had been reporting bugs to me and that I had not answered.
2159 If your bug is still in the new version, please tell me. Here is a
2160 summary of the main changes.
2162 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2163 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2164 portability, speed, and a lot more but you'll figure it out later.
2165 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2166 I wanted to release the new MSG as soon as possible and I have
2167 broken GRAS, which is the reason why, for now, only MSG is fully
2168 functional. A laconic description of these projects is available
2169 in the documentation.
2171 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2172 kernel optimized for our needs (called SURF but only the developers
2173 should use it). Hence, MSG is now roughly 30 times faster and I think
2174 that by rewriting a little bit MSG, I could event speed it up a little
2175 bit more. Beside the gain in speed, it is also much easier to encode a
2176 new platform model with SURF than it was with SG. More to come...
2178 * REVOLUTION 3: I have tried to change a little as possible the API of
2179 MSG but a few things really had to disappear. The main differences
2180 with the previous version are :
2181 1) no more m_links_t and the corresponding functions. Platforms are
2182 directly read from a XML description and cannot be hard-coded
2183 anymore. The same format is used for application deployment
2184 description. The new format is described in the documentation.
2185 Have a look in tools/platform_generation. There is a tiny script
2186 that converts from the old platform format to the new one. Concerning
2187 the application deployment format, parsing the old one is tricky.
2188 I think most of you should however be able to convert your files. If
2189 it is really an issue, I can write a C code that does the conversion.
2191 2) the toolbox tbx does not exist anymore. We now have a library
2192 with much more data-structures but without the hash-tables (we have
2193 dictionaries that are much faster).
2195 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2197 *****************************************************************************
2198 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2199 * don't feel like converting the oldies. (Mt) *
2200 *****************************************************************************
2203 Version 2.90: "the long awaited one"
2204 - Finished rewriting and debugging MSG. Rewrote the documentation.
2205 - disable GRAS for now since it needs to be ported to the newest SG
2208 - Finish the port to windows (using mingw32 for cross-compile)
2211 - Main loop and datastructures of SURF. A cpu resource object is
2212 functional. Surf can thus be used to create cpu's with variable
2213 performance on which you can execute some actions.
2215 2004-11-15 Martin Quinson
2216 - Port to ARM. Simply added the alignment and size descriptions. Should
2217 work, but the ARM machines are so slow that I didn't had the opportunity
2218 to 'make check' over there yet.
2220 2004-11-15 Arnaud Legrand
2221 - Trace manager now written. It uses a heap structure and is therefore
2222 expected to be efficient. It may however be speeded up (particularly
2223 when many events occur at the same date) by using red and black
2224 trees. One day maybe...
2225 - Max-min linear system solver written. It uses a sparse matrix
2226 structure taking advantage of its expected use. Most operations are
2227 O(1) and free/calloc are called as few as possible. The computation of
2228 the minimum could however be improved by using a red and black tree
2231 2004-11-03 Arnaud Legrand
2232 - Rename every gras_* function that was in xbt/ to its xbt_
2234 - Add a heap and a doubly-linked list to xbt
2235 - Added a dichotomy to the dictionaries. make check works as well before
2236 so I assume that the patch is correct. I do not know however if things
2237 run effectively faster than before now. :)
2239 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2240 SimGrid, and the version number is bumped to 2.x
2242 2004-10-29 Martin Quinson
2243 - Introduction of the remote errors.
2244 They are the result of a RMI/RPC on the remote machine.
2245 ErrCodes being scalar values, you can't get the host on which those
2246 errors did happen. Extending the error mechanism as in Gnome is possible.
2247 No idea yet whether it is a good idea.
2249 2004-10-28 Martin Quinson
2250 - Interface revolution: the Starred Structure Eradication.
2251 I used to do typedef struct {} toto_t; and then handle *toto_t.
2252 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2253 - ???_t is a valid type (builded with typedef)
2254 - s_toto_t is a structure (access to fields with .)
2255 - s_toto is a structure needing 'struct' keyword to be used
2256 - e_toto_t is an enum
2257 - toto_t is an 'object' (struct*)
2259 typedef struct s_toto {} s_toto_t, *toto_t;
2260 typedef enum {} e_toto_t;
2261 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2262 s_toto_t) is private.
2264 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2265 so that it can set the variable to NULL. It was so for dicts and sets,
2266 it changed for dynars.
2268 - Fix a bunch of memleaks in dict_remove
2269 - Fix a bug in sg/server_socket opening: it failed all the time.
2271 2004-10-07 Martin Quinson
2272 - Speed up dynar lookup operation a bit.
2274 gras_dynar_get is dead.
2276 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2277 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2278 (which gives you the address of the stored data).
2280 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2281 copy of the data using an affectation to do the job and not a memcpy.
2283 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2284 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2286 It does not really speedup the dynar test because they are
2287 setting elements all the time (and look them seldom). But the dict does
2288 far more lookup than setting.
2290 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2292 2004-10-05 Martin Quinson
2293 - Allow to (en/dis)able the cycle detection at run time.
2295 Whether we should check for cycle or not is now a property of each
2296 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2297 datadesc_cycle_unset allow to remove this property when previously set.
2299 Note that the cycle detection is off by default since it impacts the
2300 performance. Watch the data you feed GRAS with ;)
2302 This property is hereditary. Any element embedded in a structure having it
2303 set have it set for the time of this data exchange.
2305 You should set it both on sender and receiver side. If you don't set it on
2306 sender side, it will enter an endless loop. If you forget on receiver
2307 side, the cycles won't be recreated after communication.
2309 - Header reorganization.
2310 Kill gras_private.h, each submodule must load the headers it needs.
2312 2004-10-04 Martin Quinson
2313 - Interface revolution: do not try to survive to malloc failure.
2315 Now, gras_malloc and friends call gras_abort() on failure.
2316 As a conclusion, malloc_error is not a valid error anymore, and all
2317 functions for which it was the only gras_error_t return value are
2318 changed. They now return void, or there result directly.
2319 This simplify the API a lot.
2321 2004-09-29 Martin Quinson
2322 - Re-enable raw sockets.
2323 Created by gras_socket_{client,server}_ext;
2324 Used with gras_raw_{send,recv}
2327 It should allow to kill the last bits of gras first version soon.
2329 This is not completely satisfactory yet (duplicate code with
2330 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2333 - Simplify transport plugin (internal) interface by not passing any
2334 argument to _server and _client, but embedding them in the socket
2337 2004-09-28 Martin Quinson
2338 - Finish the port to AIX.
2339 autoconf was my problem (segfault within the malloc replacement
2340 function. No idea why)
2342 2004-09-16 Martin Quinson
2343 - Fix some size_t madness on 64bit architectures.
2345 2004-09-08 Martin Quinson
2346 - Reduce the number of system headers loaded, overload some more system
2347 calls (such as malloc to cast the result of the system one, and work
2349 - Fix and reintroduce the config support
2351 2004-09-07 Martin Quinson
2352 - Source code reorganization to allow Arnaud to surf all over there.
2353 - Allow to document the logging categories.
2354 - Remove all uppercase from logging categories and useless cleanup in names.
2356 2004-08-18 Martin Quinson
2357 Version 0.6.2 (protocol not changed; API changed)
2358 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2359 gras_error_t), and NULL when not found. Functions expecting a msgtype
2360 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2361 hopefully usefull message.
2362 - Portability to prehistoric sparcs again
2364 2004-08-17 Martin Quinson
2365 Version 0.6.1 (protocol not changed; ABI not changed)
2366 - prealloc some buffers to speed things up
2368 2004-08-11 Martin Quinson
2369 Version 0.6 (protocol not changed; ABI expended)
2370 - The parsing macro can deal with the references, provided that you add
2371 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2373 2004-08-09 Martin Quinson
2374 Version 0.5 (protocol not changed; ABI changed)
2375 - Allow to off turn the cycle detection code in data exchange at
2376 compilation time. It should be at run time, but I'm short of time (and
2377 the config stuff is still broken). That way, we keep dict out of the
2378 critical path, which is good because the performance is poor:
2379 - search not dichotomial yet
2380 - dynar give no way to access their content and memcpy everytime
2381 - In composed data description (struct, ref and so on), stop foolness of
2382 keeping the subtype's ID, but store the type itself. This keeps sets out
2383 of the critical path, which is good since they rely on dynar and
2384 dictionnaries. The only loose of that is that we cannot detect the
2385 redeclaration of a structure/union with another content (but I'm not sure
2386 the code detected well this error before anyway). We still can detect
2387 the redefinition discrepancy for the other types.
2388 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2389 it breaks the code because of type-punning used all over the place).
2390 This breaks on all non-gcc architectures (for now).
2392 All those changes (plus the buffer of last time) allow me to gain 2 order
2393 of magnitude on cruel tests consisting of 800000 array of integers on two
2394 level of a hierarchical structure (200 secondes -> 4 secondes)
2397 - the selector of reference must now return the type it points to, not
2398 the ID of this type.
2400 2004-08-06 Martin Quinson
2401 Version 0.4 (protocol changed; ABI not changed)
2402 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2403 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2404 - Implement a buffer transport, and use it by default (it relies on tcp in
2405 real life and on sg in simulation).
2406 That's a bit hackish since I had a new field to the structure to store
2407 its data without interfering with the subtype ones. Inheritance
2408 is tricky in C. And that's a kind of reverse inheritance with one class
2409 derivating two classes. Or maybe a game with java interfaces. Anyway,
2410 that's damn hard in C (at least).
2411 Moreover, I got tired while trying to ensure plugin separation and
2412 genericity in SG mode. MSG wants me to do weird things, so let's go for
2413 cruel hacks (temporarily of course ;).
2414 See comment in transport_private.h:71
2415 - do not include all the _interface headers in private but in the files
2416 which really need them (to cut the compilation time when they are
2419 2004-07-26 Martin Quinson
2420 Version 0.3 (protocol not changed; ABI changed)
2421 - Major overhault of the datadesc interface to simplify it:
2422 - shorted the function names:
2423 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2424 - add a trivial way to push/pop integers into the cbps without malloc.
2425 This allows to make really generic sub_type description, which simply
2426 pop their size of the stack.
2427 - add a function gras_datadesc_ref_pop_arr() which does what users want
2428 most of the time: Declare a dynamic array (which pops its size of the
2429 stack) and declare a reference to it. Poor name, but anyway.
2430 - kill the post-send callback, add a post-receive one
2432 2004-07-23 Martin Quinson
2433 Version 0.2 (protocol changed; ABI changed)
2434 - add some testing for cpbs in the test cases, and fix some more bugs.
2435 This invalidate again the little64 data file, since I cannot regenerate
2437 - remove an awfull optimization in the logging stuff, allowing me to:
2438 - understand it again
2439 - learn gcc how to check that the argument match the provided format
2440 - fix all errors revealed by gcc after that
2441 - internal keys of dict are not \0 terminated. Deal with it properly in
2442 loggings instead of segfaulting when the user want to see the logs :-/
2444 2004-07-22 Martin Quinson
2445 - Fix some stupid bug preventing cbps (callback postit) from working
2447 2004-07-21 Martin Quinson
2448 - Some documentation cleanups
2449 - remove the useless last argument of msgtype_declare
2450 - rename the Virtu functions to fit into the 'os' namespace
2451 - move headers src/include -> src/include/gras/ and stop fooling with
2452 gras -> . symbolic link
2453 - make distcheck is now successful
2455 2004-07-19 Martin Quinson
2457 - Build shared library also
2458 - Install html doc to the right location
2459 - stop removing maintainer files in make clean
2460 - build tests only on make check
2462 2004-07-13 Martin Quinson
2464 - No major issue in previous version => change versionning schema
2465 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2466 corresponding dataset.
2468 2004-07-11 Martin Quinson
2470 - Get it working with any kind of structure (we can compute the padding
2471 bytes remotely for all the architectures I have access to)
2472 - Implement the structure parsing macro (still not quite robust/complete)
2473 - Improvement to the remote testing toysuite
2475 2004-07-10 Martin Quinson
2476 [autoconf mechanism]
2477 - get ride of a bunch of deprecated macros
2478 - actually run the test for two-compliment, not only compile it :-/
2479 - test whether the structures get packed (and bail out if yes. Damn.
2480 Alignment is a serious matter)
2481 - test whether the structures get compacted (but respecting the alignment
2482 constraints of each types)
2483 - test whether the array fields of structures can straddle alignment boundaries
2485 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2486 (took me 2 hours to find that bug, looking at the wrong place)
2487 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2488 before sending/receiving and are used to compute the offsets of fields
2489 - Given that padding size depend even on compiler options, keep track of
2490 alignment and aligned_size only for the current architecture. Not a big
2491 deal since we send structure fields one after the other (seems
2493 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2494 program, let it work on linux/gcc/little32. portability todo.
2496 2004-07-08 Martin Quinson
2497 - import and improve remote compilation support from FAST
2498 - make sure make check works on half a dozen of machines out there
2500 2004-07-07 Martin Quinson
2501 Let's say it's version 0.0.3 ;)
2502 - Implement conversions (yuhu!)
2503 - Let it work on solaris (beside conversion, of course)
2504 - Stupid me, using rand() to generate the conversion datatests in not wise.
2506 2004-07-06 Martin Quinson
2507 - Let make dist work, since I'm gonna need it to compile on remote hosts
2508 - Let Tests/datadesc_usage write the architecture on which the file was
2509 generated as first byte.
2510 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2511 architecture descriptions.
2512 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2513 on those architectures.
2514 - Optimization: send/recv array of scalar in one shoot
2516 2004-07-05 Martin Quinson
2517 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2519 - Plug a whole bunch of memleaks
2520 - each process now have to call gras_{init,exit}. One day, their log
2521 settings will be separated
2522 - Continue the code factorisation between SG, RL and common in Transport.
2524 2004-07-04 Martin Quinson
2526 - Redistribution between SG and RL.
2527 We wanna have to accept in SG, so move accepted related parts of RL in
2528 the common part. (more precisely, the dynar of all known sockets is no
2529 more a static in transport.c, but part of the process_data)
2531 [gras_stub_generator]
2532 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2534 2004-07-03 Martin Quinson
2535 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2536 them all up in one shot)
2538 - Fix the ugly case of reference to dynamic array.
2539 - New (semi-public) function gras_datadesc_size to allow the messaging
2540 layer to malloc the needed space for the buffer.
2542 - gras_socket_close now expect the socket to close (and not its address to
2543 put NULL in it after it). This is because the socket passed to handlers
2544 is one of their argument (=> not writable).
2546 - propagate the interface cleanup from last week in datadesc, ie remove a
2547 superfluous level of indirection. User pass adress of variable
2548 containing data (both when sending and receiving), and not of a variable
2549 being a pointer to the data. Let's say that I like it better ;)
2550 The price for that is constructs like "int msg=*(int*)payload" in
2551 handlers, but it's a fine price, IMHO.
2553 - Let it work in RL (yuhu)
2555 2004-06-21 Martin Quinson
2557 - porting SG plugin and SG select to new standards (works almost).
2558 - plug memleaks and fix bugs around.
2561 - cleanup the prototype of data recv and force users to specify when they
2562 want to handle references to objects. Test case working even for cycles.
2563 - plug memleaks. Valgrind is perfectly ok with this.
2565 2004-06-12 Martin Quinson
2567 - cleanup the separation between plugin and main code in plugin creation
2569 2004-06-11 Martin Quinson
2571 - Reput hook for raw sockets, needed for BW experiments
2572 - kill a few lines of dead code
2573 [Data description] Interface cleanup
2574 - gras_datadesc_by_name returns the searched type or NULL.
2575 That way, no variable is needed to use a type desc once, which makes
2577 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2578 two parameters were strings (field name, type name), leading to
2580 [Dicos] Interface cleanup
2581 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2582 This is consistant with the dynar API.
2584 2004-04-21 Martin Quinson
2586 - Porting to new standards.
2588 - interface cleanup.
2589 There is no bag anymore, no need to take extra provision to mask the
2590 pointers behind "ID".
2591 Better splitup of functions between files create/exchange/convert.
2592 This is still a bit artificial since convert and receive are so
2593 interleaved, but anyway.
2595 - add a queued message list to procdata (the ones not matching criteria
2597 - factorize some more code between SG and RL wrt procdata
2599 - use gras_exit in example to track memleaks
2600 - get rid of gs_example now that GS is properly integrated into gras
2601 - update run_test to integrate the latest tests (datadesc)
2603 - rename WARNINGn macros to WARNn since it prooved error-prone
2605 2004-04-19 Martin Quinson
2607 - register init/exit functions within gras module mechanism
2608 - send/receive function.
2609 Convertion is not implemented, but short-cutted if not needed.
2610 struct/array elements are sent one by one (instead of block-wise), but
2611 nobody really cares (yet). Get a prototype before optimizing.
2612 - tests (using a file socket) for DD send/receive on:
2613 - base types: int, float
2614 - array: fixed size, string (ie ref to dynamic string)
2615 - structure: homogeneous, heterogeneous
2616 - chained list, graph with cycle
2617 Believe it or not, valgrind is not too unhappy with the results. The
2618 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2622 - Bugfix when using a filename explicitely (instead of '-')
2624 2004-04-09 Martin Quinson
2626 - factorize more code between RL and SG in socket creation
2627 - Complete the implementation and tests of:
2629 o file (only in RL, and mainly for debugging)
2631 I lost 3 days to design a portable address resolver, and then decided
2632 that the prototype mainly have to run on my box.
2633 Addressing portability too early may be like optimizing too early :-/
2635 - use gras_init in the Tests instead of the crappy parse_log_opt
2636 (the latter function is removed)
2637 [Conditional execution]
2638 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2639 [Code reorganisation]
2640 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2641 Build examples against the RL library, since there is no way to disable
2642 its creation for now.
2644 For information, the beginning of coding on GRAS was back in june
2645 2003. I guess that every line has been rewritten at least twice since