1 SimGrid (3.11) NOT RELEASED; urgency=low
3 TENTATIVE RELEASE GOALS for 3.11:
4 * Consider the removal of Supernovae mode if no user has manifested since the
5 release of version 3.10.
6 * Switch to tesh.pl, and kill the now unused parts of xbt that seem fragile
7 * Switch to surf++, and reintegrate the hypervisor branch on top of it
9 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
11 SimGrid (3.10-rc1) unstable; urgency=low
14 * Reintegrate Java to the main archive as desynchronizing these
15 package is not acceptable anymore (Java is now considered stable)
16 * Add explicit synchronization facilities through semaphores
17 * Bug fix: Task.setDataSize() only changed the C world, not the value
18 cached in the Java world
21 * Dramatically change the way files are handled. API and internals changed, but
22 this part of MSG was not considered as production grade either.
23 * Add explicit synchronization facilities through semaphores
24 * Add a new function MSG_host_get_process_list()
25 * Preliminary DVFS support (see examples/msg/energy/ for details)
28 * SMPI is now included directly in the libsimgrid as the windows
29 linker doesn't force us on splitting it anymore.
30 * Improvements of the SMPI replay tool:
31 - Most of the collective communications are now rooted in the same process as
32 in the original application.
33 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
34 used until now). Multiple data types can now be used in a trace.
35 - The replay tool now supports traces produce either by TAU or a modified
37 - Bug Fix: the compute part of the reduce action is now taken into account.
38 * smpirun generates the host file if needed (with given host count and platform)
39 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
40 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
41 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
42 --cfg=smpi/coll_selector:(mpich/ompi)
43 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
45 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
46 happening inside a collective SMPI call.
47 * Fix the behavior of complex data types handling.
48 * Make MPI_Wtime another synchronization point to take computations into account.
49 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
50 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
51 * Add all missing Fortran bindings, SMPI should work with Fortran 90
52 (no privatization of global variables yet)
55 * Verification of liveness properties is now available for SMPI applications
56 (in addition to MSG applications)
57 * Bugged examples using SMPI in examples/smpi/mc/
58 * Add --cfg=model-check/visited option. Allows the verification of infinite
59 programs. Detection of loops in the execution thanks to the system state
60 comparison and reduction of the state space to explore. Can be combined with
61 DPOR for safety properties.
64 * Allow to change SimGrid configuration (see --help) within the code
65 thanks to SD_config() as it can be done in MSG.
66 * Add a new function SD_task_set_amount() upon user request.
69 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
70 * Remove rule based routing (no more PCRE dependency)
71 * Add a limiter_link option to cluster tag, to specify a maximum reachable
72 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
73 * Add a loopback_bw and loopback_lat options to cluster tag.
74 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
75 routing. Now peers are encapsulated in an AS and have their own private
76 router but this is transparent.
79 * Our own implementation of getline is renamed xbt_getline, and gets
80 used even if the OS provide a getline(). This should reduce the
81 configuration complexity by using the same code on all platforms.
82 * New type: xbt_cfg_elm_boolean.
83 * Allow to use yes/no for boolean configuration options in the command line.
84 * Allow to disable SimGrid cleanups at exit from command line option.
85 There are situations where one may want a simulation to end with an exit.
86 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
87 annoying when scripting around the simulator. Adding a
88 --cfg=clean_atexit:no allows to circumvent this issue.
91 * New command line option --version, to get SimGrid version information.
92 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
94 * Supernovae builds are deprecated, and expected to be removed in the next
97 -- Mon Oct 7 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
99 SimGrid (3.9) stable; urgency=low
101 The Grasgory release: GRAS is really dead now.
103 * Complete overhaul of the internal host structures scheme.
106 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
107 was considered as experimental and badly maintained since then.
108 * Keeping it was thus a trap to our potential users, that could take
109 it instead of MSG or SMPI by mistake despite is pity state.
110 * GRAS seems to have very few users (if any), and no one volunteered
111 to maintain it further. It also induces a lot of XBT code (for
112 portability sake), that must be maintained too.
113 * For all these reasons, we killed GRAS. If someone wants to revive it
114 in the future, don't cry, our git history still remembers of GRAS.
117 * Major overhaul. Merge our documentation again as time proved that
118 splitting it was really not helping our users.
119 * Further improve the developer documentation to help newcomers
120 hacking on SimGrid itself. The user documentation (and in
121 particular, the beginner documentation) is still in a sorry state.
124 * Now works on Windows too!
125 * Much more extensive test suite, from MPICH
128 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
129 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
130 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
131 * Bug fix: let task be scheduled when the last dependency to be solved is
132 a control dependency.
133 * Remove SD_load_environment_script function.
134 Use the C sg_platf function if you want to declare a platform
138 * New function: MSG_process_get_number()
139 * Old function documented: MSG_config()
140 * Remove MSG_load_platform_script function
141 Use the C sg_platf function if you want to declare a platform
145 * Change the default value of the TCP_gamma constant (maximal size of TCP
146 congestion window) to a more realistic 4MiB value. If you notice changes in
147 your simulation results, you can fall back to the previous 20k tiny window
148 by adding --cfg=network/TCP_gamma:20000 on command line.
149 * (Hopefully) fix a bug wrt periodic availability and state traces
150 * Bug fix: use default values at start when first event in availability/state
151 trace is not at time 0.
154 * remove the "new_" part of function name sg_platf_new_trace_connect
155 (resulting in sg_platf_trace_connect), since it does not create
159 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
160 use case where it's really mandatory, and maintaining it was a pain in
162 * New: xbt_fifo_search(), search an item with a user-provided
163 comparison function instead of dumb pointer comparison.
166 * Fix the lua deployment:
167 Use `simgrid.init_application()` before deployment instead of
168 `simgrid.msg_register_application()` after.
171 * Transfer the tracing files into the corresponding modules.
173 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
175 SimGrid (3.8.1) stable; urgency=low
177 The "we are told that some people want to also *install* the simgrid
180 * Add missing file "tesh.1" to the archive.
182 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
184 SimGrid (3.8) stable; urgency=low
186 The Psssshiiiit release: SimGrid jumps into the Cloud.
189 * Add an experimental interface to manipulate VMs. They are mainly
190 process groups with very few intrinsic semantic, but they should
191 allow you to build the semantic you want easily.
192 * New function: MSG_host_set_property_value()
193 * New function: MSG_process_on_exit(). To clean memory in all cases.
194 * Bug fixes that made the host (and link) failures unusable.
195 * Add a way to auto-restart process when the host in which they are
196 executing comes back (ON_FAILURE="RESTART" on deployment file,
197 MSG_process_auto_restart_set).
198 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
199 please stop using the old ones, they are DEPRECATED.
201 * Deprecate functions MSG_global_init() / MSG_global_init_args()
202 Please use MSG_init() instead. (reducing the amount of entry
203 points in the library helps us).
204 * Make it impossible to link against the wrong version of the lib
205 * Deprecate MSG_clean(). No need to call it anymore.
206 * Function MSG_get_host_number() is not deprecated anymore.
209 * Split the doc into a user guide and a reference guide.
210 * Start a developper guide to help people hacking on SimGrid.
213 * Enable tracing by default. This modules rocks you should use it.
214 * Remove option custom_flags. Now use environment variables CFLAGS
216 * Use default cmake things to detect lua instead of home grown ones.
217 * New option "enable_mallocators" to disable mallocators, for debugging
218 purpose ("on" by default).
221 * Bug fixes around the resource failures: don't let the processes
222 survive the host they are running onto.
223 * Add an interface to auto-restart processes when the host in which they are
224 executing comes back.
225 * Ensures that SIMIX_clean is called automatically. It's not part of
226 the public interface anymore (bindings should be updated).
229 * Bug fix for when SD_Simulate is called with a positive value: be careful
230 when comparing doubles. Sometimes they are different for non significant
232 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
233 parallel task whose initial work is distributed among host according
234 to the Amdahl's law. Such tasks are created with a parameter alpha
235 that corresponds to the non-parallelizable part of the computation.
236 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
237 between two sets of workstations assuming a 1D block distribution (each
238 workstation owns a similar share of data) on both sides.
240 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
241 Data redistribution will be automatically scheduled once parent and child
242 are both scheduled. The filling of computation_amount and
243 communication_amount structures is now done seamlessly thanks to the chosen
245 * New function SD_workstation_dump to display various information
246 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
247 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
248 bandwidth on the route onto which the task is scheduled and the amount of
250 To divide the nominal bandwidth by 2, the rate then has to be :
251 rate = bandwidth/(2*amount)
252 * Compute tasks that have failed can now be rescheduled and executed again
253 (from their beginning)
254 * Increasing source code coverage (src/simdag is now covered at 95.8%
258 * Re-implement time-independent trace replay using SMPI (at the
259 smpi_smp_* level) instead of MSG. This should replace
260 examples/msg/actions/actions.c
261 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
262 hindexed and structs)
263 * Implement the exchange of non-contiguous data.
264 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
265 * Correct behavior of smpi/sender_gap and set its default value to 0
266 * Add option to asynchronously send small messages to allow better
267 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
268 specifies the size in bytes under which messages will be asynchronously sent.
269 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
270 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
273 * New C interface to define a platform: XML is now optional.
274 For more info, please check include/simgrid/platf.h.
275 * New interface to define random platforms from the C:
276 For more info, please check include/simgrid/platf_generator.h and
277 examples/msg/masterslave/masterslave_platfgen.c
278 * Export a sg_cmdline dynar containing all the arguments we got from
282 * Two new tracing options for adding comments to trace file so you
283 can track your experiments (see --help-tracing for details).
284 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
285 * Adding the SimGrid version that generated the trace file as a comment.
286 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
287 * Fix to avoid key clashes on Paje links
288 * Other minor fixes related to the Paje specification
291 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
292 and renamed to xbt_str_hash() and xbt_str_hash_ext().
293 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
294 * Greatly improve the robustness of mmalloc to user errors (such as
295 using an area after freeing it, or freeing it twice)
297 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
299 SimGrid (3.7.1) stable; urgency=low
302 * Restore the prototype of MSG_process_create_with_environment() to
303 the pre-3.7 situation by removing the kill_time argument.
304 * Add a MSG_process_set_kill_time() function instead.
307 * Fix weird behaviors when dealing with parallel tasks.
310 * Simgrid is now built as a dll.
311 * Simgrid-java now works on Windows.
312 * Simgrid-Java is now included into Windows package.
315 * First pre-build package for MacOSX.
318 * Fix compilation when using MSG_USE_DEPRECATED.
319 * Fix some compilation issues on Macs and Windows.
320 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
321 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
323 We discovered that the Lua console is broken, but we are missing the
324 manpower to fix it right now. The problem existed in 3.7 too, so we
325 are not blocking the release for that. Sorry if you depended on this
326 feature, any help would be really welcome.
328 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
330 SimGrid (3.7) stable; urgency=low
332 The "spring cleanups (before the next Big Project kicks in)" release.
335 * We can specify the SMPI latency/bandwidth factor with command line
336 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
337 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
338 You can also use the "config tag" from platform file by adding this line
339 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
340 (see "example/platforms/tag_config.xml" to use "config tag").
341 Note that the command line supersedes the platform file configuration.
342 * Change the correction factors used in LMM model, according to
343 the latest experiments described in INRIA RR-7821.
344 Accuracy should be improved this way.
345 * Use the partial invalidation optimization by default for the
346 network too. Should produce the exact same results, only faster.
347 * Major cleanup in surf to merge models and split some optimization
348 mechanisms from the core of the models. As a result you can now
349 specify which model to use (e.g., --cfg=network/model:LV08
350 --cfg=cpu/model:Cas01) and which optimization mode to use
351 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
352 Incompatible combinations should err at initialization. See
353 --help-models for the list of all models and optimization modes.
354 * The CLM03 workstation model was dropped for simplicity because it
355 used the deprecated CM02 network model. Use default instead.
356 * Rename the TCP_gamma configuration option to network/TCP_gamma
357 * Rename the coordinates configuration option to
358 network/coordinates, and document it
359 * Use now crosstraffic keyword instead of the terribly misleading
360 fullduplex keyword. It is activated by default now in the current
361 default model, use --cfg=network/crosstraffic:0 to turn it off.
362 * Ongoing refactoring the model parsing to make XML files optional
363 See include/simgrid/platf.h for details (still to be completed)
366 * Major overhaul of the documentation. Almost instructive now :/
367 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
368 functions and friends. This interface was considered as
369 deprecated since over 2 years, it's time to inform our users that it is.
370 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
371 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
372 use these (crappy) functions in your code.
373 These functions will be removed soon. Stop using them now.
374 * Deprecate MSG_get_host_{table,number}
375 Implement MSG_hosts_as_dynar() instead.
376 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
377 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
380 * Stabilize the parallel execution mode of user contexts
381 * Introduce configuration variables to control parallel execution:
382 - contexts/synchro: Synchronization mode to use when running
383 contexts in parallel (either futex, posix or busy_wait)
384 - contexts/parallel_threshold: Minimal number of user contexts
385 that must be part of a scheduling round to switch to parallel
386 execution mode (raw contexts only)
387 * Fix bugs that prevented to use suspend/resume along with
388 synchronization structures.
389 * Fix bugs in process termination that lead to invalid memory access
390 in very specific conditions.
393 * Introduce a parallel mode for the models (controlled by surf/nthreads
394 configuration item). In our tests, running the models in parallel
395 never lead to any speedups because they are so fast that the gain
396 of computing each model in parallel does not amortizes the
397 synchronization costs, even when ultra fast futexes are used.
398 This is released anyway because YMMV.
401 * Performance boost by using a swag internally to compute the set of
402 tasks that are finished and should constitute the return value of
406 * Enable it by default now that it is considered rather stable.
409 * Documentation of the tracing functions.
410 * Performance gains when tracing categorized/uncategorized resource
411 utilization by avoiding calls to get route when updating resource
412 variables. LMM constraints are being used instead.
413 * API changed to set task categories. Use MSG_task_set_category instead
414 of TRACE_msg_set_task_category, and SD_task_set_category instead
415 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
416 * Bugfix for graphicator, routes not correctly obtained, memory leaks
417 * Examples for link user variables added (see at examples/msg/tracing/)
418 * Deprecated function TRACE_msg_set_process_category completely removed
419 * Trace header updated according to the latest Paje file format
420 * Tracing network lazy updates, no longer obligate users to use full updates
421 * --cfg=tracing/platform:1 also registers power/bandwidth variables
422 * Experimental: let user code declare/set/push/pop application states for hosts
423 * API changed to allow the manual creation of graph configuration files
424 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
427 * Improve the API of Lua MSG bindings, using the Lua spirit.
428 * Each simulated process now lives in its own Lua world (globals are
429 automatically duplicated). It helps writing simulators. Will allow
430 to run Splay programs within SimGrid in the future.
431 * Add a Chord example in Lua, equivalent to the MSG one.
434 * Start the implementation of a solution to express temporal
435 properties, not only local assertions. This is still an
436 experimental work in progress, stay clear from it to be safe.
440 - Add new runtime parameters --help-logs and --help-log-categories
441 to display informations about supported logging parameters and
443 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
444 don't exists anymore.
445 * Mallocators: allow value NULL for the reset function.
447 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
448 create homogeneous dictionaries, where all the elements share the
449 same free function. Non homogeneous dictionaries will be
450 deprecated in the next release.
451 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
452 - Multi-level dictionaries are deprecated.
454 - new function xbt_dynar_search_or_negative() that is useful when
455 you have less than 2 million elements in your dynar and don't
456 want of the extra complexity of catching exceptions when the
457 element is not found.
459 - Make xbt_os_thread module (for thread portability) public.
460 Documentation is still to come, sorry.
462 - Cleanups and simplifications to make it maintainable again.
463 - Exotic features (such as memalign and valloc) were removed.
464 - The metadata were extended and improved so that the
465 model-checker becomes able to explore and inspect the heaps.
466 - This may induce a performance drop when enable_model-checking is
467 ON in cmake (even if it's not used in the simulation), but it is
468 necessary at this point to get MC working.
470 Turn model-checking OFF if simulation performance matters to you.
471 Not enabling it at runtime is not enough, disable it in cmake.
473 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
475 SimGrid (3.6.2) stable; urgency=low
477 The "Not coding new stuff allows to polish old things" release.
480 * New bindings to the NS3 packet level simulator (experimental)
481 * Use the raw (efficient) execution contexts instead of the sysv
482 (portable) ones when possible.
483 * libpcre is now mandatory in any cases since not using it led to
484 severe performance loss and possibly other issues
485 * Update the XML platforms:
486 - G5K: include the latest machine in Nancy
487 - GridPP and LCG: new platforms
488 * Documentation was partially updated, at least (more to come)
490 Bug fixes, cosmetics and small improvements
491 * Free terminated processes before the end of the simulation to avoid
492 exhausting the memory of users having very dynamic amount of
494 * Bug fix and cosmetics about canceling non-running tasks
495 * Bug fix about the dot loader's issues when using libcgraph
498 * Create an installer for windows with nsis (amd64 and win32)
499 - Add an hello world project to illustrate simgrid project creation.
500 - Embed libpcre into the Simgrid installer to avoid
501 its compilation burden
502 * The raw execution contexts should work on Apple now
503 * Port to Windows 64 bits
504 - Sysv contexts now have an implementation for this arch
505 - GRAS communication features now support this arch
506 * Drop support for borland compiler on windows
507 - this code was not maintained, and we kinda depend on gcc nowadays
508 * Fix portability issues on kfreebsd/gnu: build error about semaphores
509 * Fix portability issue on unstable ubuntu: linker became picky on
512 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
514 SimGrid (3.6.1) stable; urgency=low
516 The "Oops, we broke Macs too" release
519 * Fixed contexts detection so that raw ones are used when possible
520 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
521 leads to a strange error, with user code segfaulting sometimes when
522 the generated code is not perfectly aligned (which is not
523 controllable from the user side, depends on the amount of code)
526 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
528 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
530 SimGrid (3.6) stable; urgency=medium
532 The Summer Release, also known as the "OMG! They Killed Kenny!" version
535 * Bindings now constitute their own package, separated from the main one.
536 Rationale: reduce our maintenance nightmare by reducing the module coupling
537 They will soon be released on their own on gforge.
539 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
540 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
542 GRAS: It is not considered as stable anymore, but experimental. Sorry.
543 * It's not quite deprecated for now because we have no replacement,
544 but it may soon become the case.
547 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
548 * Fortran: New user-level cache variable to store the rank of the running
549 process. This improves performance by an order of magnitude.
550 * C: New coccinelle script to automatically locate and modify global and
551 local static variables.
552 * Improved SMPI network model with a sender-side gap to account for multiple
556 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
557 only say if a communication is finished, no matter whether it succeeded or
558 failed. You can call MSG_comm_get_status() to know the status of a finished
560 * New function MSG_task_dsend() to send a task and detach it. When a
561 communication is detached, you are never notified of its success or failure
562 and the memory is released automatically once it is finished. This function
563 is useful when you don't care about the end nor the success of a
565 * Change the prototypes of action replay. Sorry for inconvenience,
566 but this is really more efficient this way (and to adapt your code,
567 you just have to fix the initialization, that shouldn't be too long)
568 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
569 ever managed to do anything useful with it.
570 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
571 takes a dynar of communications. It returns immediately and gives the
572 index of a finished communication (if any).
573 * New example: a basic implementation of the Chord P2P algorithm.
576 * New model for multi-core CPUs. You can now use the core attribute to
577 precise the number of cores of a host. This is a basic model. Every
578 process running on the host receives at most the power provided in
579 the DTD (throughput<=power). Total throughput of process cannot exceed
581 * New peer tag. This peer tag creates a tiny AS comprising a host and a
582 router linked by an up-link and a down-link (possibly asymmetrical).
583 This kind of pattern allows to easily build last-mile model style platforms.
584 Aggregating such patterns in a rule-based AS is thus the technique of
585 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
586 * New model for Vivaldi routing. We transformed the Vivaldi network model
587 into a Vivaldi routing model (based on the rule-based model). This allows to
588 combine Vivaldi based latencies with last-mile platforms.
591 * Added a check for NaN of IEEE754 infinite in the double entries of
593 * Introduce a new context factory "raw", highly inspirated from the
594 ucontext factory, but using manually crafted functions in assembly to
595 do the work in an efficient manner.
596 * Allow to change the used context factory at run time, not only at
597 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
598 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
599 contexts at runtime (only with raw contexts or ucontexts).
600 * Completely rewrote this module to allow parallel execution of user
601 processes. Use --cfg=contexts/nthreads:N to execute user processes
602 with N parallel threads (the default is 1, meaning no parallelism).
603 * Allow to decide dynamically between sequential and parallel modes.
604 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
605 processes in parallel only when their number is greater than or equal to P
607 * Added a check for NaN of IEEE754 infinite in the double entries of
611 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
612 won't output the state of processes when interrupted with Ctrl-C
613 * Add a new function xbt_dynar_to_array that transforms a dynar into a
614 NULL-terminated array. This may solve backward compatibility issues
615 due to the change to return type of SD_simulate. See also:
616 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
617 * Add new macros with variable number of arguments.
618 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
619 - in xbt/asserts.h: xbt_assert
620 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
621 - in xbt/ex.h: THROWF and RETHROWF.
622 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
624 * Change xbt_die() to accept a format string with arguments, just like printf.
625 * New data structure: xbt_lib_t, like a dict but more general and with better
629 * New configuration options
630 Options triva/categorized and triva/uncategorized can be used to generate
631 graph configuration files for Triva visualization tool.
632 * Configuration option tracing/platform is renamed to tracing/categorized
633 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
634 * New TRACE_user_link_variable interface:
635 User provides the name of the link and the tracing variable to attach to it
636 * the declaration of tracing categories must be done after the environment
638 * simpler tracing interface, just one way to declare categories
639 TRACE_category or TRACE_category_with_color, it is up to you
640 * links in the trace file are again identified by their names
641 * trace contains the full platform hierarchy exactly as declared using the ASes
642 * Options tracing/msg/[task|process]:1 groups the process by hosts
643 for both cases, tasks and processes must have names that are unique during the simulation
644 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
645 * The experimental option tracing/msg/volume is deprecated
646 its functionality may be reincorporated if needed
648 The tracing generates a trace file with unordered timestamped events,
649 because of the way the core simulator (surf) works. A script available
650 at the tools directory (fix-paje-trace.sh) can be used to put the events
651 in order. We have changed the tracing so it can generate ordered timestamped
652 events in the final trace, but depending on the simulator (and how much time
653 is simulated) that can lead to a huge memory utilization. It is deactivated
654 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
657 * Define a SIMGRID_VERSION macro in simgrid_config.h.
658 - We are trying hard to keep the API stable, but it may happen that
659 some things change (we're a research project after all, not a
660 nuclear plant operating system). If such things should happen, you
661 could rely on that macro to adapt.
662 - current value: 30600 for 3.06.00, aka 3.6
663 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
664 a number that can be compared with SIMGRID_VERSION.
665 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
666 assertions and verbose/debug logging events are disabled at compile time.
668 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
670 SimGrid (3.5) stable; urgency=medium
673 * New feature to any SimGrid-based simulator: Model-Checking
674 Check SIN#1 for more details.
677 * New Model SMPI (three-interval linear regression for correction factors)
678 See RR-7426, available at http://hal.inria.fr/inria-00527150
679 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
680 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
681 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
682 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
683 * See also src/smpi/README
688 - Tracing API changes: TRACE_start and TRACE_end should not be called
689 by user-code. They are automatically called by simulators created
690 with SimDAG, MSG and SMPI if the toolkit is compiled with
691 tracing_enabled=ON. Categories declaration and utilization remain the
692 same for MSG and SimDag.
693 - A function was added to the tracing API to declare categories with
695 - TRACE_category_with_color (char *category, char *color)
696 where color must be in the following format
697 "%f %f %f", red, green, blue
698 and red, green, blue are float values in the interval [0, 1]
699 - User can specify NULL as color parameter, or continue calling
701 On that case, the tracing system will define random colors
702 - The following command-line options are supported:
703 --cfg=tracing/filename:msg.trace
704 --cfg=tracing:1 (activate tracing, needed to use others)
705 --cfg=tracing/platform:1 (categorized resource use)
706 --cfg=tracing/uncategorized:1 (uncategorized resource use)
707 --cfg=tracing/msg/task:1 (task creation)
708 --cfg=tracing/msg/process:1 (process creation, migration)
709 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
710 --cfg=tracing/smpi:1 (SMPI interface tracing)
711 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
712 - examples of examples/msg/tracing updated
714 - DAXLoader and DOTLoader functions can generate tasks with categories
715 - A new function to attribute a category to SD tasks:
716 TRACE_sd_set_task_category (SD_task_t task, char *category)
717 * Tracing the MPI interface implemented by SMPI
718 - Collective operations are traced with states
719 - Point-to-Point operations are traced with states/links
720 - Tracing activated by a parameter "-trace filename" passed
721 to smpirun during execution (considering that simgrid
722 is compiled with tracing enabled)
723 - To run the simulation with gdb, the simulator
724 accepts --cfg=tracing/smpi:1 to trace SMPI
725 - tesh files to check if smpi tracing is ok
726 - See examples/smpi/NAS/DT-trace
727 * GTNetS tracing re-worked
728 - adaptation to the tracing system of GTNets to cope
729 with modifications regarding the fullduplex mode
730 - new tesh files to check if gtnets tracing is ok
733 * Asynchronous communications through the functions:
734 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
735 * New function: MSG_load_platform_script()
736 to make possible using a lua script instead of XML files to set up platforms
737 * New function: MSG_set_function
738 to associate functions to processes, used when bypassing the parser
739 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
741 Platforms: Add some more examples in examples/platforms
742 * Grid'5000: see www.grid5000.fr
743 * *_30000_hosts.xml: various huge files [mainly scalability testing]
746 * Change the XML format. This is a very important modification. SimGrid
747 3.5 introduces a new hierarchical format based on the notion of
748 Autonomous Systems. Compatibility with old format is ensured
749 through the perl script provided in the install bin directory
750 bin/simgrid_update_xml.
751 It is now possible to build platforms with specific routing
752 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
753 connect several platforms together. We will try to provide soon
754 set of realistic platforms exploiting these properties (have a
755 look at examples/platforms/ for the moment).
756 * Take the opportunity of the XML format change to be a good XML citizen:
757 rename link:ctn to link_ctn and similar changes (also dealt with by
759 * Add a new routing scheme (rule-based) using regular expressions. It
760 enables to have an extremely low memory footprint when the
761 underlying routing is simple and can be compactly described. You
762 need to have libpcre4-dev (perl regular expressions) installed if
763 you want to use this routing scheme.
764 * Revive the cluster TAG and allow to easily and efficiently (both in
765 term of memory and speed) connect clusters together. Have a look
766 at teshsuite/simdag/platforms/ to see how this can be done. With
767 this tag, you can create clusters with thousands of tasks at no
768 cost (have a look at examples/platforms/).
769 Note: clusters are implemented as ASes, so there is no need for an
770 enclosing AS tag if you have only one cluster in your platform.
771 * Add new generic functions in the public interface that allows the user
772 to call SURF 'create_resource' methods from your code (same
773 functionality as the XML bypass mechanism but with a much lighter
775 * Add a new model (enabled through command line --cfg=network/model:SMPI)
776 that uses a piecewise linear approximation to produce better
777 results when exchanging small messages.
778 * Add a new parameter to handle correctly full duplex link and account
779 for interferences between uplink and downlink communications
780 (activate with --cfg=fullduplex:1).
783 * Rename the SD_READY (all dependencies are satisfied and task is
784 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
785 dependencies are satisfied) state.
786 This prevents a confusion between the notion of "ready to schedule"
787 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
788 simulated" (SD_RUNNABLE) used by the simulation kernel.
789 * Change the way a task is considered as ready. Instead of removing
790 dependencies when a task is done, a counter is decreased. This way,
791 it is always possible to reach ancestors thanks to the
792 SD_taks_get_parents function (even after the end of the simulation.)
793 * Change the return type of SD_Simulate from (SD_task_t*) into
794 xbt_dynar_t. This function was in handling a dynar internally and
795 converted it into a NULL terminated array for historical reasons.
796 * New function SD_dotload(char*) to load a DAG described in dot
797 format. This loader and the corresponding examples require the
798 installation of the graphviz library.
799 * Fix a bug in the management of tasks of size 0 in the surf network
800 models. This problem was only visible with SIMDAG and you should
801 thus disregard results produced with earlier versions if you
802 relied on this feature (some tasks were blocked because of this).
803 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
804 even though your applications were doing only point-to-point
805 communications and sequential computations. Now you can really use any
806 model you want (of course, if you create real parallel tasks, which are
807 not implemented in most models beside ptaskL07, this will abort).
808 * Add an example that schedules a DAX on an heterogeneous platform
809 using a Min-Min strategy.
810 * New function SD_workstation_get_current_task() that returns the kind
811 of task currently running on a workstation in the sequential access
813 * Raise some warnings when unexecuted tasks remains at the end of the
814 simulation. This is usually caused by cycles in the DAG.
817 * New function: SIMIX_process_set_function() called by MSG_set_function
818 * Change the underlying waiting queue in semaphores so that a process
819 can wait on several of them simultaneously (as in waitany).
820 * Fix the way to handle tokens in semaphores so that all access patterns
821 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
822 * kill the dirty pimple SIMIX_message_sizes_output()
823 Please use (proper) visualization instead
826 * New data container: setset (set of sets of elements)
827 * New module: mmalloc (mapped malloc, allowing to have several
828 independent segments of malloc)
829 * New function: xbt_dict_cursor_set_data()
830 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
831 * New function: xbt_dynar_is_empty()
832 * New function: xbt_fifo_get_last_item()
833 * Fix xbt_dynar_shrink(): use the right element size.
834 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
835 uninitialized areas during expand.
836 * Fix semaphores: previous implementation was severely broken.
837 * Use library init/fini functions for our initialization.
838 - you can use logs and other feature as soon as you want in your
839 code (even before the xbt_init / MSG_init)
840 - xbt_exit is now a no-op and produce a warning when used.
843 * Port GRAS to new SIMIX mechanisms. This allows gras users to
844 benefit from the latest improvement to the simulation kernel.
845 * Kill measurement sockets for now. If you rely on them, sorry. This
846 release is not for you. This feature will be reintroduced in the
847 future, but we cannot delay the release any further.
848 * New function: gras_msgtype_get_name().
849 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
850 * Fix (at last) the pmm example: it should not randomly fail anymore.
852 Build chain: bug fixes and overall polishing
853 * Cmake is now stable enough. Hence, we killed the autotools.
854 * Port to windows ( TM :)
855 * Fix the 'make install' target.
856 No need to use 'make install-simgrid' anymore
857 * Introduce a 'make dist' target compiling a *source* archive
858 'make package' compiles a binary archive
859 * Compile java files only on need
860 * Add --cd and --setenv command line options to tesh
861 * Out of source builds are not fully supported yet, but we are close
862 * Enable supernovae and optimization flags by default for our users
865 * Add layer to set up environment directly from lua, without XML.
866 * The effect of gras_stub_generator can be achieved through
867 lua too (check examples/gras/console/ping_generator.lua)
869 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
871 SimGrid (3.4.1) stable; urgency=low
873 The "Polishing easter eggs is probably a good idea" release.
874 This is a bug fixes release only.
878 * Fix a bug preventing the tasks from begin garbage collected.
881 * Fix a bug occuring when a host involved in a communication fails.
882 This was not detected properly by the other peer involved in the
883 communication. Now, it's reported as a network error.
886 * Warn the user about loop dependencies in data flow of DAX files
887 * Obey the control-flow dependencies of DAX files
890 * Add option "enable_smpi" allowing to not compile SMPI.
891 Probably useful for the (Mac) users experiencing a build error here
892 * Improve the detection of lua5.1 and ruby1.8
894 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
896 SimGrid (3.4) stable; urgency=low
898 The "Easter in Cargese" release. Also known as (major changes):
900 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
901 ~> bindings were greatly improved
902 ~> new tracing infrastructure for better visualization introduced
904 * the "Welcome to configury modernity" release.
905 ~> we switched from autotools to cmake, and improved our cdash
908 A more detailled list of changes follow (full detail in svn log).
910 Java Bindings: Various Cleanups
911 * (install java-gcj-compat-dev on debian-like to use them)
912 * Remove put/get: no need to export deprecated interface in Java
913 Use send/receive instead.
914 * Cleanup the examples and add a README per directory
915 * Remove example autoDestination (that's the only way to go now)
916 * Remove example explicitDestination (was a plain copy of basic)
917 * Make JniException a runtime exception, so that there is no need to
918 declare the fact that you may encounter such a beast. I guess that
919 nobody will ever want to survive such error.
920 * Create specific errors for each MSG case of failure:
921 host failure, transfer failure, timeout, task cancelled
922 * Cleanup the exceptions that may get thrown by each function
923 * Other internal cleanups in Java bindings. Performance still bad :/
924 Ruby and Lua Bindings: create them
925 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
926 * That's new and great, you should try them out.
927 Same functionalities than Java bindings, only even less polished
929 * Kill the useless "rate" argument of SD_task_get_execution_time()
930 Everyone used to provide -1 as a value, it was not used, and the
931 semantic of a possible use wasn't even clear.
932 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
933 either as comm costs or compute costs to mean that there is no
934 such thing for that specific task.
935 * Add a SD_task_set_name() function
936 * Fix SD_task_unschedule() on typed tasks
937 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
938 * In DAX loader, accept useless 'level' attributes to <job> since
939 LIGO DAGs have them (seem to be to ease graphical representation).
941 * Add an example masterslave_mailbox.c using send/receive and not
942 the deprecated put/get interface.
943 * Kill the MSG_paje_output() function. It's a noop since 2 years.
944 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
946 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
947 (declare MSG_USE_DEPRECATED to still have the old name)
948 * Add a MSG_task_set_data() function
949 * About trace replay (see examples/msg/actions):
951 - Allow to work with splitted trace files for each process
952 Give the specific trace file as argument of each process,
953 and call MSG_action_trace_run(NULL)
954 You can still have one merged file for all processes.
955 - Fix implementation of collective operations
956 * Allow task_execute() on 0-sized tasks (closes #10063)
958 * This is the first release of SimGrid where SMPI is not considered
959 beta anymore (even if some corners should still be improved)
960 * Port over the new SIMIX_network submodule (internal refactoring)
961 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
962 * Implement more missing elements of the standard:
964 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
966 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
967 MPI_Scan() MPI_Get_processor_name()
968 - Added implementation of missing case for Alltoall (warning: it's
969 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
970 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
971 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
972 SMPI_MPI_Allgatherv()
974 - MPI_Waitsome() was broken
975 - Allow relative includes in smpicc
976 - Command line cfg argument 'reference_speed' was ignored...
977 - Some functions did not properly lead to auto-benching of user code
978 - smpicc passes -O2 by default (just like openmpi one)
980 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
981 * Bug fixes about timeouts during communications
982 * add SIMIX_message_sizes_output() as a pimple to write to file the
983 amount of messages per size. Use gnuplot to get histogram.
984 Pimple because that's the only user-visible function of simix,
985 defined directly in xbt.h (irk, sorry)
987 - Add a SIMIX_sem_get_capacity() function
988 - Fix interactions with processe resume/suspende
989 - release_forever() was stupidly broken
990 - Fix SIMIX_display_process_status() for processes in a semaphore
991 - Make SIMIX_sem_block_onto() user-visible
992 * Refactoring context stuff:
993 - Use pseudo-OOP for better modularity
994 - reimplement SIMIX_process_kill() without process_schedule() so
995 that the latter can take as invariant that it is called from
997 - Merge context_start into context_new for sake of simplicity
999 * Add a Vivaldi network model, coded live during SUD'10 ;)
1000 * Rename configuration variables to start a hierarchy:
1001 o cpu_model -> cpu/model
1002 o network_model -> network/model
1003 o workstation_model -> workstation/model
1004 * New configuration variables:
1005 o network/bandwidth_factor: correction to bandwith
1006 o network/latency_factor: correction to latency
1007 o netwotk/weight_S: correction to the weight of competing streams
1008 * Add a long description to the models, that users can see with such
1009 argument on the command line: --cfg=cpu/model:help
1010 * --help-models display the long description of all known models
1012 * config: add the ability to set a default value after registration
1013 Does not override any previously set value (e.g. from cmd line)
1014 * dict: allow to have integer key and data.
1015 When so, you need to use the following functions
1016 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1017 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1018 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1019 In contrary to regular dicts, the key is not malloced before copy.
1020 Mixing scalar and regular elements in the same dict is not tested
1022 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1023 Tracing for Visualization:
1024 * SimGrid is now instrumented in order to generate a trace file for
1025 visualization analysis: to use it, need to compile SimGrid with the
1026 "tracing" option enabled, and instrument the program using SimGrid with
1027 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1028 (among other functions).
1029 * The instrumentation only traces the platform utilization for now
1030 * Documentation to use the tracing functions and how to analyze the
1031 traces with the Triva tool is written.
1032 * More information about: SimGrid FAQ (in the section Tracing Simulations
1035 * We moved to cmake as default build system. Autotools support will
1036 be dropped soon. Check the FAQ for more info about how to use it.
1037 * Greatly improved our cdash/ctest interactions
1038 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1039 * Added memory checking tests with valgrind; lot of memleak fixing.
1040 This may be the first release of simgrid with so few memory issues
1041 * Added code coverage tests.
1042 Our coverage is still improvable, but at least we see it on cdash.
1044 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1046 SimGrid (3.3.4) stable; urgency=low
1048 The "Desktop Grid needs love too" release (also called Xmas release).
1050 Models improvements:
1051 * Major speedup in the maxmin system solving by using lazy evaluation
1052 Instead of solving completely the maxmin system at each iteration,
1053 only invalidate (and recompute) the modified parts.
1054 This new feature is enabled in default models but you can try to
1055 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1056 * Cas01 IMproved as default CPU model
1057 This CPU model is the same Cas01 model, but it uses the
1058 maxmin-selective-update flag and a heap structure to manage
1059 actions on SURF kernel.
1060 It reduces the complexity to find the next action to finish and,
1061 consequently, it's faster than the old Cas01.
1062 This is the new default CPU model (Cas01).
1063 * Rename the old Cas01 model to Cas01_fullupdate
1064 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1065 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1066 * CpuTI (CPU Trace Integration)
1067 A new CPU model whose objective is simulate faster when using
1068 availability trace files.
1069 Instead of using a full featured, over engineered maxmin system for
1070 CPU modeling, this model does the pre-integration of traces files
1071 to calculate the amount of CPU power available, and so, executes
1072 faster than the old CPU models.
1073 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1074 * Use LV08 as default network model since it gives better accuracy
1075 for small messages and shouldn't change things for big ones.
1076 Use --cfg=network_model:CM02 to get the previous behavior.
1079 ******************************************
1080 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1081 ******************************************
1082 * The new CPU model may changes simulations!
1083 The point is that events occurring at the exact same timestamp
1084 are not scheduled in the same order with the old and new
1085 version. This may be enough to completely change the execution
1086 of simulations in some cases.
1087 * The new network model will change simulations!
1088 This new model is more realistic than the previous one, so you
1089 should consider redoing your old experiments with this model.
1090 Sorry for the inconvenience.
1093 * Introduce the supernovae compilation mode
1094 When compiled that way, the whole SimGrid (or almost) is put in a
1095 single compilation unit and compiled in one shoot.
1096 This is to help gcc which has difficulties to inline stuff from one
1098 The speedup seem to be above 15%, althrough more tests are needed on
1099 amd64 to confirm that gain.
1102 * Port of MSG's mailbox on top of SIMIX network
1103 The put/get mechanism was greatly simplified on the way.
1106 * New SIMIX network module. Provides:
1107 - Mailbox: rendez-vous mecanism to find with who you want to speak
1108 - Synchronous send/recv: easier and hopefully faster since the
1109 logic is handled in the maestro process directly now
1110 - Asynchronous send/recv: you dreamt of it? It's here now
1111 Too bad that nobody cared enough to propagate the change to MSG.
1112 * Add semaphores as SIMIX synchronization mechanism.
1115 * new function SD_daxload(char*) to load a DAX file
1116 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1117 * Introduce typed tasks. Specify its kind and cost at creation.
1118 At scheduling, just give where it should be placed, and the cost
1119 for each involved resource is automatically computed.
1120 Existing constructors so far (more to come of course):
1121 - SD_task_create_comm_e2e() for end-to-end communication
1122 - SD_task_create_comp_seq() for sequential computation
1123 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1124 * new function SD_task_dump() for debuging display
1125 * new function SD_task_dotty(task,FILE*) writing to file the info
1126 about the task in dotty format
1127 * SD_task_dependency_exists() can now cope with having one of its
1128 arguments NULL. If so, it tests whether the other argument has any
1130 * Add getters on list of preceding/following tasks:
1131 SD_task_get_parents(task) and SD_task_get_children(task)
1132 * Add getters on amount of workstations and list:
1133 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1134 * Add getter on task kind: SD_task_get_kind(task)
1135 * Update the start_time and finish_time of tasks on completion/failure
1136 * Bugfix: Remove task from state swags when destroyed
1139 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1142 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1143 dynar with one separate thread per value of the dynar.
1144 * Change the prototype of xbt_thread_create(), sorry.
1145 Added a boolean parameter indicating whether we want to join this
1146 thread (used in SG only for now)
1147 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1150 * GTNetS wrappers should now be usable again (and betterly tested too)
1151 * Fix a major regression from 3.2 where the timeout provided to
1152 MSG_task_put_with_timeout() was used as absolute time before which
1153 the comm should be done.
1154 * Start to fix the <cluster> tag.
1155 - Internal links should be good now (beside of the loopback, which
1156 use the private link instead)
1157 - paths to the external world is still rather broken
1158 - the <route:multi> tag is just broken. Actually that's brain-dead.
1159 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1160 to make it less stupid
1161 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1162 * Fix a source-level compatibility glitch from 3.2: after defining
1163 MSG_USE_DEPRECATED, you can use the old name
1164 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1165 * Allow to compile from the SVN with automake 1.11
1166 * Fix some problems when using the "start_time" tag in deployment XMLs.
1167 * Fix #8569: XBT/synchro.h has redundant declarations
1168 * Fix #8563: MSG return values and exceptions
1169 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1170 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1171 Thanks to Robson Peixoto.
1172 * Fix a vicious bug in dictionaries inducing that some elements were
1173 not freed on xbt_dict_free()
1175 Portability report of this version:
1176 * Main portability targets:
1177 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1178 on (amd64/i386/ia64)
1179 - mac leopard on i386
1180 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1181 but nothing critical.
1182 * Other platforms: windows, AIX and others were not tested for this release
1184 Timing report of this version:
1185 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1186 scenario parameters). From 8h to a few seconds in desktop grid settings.
1187 * Supernovae brings about 25% speedup on i386.
1189 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1191 SimGrid (3.3.3) stable; urgency=low
1193 The "Need for Speed" release.
1195 The timings done to validate the 3.3.2 were faulty.
1196 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1198 The problem was a conversion from a manually handled vector to
1199 xbt_dynar_t on the critical path.
1200 xbt_dynar_foreach calls functions, inducing stack management crap.
1202 We inlined these functions and xbt_dynar_foreach is now breath taking.
1203 We also inlined xbt_swag_belong on the way.
1205 Here are some approximate speedup measurements (on master/slaves
1206 simulations lasting between 10s and 20s each):
1207 3.3.1 -> 3.3.2: about same performance
1208 3.3.2 -> 3.3.3: 40% speedup
1209 3.3.1 -> 3.3.3: 40% speedup
1210 3.3.1 with inline patch -> 3.3.3: 30% speedup
1212 Our reading is that the refactoring which occurred in 3.3.2 made us
1213 suffer much more from the xbt_dynar_foreach low performance, but
1214 once we solved this, this refactoring proved to be very performance
1215 effective. From the 40% speedup, somehow, 10% are due to the
1216 inlining and 30% to the refactoring.
1218 That's a pitty that gcc cannot inline functions placed in other files
1219 alone. We have to choose between:
1220 - break the encapsulation (by putting private data structures and
1221 accessors in headers files to help gcc)
1222 - live with low performance
1223 - switch to a decent compiler such as icc (not quite possible).
1225 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1227 SimGrid (3.3.2) stable; urgency=low
1229 The "Simplicity does not preceed complexity, but follows it" release.
1231 The main contributors of this release were (lexical order):
1232 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1235 * Extract the routing logic into its own object.
1236 (was dupplicated in network.c and workstation_LV07.c;
1237 Allows to implement other ways of storing that info)
1238 => kill now useless network_card concept
1239 - Use dynar to represent routes (instead of void** + int*)
1240 - kill link_set (use surf_network_model->resource_set instead)
1241 - Add a command-line option to choose the routing schema to use
1242 - Add three new models:
1243 * Floyd (shortest path computed at initialization)
1244 * Dijikstra (shortest path recomputed all the time)
1245 * Cached Dijikstra (shortest path computed on need)
1246 All these models where contributed by Silas De Munck, and are
1247 described in his ICCS09 paper.
1249 * Simplify model declaration
1250 (less redirections, less function to write when defining a model)
1251 - Factorize stuff between models:
1254 surf_model_resource_set(model)
1255 surf_model_resource_by_name(model, name)
1256 - Unify the types of models in s_surf_model_t (using an union)
1257 - Embeed fields of common_public directly into s_surf_model_t
1258 - Rename model methods:
1259 action_free ~> action_unref
1260 action_change_state ~> action_state_set
1261 action_get_state ~> action_state_get
1262 - Change model methods into functions :
1263 (model)->common_public->action_use ~> surf_action_ref
1265 * Implement a generic resource; use it as ancestor to specific ones
1266 (allows to kill duplicated code in models)
1267 Drawback: timer command don't need no name nor properties;
1268 workstation_CLM03 don't need no properties
1269 (but I guess we can live with those few bytes wasted)
1271 * Improve the action object model
1272 - implement a constructor avoiding dupplicated code about field
1273 initialization in generic_action part.
1275 * Kill the SDP model: it has an external dependency, is deprecated
1276 in flavor of modern lmm models, and didn't compile since a while
1279 * Relocation of the context module from XBT to SIMIX.
1280 (the context were decoupled from the simix processes, duplicating a lot of code)
1281 => a lot of code was factorized
1282 - less overhead is introduced during scheduling
1283 - simpler API for the context factory
1284 - the logic for process creation,destruction and manipulation was simplified
1285 * Simplification of the s_smx_process_t data structure.
1286 => accesing the simix level data associated to a process is faster now,
1287 and the code is a lot more readable.
1290 * Implement some more MPI primitives:
1291 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1292 -implementation: Bcast: flat or 2-ary tree (default),
1293 Barrier: 4-ary tree,
1295 Allreduce: Reduce then Bcast
1296 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1297 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1298 Alltoallv: flat tree, like ompi
1300 * Add support for optimized collectives (Bcast is now binomial by default)
1301 * Port smpirun and smpicc to OS X
1304 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1305 Shout out if you used it.
1308 * Display the list of still queued messages in SG mode when existing
1312 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1313 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1314 * Add xbt_str_from_file(FILE*)
1315 * Add xbt_dict_get_key achieving a linear reverse search
1316 * Remove the context module
1318 Portability report of this version:
1319 * Main portability targets:
1320 - Linux(debian)/x86/context
1321 - Linux(debian)/x86/pthreads
1322 - Linux(debian)/amd64/context
1323 - Linux(debian)/amd64/pthreads
1324 On these, we still have the eratic breakages of gras/pmm and
1325 amok/saturate_sg reported in previous version. We still think
1326 that the tests are the cause of the fault, not the tested code.
1328 - Mac OSX Leopard/x86/context
1329 Still false negative in tesh autotesting.
1330 Smpi still fails, but this time because readlink does not accept -f
1331 Everything seems to work properly beside of that.
1334 - AIX version 5.3 (only tested contexts this time)
1335 Smpi still fails there because mktemp is not installed.
1336 Everything seems to work properly beside of that.
1338 I managed to compile it for the first time, but several breakages.
1339 Won't delay the release for this exotic platform.
1341 * Windows: it's still lagging behind. If you want to help, please
1344 Timing report of this version:
1345 This version seem to be more than 5% faster than 3.3.1 (on linux
1346 64bits with contextes). The gain is less than expected, we are
1347 investigating this for next release.
1349 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1351 SimGrid (3.3.1) stable; urgency=low
1354 * Implement a --cfg-help to show existing configuration variables
1355 * Build chain do not require doxygen in maintainer mode
1358 * fix a bug on struct sizeof computation, which prevented the
1359 exchange of arrays of structs in some conditions
1360 - added a regression test about this in datadesc_usage
1361 * Allow the exchange of 0-long dynamic vectors.
1362 - for that, use -1 as indicator of dynamic size instead of 0
1363 - This implied to change any size from unsigned long to long,
1364 reducing a bit communication abilities, but I guess that with
1365 64bits being quite common, this is more than enough.
1366 - This also induce a protocol change, thus bumping network protocol
1367 version from 0 to 1 (if we have external users, we have to get
1368 clean on that point too ;)
1369 - added two regression tests about this in datadesc_usage
1370 * Be more verbose when propagating local exceptions
1371 This helps debugging.
1372 * Display the status of simulated processes when receiving SIGINT in
1376 * Allow to control the simulation from a trace file.
1377 New functions MSG_action_register() and MSG_action_trace_run()
1378 The first one allows to associate a function execution to each
1379 kind of action while the second one parses a trace file and
1380 triggers the corresponding actions within the system.
1381 For now, only a toy example is provided in examples/msg/actions
1382 * Add an exemple of process migration in examples/msg/migration
1383 * Fix a bug in task exchange which broke MSG_task_get_sender()
1384 Add a teshsuite regression test for that.
1385 [Bug: if MSG_task_get_sender() is called after sender exit,
1387 * Fix a bug which prevented suspend/resume to work properly
1388 * Display the status of simulated processes when receiving SIGINT
1389 This fixes a regression of v3.3. due to the introduction of SIMIX
1390 * Bug fixing in failure management:
1391 - trace could not start by a failure at time 0
1392 - failure during communications were not working
1395 * Add SIMIX_process_set_name() to change the name of the current
1396 process in the log messages.
1397 * Store smx_hosts in a dict since we only retrieve them by name
1398 * Move the configuration infrastructure to surf
1401 * Move the configuration infrastructure to surf
1404 * Massive internal cleanups:
1405 - Store internal structures on processes instead of hosts (allows
1406 to have more than one process per host, in addition of being more
1408 - Cleanup the initialization/finalization process
1409 - Kill a whole bunch of unneeded synchronization:
1410 processes run in exclusive manner within the simulator
1411 - Move queues from global tables to process data fields
1413 - now accept -platform and -hostfile arguments
1414 - Pass the right rank value to processes according to the hostfile
1415 * Compile the examples by default, and use them as regression tests
1416 * Implement MPI_Wtime()
1417 * Change the reference speed to a command line option
1420 * TCP_gamma can now be specified as command line option using
1421 --cfg=TCP_gamma:10000000.0
1422 * Change the --surf-path cmd line option into --cfg=path:
1425 * Also include strbuff from xbt.h public header
1426 * xbt_ex_display(): do not free the exception after displaying
1427 This allows to do more with the given exception afterward.
1428 Users should call xbt_ex_free() themselves.
1432 Portability report of this version:
1433 * Main portability targets:
1434 - Linux(debian)/x86/context
1435 - Linux(debian)/x86/pthreads
1436 - Linux(debian)/amd64/context
1437 - Linux(debian)/amd64/pthreads
1438 These targets fail about 1/10 of times on gras/pmm, but we believe
1439 that this is because of the test, not because of simgrid.
1440 amok/saturate_sg fails even more rarely, and the test may not be
1443 - Mac OSX Leopard/x86/context
1444 The test suite still spits tons of errors because some obscure
1445 force prevents us from removing the temporary directories
1446 arguing that they still contain some metadata I've never heard of.
1447 Smpi fails because seq is not installed.
1448 Everything seems to work properly beside of that.
1451 - AIX version 5.3 (both contexts and pthread)
1452 Smpi still fails there because mktemp is not installed.
1453 XML inclusions seems rosty on AIX.
1455 * Windows: it's still lagging behind. If you want to help, please
1458 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1460 SimGrid (3.3) stable; urgency=high
1464 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1465 [Malek Cherier & Mt]
1467 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1470 Until now, GRAS were using MSG as an interface to SURF. It was
1471 quite difficult because both interface have several differences
1472 (MSG channels vs GRAS sockets were the most notable point).
1474 This also opens the gate to SMPI (which should occur soon) and speed
1475 up simulations by to 40% (even if it were not the main goal).
1477 **************************************
1478 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1479 **************************************
1480 The point is that events occuring at the exact same timestamp are
1481 not scheduled in the same order with the old and new version. This
1482 may be enough to completely change the execution of simulations in
1483 some cases. Sorry for the inconvenience.
1485 * Cleanup and upgrade the XML format to push further scalability
1486 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1488 * Improve the testing infrastructure with tesh. Now a very large part of
1489 the code is tested not only by being run but also by checking that the
1490 output match an expected output [Mt].
1492 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1493 is really less memory-demanding, which should allow you to use
1494 larger files in SimGrid [AL].
1496 * Inform valgrind about our contextes, so that it becomes usable
1497 with the default (and more effecient) version of SimGrid
1498 [contributed by Sékou Diakite, many thanks]
1501 * Introduce a listener thread in charge of receiving incoming
1502 messages from the network. It allows to overlap communication and
1503 computation but most notably, it removes some stupid deadlocks due
1504 to the fact that so far, a process could not send and receive at
1505 the same time. This made most non trivial communication schema
1507 * Convert the PIDs from long int to int to match the MSG ones (and
1508 linux ones too) [Mt]
1509 * New function: gras_agent_spawn() to launch a new process on
1510 current host. Only working in simulation for now. [Mt]
1511 * New function: gras_os_hostport() returning a constant form (ie,
1512 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1515 * Make the backtrace of exceptions more human readable [Mt]
1516 * New module: xbt/str [Mt]
1517 a ton of string utility functions (split, join, printf to a newly
1518 allocated buffer, trim, etc)
1519 * New module: xbt/hash [Mt]
1520 SHA1 hashing algorithm (more to come if needed)
1521 * New module: xbt/synchro [Mt]
1522 synchronization tools (mutex and conditions) working the same way
1523 in simulation and in real life (mainly useful for GRAS, but not
1525 * New module: xbt/queue [Mt]
1526 classical producer/consumer synchronization scheme
1527 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1528 (using the classical functions will get serialized) [Mt]
1529 * Make dictionary internal table dynamic. No need to specify its size
1530 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1532 * Make sure the log channels are organized as a tree under windows
1533 (because of ANSI C compatibility issue, any channel were child of
1534 root directly) [Mt].
1537 * Cleaned many thing in surf and fixed a few bugs [AL].
1538 * Add a nice command line configuration mechanism to compose models [AL].
1539 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1540 the previous one (KCCFLN05). It relies on something that looks like
1541 a max-min sharing mechanism but cannot be written as such. A new solver
1542 was thus designed [AL].
1543 * Add a new solver to lmm. Based on Lagrange optimization and
1544 gradient-based descent, it enables to efficiently maximise systems s.a
1546 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1548 This solver enables to propose two new network models for TCP Reno and
1549 TCP Vegas based on Low's work. These models still need to be fully
1550 tested though [Pedro Velho].
1553 * Bug fix in SD_simulate. Now the time bound given as argument is
1555 * Use the new parallel task model (ptask_L07) as default.
1556 * Use the SURF command line configuration mechanism.
1557 * 0-size tasks (for synchronization) should now work.
1559 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1561 SimGrid (3.2) stable; urgency=high
1565 We still experience issues on this platform, but we believe that at
1566 least MSG is usable.
1568 GRAS API BREAKAGE (for simplification purpose, sorry):
1569 * the gras_msgtype_by_name is not used anymore. Instead of
1570 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1571 you can write (and must)
1572 gras_msg_send(toserver, "request", &request);
1573 - If you still want to pass a gras_msgtype_t to the function (to cache
1574 the type and avoid the lookup time), use the gras_msg_send_() variant.
1575 - Impacted functions:
1576 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1577 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1578 * The callbacks are now expected to return 0 when everything went well
1579 (just like the main() function)
1581 GRAS new features and improvements:
1582 * New module mecanism where user code can use per process globals [Mt]
1583 This is similar to gras_userdata_*() functions, but for libraries. It
1584 factorize some code developped over and over in the examples and AMOK.
1585 It has still to be documented and used (only amok/peermanagement is
1587 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1588 loss when they were small enough to fit into the buffer and sent quickly
1589 enough so that they can all get received in one shoot.
1590 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1591 if not found. Use the *_or_null() variant for the old semantic.
1592 * In gras_msg_handle, do not discard messages without callback.
1593 They are probably messages to be explicitly awaited later (ie, proofs of
1594 mis-synchronization in userland since they are sent before being awaited)
1595 No big deal usually.
1596 * gras_socket_meas_send/recv: semantic changed!
1597 The numerical arguments used to be (1) the total amount of data to send
1598 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1599 messages. This was need for the fool willing to send more than MAXINT
1600 bytes on quite fat pipes.
1603 * Do really rename the hostmanagement module to peermanagement. [Mt]
1604 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1605 but this is rather new and this was documented in the module
1606 documentation (poor excuses, I admit)
1607 * Bandwidth measurement semantic changed! This follows the changes to
1608 gras_socket_meas_send/recv explained above.
1611 * A sequential mode has been added to the workstations. When a workstation
1612 is in sequential mode, it can execute only one task, and the other tasks
1613 are waiting in a FIFO. [Christophe Thiery]
1616 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1617 model for SIMDAG. [Christophe Thiery]
1618 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1619 equal to 0, which caused some bad behaviors in
1620 saturated_constraint_set_update. I now use a threshold mechanism like in
1624 * When running manually src/testall, you select specific units [Mt]
1625 testall is the result of our cunit mecanism, and should replace all
1626 the scripty thingy around since bash don't run easily on billware.
1628 * A mallocator system has been added. [Christophe Thiery]
1629 Mallocators allow you to recycle your unused objects instead of freeing them
1630 and allocating new ones.
1632 Documentation update:
1633 * FAQ reworking + New FAQs:
1634 - "Valgrind spits tons of errors!" [Mt]
1635 - "How to repport bugs" [Mt]
1636 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1637 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1638 - Communication time measurement within MSG [AL]
1639 - I experience weird communication times when I change the latency [AL]
1640 * GRAS tutorial [Mt]
1642 - an introduction to the framework and to the used communication model
1643 - an initiatic tour introducing the most proheminent features:
1645 . Lesson 0: Installing GRAS
1646 . Lesson 1: Setting up your own project
1647 o Part 2: Message passing
1648 . Lesson 2: Exchanging simple messages
1649 . Lesson 3: Passing arguments to the processes (in SG)
1650 . Lesson 4: Attaching callbacks to messages
1651 . Lesson 5: Using globals in processes
1652 . Lesson 6: Logging informations properly
1653 . Lesson 7: Using internal timers
1654 . Lesson 8: Handling errors through exceptions
1655 . Lesson 9: Exchanging simple data
1656 . Lesson 10: Remote Procedure Calling (RPC)
1657 . Lesson 11: Explicitely waiting for messages
1658 . Recapping of message passing features in GRAS
1659 - A HOWTO section containing:
1660 o HOWTO design a GRAS application
1661 More are due, of course. They will come latter. In the meanwhile, you can
1662 check the examples which are still here.
1664 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1666 SimGrid (3.1) stable; urgency=high
1670 There was a stack corruption somewhere, visible only when optimizing
1671 with these versions. [Vince]
1674 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1675 module that enabled to play with Directed Acyclic Graphs. It is built
1676 directly on top of SURF and provides an API rather close to the old
1677 SG. Some old codes using SG are currently under rewrite to check that
1678 all needful functions are provided. [Christophe Thiery]
1681 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1682 extension of the classical CLM03 model that gracefully handles
1683 failures. This is now the default model for MSG and GRAS. It doesn't
1684 handle parallel tasks yet though. [AL]
1685 * Bug fix: Weights were not correctly set in the network part.
1686 WARNING: This may have resulted in incorrect results with simulations
1687 where there are more than one flow on a given link. [AL]
1690 * After a (long ?) discussion on simgrid-devel, we have decided that the
1691 convention we had on units was stupid. That is why it has been decided
1692 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1693 WARNING : This means that all previous platform files will not work as
1694 such with this version! A warning is issued to ask users to update
1696 A conversion script can be found in the contrib module of the CVS, under
1697 the name contrib/platform_generation/surfxml_update.pl [MQ]
1700 * Bug fix: Processes were started in reverse order, wrt deployment file.
1701 WARNING: if your code relies on this bug, please fix it. [AL]
1702 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1703 being executed on two different locations. [AL]
1704 * Bug fix: Failures are now better supported thanks to Derrick's tests
1705 (there was many failure situations I hadn't thought of and that weren't
1706 correctly handled). [AL]
1707 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1711 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1712 Exception occurring on server-side are propagated back to client (!).
1714 API CHANGE: the callback changed their prototype. Change:
1715 int my_handler(gras_socket_t expeditor, void *payload_data) {
1717 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1718 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1720 * New! function: gras_msg_handleall to deal with all messages arriving
1721 within a given period.
1722 * New! function: gras_socket_server_range to get a server socket in a
1723 range of port numbers (ease to avoid port number conflicts) [MQ]
1724 * New! gras processes display their backtrace when they get a SIGUSR1
1725 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1726 Sweet to debug RL processes [MQ]
1730 - Do not force experiment sizes to be expressed in kb, or it becomes
1731 impossible to measure the latency this way (needs one byte-long tests)
1732 WARNING: this changes the amok_bw_* function semantic. [MQ]
1733 - Implements the link saturation stuff. [MQ]
1734 * Peer management module:
1735 New! module factorizing code that we wrote over and over [MQ].
1738 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1739 - Test units are placed directly into the library code, they get extracted
1740 automatically and placed into the src/testall binary.
1741 - Convert most of the XBT tests to this system.
1742 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1743 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1744 live in the same C code. This required to change a little bit the API
1745 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1746 * New module: sparse graph structure with basic algorithms (this is work
1747 in progress and the API is not considered to be frozen yet). [AL]
1748 * Display more information on backtraces: source line & function names are
1749 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1750 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1751 * Reworked a little bit some #include statements to load only required
1752 headers. Some user code that relied on SimGrid to include stdlib or
1753 stdio may need to include it by themselves. [AL]
1754 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1756 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1757 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1758 but I guess that almost nobody used those parts. [MQ]
1760 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1762 SimGrid (3.0.1) stable; urgency=low
1765 * Unfortunately, I had missed 5 misnamed functions:
1766 xbt_fifo_item_t xbt_fifo_newitem(void);
1767 void xbt_fifo_freeitem(xbt_fifo_item_t);
1768 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1769 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1770 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1771 They're now deprecated. Please use their new versions:
1772 xbt_fifo_item_t xbt_fifo_new_item(void);
1773 void xbt_fifo_free_item(xbt_fifo_item_t);
1774 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1775 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1776 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1778 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1779 * Documentation: xbt_log module unmercifully reworked [MQ]
1780 * Bugfix: there was a problem with the ending of contexts with
1781 the pthread backend. It caused some weird deadlock or behavior
1782 depending on the pthread implementation. [AL]
1783 * Bugfix: get the exceptions raised in the simulator repport where
1784 and why they come from when they are not catched in time [AL, MQ]
1787 * Bugfix: Do repport the error when two non-connected hosts try to
1788 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1791 * Add additionnal checkings on communications. Assert that two
1792 communicating hosts are connected by a set of links... [AL]
1795 * Add additionnal checkings on channel values in communication [AL]
1796 * New: MSG_task_get_source to see on which host a task was generated [HC]
1797 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1798 the number of tasks waiting to be received on channel and sent
1800 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1801 waits for the first task coming from a given host.. [AL]
1803 GRAS new functionnalities: [MQ]
1804 * Enhance the parsing macro to allow the size of multidimentional objects
1805 to be given thru annotations.
1806 * New example (and documentation): Matrix Multiplication a la RPC
1807 (as when I was young!) and fix a bunch of bugs found on the way.
1809 GRAS performance improvements: [MQ]
1811 * Reduce the amount of cbps creation/destruction by making it static to
1812 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1815 * Change libdata to a set so that we can search for stuff by ID (and thus
1816 reduce the insane amount of dict lookups)
1819 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1820 RL communication on top of SG since it's so uneffective.
1821 It may also be used for inter-thread communication in RL, one day.
1822 * Use gras_datadesc_copy() to exchange messages on top of SG
1824 - improve message exchange performance on top of SG
1825 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1826 * Don't exchange on the network the size of the used part of buffer,
1827 instead, specify the possible buffer size to read().
1829 - reduces the amount of read/write calls (one pair per exchange)
1830 - reduces the amount of exchanged data (the size)
1831 - allows to retrieve all arrived data on receiver side, if we don't need
1832 it right now (subsequent read will peek the buffer)
1833 - allows the receiver to proceed with the begining of the stream before
1834 everything is arrived
1835 - make it possible to build an iov transport (using readv/writev)
1837 - take care of the data with non-stable storage (like stacked data),
1839 * If possible, TCP send uses vector I/O (when writev() is here)
1840 - Don't use it for receive since we send data sizes and data on the
1841 same stream, so we wouldn't be able to chain large amount of chunks
1842 before having to flush the stuff to read the size.
1843 * Rework the transport plugin mecanism to simplify it and reduce the
1844 amount of pointer dereferencement when searching for the right function
1847 * I guess that now, we do almost as few system calls as possible while
1848 doing as few data copy as possible.
1850 To improve it further, we could try to send all the sizes first and then
1851 all the data (to use iov on receiving size), but it's only a partial
1852 solution: when you have 2 dimensional data, the sizes of the second
1853 dimension is data of the first dimension, so you need 3 streams.
1855 I'm not sure the potential performance gains justify the coding burden.
1857 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1859 SimGrid (3.00) stable; urgency=high
1862 * New! Give the possibility to hijack the surf parser and thus bypass
1863 MSG_create_environment and MSG_launch_application. Have a look at
1864 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1866 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1868 SimGrid (2.96) unstable; urgency=low
1873 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1874 This deprecates the xbt_error_t mecanisms.
1875 It modifies (simplifies) all XBT and GRAS API.
1876 MSG API keeps unchanged (exceptions raised by XBT are catched from
1877 within MSG and masked with existing error handling facilities)
1880 * New! Add a FATPIPE model. [AL]
1881 * New! Add a parallel task model. [AL]
1882 * New! Add automatically a loopback interface (in the default
1883 network model) if none was precised.
1886 * Bugfix: MSG_process_resume now works with the current running process.
1888 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1889 * Modification of MSG_task_get_compute_duration. Once a task has been
1890 processed, the value returned by this function is now equal to 0. [AL]
1891 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1892 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1893 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1894 of a m_task. For now, MSG_task_cancel only works with computation
1896 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1897 of the processor (in Mflop/s) regardless of the current load on the
1899 * API Change: use proper naming convention for MSG_getClock and
1900 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1902 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1903 This function changes the priority of a computation task. This priority
1904 doesn't affect the transfer rate. A priority of 2 will make a task
1905 receive two times more cpu power than the other ones. This function
1906 has been added to suit the needs of Nguyen The Loc and hasn't been that
1907 much tested yet. So if it fails, please report it and send me your code.
1909 * API Change: removed all functions and types that were marked "deprecated"
1910 since many months. Renamed MSG_global_init_args to MSG_global_init.
1912 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1914 SimGrid (2.95) unstable; urgency=low
1917 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1918 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1919 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1922 * Bugfix: Allow absolute paths for platform description files [MQ]
1923 * Bugfix: do free the variables after use. Leads to drastic performance
1925 * Implement max_duration (ie, timeouts) on resources [AL]
1928 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1930 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1931 the needed support (timeouts) [AL]
1934 * Implement measurement sockets. You can now get the bandwidth between two
1935 hosts thanks to AMOK (see below). [MQ]
1936 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1937 dynar over the network (yeah) [MQ]
1938 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1939 * simple/static token-ring example. [Alexandre Colucci and MQ]
1940 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1941 in place before [MQ]
1944 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1945 type parsing. "short" and "long" modifiers were also ignored; other
1946 modifier (such as reference level) are still ignored. [MQ]
1947 * Embeed the buffer size within the buffer itself on SG. [MQ]
1948 That way, send() are atomic and cannot get intermixed anymore (at least
1949 the ones which are less than 100k; bigger messages still have the issue)
1950 * Array size pushed by the field, not by the field type (or each
1951 and every long int will push stuff to the cbps) [MQ]
1952 * use select() to sleep since it allows to portably sleep less than one
1955 GRAS (minor cleanups)
1956 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1959 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1960 * type category 'ignored' killed as it was never used and were difficult
1962 * whether a type can cycle or not is now a flag, leaving room for more
1963 flags (as "ignored", if we feel the need one day ;)
1964 * Rename raw sockets to measurement sockets since "raw" has another
1965 meaning in networking community.
1968 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1969 and offers features not belonging to GRAS but that most applications
1970 need. One day, it may be a set of plugins loadable at runtime.
1971 * New module: bandwidth
1972 bandwidth measurement between arbitrary nodes running this module. [MQ]
1974 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1976 SimGrid (2.94) unstable; urgency=low
1978 The first beta release of SimGrid 3 !
1982 * Update the main page and the FAQ. Adding references to gforge.
1985 * Add a gras_os_getpid function.
1988 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1989 * Extend the logs so that they also print PID, hostname, date, ... if
1991 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1992 and kill the old version which were in testsuite/
1993 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1997 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1998 for users. Instead, each programming environment (GRAS, MSG,...) use it
1999 when needed to provide such a feature to users.
2000 Don't shortcut the mecanism or you will also shortcut the virtualization
2001 you need on the simulator.
2005 * Cleanups in configury with regard to compile optimization/warning flags.
2006 Also add -fno-loop-optimize to any powerpc since it's the optimization
2007 killing gcc (< 3.4.0).
2008 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2009 needed by MSG examples complications
2010 * Borrow configury beautifications from PHP
2013 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2014 warning (thanks loris for stumbling into this one).
2015 * Bugfix: stop loading private headers (gras_config.h) from the public
2019 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2020 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2021 RL would imply a DNS resolver, which is *hard* to do in a portable way
2022 (and therefore delayed).
2023 * Implement a real timer mecanism and use it in timing macros. This allows
2024 to avoid rounding errors and get a 0.000005 sec precision in timing
2025 macros. While I was at it, various cleanups:
2026 - allow to declare more than one timed section per file (fix a stupid bug)
2027 - move some private declaration to the right place
2028 - merge conditional execution and timing macros into emulation module
2029 - document the module
2030 - make sure the module cleanups its mess on gras_exit
2031 * Documentation improvements:
2032 - (new) how to compile applications using GRAS
2033 - (new) emulation support (timing macros)
2035 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2037 SimGrid (2.93) unstable; urgency=low
2039 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2042 - Use Paje properly where used. Still to be sanitized properly.
2043 - Portability fix: Add an implementation of the contexts using pthread
2047 - Add xbt_procname(): returns the name of the current process.
2048 Use it to show the current process's name in all logging.
2050 - fix detection of older flex version and the reaction, since we do
2051 depend on modern ones (we use lex_destroy)
2052 - Better separation of SG and RL in the libs: remove all simulation code
2053 from libgras. As a result, this lib is now only 200k when stripped.
2054 Some of the xbt modules may also be duplicated (two sets and such) and
2055 should be cleaned/killed before SG3.
2056 - Insist on using xlC on AIX because of weird problems involving gcc there.
2057 - Cleanup the make remote stuff. This is now done by scripts
2058 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2059 mainly for our private use, but we're working on changing them to user
2062 - Bugfix: flush the socket on close only if there is some *output*.
2063 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2065 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2068 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2070 SimGrid (2.92) unstable; urgency=low
2072 Alpha 3 on the path to SimGrid 3
2076 - New! First try of benchmarking macros.
2077 - New! First try so that gras_stub_generator generate deployment and
2078 remote compilation helpers.
2080 - Bugfix: Initialization fix in msg_test.
2084 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2086 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2088 - API Change: gras_os_sleep to take the amount of seconds as a double.
2089 Accepting an int was error prone since it was the only location where
2090 seconds were coded as such. It leaded to damn rounding errors.
2091 - Bugfix: Hard to belive that timers ever worked before this.
2093 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2095 SimGrid (2.91) unstable; urgency=low
2097 Alpha 2 on the path to SimGrid 3
2101 - Bug fix in the lmm_solver.
2103 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2104 through the function MSG_paje_output.
2105 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2106 - It is possible to bound the rate of a communication in MSG with
2107 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2108 to write it in the changelog).
2109 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2114 - Various cleanups to the autotools stuff
2115 - Begin to move Gras examples to examples/gras/
2116 - Let make distcheck work again (yeah!)
2118 - documentation overhauled using doxygen.
2119 gtk-doc-tools is dead in SimGrid now.
2120 - Automatically extract all existing logging categories, and add the list
2121 to the documentation (long standing one, to say the less)
2123 - Cleanup the known architecture table. Reorder the entries to group what
2124 should be, and use a more consistent naming scheme.
2125 (some of the test dataset are still to be regenerated)
2126 - New! Allow library to register globals on each process just as userdata
2128 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2129 lookup time (for now).
2130 Use it in msg and trp.
2131 This cleans a lot the internals and helps enforcing privacy of the
2132 headers between the gras components.
2133 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2134 - Bugfix: gras_os_time was delirious in RL.
2135 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2137 - Reenable GRAS now that it works.
2139 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2141 SimGrid (2.90) unstable; urgency=low
2143 Alpha 1 on the path to SimGrid 3
2145 * It is a long time since the last release of SimGrid. I'm sorry about
2146 that but as I had told you, I was rewriting a lot of things. I apologize
2147 to those who had been reporting bugs to me and that I had not answered.
2148 If your bug is still in the new version, please tell me. Here is a
2149 summary of the main changes.
2151 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2152 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2153 portability, speed, and a lot more but you'll figure it out later.
2154 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2155 I wanted to release the new MSG as soon as possible and I have
2156 broken GRAS, which is the reason why, for now, only MSG is fully
2157 functional. A laconic description of these projects is available
2158 in the documentation.
2160 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2161 kernel optimized for our needs (called SURF but only the developers
2162 should use it). Hence, MSG is now roughly 30 times faster and I think
2163 that by rewriting a little bit MSG, I could event speed it up a little
2164 bit more. Beside the gain in speed, it is also much easier to encode a
2165 new platform model with SURF than it was with SG. More to come...
2167 * REVOLUTION 3: I have tried to change a little as possible the API of
2168 MSG but a few things really had to disappear. The main differences
2169 with the previous version are :
2170 1) no more m_links_t and the corresponding functions. Platforms are
2171 directly read from a XML description and cannot be hard-coded
2172 anymore. The same format is used for application deployment
2173 description. The new format is described in the documentation.
2174 Have a look in tools/platform_generation. There is a tiny script
2175 that converts from the old platform format to the new one. Concerning
2176 the application deployment format, parsing the old one is tricky.
2177 I think most of you should however be able to convert your files. If
2178 it is really an issue, I can write a C code that does the conversion.
2180 2) the toolbox tbx does not exist anymore. We now have a library
2181 with much more data-structures but without the hash-tables (we have
2182 dictionaries that are much faster).
2184 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2186 *****************************************************************************
2187 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2188 * don't feel like converting the oldies. (Mt) *
2189 *****************************************************************************
2192 Version 2.90: "the long awaited one"
2193 - Finished rewriting and debugging MSG. Rewrote the documentation.
2194 - disable GRAS for now since it needs to be ported to the newest SG
2197 - Finish the port to windows (using mingw32 for cross-compile)
2200 - Main loop and datastructures of SURF. A cpu resource object is
2201 functional. Surf can thus be used to create cpu's with variable
2202 performance on which you can execute some actions.
2204 2004-11-15 Martin Quinson
2205 - Port to ARM. Simply added the alignment and size descriptions. Should
2206 work, but the ARM machines are so slow that I didn't had the opportunity
2207 to 'make check' over there yet.
2209 2004-11-15 Arnaud Legrand
2210 - Trace manager now written. It uses a heap structure and is therefore
2211 expected to be efficient. It may however be speeded up (particularly
2212 when many events occur at the same date) by using red and black
2213 trees. One day maybe...
2214 - Max-min linear system solver written. It uses a sparse matrix
2215 structure taking advantage of its expected use. Most operations are
2216 O(1) and free/calloc are called as few as possible. The computation of
2217 the minimum could however be improved by using a red and black tree
2220 2004-11-03 Arnaud Legrand
2221 - Rename every gras_* function that was in xbt/ to its xbt_
2223 - Add a heap and a doubly-linked list to xbt
2224 - Added a dichotomy to the dictionaries. make check works as well before
2225 so I assume that the patch is correct. I do not know however if things
2226 run effectively faster than before now. :)
2228 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2229 SimGrid, and the version number is bumped to 2.x
2231 2004-10-29 Martin Quinson
2232 - Introduction of the remote errors.
2233 They are the result of a RMI/RPC on the remote machine.
2234 ErrCodes being scalar values, you can't get the host on which those
2235 errors did happen. Extending the error mechanism as in Gnome is possible.
2236 No idea yet whether it is a good idea.
2238 2004-10-28 Martin Quinson
2239 - Interface revolution: the Starred Structure Eradication.
2240 I used to do typedef struct {} toto_t; and then handle *toto_t.
2241 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2242 - ???_t is a valid type (builded with typedef)
2243 - s_toto_t is a structure (access to fields with .)
2244 - s_toto is a structure needing 'struct' keyword to be used
2245 - e_toto_t is an enum
2246 - toto_t is an 'object' (struct*)
2248 typedef struct s_toto {} s_toto_t, *toto_t;
2249 typedef enum {} e_toto_t;
2250 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2251 s_toto_t) is private.
2253 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2254 so that it can set the variable to NULL. It was so for dicts and sets,
2255 it changed for dynars.
2257 - Fix a bunch of memleaks in dict_remove
2258 - Fix a bug in sg/server_socket opening: it failed all the time.
2260 2004-10-07 Martin Quinson
2261 - Speed up dynar lookup operation a bit.
2263 gras_dynar_get is dead.
2265 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2266 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2267 (which gives you the address of the stored data).
2269 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2270 copy of the data using an affectation to do the job and not a memcpy.
2272 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2273 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2275 It does not really speedup the dynar test because they are
2276 setting elements all the time (and look them seldom). But the dict does
2277 far more lookup than setting.
2279 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2281 2004-10-05 Martin Quinson
2282 - Allow to (en/dis)able the cycle detection at run time.
2284 Whether we should check for cycle or not is now a property of each
2285 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2286 datadesc_cycle_unset allow to remove this property when previously set.
2288 Note that the cycle detection is off by default since it impacts the
2289 performance. Watch the data you feed GRAS with ;)
2291 This property is hereditary. Any element embedded in a structure having it
2292 set have it set for the time of this data exchange.
2294 You should set it both on sender and receiver side. If you don't set it on
2295 sender side, it will enter an endless loop. If you forget on receiver
2296 side, the cycles won't be recreated after communication.
2298 - Header reorganization.
2299 Kill gras_private.h, each submodule must load the headers it needs.
2301 2004-10-04 Martin Quinson
2302 - Interface revolution: do not try to survive to malloc failure.
2304 Now, gras_malloc and friends call gras_abort() on failure.
2305 As a conclusion, malloc_error is not a valid error anymore, and all
2306 functions for which it was the only gras_error_t return value are
2307 changed. They now return void, or there result directly.
2308 This simplify the API a lot.
2310 2004-09-29 Martin Quinson
2311 - Re-enable raw sockets.
2312 Created by gras_socket_{client,server}_ext;
2313 Used with gras_raw_{send,recv}
2316 It should allow to kill the last bits of gras first version soon.
2318 This is not completely satisfactory yet (duplicate code with
2319 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2322 - Simplify transport plugin (internal) interface by not passing any
2323 argument to _server and _client, but embedding them in the socket
2326 2004-09-28 Martin Quinson
2327 - Finish the port to AIX.
2328 autoconf was my problem (segfault within the malloc replacement
2329 function. No idea why)
2331 2004-09-16 Martin Quinson
2332 - Fix some size_t madness on 64bit architectures.
2334 2004-09-08 Martin Quinson
2335 - Reduce the number of system headers loaded, overload some more system
2336 calls (such as malloc to cast the result of the system one, and work
2338 - Fix and reintroduce the config support
2340 2004-09-07 Martin Quinson
2341 - Source code reorganization to allow Arnaud to surf all over there.
2342 - Allow to document the logging categories.
2343 - Remove all uppercase from logging categories and useless cleanup in names.
2345 2004-08-18 Martin Quinson
2346 Version 0.6.2 (protocol not changed; API changed)
2347 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2348 gras_error_t), and NULL when not found. Functions expecting a msgtype
2349 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2350 hopefully usefull message.
2351 - Portability to prehistoric sparcs again
2353 2004-08-17 Martin Quinson
2354 Version 0.6.1 (protocol not changed; ABI not changed)
2355 - prealloc some buffers to speed things up
2357 2004-08-11 Martin Quinson
2358 Version 0.6 (protocol not changed; ABI expended)
2359 - The parsing macro can deal with the references, provided that you add
2360 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2362 2004-08-09 Martin Quinson
2363 Version 0.5 (protocol not changed; ABI changed)
2364 - Allow to off turn the cycle detection code in data exchange at
2365 compilation time. It should be at run time, but I'm short of time (and
2366 the config stuff is still broken). That way, we keep dict out of the
2367 critical path, which is good because the performance is poor:
2368 - search not dichotomial yet
2369 - dynar give no way to access their content and memcpy everytime
2370 - In composed data description (struct, ref and so on), stop foolness of
2371 keeping the subtype's ID, but store the type itself. This keeps sets out
2372 of the critical path, which is good since they rely on dynar and
2373 dictionnaries. The only loose of that is that we cannot detect the
2374 redeclaration of a structure/union with another content (but I'm not sure
2375 the code detected well this error before anyway). We still can detect
2376 the redefinition discrepancy for the other types.
2377 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2378 it breaks the code because of type-punning used all over the place).
2379 This breaks on all non-gcc architectures (for now).
2381 All those changes (plus the buffer of last time) allow me to gain 2 order
2382 of magnitude on cruel tests consisting of 800000 array of integers on two
2383 level of a hierarchical structure (200 secondes -> 4 secondes)
2386 - the selector of reference must now return the type it points to, not
2387 the ID of this type.
2389 2004-08-06 Martin Quinson
2390 Version 0.4 (protocol changed; ABI not changed)
2391 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2392 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2393 - Implement a buffer transport, and use it by default (it relies on tcp in
2394 real life and on sg in simulation).
2395 That's a bit hackish since I had a new field to the structure to store
2396 its data without interfering with the subtype ones. Inheritance
2397 is tricky in C. And that's a kind of reverse inheritance with one class
2398 derivating two classes. Or maybe a game with java interfaces. Anyway,
2399 that's damn hard in C (at least).
2400 Moreover, I got tired while trying to ensure plugin separation and
2401 genericity in SG mode. MSG wants me to do weird things, so let's go for
2402 cruel hacks (temporarily of course ;).
2403 See comment in transport_private.h:71
2404 - do not include all the _interface headers in private but in the files
2405 which really need them (to cut the compilation time when they are
2408 2004-07-26 Martin Quinson
2409 Version 0.3 (protocol not changed; ABI changed)
2410 - Major overhault of the datadesc interface to simplify it:
2411 - shorted the function names:
2412 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2413 - add a trivial way to push/pop integers into the cbps without malloc.
2414 This allows to make really generic sub_type description, which simply
2415 pop their size of the stack.
2416 - add a function gras_datadesc_ref_pop_arr() which does what users want
2417 most of the time: Declare a dynamic array (which pops its size of the
2418 stack) and declare a reference to it. Poor name, but anyway.
2419 - kill the post-send callback, add a post-receive one
2421 2004-07-23 Martin Quinson
2422 Version 0.2 (protocol changed; ABI changed)
2423 - add some testing for cpbs in the test cases, and fix some more bugs.
2424 This invalidate again the little64 data file, since I cannot regenerate
2426 - remove an awfull optimization in the logging stuff, allowing me to:
2427 - understand it again
2428 - learn gcc how to check that the argument match the provided format
2429 - fix all errors revealed by gcc after that
2430 - internal keys of dict are not \0 terminated. Deal with it properly in
2431 loggings instead of segfaulting when the user want to see the logs :-/
2433 2004-07-22 Martin Quinson
2434 - Fix some stupid bug preventing cbps (callback postit) from working
2436 2004-07-21 Martin Quinson
2437 - Some documentation cleanups
2438 - remove the useless last argument of msgtype_declare
2439 - rename the Virtu functions to fit into the 'os' namespace
2440 - move headers src/include -> src/include/gras/ and stop fooling with
2441 gras -> . symbolic link
2442 - make distcheck is now successful
2444 2004-07-19 Martin Quinson
2446 - Build shared library also
2447 - Install html doc to the right location
2448 - stop removing maintainer files in make clean
2449 - build tests only on make check
2451 2004-07-13 Martin Quinson
2453 - No major issue in previous version => change versionning schema
2454 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2455 corresponding dataset.
2457 2004-07-11 Martin Quinson
2459 - Get it working with any kind of structure (we can compute the padding
2460 bytes remotely for all the architectures I have access to)
2461 - Implement the structure parsing macro (still not quite robust/complete)
2462 - Improvement to the remote testing toysuite
2464 2004-07-10 Martin Quinson
2465 [autoconf mechanism]
2466 - get ride of a bunch of deprecated macros
2467 - actually run the test for two-compliment, not only compile it :-/
2468 - test whether the structures get packed (and bail out if yes. Damn.
2469 Alignment is a serious matter)
2470 - test whether the structures get compacted (but respecting the alignment
2471 constraints of each types)
2472 - test whether the array fields of structures can straddle alignment boundaries
2474 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2475 (took me 2 hours to find that bug, looking at the wrong place)
2476 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2477 before sending/receiving and are used to compute the offsets of fields
2478 - Given that padding size depend even on compiler options, keep track of
2479 alignment and aligned_size only for the current architecture. Not a big
2480 deal since we send structure fields one after the other (seems
2482 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2483 program, let it work on linux/gcc/little32. portability todo.
2485 2004-07-08 Martin Quinson
2486 - import and improve remote compilation support from FAST
2487 - make sure make check works on half a dozen of machines out there
2489 2004-07-07 Martin Quinson
2490 Let's say it's version 0.0.3 ;)
2491 - Implement conversions (yuhu!)
2492 - Let it work on solaris (beside conversion, of course)
2493 - Stupid me, using rand() to generate the conversion datatests in not wise.
2495 2004-07-06 Martin Quinson
2496 - Let make dist work, since I'm gonna need it to compile on remote hosts
2497 - Let Tests/datadesc_usage write the architecture on which the file was
2498 generated as first byte.
2499 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2500 architecture descriptions.
2501 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2502 on those architectures.
2503 - Optimization: send/recv array of scalar in one shoot
2505 2004-07-05 Martin Quinson
2506 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2508 - Plug a whole bunch of memleaks
2509 - each process now have to call gras_{init,exit}. One day, their log
2510 settings will be separated
2511 - Continue the code factorisation between SG, RL and common in Transport.
2513 2004-07-04 Martin Quinson
2515 - Redistribution between SG and RL.
2516 We wanna have to accept in SG, so move accepted related parts of RL in
2517 the common part. (more precisely, the dynar of all known sockets is no
2518 more a static in transport.c, but part of the process_data)
2520 [gras_stub_generator]
2521 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2523 2004-07-03 Martin Quinson
2524 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2525 them all up in one shot)
2527 - Fix the ugly case of reference to dynamic array.
2528 - New (semi-public) function gras_datadesc_size to allow the messaging
2529 layer to malloc the needed space for the buffer.
2531 - gras_socket_close now expect the socket to close (and not its address to
2532 put NULL in it after it). This is because the socket passed to handlers
2533 is one of their argument (=> not writable).
2535 - propagate the interface cleanup from last week in datadesc, ie remove a
2536 superfluous level of indirection. User pass adress of variable
2537 containing data (both when sending and receiving), and not of a variable
2538 being a pointer to the data. Let's say that I like it better ;)
2539 The price for that is constructs like "int msg=*(int*)payload" in
2540 handlers, but it's a fine price, IMHO.
2542 - Let it work in RL (yuhu)
2544 2004-06-21 Martin Quinson
2546 - porting SG plugin and SG select to new standards (works almost).
2547 - plug memleaks and fix bugs around.
2550 - cleanup the prototype of data recv and force users to specify when they
2551 want to handle references to objects. Test case working even for cycles.
2552 - plug memleaks. Valgrind is perfectly ok with this.
2554 2004-06-12 Martin Quinson
2556 - cleanup the separation between plugin and main code in plugin creation
2558 2004-06-11 Martin Quinson
2560 - Reput hook for raw sockets, needed for BW experiments
2561 - kill a few lines of dead code
2562 [Data description] Interface cleanup
2563 - gras_datadesc_by_name returns the searched type or NULL.
2564 That way, no variable is needed to use a type desc once, which makes
2566 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2567 two parameters were strings (field name, type name), leading to
2569 [Dicos] Interface cleanup
2570 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2571 This is consistant with the dynar API.
2573 2004-04-21 Martin Quinson
2575 - Porting to new standards.
2577 - interface cleanup.
2578 There is no bag anymore, no need to take extra provision to mask the
2579 pointers behind "ID".
2580 Better splitup of functions between files create/exchange/convert.
2581 This is still a bit artificial since convert and receive are so
2582 interleaved, but anyway.
2584 - add a queued message list to procdata (the ones not matching criteria
2586 - factorize some more code between SG and RL wrt procdata
2588 - use gras_exit in example to track memleaks
2589 - get rid of gs_example now that GS is properly integrated into gras
2590 - update run_test to integrate the latest tests (datadesc)
2592 - rename WARNINGn macros to WARNn since it prooved error-prone
2594 2004-04-19 Martin Quinson
2596 - register init/exit functions within gras module mechanism
2597 - send/receive function.
2598 Convertion is not implemented, but short-cutted if not needed.
2599 struct/array elements are sent one by one (instead of block-wise), but
2600 nobody really cares (yet). Get a prototype before optimizing.
2601 - tests (using a file socket) for DD send/receive on:
2602 - base types: int, float
2603 - array: fixed size, string (ie ref to dynamic string)
2604 - structure: homogeneous, heterogeneous
2605 - chained list, graph with cycle
2606 Believe it or not, valgrind is not too unhappy with the results. The
2607 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2611 - Bugfix when using a filename explicitely (instead of '-')
2613 2004-04-09 Martin Quinson
2615 - factorize more code between RL and SG in socket creation
2616 - Complete the implementation and tests of:
2618 o file (only in RL, and mainly for debugging)
2620 I lost 3 days to design a portable address resolver, and then decided
2621 that the prototype mainly have to run on my box.
2622 Addressing portability too early may be like optimizing too early :-/
2624 - use gras_init in the Tests instead of the crappy parse_log_opt
2625 (the latter function is removed)
2626 [Conditional execution]
2627 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2628 [Code reorganisation]
2629 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2630 Build examples against the RL library, since there is no way to disable
2631 its creation for now.
2633 For information, the beginning of coding on GRAS was back in june
2634 2003. I guess that every line has been rewritten at least twice since