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 * Clean up CMake files (may need a full rewrite).
8 Non exhaustive list of subgoals:
9 - Use genuine cmake mechanisms and variables when available,
10 instead of reinventing the wheel.
11 - Correctly determine system and architecture (e.g. x32).
12 - Correctly determine compiler type and version (e.g. clang).
13 - Correctly set compiler flags according to compiler type and version.
14 - Correctly set compiler flags for C++, Java, and Fortran compilers too.
15 - Use git to generate the dist archive. Either use git-archive to
16 generate the tarball, or keep using cmake -E tar, but use git-ls-files
17 to get the list of files to pack.
18 - For Java, add an option to enable/disable the inclusion of the native
19 libraries in the jar file, and avoid to duplicate these files.
22 * [FIXME: elaborate] Merge branch hypervisor
25 * [FIXME: elaborate] Rewrite Surf in C++
27 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
29 SimGrid (3.10) stable; urgency=low
31 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
34 * Reintegrate Java to the main archive as desynchronizing these
35 package is not acceptable anymore (Java is now considered stable)
36 * Add explicit synchronization facilities through semaphores
37 * Bug fix: Task.setDataSize() only changed the C world, not the value
38 cached in the Java world
41 * Dramatically change the way files are handled. API and internals changed, but
42 this part of MSG was not considered as production grade either.
43 * Add explicit synchronization facilities through semaphores
44 * Add a new function MSG_host_get_process_list()
45 * Preliminary DVFS support (see examples/msg/energy/ for details)
48 * SMPI is now included directly in the libsimgrid as the windows
49 linker doesn't force us on splitting it anymore.
50 * Improvements of the SMPI replay tool:
51 - Most of the collective communications are now rooted in the same process as
52 in the original application.
53 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
54 used until now). Multiple data types can now be used in a trace.
55 - The replay tool now supports traces produce either by TAU or a modified
57 - Bug Fix: the compute part of the reduce action is now taken into account.
58 - Gatherv collective is now supported
59 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
60 of smpirun outputs time independent traces for the current run. One file
61 is created per process. If too many processes are simulated, this behavior
62 can be changed to one file for all processes by using the
63 tracing/smpi/format/ti_one_file flag
64 * smpirun generates the host file if needed (with given host count and platform)
65 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
66 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
67 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
68 --cfg=smpi/coll_selector:(mpich/ompi)
69 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
71 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
72 happening inside a collective SMPI call.
73 * Fix the behavior of complex data types handling.
74 * Make MPI_Wtime another synchronization point to take computations into account.
75 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
76 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
77 * Add all missing Fortran bindings, SMPI should work with Fortran 90
78 (no privatization of global variables yet)
79 * Preliminary DVFS support (see examples/smpi/energy/ for details)
82 * Verification of liveness properties is now available for SMPI applications
83 (in addition to MSG applications)
84 * Bugged examples using SMPI in examples/smpi/mc/
85 * Add --cfg=model-check/visited option. Allows the verification of infinite
86 programs. Detection of loops in the execution thanks to the system state
87 comparison and reduction of the state space to explore. Can be combined with
88 DPOR for safety properties.
91 * Allow to change SimGrid configuration (see --help) within the code
92 thanks to SD_config() as it can be done in MSG.
93 * Add a new function SD_task_set_amount() upon user request.
96 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
97 * Remove rule based routing (no more PCRE dependency)
98 * Add a limiter_link option to cluster tag, to specify a maximum reachable
99 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
100 * Add a loopback_bw and loopback_lat options to cluster tag.
101 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
102 routing. Now peers are encapsulated in an AS and have their own private
103 router but this is transparent.
106 * Our own implementation of getline is renamed xbt_getline, and gets
107 used even if the OS provide a getline(). This should reduce the
108 configuration complexity by using the same code on all platforms.
109 * New type: xbt_cfg_elm_boolean.
110 * Allow to use yes/no for boolean configuration options in the command line.
111 * Allow to disable SimGrid cleanups at exit from command line option.
112 There are situations where one may want a simulation to end with an exit.
113 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
114 annoying when scripting around the simulator. Adding a
115 --cfg=clean_atexit:no allows to circumvent this issue.
118 * Lots of memory leaks were corrected in this release.
119 There is no known memory leaks anymore, in all of our 600+ tests.
120 * New command line option --version, to get SimGrid version information.
121 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
123 * Supernovae builds are deprecated, and expected to be removed in the next
126 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
128 SimGrid (3.9) stable; urgency=low
130 The Grasgory release: GRAS is really dead now.
132 * Complete overhaul of the internal host structures scheme.
135 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
136 was considered as experimental and badly maintained since then.
137 * Keeping it was thus a trap to our potential users, that could take
138 it instead of MSG or SMPI by mistake despite is pity state.
139 * GRAS seems to have very few users (if any), and no one volunteered
140 to maintain it further. It also induces a lot of XBT code (for
141 portability sake), that must be maintained too.
142 * For all these reasons, we killed GRAS. If someone wants to revive it
143 in the future, don't cry, our git history still remembers of GRAS.
146 * Major overhaul. Merge our documentation again as time proved that
147 splitting it was really not helping our users.
148 * Further improve the developer documentation to help newcomers
149 hacking on SimGrid itself. The user documentation (and in
150 particular, the beginner documentation) is still in a sorry state.
153 * Now works on Windows too!
154 * Much more extensive test suite, from MPICH
157 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
158 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
159 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
160 * Bug fix: let task be scheduled when the last dependency to be solved is
161 a control dependency.
162 * Remove SD_load_environment_script function.
163 Use the C sg_platf function if you want to declare a platform
167 * New function: MSG_process_get_number()
168 * Old function documented: MSG_config()
169 * Remove MSG_load_platform_script function
170 Use the C sg_platf function if you want to declare a platform
174 * Change the default value of the TCP_gamma constant (maximal size of TCP
175 congestion window) to a more realistic 4MiB value. If you notice changes in
176 your simulation results, you can fall back to the previous 20k tiny window
177 by adding --cfg=network/TCP_gamma:20000 on command line.
178 * (Hopefully) fix a bug wrt periodic availability and state traces
179 * Bug fix: use default values at start when first event in availability/state
180 trace is not at time 0.
183 * remove the "new_" part of function name sg_platf_new_trace_connect
184 (resulting in sg_platf_trace_connect), since it does not create
188 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
189 use case where it's really mandatory, and maintaining it was a pain in
191 * New: xbt_fifo_search(), search an item with a user-provided
192 comparison function instead of dumb pointer comparison.
195 * Fix the lua deployment:
196 Use `simgrid.init_application()` before deployment instead of
197 `simgrid.msg_register_application()` after.
200 * Transfer the tracing files into the corresponding modules.
202 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
204 SimGrid (3.8.1) stable; urgency=low
206 The "we are told that some people want to also *install* the simgrid
209 * Add missing file "tesh.1" to the archive.
211 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
213 SimGrid (3.8) stable; urgency=low
215 The Psssshiiiit release: SimGrid jumps into the Cloud.
218 * Add an experimental interface to manipulate VMs. They are mainly
219 process groups with very few intrinsic semantic, but they should
220 allow you to build the semantic you want easily.
221 * New function: MSG_host_set_property_value()
222 * New function: MSG_process_on_exit(). To clean memory in all cases.
223 * Bug fixes that made the host (and link) failures unusable.
224 * Add a way to auto-restart process when the host in which they are
225 executing comes back (ON_FAILURE="RESTART" on deployment file,
226 MSG_process_auto_restart_set).
227 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
228 please stop using the old ones, they are DEPRECATED.
230 * Deprecate functions MSG_global_init() / MSG_global_init_args()
231 Please use MSG_init() instead. (reducing the amount of entry
232 points in the library helps us).
233 * Make it impossible to link against the wrong version of the lib
234 * Deprecate MSG_clean(). No need to call it anymore.
235 * Function MSG_get_host_number() is not deprecated anymore.
238 * Split the doc into a user guide and a reference guide.
239 * Start a developper guide to help people hacking on SimGrid.
242 * Enable tracing by default. This modules rocks you should use it.
243 * Remove option custom_flags. Now use environment variables CFLAGS
245 * Use default cmake things to detect lua instead of home grown ones.
246 * New option "enable_mallocators" to disable mallocators, for debugging
247 purpose ("on" by default).
250 * Bug fixes around the resource failures: don't let the processes
251 survive the host they are running onto.
252 * Add an interface to auto-restart processes when the host in which they are
253 executing comes back.
254 * Ensures that SIMIX_clean is called automatically. It's not part of
255 the public interface anymore (bindings should be updated).
258 * Bug fix for when SD_Simulate is called with a positive value: be careful
259 when comparing doubles. Sometimes they are different for non significant
261 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
262 parallel task whose initial work is distributed among host according
263 to the Amdahl's law. Such tasks are created with a parameter alpha
264 that corresponds to the non-parallelizable part of the computation.
265 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
266 between two sets of workstations assuming a 1D block distribution (each
267 workstation owns a similar share of data) on both sides.
269 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
270 Data redistribution will be automatically scheduled once parent and child
271 are both scheduled. The filling of computation_amount and
272 communication_amount structures is now done seamlessly thanks to the chosen
274 * New function SD_workstation_dump to display various information
275 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
276 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
277 bandwidth on the route onto which the task is scheduled and the amount of
279 To divide the nominal bandwidth by 2, the rate then has to be :
280 rate = bandwidth/(2*amount)
281 * Compute tasks that have failed can now be rescheduled and executed again
282 (from their beginning)
283 * Increasing source code coverage (src/simdag is now covered at 95.8%
287 * Re-implement time-independent trace replay using SMPI (at the
288 smpi_smp_* level) instead of MSG. This should replace
289 examples/msg/actions/actions.c
290 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
291 hindexed and structs)
292 * Implement the exchange of non-contiguous data.
293 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
294 * Correct behavior of smpi/sender_gap and set its default value to 0
295 * Add option to asynchronously send small messages to allow better
296 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
297 specifies the size in bytes under which messages will be asynchronously sent.
298 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
299 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
302 * New C interface to define a platform: XML is now optional.
303 For more info, please check include/simgrid/platf.h.
304 * New interface to define random platforms from the C:
305 For more info, please check include/simgrid/platf_generator.h and
306 examples/msg/masterslave/masterslave_platfgen.c
307 * Export a sg_cmdline dynar containing all the arguments we got from
311 * Two new tracing options for adding comments to trace file so you
312 can track your experiments (see --help-tracing for details).
313 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
314 * Adding the SimGrid version that generated the trace file as a comment.
315 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
316 * Fix to avoid key clashes on Paje links
317 * Other minor fixes related to the Paje specification
320 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
321 and renamed to xbt_str_hash() and xbt_str_hash_ext().
322 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
323 * Greatly improve the robustness of mmalloc to user errors (such as
324 using an area after freeing it, or freeing it twice)
326 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
328 SimGrid (3.7.1) stable; urgency=low
331 * Restore the prototype of MSG_process_create_with_environment() to
332 the pre-3.7 situation by removing the kill_time argument.
333 * Add a MSG_process_set_kill_time() function instead.
336 * Fix weird behaviors when dealing with parallel tasks.
339 * Simgrid is now built as a dll.
340 * Simgrid-java now works on Windows.
341 * Simgrid-Java is now included into Windows package.
344 * First pre-build package for MacOSX.
347 * Fix compilation when using MSG_USE_DEPRECATED.
348 * Fix some compilation issues on Macs and Windows.
349 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
350 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
352 We discovered that the Lua console is broken, but we are missing the
353 manpower to fix it right now. The problem existed in 3.7 too, so we
354 are not blocking the release for that. Sorry if you depended on this
355 feature, any help would be really welcome.
357 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
359 SimGrid (3.7) stable; urgency=low
361 The "spring cleanups (before the next Big Project kicks in)" release.
364 * We can specify the SMPI latency/bandwidth factor with command line
365 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
366 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
367 You can also use the "config tag" from platform file by adding this line
368 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
369 (see "example/platforms/tag_config.xml" to use "config tag").
370 Note that the command line supersedes the platform file configuration.
371 * Change the correction factors used in LMM model, according to
372 the latest experiments described in INRIA RR-7821.
373 Accuracy should be improved this way.
374 * Use the partial invalidation optimization by default for the
375 network too. Should produce the exact same results, only faster.
376 * Major cleanup in surf to merge models and split some optimization
377 mechanisms from the core of the models. As a result you can now
378 specify which model to use (e.g., --cfg=network/model:LV08
379 --cfg=cpu/model:Cas01) and which optimization mode to use
380 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
381 Incompatible combinations should err at initialization. See
382 --help-models for the list of all models and optimization modes.
383 * The CLM03 workstation model was dropped for simplicity because it
384 used the deprecated CM02 network model. Use default instead.
385 * Rename the TCP_gamma configuration option to network/TCP_gamma
386 * Rename the coordinates configuration option to
387 network/coordinates, and document it
388 * Use now crosstraffic keyword instead of the terribly misleading
389 fullduplex keyword. It is activated by default now in the current
390 default model, use --cfg=network/crosstraffic:0 to turn it off.
391 * Ongoing refactoring the model parsing to make XML files optional
392 See include/simgrid/platf.h for details (still to be completed)
395 * Major overhaul of the documentation. Almost instructive now :/
396 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
397 functions and friends. This interface was considered as
398 deprecated since over 2 years, it's time to inform our users that it is.
399 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
400 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
401 use these (crappy) functions in your code.
402 These functions will be removed soon. Stop using them now.
403 * Deprecate MSG_get_host_{table,number}
404 Implement MSG_hosts_as_dynar() instead.
405 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
406 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
409 * Stabilize the parallel execution mode of user contexts
410 * Introduce configuration variables to control parallel execution:
411 - contexts/synchro: Synchronization mode to use when running
412 contexts in parallel (either futex, posix or busy_wait)
413 - contexts/parallel_threshold: Minimal number of user contexts
414 that must be part of a scheduling round to switch to parallel
415 execution mode (raw contexts only)
416 * Fix bugs that prevented to use suspend/resume along with
417 synchronization structures.
418 * Fix bugs in process termination that lead to invalid memory access
419 in very specific conditions.
422 * Introduce a parallel mode for the models (controlled by surf/nthreads
423 configuration item). In our tests, running the models in parallel
424 never lead to any speedups because they are so fast that the gain
425 of computing each model in parallel does not amortizes the
426 synchronization costs, even when ultra fast futexes are used.
427 This is released anyway because YMMV.
430 * Performance boost by using a swag internally to compute the set of
431 tasks that are finished and should constitute the return value of
435 * Enable it by default now that it is considered rather stable.
438 * Documentation of the tracing functions.
439 * Performance gains when tracing categorized/uncategorized resource
440 utilization by avoiding calls to get route when updating resource
441 variables. LMM constraints are being used instead.
442 * API changed to set task categories. Use MSG_task_set_category instead
443 of TRACE_msg_set_task_category, and SD_task_set_category instead
444 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
445 * Bugfix for graphicator, routes not correctly obtained, memory leaks
446 * Examples for link user variables added (see at examples/msg/tracing/)
447 * Deprecated function TRACE_msg_set_process_category completely removed
448 * Trace header updated according to the latest Paje file format
449 * Tracing network lazy updates, no longer obligate users to use full updates
450 * --cfg=tracing/platform:1 also registers power/bandwidth variables
451 * Experimental: let user code declare/set/push/pop application states for hosts
452 * API changed to allow the manual creation of graph configuration files
453 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
456 * Improve the API of Lua MSG bindings, using the Lua spirit.
457 * Each simulated process now lives in its own Lua world (globals are
458 automatically duplicated). It helps writing simulators. Will allow
459 to run Splay programs within SimGrid in the future.
460 * Add a Chord example in Lua, equivalent to the MSG one.
463 * Start the implementation of a solution to express temporal
464 properties, not only local assertions. This is still an
465 experimental work in progress, stay clear from it to be safe.
469 - Add new runtime parameters --help-logs and --help-log-categories
470 to display informations about supported logging parameters and
472 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
473 don't exists anymore.
474 * Mallocators: allow value NULL for the reset function.
476 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
477 create homogeneous dictionaries, where all the elements share the
478 same free function. Non homogeneous dictionaries will be
479 deprecated in the next release.
480 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
481 - Multi-level dictionaries are deprecated.
483 - new function xbt_dynar_search_or_negative() that is useful when
484 you have less than 2 million elements in your dynar and don't
485 want of the extra complexity of catching exceptions when the
486 element is not found.
488 - Make xbt_os_thread module (for thread portability) public.
489 Documentation is still to come, sorry.
491 - Cleanups and simplifications to make it maintainable again.
492 - Exotic features (such as memalign and valloc) were removed.
493 - The metadata were extended and improved so that the
494 model-checker becomes able to explore and inspect the heaps.
495 - This may induce a performance drop when enable_model-checking is
496 ON in cmake (even if it's not used in the simulation), but it is
497 necessary at this point to get MC working.
499 Turn model-checking OFF if simulation performance matters to you.
500 Not enabling it at runtime is not enough, disable it in cmake.
502 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
504 SimGrid (3.6.2) stable; urgency=low
506 The "Not coding new stuff allows to polish old things" release.
509 * New bindings to the NS3 packet level simulator (experimental)
510 * Use the raw (efficient) execution contexts instead of the sysv
511 (portable) ones when possible.
512 * libpcre is now mandatory in any cases since not using it led to
513 severe performance loss and possibly other issues
514 * Update the XML platforms:
515 - G5K: include the latest machine in Nancy
516 - GridPP and LCG: new platforms
517 * Documentation was partially updated, at least (more to come)
519 Bug fixes, cosmetics and small improvements
520 * Free terminated processes before the end of the simulation to avoid
521 exhausting the memory of users having very dynamic amount of
523 * Bug fix and cosmetics about canceling non-running tasks
524 * Bug fix about the dot loader's issues when using libcgraph
527 * Create an installer for windows with nsis (amd64 and win32)
528 - Add an hello world project to illustrate simgrid project creation.
529 - Embed libpcre into the Simgrid installer to avoid
530 its compilation burden
531 * The raw execution contexts should work on Apple now
532 * Port to Windows 64 bits
533 - Sysv contexts now have an implementation for this arch
534 - GRAS communication features now support this arch
535 * Drop support for borland compiler on windows
536 - this code was not maintained, and we kinda depend on gcc nowadays
537 * Fix portability issues on kfreebsd/gnu: build error about semaphores
538 * Fix portability issue on unstable ubuntu: linker became picky on
541 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
543 SimGrid (3.6.1) stable; urgency=low
545 The "Oops, we broke Macs too" release
548 * Fixed contexts detection so that raw ones are used when possible
549 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
550 leads to a strange error, with user code segfaulting sometimes when
551 the generated code is not perfectly aligned (which is not
552 controllable from the user side, depends on the amount of code)
555 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
557 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
559 SimGrid (3.6) stable; urgency=medium
561 The Summer Release, also known as the "OMG! They Killed Kenny!" version
564 * Bindings now constitute their own package, separated from the main one.
565 Rationale: reduce our maintenance nightmare by reducing the module coupling
566 They will soon be released on their own on gforge.
568 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
569 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
571 GRAS: It is not considered as stable anymore, but experimental. Sorry.
572 * It's not quite deprecated for now because we have no replacement,
573 but it may soon become the case.
576 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
577 * Fortran: New user-level cache variable to store the rank of the running
578 process. This improves performance by an order of magnitude.
579 * C: New coccinelle script to automatically locate and modify global and
580 local static variables.
581 * Improved SMPI network model with a sender-side gap to account for multiple
585 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
586 only say if a communication is finished, no matter whether it succeeded or
587 failed. You can call MSG_comm_get_status() to know the status of a finished
589 * New function MSG_task_dsend() to send a task and detach it. When a
590 communication is detached, you are never notified of its success or failure
591 and the memory is released automatically once it is finished. This function
592 is useful when you don't care about the end nor the success of a
594 * Change the prototypes of action replay. Sorry for inconvenience,
595 but this is really more efficient this way (and to adapt your code,
596 you just have to fix the initialization, that shouldn't be too long)
597 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
598 ever managed to do anything useful with it.
599 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
600 takes a dynar of communications. It returns immediately and gives the
601 index of a finished communication (if any).
602 * New example: a basic implementation of the Chord P2P algorithm.
605 * New model for multi-core CPUs. You can now use the core attribute to
606 precise the number of cores of a host. This is a basic model. Every
607 process running on the host receives at most the power provided in
608 the DTD (throughput<=power). Total throughput of process cannot exceed
610 * New peer tag. This peer tag creates a tiny AS comprising a host and a
611 router linked by an up-link and a down-link (possibly asymmetrical).
612 This kind of pattern allows to easily build last-mile model style platforms.
613 Aggregating such patterns in a rule-based AS is thus the technique of
614 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
615 * New model for Vivaldi routing. We transformed the Vivaldi network model
616 into a Vivaldi routing model (based on the rule-based model). This allows to
617 combine Vivaldi based latencies with last-mile platforms.
620 * Added a check for NaN of IEEE754 infinite in the double entries of
622 * Introduce a new context factory "raw", highly inspirated from the
623 ucontext factory, but using manually crafted functions in assembly to
624 do the work in an efficient manner.
625 * Allow to change the used context factory at run time, not only at
626 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
627 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
628 contexts at runtime (only with raw contexts or ucontexts).
629 * Completely rewrote this module to allow parallel execution of user
630 processes. Use --cfg=contexts/nthreads:N to execute user processes
631 with N parallel threads (the default is 1, meaning no parallelism).
632 * Allow to decide dynamically between sequential and parallel modes.
633 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
634 processes in parallel only when their number is greater than or equal to P
636 * Added a check for NaN of IEEE754 infinite in the double entries of
640 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
641 won't output the state of processes when interrupted with Ctrl-C
642 * Add a new function xbt_dynar_to_array that transforms a dynar into a
643 NULL-terminated array. This may solve backward compatibility issues
644 due to the change to return type of SD_simulate. See also:
645 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
646 * Add new macros with variable number of arguments.
647 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
648 - in xbt/asserts.h: xbt_assert
649 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
650 - in xbt/ex.h: THROWF and RETHROWF.
651 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
653 * Change xbt_die() to accept a format string with arguments, just like printf.
654 * New data structure: xbt_lib_t, like a dict but more general and with better
658 * New configuration options
659 Options triva/categorized and triva/uncategorized can be used to generate
660 graph configuration files for Triva visualization tool.
661 * Configuration option tracing/platform is renamed to tracing/categorized
662 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
663 * New TRACE_user_link_variable interface:
664 User provides the name of the link and the tracing variable to attach to it
665 * the declaration of tracing categories must be done after the environment
667 * simpler tracing interface, just one way to declare categories
668 TRACE_category or TRACE_category_with_color, it is up to you
669 * links in the trace file are again identified by their names
670 * trace contains the full platform hierarchy exactly as declared using the ASes
671 * Options tracing/msg/[task|process]:1 groups the process by hosts
672 for both cases, tasks and processes must have names that are unique during the simulation
673 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
674 * The experimental option tracing/msg/volume is deprecated
675 its functionality may be reincorporated if needed
677 The tracing generates a trace file with unordered timestamped events,
678 because of the way the core simulator (surf) works. A script available
679 at the tools directory (fix-paje-trace.sh) can be used to put the events
680 in order. We have changed the tracing so it can generate ordered timestamped
681 events in the final trace, but depending on the simulator (and how much time
682 is simulated) that can lead to a huge memory utilization. It is deactivated
683 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
686 * Define a SIMGRID_VERSION macro in simgrid_config.h.
687 - We are trying hard to keep the API stable, but it may happen that
688 some things change (we're a research project after all, not a
689 nuclear plant operating system). If such things should happen, you
690 could rely on that macro to adapt.
691 - current value: 30600 for 3.06.00, aka 3.6
692 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
693 a number that can be compared with SIMGRID_VERSION.
694 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
695 assertions and verbose/debug logging events are disabled at compile time.
697 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
699 SimGrid (3.5) stable; urgency=medium
702 * New feature to any SimGrid-based simulator: Model-Checking
703 Check SIN#1 for more details.
706 * New Model SMPI (three-interval linear regression for correction factors)
707 See RR-7426, available at http://hal.inria.fr/inria-00527150
708 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
709 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
710 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
711 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
712 * See also src/smpi/README
717 - Tracing API changes: TRACE_start and TRACE_end should not be called
718 by user-code. They are automatically called by simulators created
719 with SimDAG, MSG and SMPI if the toolkit is compiled with
720 tracing_enabled=ON. Categories declaration and utilization remain the
721 same for MSG and SimDag.
722 - A function was added to the tracing API to declare categories with
724 - TRACE_category_with_color (char *category, char *color)
725 where color must be in the following format
726 "%f %f %f", red, green, blue
727 and red, green, blue are float values in the interval [0, 1]
728 - User can specify NULL as color parameter, or continue calling
730 On that case, the tracing system will define random colors
731 - The following command-line options are supported:
732 --cfg=tracing/filename:msg.trace
733 --cfg=tracing:1 (activate tracing, needed to use others)
734 --cfg=tracing/platform:1 (categorized resource use)
735 --cfg=tracing/uncategorized:1 (uncategorized resource use)
736 --cfg=tracing/msg/task:1 (task creation)
737 --cfg=tracing/msg/process:1 (process creation, migration)
738 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
739 --cfg=tracing/smpi:1 (SMPI interface tracing)
740 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
741 - examples of examples/msg/tracing updated
743 - DAXLoader and DOTLoader functions can generate tasks with categories
744 - A new function to attribute a category to SD tasks:
745 TRACE_sd_set_task_category (SD_task_t task, char *category)
746 * Tracing the MPI interface implemented by SMPI
747 - Collective operations are traced with states
748 - Point-to-Point operations are traced with states/links
749 - Tracing activated by a parameter "-trace filename" passed
750 to smpirun during execution (considering that simgrid
751 is compiled with tracing enabled)
752 - To run the simulation with gdb, the simulator
753 accepts --cfg=tracing/smpi:1 to trace SMPI
754 - tesh files to check if smpi tracing is ok
755 - See examples/smpi/NAS/DT-trace
756 * GTNetS tracing re-worked
757 - adaptation to the tracing system of GTNets to cope
758 with modifications regarding the fullduplex mode
759 - new tesh files to check if gtnets tracing is ok
762 * Asynchronous communications through the functions:
763 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
764 * New function: MSG_load_platform_script()
765 to make possible using a lua script instead of XML files to set up platforms
766 * New function: MSG_set_function
767 to associate functions to processes, used when bypassing the parser
768 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
770 Platforms: Add some more examples in examples/platforms
771 * Grid'5000: see www.grid5000.fr
772 * *_30000_hosts.xml: various huge files [mainly scalability testing]
775 * Change the XML format. This is a very important modification. SimGrid
776 3.5 introduces a new hierarchical format based on the notion of
777 Autonomous Systems. Compatibility with old format is ensured
778 through the perl script provided in the install bin directory
779 bin/simgrid_update_xml.
780 It is now possible to build platforms with specific routing
781 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
782 connect several platforms together. We will try to provide soon
783 set of realistic platforms exploiting these properties (have a
784 look at examples/platforms/ for the moment).
785 * Take the opportunity of the XML format change to be a good XML citizen:
786 rename link:ctn to link_ctn and similar changes (also dealt with by
788 * Add a new routing scheme (rule-based) using regular expressions. It
789 enables to have an extremely low memory footprint when the
790 underlying routing is simple and can be compactly described. You
791 need to have libpcre4-dev (perl regular expressions) installed if
792 you want to use this routing scheme.
793 * Revive the cluster TAG and allow to easily and efficiently (both in
794 term of memory and speed) connect clusters together. Have a look
795 at teshsuite/simdag/platforms/ to see how this can be done. With
796 this tag, you can create clusters with thousands of tasks at no
797 cost (have a look at examples/platforms/).
798 Note: clusters are implemented as ASes, so there is no need for an
799 enclosing AS tag if you have only one cluster in your platform.
800 * Add new generic functions in the public interface that allows the user
801 to call SURF 'create_resource' methods from your code (same
802 functionality as the XML bypass mechanism but with a much lighter
804 * Add a new model (enabled through command line --cfg=network/model:SMPI)
805 that uses a piecewise linear approximation to produce better
806 results when exchanging small messages.
807 * Add a new parameter to handle correctly full duplex link and account
808 for interferences between uplink and downlink communications
809 (activate with --cfg=fullduplex:1).
812 * Rename the SD_READY (all dependencies are satisfied and task is
813 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
814 dependencies are satisfied) state.
815 This prevents a confusion between the notion of "ready to schedule"
816 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
817 simulated" (SD_RUNNABLE) used by the simulation kernel.
818 * Change the way a task is considered as ready. Instead of removing
819 dependencies when a task is done, a counter is decreased. This way,
820 it is always possible to reach ancestors thanks to the
821 SD_taks_get_parents function (even after the end of the simulation.)
822 * Change the return type of SD_Simulate from (SD_task_t*) into
823 xbt_dynar_t. This function was in handling a dynar internally and
824 converted it into a NULL terminated array for historical reasons.
825 * New function SD_dotload(char*) to load a DAG described in dot
826 format. This loader and the corresponding examples require the
827 installation of the graphviz library.
828 * Fix a bug in the management of tasks of size 0 in the surf network
829 models. This problem was only visible with SIMDAG and you should
830 thus disregard results produced with earlier versions if you
831 relied on this feature (some tasks were blocked because of this).
832 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
833 even though your applications were doing only point-to-point
834 communications and sequential computations. Now you can really use any
835 model you want (of course, if you create real parallel tasks, which are
836 not implemented in most models beside ptaskL07, this will abort).
837 * Add an example that schedules a DAX on an heterogeneous platform
838 using a Min-Min strategy.
839 * New function SD_workstation_get_current_task() that returns the kind
840 of task currently running on a workstation in the sequential access
842 * Raise some warnings when unexecuted tasks remains at the end of the
843 simulation. This is usually caused by cycles in the DAG.
846 * New function: SIMIX_process_set_function() called by MSG_set_function
847 * Change the underlying waiting queue in semaphores so that a process
848 can wait on several of them simultaneously (as in waitany).
849 * Fix the way to handle tokens in semaphores so that all access patterns
850 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
851 * kill the dirty pimple SIMIX_message_sizes_output()
852 Please use (proper) visualization instead
855 * New data container: setset (set of sets of elements)
856 * New module: mmalloc (mapped malloc, allowing to have several
857 independent segments of malloc)
858 * New function: xbt_dict_cursor_set_data()
859 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
860 * New function: xbt_dynar_is_empty()
861 * New function: xbt_fifo_get_last_item()
862 * Fix xbt_dynar_shrink(): use the right element size.
863 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
864 uninitialized areas during expand.
865 * Fix semaphores: previous implementation was severely broken.
866 * Use library init/fini functions for our initialization.
867 - you can use logs and other feature as soon as you want in your
868 code (even before the xbt_init / MSG_init)
869 - xbt_exit is now a no-op and produce a warning when used.
872 * Port GRAS to new SIMIX mechanisms. This allows gras users to
873 benefit from the latest improvement to the simulation kernel.
874 * Kill measurement sockets for now. If you rely on them, sorry. This
875 release is not for you. This feature will be reintroduced in the
876 future, but we cannot delay the release any further.
877 * New function: gras_msgtype_get_name().
878 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
879 * Fix (at last) the pmm example: it should not randomly fail anymore.
881 Build chain: bug fixes and overall polishing
882 * Cmake is now stable enough. Hence, we killed the autotools.
883 * Port to windows ( TM :)
884 * Fix the 'make install' target.
885 No need to use 'make install-simgrid' anymore
886 * Introduce a 'make dist' target compiling a *source* archive
887 'make package' compiles a binary archive
888 * Compile java files only on need
889 * Add --cd and --setenv command line options to tesh
890 * Out of source builds are not fully supported yet, but we are close
891 * Enable supernovae and optimization flags by default for our users
894 * Add layer to set up environment directly from lua, without XML.
895 * The effect of gras_stub_generator can be achieved through
896 lua too (check examples/gras/console/ping_generator.lua)
898 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
900 SimGrid (3.4.1) stable; urgency=low
902 The "Polishing easter eggs is probably a good idea" release.
903 This is a bug fixes release only.
907 * Fix a bug preventing the tasks from begin garbage collected.
910 * Fix a bug occuring when a host involved in a communication fails.
911 This was not detected properly by the other peer involved in the
912 communication. Now, it's reported as a network error.
915 * Warn the user about loop dependencies in data flow of DAX files
916 * Obey the control-flow dependencies of DAX files
919 * Add option "enable_smpi" allowing to not compile SMPI.
920 Probably useful for the (Mac) users experiencing a build error here
921 * Improve the detection of lua5.1 and ruby1.8
923 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
925 SimGrid (3.4) stable; urgency=low
927 The "Easter in Cargese" release. Also known as (major changes):
929 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
930 ~> bindings were greatly improved
931 ~> new tracing infrastructure for better visualization introduced
933 * the "Welcome to configury modernity" release.
934 ~> we switched from autotools to cmake, and improved our cdash
937 A more detailled list of changes follow (full detail in svn log).
939 Java Bindings: Various Cleanups
940 * (install java-gcj-compat-dev on debian-like to use them)
941 * Remove put/get: no need to export deprecated interface in Java
942 Use send/receive instead.
943 * Cleanup the examples and add a README per directory
944 * Remove example autoDestination (that's the only way to go now)
945 * Remove example explicitDestination (was a plain copy of basic)
946 * Make JniException a runtime exception, so that there is no need to
947 declare the fact that you may encounter such a beast. I guess that
948 nobody will ever want to survive such error.
949 * Create specific errors for each MSG case of failure:
950 host failure, transfer failure, timeout, task cancelled
951 * Cleanup the exceptions that may get thrown by each function
952 * Other internal cleanups in Java bindings. Performance still bad :/
953 Ruby and Lua Bindings: create them
954 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
955 * That's new and great, you should try them out.
956 Same functionalities than Java bindings, only even less polished
958 * Kill the useless "rate" argument of SD_task_get_execution_time()
959 Everyone used to provide -1 as a value, it was not used, and the
960 semantic of a possible use wasn't even clear.
961 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
962 either as comm costs or compute costs to mean that there is no
963 such thing for that specific task.
964 * Add a SD_task_set_name() function
965 * Fix SD_task_unschedule() on typed tasks
966 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
967 * In DAX loader, accept useless 'level' attributes to <job> since
968 LIGO DAGs have them (seem to be to ease graphical representation).
970 * Add an example masterslave_mailbox.c using send/receive and not
971 the deprecated put/get interface.
972 * Kill the MSG_paje_output() function. It's a noop since 2 years.
973 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
975 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
976 (declare MSG_USE_DEPRECATED to still have the old name)
977 * Add a MSG_task_set_data() function
978 * About trace replay (see examples/msg/actions):
980 - Allow to work with splitted trace files for each process
981 Give the specific trace file as argument of each process,
982 and call MSG_action_trace_run(NULL)
983 You can still have one merged file for all processes.
984 - Fix implementation of collective operations
985 * Allow task_execute() on 0-sized tasks (closes #10063)
987 * This is the first release of SimGrid where SMPI is not considered
988 beta anymore (even if some corners should still be improved)
989 * Port over the new SIMIX_network submodule (internal refactoring)
990 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
991 * Implement more missing elements of the standard:
993 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
995 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
996 MPI_Scan() MPI_Get_processor_name()
997 - Added implementation of missing case for Alltoall (warning: it's
998 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
999 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1000 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1001 SMPI_MPI_Allgatherv()
1002 * Bug fixes include:
1003 - MPI_Waitsome() was broken
1004 - Allow relative includes in smpicc
1005 - Command line cfg argument 'reference_speed' was ignored...
1006 - Some functions did not properly lead to auto-benching of user code
1007 - smpicc passes -O2 by default (just like openmpi one)
1009 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1010 * Bug fixes about timeouts during communications
1011 * add SIMIX_message_sizes_output() as a pimple to write to file the
1012 amount of messages per size. Use gnuplot to get histogram.
1013 Pimple because that's the only user-visible function of simix,
1014 defined directly in xbt.h (irk, sorry)
1016 - Add a SIMIX_sem_get_capacity() function
1017 - Fix interactions with processe resume/suspende
1018 - release_forever() was stupidly broken
1019 - Fix SIMIX_display_process_status() for processes in a semaphore
1020 - Make SIMIX_sem_block_onto() user-visible
1021 * Refactoring context stuff:
1022 - Use pseudo-OOP for better modularity
1023 - reimplement SIMIX_process_kill() without process_schedule() so
1024 that the latter can take as invariant that it is called from
1026 - Merge context_start into context_new for sake of simplicity
1028 * Add a Vivaldi network model, coded live during SUD'10 ;)
1029 * Rename configuration variables to start a hierarchy:
1030 o cpu_model -> cpu/model
1031 o network_model -> network/model
1032 o workstation_model -> workstation/model
1033 * New configuration variables:
1034 o network/bandwidth_factor: correction to bandwith
1035 o network/latency_factor: correction to latency
1036 o netwotk/weight_S: correction to the weight of competing streams
1037 * Add a long description to the models, that users can see with such
1038 argument on the command line: --cfg=cpu/model:help
1039 * --help-models display the long description of all known models
1041 * config: add the ability to set a default value after registration
1042 Does not override any previously set value (e.g. from cmd line)
1043 * dict: allow to have integer key and data.
1044 When so, you need to use the following functions
1045 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1046 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1047 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1048 In contrary to regular dicts, the key is not malloced before copy.
1049 Mixing scalar and regular elements in the same dict is not tested
1051 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1052 Tracing for Visualization:
1053 * SimGrid is now instrumented in order to generate a trace file for
1054 visualization analysis: to use it, need to compile SimGrid with the
1055 "tracing" option enabled, and instrument the program using SimGrid with
1056 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1057 (among other functions).
1058 * The instrumentation only traces the platform utilization for now
1059 * Documentation to use the tracing functions and how to analyze the
1060 traces with the Triva tool is written.
1061 * More information about: SimGrid FAQ (in the section Tracing Simulations
1064 * We moved to cmake as default build system. Autotools support will
1065 be dropped soon. Check the FAQ for more info about how to use it.
1066 * Greatly improved our cdash/ctest interactions
1067 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1068 * Added memory checking tests with valgrind; lot of memleak fixing.
1069 This may be the first release of simgrid with so few memory issues
1070 * Added code coverage tests.
1071 Our coverage is still improvable, but at least we see it on cdash.
1073 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1075 SimGrid (3.3.4) stable; urgency=low
1077 The "Desktop Grid needs love too" release (also called Xmas release).
1079 Models improvements:
1080 * Major speedup in the maxmin system solving by using lazy evaluation
1081 Instead of solving completely the maxmin system at each iteration,
1082 only invalidate (and recompute) the modified parts.
1083 This new feature is enabled in default models but you can try to
1084 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1085 * Cas01 IMproved as default CPU model
1086 This CPU model is the same Cas01 model, but it uses the
1087 maxmin-selective-update flag and a heap structure to manage
1088 actions on SURF kernel.
1089 It reduces the complexity to find the next action to finish and,
1090 consequently, it's faster than the old Cas01.
1091 This is the new default CPU model (Cas01).
1092 * Rename the old Cas01 model to Cas01_fullupdate
1093 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1094 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1095 * CpuTI (CPU Trace Integration)
1096 A new CPU model whose objective is simulate faster when using
1097 availability trace files.
1098 Instead of using a full featured, over engineered maxmin system for
1099 CPU modeling, this model does the pre-integration of traces files
1100 to calculate the amount of CPU power available, and so, executes
1101 faster than the old CPU models.
1102 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1103 * Use LV08 as default network model since it gives better accuracy
1104 for small messages and shouldn't change things for big ones.
1105 Use --cfg=network_model:CM02 to get the previous behavior.
1108 ******************************************
1109 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1110 ******************************************
1111 * The new CPU model may changes simulations!
1112 The point is that events occurring at the exact same timestamp
1113 are not scheduled in the same order with the old and new
1114 version. This may be enough to completely change the execution
1115 of simulations in some cases.
1116 * The new network model will change simulations!
1117 This new model is more realistic than the previous one, so you
1118 should consider redoing your old experiments with this model.
1119 Sorry for the inconvenience.
1122 * Introduce the supernovae compilation mode
1123 When compiled that way, the whole SimGrid (or almost) is put in a
1124 single compilation unit and compiled in one shoot.
1125 This is to help gcc which has difficulties to inline stuff from one
1127 The speedup seem to be above 15%, althrough more tests are needed on
1128 amd64 to confirm that gain.
1131 * Port of MSG's mailbox on top of SIMIX network
1132 The put/get mechanism was greatly simplified on the way.
1135 * New SIMIX network module. Provides:
1136 - Mailbox: rendez-vous mecanism to find with who you want to speak
1137 - Synchronous send/recv: easier and hopefully faster since the
1138 logic is handled in the maestro process directly now
1139 - Asynchronous send/recv: you dreamt of it? It's here now
1140 Too bad that nobody cared enough to propagate the change to MSG.
1141 * Add semaphores as SIMIX synchronization mechanism.
1144 * new function SD_daxload(char*) to load a DAX file
1145 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1146 * Introduce typed tasks. Specify its kind and cost at creation.
1147 At scheduling, just give where it should be placed, and the cost
1148 for each involved resource is automatically computed.
1149 Existing constructors so far (more to come of course):
1150 - SD_task_create_comm_e2e() for end-to-end communication
1151 - SD_task_create_comp_seq() for sequential computation
1152 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1153 * new function SD_task_dump() for debuging display
1154 * new function SD_task_dotty(task,FILE*) writing to file the info
1155 about the task in dotty format
1156 * SD_task_dependency_exists() can now cope with having one of its
1157 arguments NULL. If so, it tests whether the other argument has any
1159 * Add getters on list of preceding/following tasks:
1160 SD_task_get_parents(task) and SD_task_get_children(task)
1161 * Add getters on amount of workstations and list:
1162 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1163 * Add getter on task kind: SD_task_get_kind(task)
1164 * Update the start_time and finish_time of tasks on completion/failure
1165 * Bugfix: Remove task from state swags when destroyed
1168 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1171 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1172 dynar with one separate thread per value of the dynar.
1173 * Change the prototype of xbt_thread_create(), sorry.
1174 Added a boolean parameter indicating whether we want to join this
1175 thread (used in SG only for now)
1176 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1179 * GTNetS wrappers should now be usable again (and betterly tested too)
1180 * Fix a major regression from 3.2 where the timeout provided to
1181 MSG_task_put_with_timeout() was used as absolute time before which
1182 the comm should be done.
1183 * Start to fix the <cluster> tag.
1184 - Internal links should be good now (beside of the loopback, which
1185 use the private link instead)
1186 - paths to the external world is still rather broken
1187 - the <route:multi> tag is just broken. Actually that's brain-dead.
1188 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1189 to make it less stupid
1190 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1191 * Fix a source-level compatibility glitch from 3.2: after defining
1192 MSG_USE_DEPRECATED, you can use the old name
1193 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1194 * Allow to compile from the SVN with automake 1.11
1195 * Fix some problems when using the "start_time" tag in deployment XMLs.
1196 * Fix #8569: XBT/synchro.h has redundant declarations
1197 * Fix #8563: MSG return values and exceptions
1198 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1199 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1200 Thanks to Robson Peixoto.
1201 * Fix a vicious bug in dictionaries inducing that some elements were
1202 not freed on xbt_dict_free()
1204 Portability report of this version:
1205 * Main portability targets:
1206 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1207 on (amd64/i386/ia64)
1208 - mac leopard on i386
1209 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1210 but nothing critical.
1211 * Other platforms: windows, AIX and others were not tested for this release
1213 Timing report of this version:
1214 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1215 scenario parameters). From 8h to a few seconds in desktop grid settings.
1216 * Supernovae brings about 25% speedup on i386.
1218 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1220 SimGrid (3.3.3) stable; urgency=low
1222 The "Need for Speed" release.
1224 The timings done to validate the 3.3.2 were faulty.
1225 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1227 The problem was a conversion from a manually handled vector to
1228 xbt_dynar_t on the critical path.
1229 xbt_dynar_foreach calls functions, inducing stack management crap.
1231 We inlined these functions and xbt_dynar_foreach is now breath taking.
1232 We also inlined xbt_swag_belong on the way.
1234 Here are some approximate speedup measurements (on master/slaves
1235 simulations lasting between 10s and 20s each):
1236 3.3.1 -> 3.3.2: about same performance
1237 3.3.2 -> 3.3.3: 40% speedup
1238 3.3.1 -> 3.3.3: 40% speedup
1239 3.3.1 with inline patch -> 3.3.3: 30% speedup
1241 Our reading is that the refactoring which occurred in 3.3.2 made us
1242 suffer much more from the xbt_dynar_foreach low performance, but
1243 once we solved this, this refactoring proved to be very performance
1244 effective. From the 40% speedup, somehow, 10% are due to the
1245 inlining and 30% to the refactoring.
1247 That's a pitty that gcc cannot inline functions placed in other files
1248 alone. We have to choose between:
1249 - break the encapsulation (by putting private data structures and
1250 accessors in headers files to help gcc)
1251 - live with low performance
1252 - switch to a decent compiler such as icc (not quite possible).
1254 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1256 SimGrid (3.3.2) stable; urgency=low
1258 The "Simplicity does not preceed complexity, but follows it" release.
1260 The main contributors of this release were (lexical order):
1261 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1264 * Extract the routing logic into its own object.
1265 (was dupplicated in network.c and workstation_LV07.c;
1266 Allows to implement other ways of storing that info)
1267 => kill now useless network_card concept
1268 - Use dynar to represent routes (instead of void** + int*)
1269 - kill link_set (use surf_network_model->resource_set instead)
1270 - Add a command-line option to choose the routing schema to use
1271 - Add three new models:
1272 * Floyd (shortest path computed at initialization)
1273 * Dijikstra (shortest path recomputed all the time)
1274 * Cached Dijikstra (shortest path computed on need)
1275 All these models where contributed by Silas De Munck, and are
1276 described in his ICCS09 paper.
1278 * Simplify model declaration
1279 (less redirections, less function to write when defining a model)
1280 - Factorize stuff between models:
1283 surf_model_resource_set(model)
1284 surf_model_resource_by_name(model, name)
1285 - Unify the types of models in s_surf_model_t (using an union)
1286 - Embeed fields of common_public directly into s_surf_model_t
1287 - Rename model methods:
1288 action_free ~> action_unref
1289 action_change_state ~> action_state_set
1290 action_get_state ~> action_state_get
1291 - Change model methods into functions :
1292 (model)->common_public->action_use ~> surf_action_ref
1294 * Implement a generic resource; use it as ancestor to specific ones
1295 (allows to kill duplicated code in models)
1296 Drawback: timer command don't need no name nor properties;
1297 workstation_CLM03 don't need no properties
1298 (but I guess we can live with those few bytes wasted)
1300 * Improve the action object model
1301 - implement a constructor avoiding dupplicated code about field
1302 initialization in generic_action part.
1304 * Kill the SDP model: it has an external dependency, is deprecated
1305 in flavor of modern lmm models, and didn't compile since a while
1308 * Relocation of the context module from XBT to SIMIX.
1309 (the context were decoupled from the simix processes, duplicating a lot of code)
1310 => a lot of code was factorized
1311 - less overhead is introduced during scheduling
1312 - simpler API for the context factory
1313 - the logic for process creation,destruction and manipulation was simplified
1314 * Simplification of the s_smx_process_t data structure.
1315 => accesing the simix level data associated to a process is faster now,
1316 and the code is a lot more readable.
1319 * Implement some more MPI primitives:
1320 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1321 -implementation: Bcast: flat or 2-ary tree (default),
1322 Barrier: 4-ary tree,
1324 Allreduce: Reduce then Bcast
1325 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1326 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1327 Alltoallv: flat tree, like ompi
1329 * Add support for optimized collectives (Bcast is now binomial by default)
1330 * Port smpirun and smpicc to OS X
1333 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1334 Shout out if you used it.
1337 * Display the list of still queued messages in SG mode when existing
1341 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1342 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1343 * Add xbt_str_from_file(FILE*)
1344 * Add xbt_dict_get_key achieving a linear reverse search
1345 * Remove the context module
1347 Portability report of this version:
1348 * Main portability targets:
1349 - Linux(debian)/x86/context
1350 - Linux(debian)/x86/pthreads
1351 - Linux(debian)/amd64/context
1352 - Linux(debian)/amd64/pthreads
1353 On these, we still have the eratic breakages of gras/pmm and
1354 amok/saturate_sg reported in previous version. We still think
1355 that the tests are the cause of the fault, not the tested code.
1357 - Mac OSX Leopard/x86/context
1358 Still false negative in tesh autotesting.
1359 Smpi still fails, but this time because readlink does not accept -f
1360 Everything seems to work properly beside of that.
1363 - AIX version 5.3 (only tested contexts this time)
1364 Smpi still fails there because mktemp is not installed.
1365 Everything seems to work properly beside of that.
1367 I managed to compile it for the first time, but several breakages.
1368 Won't delay the release for this exotic platform.
1370 * Windows: it's still lagging behind. If you want to help, please
1373 Timing report of this version:
1374 This version seem to be more than 5% faster than 3.3.1 (on linux
1375 64bits with contextes). The gain is less than expected, we are
1376 investigating this for next release.
1378 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1380 SimGrid (3.3.1) stable; urgency=low
1383 * Implement a --cfg-help to show existing configuration variables
1384 * Build chain do not require doxygen in maintainer mode
1387 * fix a bug on struct sizeof computation, which prevented the
1388 exchange of arrays of structs in some conditions
1389 - added a regression test about this in datadesc_usage
1390 * Allow the exchange of 0-long dynamic vectors.
1391 - for that, use -1 as indicator of dynamic size instead of 0
1392 - This implied to change any size from unsigned long to long,
1393 reducing a bit communication abilities, but I guess that with
1394 64bits being quite common, this is more than enough.
1395 - This also induce a protocol change, thus bumping network protocol
1396 version from 0 to 1 (if we have external users, we have to get
1397 clean on that point too ;)
1398 - added two regression tests about this in datadesc_usage
1399 * Be more verbose when propagating local exceptions
1400 This helps debugging.
1401 * Display the status of simulated processes when receiving SIGINT in
1405 * Allow to control the simulation from a trace file.
1406 New functions MSG_action_register() and MSG_action_trace_run()
1407 The first one allows to associate a function execution to each
1408 kind of action while the second one parses a trace file and
1409 triggers the corresponding actions within the system.
1410 For now, only a toy example is provided in examples/msg/actions
1411 * Add an exemple of process migration in examples/msg/migration
1412 * Fix a bug in task exchange which broke MSG_task_get_sender()
1413 Add a teshsuite regression test for that.
1414 [Bug: if MSG_task_get_sender() is called after sender exit,
1416 * Fix a bug which prevented suspend/resume to work properly
1417 * Display the status of simulated processes when receiving SIGINT
1418 This fixes a regression of v3.3. due to the introduction of SIMIX
1419 * Bug fixing in failure management:
1420 - trace could not start by a failure at time 0
1421 - failure during communications were not working
1424 * Add SIMIX_process_set_name() to change the name of the current
1425 process in the log messages.
1426 * Store smx_hosts in a dict since we only retrieve them by name
1427 * Move the configuration infrastructure to surf
1430 * Move the configuration infrastructure to surf
1433 * Massive internal cleanups:
1434 - Store internal structures on processes instead of hosts (allows
1435 to have more than one process per host, in addition of being more
1437 - Cleanup the initialization/finalization process
1438 - Kill a whole bunch of unneeded synchronization:
1439 processes run in exclusive manner within the simulator
1440 - Move queues from global tables to process data fields
1442 - now accept -platform and -hostfile arguments
1443 - Pass the right rank value to processes according to the hostfile
1444 * Compile the examples by default, and use them as regression tests
1445 * Implement MPI_Wtime()
1446 * Change the reference speed to a command line option
1449 * TCP_gamma can now be specified as command line option using
1450 --cfg=TCP_gamma:10000000.0
1451 * Change the --surf-path cmd line option into --cfg=path:
1454 * Also include strbuff from xbt.h public header
1455 * xbt_ex_display(): do not free the exception after displaying
1456 This allows to do more with the given exception afterward.
1457 Users should call xbt_ex_free() themselves.
1461 Portability report of this version:
1462 * Main portability targets:
1463 - Linux(debian)/x86/context
1464 - Linux(debian)/x86/pthreads
1465 - Linux(debian)/amd64/context
1466 - Linux(debian)/amd64/pthreads
1467 These targets fail about 1/10 of times on gras/pmm, but we believe
1468 that this is because of the test, not because of simgrid.
1469 amok/saturate_sg fails even more rarely, and the test may not be
1472 - Mac OSX Leopard/x86/context
1473 The test suite still spits tons of errors because some obscure
1474 force prevents us from removing the temporary directories
1475 arguing that they still contain some metadata I've never heard of.
1476 Smpi fails because seq is not installed.
1477 Everything seems to work properly beside of that.
1480 - AIX version 5.3 (both contexts and pthread)
1481 Smpi still fails there because mktemp is not installed.
1482 XML inclusions seems rosty on AIX.
1484 * Windows: it's still lagging behind. If you want to help, please
1487 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1489 SimGrid (3.3) stable; urgency=high
1493 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1494 [Malek Cherier & Mt]
1496 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1499 Until now, GRAS were using MSG as an interface to SURF. It was
1500 quite difficult because both interface have several differences
1501 (MSG channels vs GRAS sockets were the most notable point).
1503 This also opens the gate to SMPI (which should occur soon) and speed
1504 up simulations by to 40% (even if it were not the main goal).
1506 **************************************
1507 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1508 **************************************
1509 The point is that events occuring at the exact same timestamp are
1510 not scheduled in the same order with the old and new version. This
1511 may be enough to completely change the execution of simulations in
1512 some cases. Sorry for the inconvenience.
1514 * Cleanup and upgrade the XML format to push further scalability
1515 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1517 * Improve the testing infrastructure with tesh. Now a very large part of
1518 the code is tested not only by being run but also by checking that the
1519 output match an expected output [Mt].
1521 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1522 is really less memory-demanding, which should allow you to use
1523 larger files in SimGrid [AL].
1525 * Inform valgrind about our contextes, so that it becomes usable
1526 with the default (and more effecient) version of SimGrid
1527 [contributed by Sékou Diakite, many thanks]
1530 * Introduce a listener thread in charge of receiving incoming
1531 messages from the network. It allows to overlap communication and
1532 computation but most notably, it removes some stupid deadlocks due
1533 to the fact that so far, a process could not send and receive at
1534 the same time. This made most non trivial communication schema
1536 * Convert the PIDs from long int to int to match the MSG ones (and
1537 linux ones too) [Mt]
1538 * New function: gras_agent_spawn() to launch a new process on
1539 current host. Only working in simulation for now. [Mt]
1540 * New function: gras_os_hostport() returning a constant form (ie,
1541 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1544 * Make the backtrace of exceptions more human readable [Mt]
1545 * New module: xbt/str [Mt]
1546 a ton of string utility functions (split, join, printf to a newly
1547 allocated buffer, trim, etc)
1548 * New module: xbt/hash [Mt]
1549 SHA1 hashing algorithm (more to come if needed)
1550 * New module: xbt/synchro [Mt]
1551 synchronization tools (mutex and conditions) working the same way
1552 in simulation and in real life (mainly useful for GRAS, but not
1554 * New module: xbt/queue [Mt]
1555 classical producer/consumer synchronization scheme
1556 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1557 (using the classical functions will get serialized) [Mt]
1558 * Make dictionary internal table dynamic. No need to specify its size
1559 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1561 * Make sure the log channels are organized as a tree under windows
1562 (because of ANSI C compatibility issue, any channel were child of
1563 root directly) [Mt].
1566 * Cleaned many thing in surf and fixed a few bugs [AL].
1567 * Add a nice command line configuration mechanism to compose models [AL].
1568 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1569 the previous one (KCCFLN05). It relies on something that looks like
1570 a max-min sharing mechanism but cannot be written as such. A new solver
1571 was thus designed [AL].
1572 * Add a new solver to lmm. Based on Lagrange optimization and
1573 gradient-based descent, it enables to efficiently maximise systems s.a
1575 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1577 This solver enables to propose two new network models for TCP Reno and
1578 TCP Vegas based on Low's work. These models still need to be fully
1579 tested though [Pedro Velho].
1582 * Bug fix in SD_simulate. Now the time bound given as argument is
1584 * Use the new parallel task model (ptask_L07) as default.
1585 * Use the SURF command line configuration mechanism.
1586 * 0-size tasks (for synchronization) should now work.
1588 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1590 SimGrid (3.2) stable; urgency=high
1594 We still experience issues on this platform, but we believe that at
1595 least MSG is usable.
1597 GRAS API BREAKAGE (for simplification purpose, sorry):
1598 * the gras_msgtype_by_name is not used anymore. Instead of
1599 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1600 you can write (and must)
1601 gras_msg_send(toserver, "request", &request);
1602 - If you still want to pass a gras_msgtype_t to the function (to cache
1603 the type and avoid the lookup time), use the gras_msg_send_() variant.
1604 - Impacted functions:
1605 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1606 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1607 * The callbacks are now expected to return 0 when everything went well
1608 (just like the main() function)
1610 GRAS new features and improvements:
1611 * New module mecanism where user code can use per process globals [Mt]
1612 This is similar to gras_userdata_*() functions, but for libraries. It
1613 factorize some code developped over and over in the examples and AMOK.
1614 It has still to be documented and used (only amok/peermanagement is
1616 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1617 loss when they were small enough to fit into the buffer and sent quickly
1618 enough so that they can all get received in one shoot.
1619 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1620 if not found. Use the *_or_null() variant for the old semantic.
1621 * In gras_msg_handle, do not discard messages without callback.
1622 They are probably messages to be explicitly awaited later (ie, proofs of
1623 mis-synchronization in userland since they are sent before being awaited)
1624 No big deal usually.
1625 * gras_socket_meas_send/recv: semantic changed!
1626 The numerical arguments used to be (1) the total amount of data to send
1627 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1628 messages. This was need for the fool willing to send more than MAXINT
1629 bytes on quite fat pipes.
1632 * Do really rename the hostmanagement module to peermanagement. [Mt]
1633 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1634 but this is rather new and this was documented in the module
1635 documentation (poor excuses, I admit)
1636 * Bandwidth measurement semantic changed! This follows the changes to
1637 gras_socket_meas_send/recv explained above.
1640 * A sequential mode has been added to the workstations. When a workstation
1641 is in sequential mode, it can execute only one task, and the other tasks
1642 are waiting in a FIFO. [Christophe Thiery]
1645 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1646 model for SIMDAG. [Christophe Thiery]
1647 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1648 equal to 0, which caused some bad behaviors in
1649 saturated_constraint_set_update. I now use a threshold mechanism like in
1653 * When running manually src/testall, you select specific units [Mt]
1654 testall is the result of our cunit mecanism, and should replace all
1655 the scripty thingy around since bash don't run easily on billware.
1657 * A mallocator system has been added. [Christophe Thiery]
1658 Mallocators allow you to recycle your unused objects instead of freeing them
1659 and allocating new ones.
1661 Documentation update:
1662 * FAQ reworking + New FAQs:
1663 - "Valgrind spits tons of errors!" [Mt]
1664 - "How to repport bugs" [Mt]
1665 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1666 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1667 - Communication time measurement within MSG [AL]
1668 - I experience weird communication times when I change the latency [AL]
1669 * GRAS tutorial [Mt]
1671 - an introduction to the framework and to the used communication model
1672 - an initiatic tour introducing the most proheminent features:
1674 . Lesson 0: Installing GRAS
1675 . Lesson 1: Setting up your own project
1676 o Part 2: Message passing
1677 . Lesson 2: Exchanging simple messages
1678 . Lesson 3: Passing arguments to the processes (in SG)
1679 . Lesson 4: Attaching callbacks to messages
1680 . Lesson 5: Using globals in processes
1681 . Lesson 6: Logging informations properly
1682 . Lesson 7: Using internal timers
1683 . Lesson 8: Handling errors through exceptions
1684 . Lesson 9: Exchanging simple data
1685 . Lesson 10: Remote Procedure Calling (RPC)
1686 . Lesson 11: Explicitely waiting for messages
1687 . Recapping of message passing features in GRAS
1688 - A HOWTO section containing:
1689 o HOWTO design a GRAS application
1690 More are due, of course. They will come latter. In the meanwhile, you can
1691 check the examples which are still here.
1693 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1695 SimGrid (3.1) stable; urgency=high
1699 There was a stack corruption somewhere, visible only when optimizing
1700 with these versions. [Vince]
1703 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1704 module that enabled to play with Directed Acyclic Graphs. It is built
1705 directly on top of SURF and provides an API rather close to the old
1706 SG. Some old codes using SG are currently under rewrite to check that
1707 all needful functions are provided. [Christophe Thiery]
1710 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1711 extension of the classical CLM03 model that gracefully handles
1712 failures. This is now the default model for MSG and GRAS. It doesn't
1713 handle parallel tasks yet though. [AL]
1714 * Bug fix: Weights were not correctly set in the network part.
1715 WARNING: This may have resulted in incorrect results with simulations
1716 where there are more than one flow on a given link. [AL]
1719 * After a (long ?) discussion on simgrid-devel, we have decided that the
1720 convention we had on units was stupid. That is why it has been decided
1721 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1722 WARNING : This means that all previous platform files will not work as
1723 such with this version! A warning is issued to ask users to update
1725 A conversion script can be found in the contrib module of the CVS, under
1726 the name contrib/platform_generation/surfxml_update.pl [MQ]
1729 * Bug fix: Processes were started in reverse order, wrt deployment file.
1730 WARNING: if your code relies on this bug, please fix it. [AL]
1731 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1732 being executed on two different locations. [AL]
1733 * Bug fix: Failures are now better supported thanks to Derrick's tests
1734 (there was many failure situations I hadn't thought of and that weren't
1735 correctly handled). [AL]
1736 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1740 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1741 Exception occurring on server-side are propagated back to client (!).
1743 API CHANGE: the callback changed their prototype. Change:
1744 int my_handler(gras_socket_t expeditor, void *payload_data) {
1746 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1747 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1749 * New! function: gras_msg_handleall to deal with all messages arriving
1750 within a given period.
1751 * New! function: gras_socket_server_range to get a server socket in a
1752 range of port numbers (ease to avoid port number conflicts) [MQ]
1753 * New! gras processes display their backtrace when they get a SIGUSR1
1754 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1755 Sweet to debug RL processes [MQ]
1759 - Do not force experiment sizes to be expressed in kb, or it becomes
1760 impossible to measure the latency this way (needs one byte-long tests)
1761 WARNING: this changes the amok_bw_* function semantic. [MQ]
1762 - Implements the link saturation stuff. [MQ]
1763 * Peer management module:
1764 New! module factorizing code that we wrote over and over [MQ].
1767 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1768 - Test units are placed directly into the library code, they get extracted
1769 automatically and placed into the src/testall binary.
1770 - Convert most of the XBT tests to this system.
1771 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1772 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1773 live in the same C code. This required to change a little bit the API
1774 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1775 * New module: sparse graph structure with basic algorithms (this is work
1776 in progress and the API is not considered to be frozen yet). [AL]
1777 * Display more information on backtraces: source line & function names are
1778 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1779 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1780 * Reworked a little bit some #include statements to load only required
1781 headers. Some user code that relied on SimGrid to include stdlib or
1782 stdio may need to include it by themselves. [AL]
1783 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1785 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1786 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1787 but I guess that almost nobody used those parts. [MQ]
1789 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1791 SimGrid (3.0.1) stable; urgency=low
1794 * Unfortunately, I had missed 5 misnamed functions:
1795 xbt_fifo_item_t xbt_fifo_newitem(void);
1796 void xbt_fifo_freeitem(xbt_fifo_item_t);
1797 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1798 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1799 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1800 They're now deprecated. Please use their new versions:
1801 xbt_fifo_item_t xbt_fifo_new_item(void);
1802 void xbt_fifo_free_item(xbt_fifo_item_t);
1803 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1804 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1805 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1807 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1808 * Documentation: xbt_log module unmercifully reworked [MQ]
1809 * Bugfix: there was a problem with the ending of contexts with
1810 the pthread backend. It caused some weird deadlock or behavior
1811 depending on the pthread implementation. [AL]
1812 * Bugfix: get the exceptions raised in the simulator repport where
1813 and why they come from when they are not catched in time [AL, MQ]
1816 * Bugfix: Do repport the error when two non-connected hosts try to
1817 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1820 * Add additionnal checkings on communications. Assert that two
1821 communicating hosts are connected by a set of links... [AL]
1824 * Add additionnal checkings on channel values in communication [AL]
1825 * New: MSG_task_get_source to see on which host a task was generated [HC]
1826 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1827 the number of tasks waiting to be received on channel and sent
1829 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1830 waits for the first task coming from a given host.. [AL]
1832 GRAS new functionnalities: [MQ]
1833 * Enhance the parsing macro to allow the size of multidimentional objects
1834 to be given thru annotations.
1835 * New example (and documentation): Matrix Multiplication a la RPC
1836 (as when I was young!) and fix a bunch of bugs found on the way.
1838 GRAS performance improvements: [MQ]
1840 * Reduce the amount of cbps creation/destruction by making it static to
1841 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1844 * Change libdata to a set so that we can search for stuff by ID (and thus
1845 reduce the insane amount of dict lookups)
1848 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1849 RL communication on top of SG since it's so uneffective.
1850 It may also be used for inter-thread communication in RL, one day.
1851 * Use gras_datadesc_copy() to exchange messages on top of SG
1853 - improve message exchange performance on top of SG
1854 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1855 * Don't exchange on the network the size of the used part of buffer,
1856 instead, specify the possible buffer size to read().
1858 - reduces the amount of read/write calls (one pair per exchange)
1859 - reduces the amount of exchanged data (the size)
1860 - allows to retrieve all arrived data on receiver side, if we don't need
1861 it right now (subsequent read will peek the buffer)
1862 - allows the receiver to proceed with the begining of the stream before
1863 everything is arrived
1864 - make it possible to build an iov transport (using readv/writev)
1866 - take care of the data with non-stable storage (like stacked data),
1868 * If possible, TCP send uses vector I/O (when writev() is here)
1869 - Don't use it for receive since we send data sizes and data on the
1870 same stream, so we wouldn't be able to chain large amount of chunks
1871 before having to flush the stuff to read the size.
1872 * Rework the transport plugin mecanism to simplify it and reduce the
1873 amount of pointer dereferencement when searching for the right function
1876 * I guess that now, we do almost as few system calls as possible while
1877 doing as few data copy as possible.
1879 To improve it further, we could try to send all the sizes first and then
1880 all the data (to use iov on receiving size), but it's only a partial
1881 solution: when you have 2 dimensional data, the sizes of the second
1882 dimension is data of the first dimension, so you need 3 streams.
1884 I'm not sure the potential performance gains justify the coding burden.
1886 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1888 SimGrid (3.00) stable; urgency=high
1891 * New! Give the possibility to hijack the surf parser and thus bypass
1892 MSG_create_environment and MSG_launch_application. Have a look at
1893 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1895 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1897 SimGrid (2.96) unstable; urgency=low
1902 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1903 This deprecates the xbt_error_t mecanisms.
1904 It modifies (simplifies) all XBT and GRAS API.
1905 MSG API keeps unchanged (exceptions raised by XBT are catched from
1906 within MSG and masked with existing error handling facilities)
1909 * New! Add a FATPIPE model. [AL]
1910 * New! Add a parallel task model. [AL]
1911 * New! Add automatically a loopback interface (in the default
1912 network model) if none was precised.
1915 * Bugfix: MSG_process_resume now works with the current running process.
1917 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1918 * Modification of MSG_task_get_compute_duration. Once a task has been
1919 processed, the value returned by this function is now equal to 0. [AL]
1920 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1921 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1922 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1923 of a m_task. For now, MSG_task_cancel only works with computation
1925 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1926 of the processor (in Mflop/s) regardless of the current load on the
1928 * API Change: use proper naming convention for MSG_getClock and
1929 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1931 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1932 This function changes the priority of a computation task. This priority
1933 doesn't affect the transfer rate. A priority of 2 will make a task
1934 receive two times more cpu power than the other ones. This function
1935 has been added to suit the needs of Nguyen The Loc and hasn't been that
1936 much tested yet. So if it fails, please report it and send me your code.
1938 * API Change: removed all functions and types that were marked "deprecated"
1939 since many months. Renamed MSG_global_init_args to MSG_global_init.
1941 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1943 SimGrid (2.95) unstable; urgency=low
1946 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1947 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1948 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1951 * Bugfix: Allow absolute paths for platform description files [MQ]
1952 * Bugfix: do free the variables after use. Leads to drastic performance
1954 * Implement max_duration (ie, timeouts) on resources [AL]
1957 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1959 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1960 the needed support (timeouts) [AL]
1963 * Implement measurement sockets. You can now get the bandwidth between two
1964 hosts thanks to AMOK (see below). [MQ]
1965 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1966 dynar over the network (yeah) [MQ]
1967 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1968 * simple/static token-ring example. [Alexandre Colucci and MQ]
1969 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1970 in place before [MQ]
1973 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1974 type parsing. "short" and "long" modifiers were also ignored; other
1975 modifier (such as reference level) are still ignored. [MQ]
1976 * Embeed the buffer size within the buffer itself on SG. [MQ]
1977 That way, send() are atomic and cannot get intermixed anymore (at least
1978 the ones which are less than 100k; bigger messages still have the issue)
1979 * Array size pushed by the field, not by the field type (or each
1980 and every long int will push stuff to the cbps) [MQ]
1981 * use select() to sleep since it allows to portably sleep less than one
1984 GRAS (minor cleanups)
1985 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1988 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1989 * type category 'ignored' killed as it was never used and were difficult
1991 * whether a type can cycle or not is now a flag, leaving room for more
1992 flags (as "ignored", if we feel the need one day ;)
1993 * Rename raw sockets to measurement sockets since "raw" has another
1994 meaning in networking community.
1997 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1998 and offers features not belonging to GRAS but that most applications
1999 need. One day, it may be a set of plugins loadable at runtime.
2000 * New module: bandwidth
2001 bandwidth measurement between arbitrary nodes running this module. [MQ]
2003 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2005 SimGrid (2.94) unstable; urgency=low
2007 The first beta release of SimGrid 3 !
2011 * Update the main page and the FAQ. Adding references to gforge.
2014 * Add a gras_os_getpid function.
2017 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2018 * Extend the logs so that they also print PID, hostname, date, ... if
2020 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2021 and kill the old version which were in testsuite/
2022 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2026 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2027 for users. Instead, each programming environment (GRAS, MSG,...) use it
2028 when needed to provide such a feature to users.
2029 Don't shortcut the mecanism or you will also shortcut the virtualization
2030 you need on the simulator.
2034 * Cleanups in configury with regard to compile optimization/warning flags.
2035 Also add -fno-loop-optimize to any powerpc since it's the optimization
2036 killing gcc (< 3.4.0).
2037 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2038 needed by MSG examples complications
2039 * Borrow configury beautifications from PHP
2042 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2043 warning (thanks loris for stumbling into this one).
2044 * Bugfix: stop loading private headers (gras_config.h) from the public
2048 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2049 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2050 RL would imply a DNS resolver, which is *hard* to do in a portable way
2051 (and therefore delayed).
2052 * Implement a real timer mecanism and use it in timing macros. This allows
2053 to avoid rounding errors and get a 0.000005 sec precision in timing
2054 macros. While I was at it, various cleanups:
2055 - allow to declare more than one timed section per file (fix a stupid bug)
2056 - move some private declaration to the right place
2057 - merge conditional execution and timing macros into emulation module
2058 - document the module
2059 - make sure the module cleanups its mess on gras_exit
2060 * Documentation improvements:
2061 - (new) how to compile applications using GRAS
2062 - (new) emulation support (timing macros)
2064 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2066 SimGrid (2.93) unstable; urgency=low
2068 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2071 - Use Paje properly where used. Still to be sanitized properly.
2072 - Portability fix: Add an implementation of the contexts using pthread
2076 - Add xbt_procname(): returns the name of the current process.
2077 Use it to show the current process's name in all logging.
2079 - fix detection of older flex version and the reaction, since we do
2080 depend on modern ones (we use lex_destroy)
2081 - Better separation of SG and RL in the libs: remove all simulation code
2082 from libgras. As a result, this lib is now only 200k when stripped.
2083 Some of the xbt modules may also be duplicated (two sets and such) and
2084 should be cleaned/killed before SG3.
2085 - Insist on using xlC on AIX because of weird problems involving gcc there.
2086 - Cleanup the make remote stuff. This is now done by scripts
2087 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2088 mainly for our private use, but we're working on changing them to user
2091 - Bugfix: flush the socket on close only if there is some *output*.
2092 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2094 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2097 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2099 SimGrid (2.92) unstable; urgency=low
2101 Alpha 3 on the path to SimGrid 3
2105 - New! First try of benchmarking macros.
2106 - New! First try so that gras_stub_generator generate deployment and
2107 remote compilation helpers.
2109 - Bugfix: Initialization fix in msg_test.
2113 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2115 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2117 - API Change: gras_os_sleep to take the amount of seconds as a double.
2118 Accepting an int was error prone since it was the only location where
2119 seconds were coded as such. It leaded to damn rounding errors.
2120 - Bugfix: Hard to belive that timers ever worked before this.
2122 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2124 SimGrid (2.91) unstable; urgency=low
2126 Alpha 2 on the path to SimGrid 3
2130 - Bug fix in the lmm_solver.
2132 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2133 through the function MSG_paje_output.
2134 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2135 - It is possible to bound the rate of a communication in MSG with
2136 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2137 to write it in the changelog).
2138 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2143 - Various cleanups to the autotools stuff
2144 - Begin to move Gras examples to examples/gras/
2145 - Let make distcheck work again (yeah!)
2147 - documentation overhauled using doxygen.
2148 gtk-doc-tools is dead in SimGrid now.
2149 - Automatically extract all existing logging categories, and add the list
2150 to the documentation (long standing one, to say the less)
2152 - Cleanup the known architecture table. Reorder the entries to group what
2153 should be, and use a more consistent naming scheme.
2154 (some of the test dataset are still to be regenerated)
2155 - New! Allow library to register globals on each process just as userdata
2157 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2158 lookup time (for now).
2159 Use it in msg and trp.
2160 This cleans a lot the internals and helps enforcing privacy of the
2161 headers between the gras components.
2162 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2163 - Bugfix: gras_os_time was delirious in RL.
2164 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2166 - Reenable GRAS now that it works.
2168 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2170 SimGrid (2.90) unstable; urgency=low
2172 Alpha 1 on the path to SimGrid 3
2174 * It is a long time since the last release of SimGrid. I'm sorry about
2175 that but as I had told you, I was rewriting a lot of things. I apologize
2176 to those who had been reporting bugs to me and that I had not answered.
2177 If your bug is still in the new version, please tell me. Here is a
2178 summary of the main changes.
2180 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2181 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2182 portability, speed, and a lot more but you'll figure it out later.
2183 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2184 I wanted to release the new MSG as soon as possible and I have
2185 broken GRAS, which is the reason why, for now, only MSG is fully
2186 functional. A laconic description of these projects is available
2187 in the documentation.
2189 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2190 kernel optimized for our needs (called SURF but only the developers
2191 should use it). Hence, MSG is now roughly 30 times faster and I think
2192 that by rewriting a little bit MSG, I could event speed it up a little
2193 bit more. Beside the gain in speed, it is also much easier to encode a
2194 new platform model with SURF than it was with SG. More to come...
2196 * REVOLUTION 3: I have tried to change a little as possible the API of
2197 MSG but a few things really had to disappear. The main differences
2198 with the previous version are :
2199 1) no more m_links_t and the corresponding functions. Platforms are
2200 directly read from a XML description and cannot be hard-coded
2201 anymore. The same format is used for application deployment
2202 description. The new format is described in the documentation.
2203 Have a look in tools/platform_generation. There is a tiny script
2204 that converts from the old platform format to the new one. Concerning
2205 the application deployment format, parsing the old one is tricky.
2206 I think most of you should however be able to convert your files. If
2207 it is really an issue, I can write a C code that does the conversion.
2209 2) the toolbox tbx does not exist anymore. We now have a library
2210 with much more data-structures but without the hash-tables (we have
2211 dictionaries that are much faster).
2213 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2215 *****************************************************************************
2216 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2217 * don't feel like converting the oldies. (Mt) *
2218 *****************************************************************************
2221 Version 2.90: "the long awaited one"
2222 - Finished rewriting and debugging MSG. Rewrote the documentation.
2223 - disable GRAS for now since it needs to be ported to the newest SG
2226 - Finish the port to windows (using mingw32 for cross-compile)
2229 - Main loop and datastructures of SURF. A cpu resource object is
2230 functional. Surf can thus be used to create cpu's with variable
2231 performance on which you can execute some actions.
2233 2004-11-15 Martin Quinson
2234 - Port to ARM. Simply added the alignment and size descriptions. Should
2235 work, but the ARM machines are so slow that I didn't had the opportunity
2236 to 'make check' over there yet.
2238 2004-11-15 Arnaud Legrand
2239 - Trace manager now written. It uses a heap structure and is therefore
2240 expected to be efficient. It may however be speeded up (particularly
2241 when many events occur at the same date) by using red and black
2242 trees. One day maybe...
2243 - Max-min linear system solver written. It uses a sparse matrix
2244 structure taking advantage of its expected use. Most operations are
2245 O(1) and free/calloc are called as few as possible. The computation of
2246 the minimum could however be improved by using a red and black tree
2249 2004-11-03 Arnaud Legrand
2250 - Rename every gras_* function that was in xbt/ to its xbt_
2252 - Add a heap and a doubly-linked list to xbt
2253 - Added a dichotomy to the dictionaries. make check works as well before
2254 so I assume that the patch is correct. I do not know however if things
2255 run effectively faster than before now. :)
2257 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2258 SimGrid, and the version number is bumped to 2.x
2260 2004-10-29 Martin Quinson
2261 - Introduction of the remote errors.
2262 They are the result of a RMI/RPC on the remote machine.
2263 ErrCodes being scalar values, you can't get the host on which those
2264 errors did happen. Extending the error mechanism as in Gnome is possible.
2265 No idea yet whether it is a good idea.
2267 2004-10-28 Martin Quinson
2268 - Interface revolution: the Starred Structure Eradication.
2269 I used to do typedef struct {} toto_t; and then handle *toto_t.
2270 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2271 - ???_t is a valid type (builded with typedef)
2272 - s_toto_t is a structure (access to fields with .)
2273 - s_toto is a structure needing 'struct' keyword to be used
2274 - e_toto_t is an enum
2275 - toto_t is an 'object' (struct*)
2277 typedef struct s_toto {} s_toto_t, *toto_t;
2278 typedef enum {} e_toto_t;
2279 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2280 s_toto_t) is private.
2282 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2283 so that it can set the variable to NULL. It was so for dicts and sets,
2284 it changed for dynars.
2286 - Fix a bunch of memleaks in dict_remove
2287 - Fix a bug in sg/server_socket opening: it failed all the time.
2289 2004-10-07 Martin Quinson
2290 - Speed up dynar lookup operation a bit.
2292 gras_dynar_get is dead.
2294 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2295 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2296 (which gives you the address of the stored data).
2298 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2299 copy of the data using an affectation to do the job and not a memcpy.
2301 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2302 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2304 It does not really speedup the dynar test because they are
2305 setting elements all the time (and look them seldom). But the dict does
2306 far more lookup than setting.
2308 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2310 2004-10-05 Martin Quinson
2311 - Allow to (en/dis)able the cycle detection at run time.
2313 Whether we should check for cycle or not is now a property of each
2314 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2315 datadesc_cycle_unset allow to remove this property when previously set.
2317 Note that the cycle detection is off by default since it impacts the
2318 performance. Watch the data you feed GRAS with ;)
2320 This property is hereditary. Any element embedded in a structure having it
2321 set have it set for the time of this data exchange.
2323 You should set it both on sender and receiver side. If you don't set it on
2324 sender side, it will enter an endless loop. If you forget on receiver
2325 side, the cycles won't be recreated after communication.
2327 - Header reorganization.
2328 Kill gras_private.h, each submodule must load the headers it needs.
2330 2004-10-04 Martin Quinson
2331 - Interface revolution: do not try to survive to malloc failure.
2333 Now, gras_malloc and friends call gras_abort() on failure.
2334 As a conclusion, malloc_error is not a valid error anymore, and all
2335 functions for which it was the only gras_error_t return value are
2336 changed. They now return void, or there result directly.
2337 This simplify the API a lot.
2339 2004-09-29 Martin Quinson
2340 - Re-enable raw sockets.
2341 Created by gras_socket_{client,server}_ext;
2342 Used with gras_raw_{send,recv}
2345 It should allow to kill the last bits of gras first version soon.
2347 This is not completely satisfactory yet (duplicate code with
2348 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2351 - Simplify transport plugin (internal) interface by not passing any
2352 argument to _server and _client, but embedding them in the socket
2355 2004-09-28 Martin Quinson
2356 - Finish the port to AIX.
2357 autoconf was my problem (segfault within the malloc replacement
2358 function. No idea why)
2360 2004-09-16 Martin Quinson
2361 - Fix some size_t madness on 64bit architectures.
2363 2004-09-08 Martin Quinson
2364 - Reduce the number of system headers loaded, overload some more system
2365 calls (such as malloc to cast the result of the system one, and work
2367 - Fix and reintroduce the config support
2369 2004-09-07 Martin Quinson
2370 - Source code reorganization to allow Arnaud to surf all over there.
2371 - Allow to document the logging categories.
2372 - Remove all uppercase from logging categories and useless cleanup in names.
2374 2004-08-18 Martin Quinson
2375 Version 0.6.2 (protocol not changed; API changed)
2376 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2377 gras_error_t), and NULL when not found. Functions expecting a msgtype
2378 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2379 hopefully usefull message.
2380 - Portability to prehistoric sparcs again
2382 2004-08-17 Martin Quinson
2383 Version 0.6.1 (protocol not changed; ABI not changed)
2384 - prealloc some buffers to speed things up
2386 2004-08-11 Martin Quinson
2387 Version 0.6 (protocol not changed; ABI expended)
2388 - The parsing macro can deal with the references, provided that you add
2389 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2391 2004-08-09 Martin Quinson
2392 Version 0.5 (protocol not changed; ABI changed)
2393 - Allow to off turn the cycle detection code in data exchange at
2394 compilation time. It should be at run time, but I'm short of time (and
2395 the config stuff is still broken). That way, we keep dict out of the
2396 critical path, which is good because the performance is poor:
2397 - search not dichotomial yet
2398 - dynar give no way to access their content and memcpy everytime
2399 - In composed data description (struct, ref and so on), stop foolness of
2400 keeping the subtype's ID, but store the type itself. This keeps sets out
2401 of the critical path, which is good since they rely on dynar and
2402 dictionnaries. The only loose of that is that we cannot detect the
2403 redeclaration of a structure/union with another content (but I'm not sure
2404 the code detected well this error before anyway). We still can detect
2405 the redefinition discrepancy for the other types.
2406 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2407 it breaks the code because of type-punning used all over the place).
2408 This breaks on all non-gcc architectures (for now).
2410 All those changes (plus the buffer of last time) allow me to gain 2 order
2411 of magnitude on cruel tests consisting of 800000 array of integers on two
2412 level of a hierarchical structure (200 secondes -> 4 secondes)
2415 - the selector of reference must now return the type it points to, not
2416 the ID of this type.
2418 2004-08-06 Martin Quinson
2419 Version 0.4 (protocol changed; ABI not changed)
2420 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2421 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2422 - Implement a buffer transport, and use it by default (it relies on tcp in
2423 real life and on sg in simulation).
2424 That's a bit hackish since I had a new field to the structure to store
2425 its data without interfering with the subtype ones. Inheritance
2426 is tricky in C. And that's a kind of reverse inheritance with one class
2427 derivating two classes. Or maybe a game with java interfaces. Anyway,
2428 that's damn hard in C (at least).
2429 Moreover, I got tired while trying to ensure plugin separation and
2430 genericity in SG mode. MSG wants me to do weird things, so let's go for
2431 cruel hacks (temporarily of course ;).
2432 See comment in transport_private.h:71
2433 - do not include all the _interface headers in private but in the files
2434 which really need them (to cut the compilation time when they are
2437 2004-07-26 Martin Quinson
2438 Version 0.3 (protocol not changed; ABI changed)
2439 - Major overhault of the datadesc interface to simplify it:
2440 - shorted the function names:
2441 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2442 - add a trivial way to push/pop integers into the cbps without malloc.
2443 This allows to make really generic sub_type description, which simply
2444 pop their size of the stack.
2445 - add a function gras_datadesc_ref_pop_arr() which does what users want
2446 most of the time: Declare a dynamic array (which pops its size of the
2447 stack) and declare a reference to it. Poor name, but anyway.
2448 - kill the post-send callback, add a post-receive one
2450 2004-07-23 Martin Quinson
2451 Version 0.2 (protocol changed; ABI changed)
2452 - add some testing for cpbs in the test cases, and fix some more bugs.
2453 This invalidate again the little64 data file, since I cannot regenerate
2455 - remove an awfull optimization in the logging stuff, allowing me to:
2456 - understand it again
2457 - learn gcc how to check that the argument match the provided format
2458 - fix all errors revealed by gcc after that
2459 - internal keys of dict are not \0 terminated. Deal with it properly in
2460 loggings instead of segfaulting when the user want to see the logs :-/
2462 2004-07-22 Martin Quinson
2463 - Fix some stupid bug preventing cbps (callback postit) from working
2465 2004-07-21 Martin Quinson
2466 - Some documentation cleanups
2467 - remove the useless last argument of msgtype_declare
2468 - rename the Virtu functions to fit into the 'os' namespace
2469 - move headers src/include -> src/include/gras/ and stop fooling with
2470 gras -> . symbolic link
2471 - make distcheck is now successful
2473 2004-07-19 Martin Quinson
2475 - Build shared library also
2476 - Install html doc to the right location
2477 - stop removing maintainer files in make clean
2478 - build tests only on make check
2480 2004-07-13 Martin Quinson
2482 - No major issue in previous version => change versionning schema
2483 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2484 corresponding dataset.
2486 2004-07-11 Martin Quinson
2488 - Get it working with any kind of structure (we can compute the padding
2489 bytes remotely for all the architectures I have access to)
2490 - Implement the structure parsing macro (still not quite robust/complete)
2491 - Improvement to the remote testing toysuite
2493 2004-07-10 Martin Quinson
2494 [autoconf mechanism]
2495 - get ride of a bunch of deprecated macros
2496 - actually run the test for two-compliment, not only compile it :-/
2497 - test whether the structures get packed (and bail out if yes. Damn.
2498 Alignment is a serious matter)
2499 - test whether the structures get compacted (but respecting the alignment
2500 constraints of each types)
2501 - test whether the array fields of structures can straddle alignment boundaries
2503 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2504 (took me 2 hours to find that bug, looking at the wrong place)
2505 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2506 before sending/receiving and are used to compute the offsets of fields
2507 - Given that padding size depend even on compiler options, keep track of
2508 alignment and aligned_size only for the current architecture. Not a big
2509 deal since we send structure fields one after the other (seems
2511 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2512 program, let it work on linux/gcc/little32. portability todo.
2514 2004-07-08 Martin Quinson
2515 - import and improve remote compilation support from FAST
2516 - make sure make check works on half a dozen of machines out there
2518 2004-07-07 Martin Quinson
2519 Let's say it's version 0.0.3 ;)
2520 - Implement conversions (yuhu!)
2521 - Let it work on solaris (beside conversion, of course)
2522 - Stupid me, using rand() to generate the conversion datatests in not wise.
2524 2004-07-06 Martin Quinson
2525 - Let make dist work, since I'm gonna need it to compile on remote hosts
2526 - Let Tests/datadesc_usage write the architecture on which the file was
2527 generated as first byte.
2528 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2529 architecture descriptions.
2530 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2531 on those architectures.
2532 - Optimization: send/recv array of scalar in one shoot
2534 2004-07-05 Martin Quinson
2535 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2537 - Plug a whole bunch of memleaks
2538 - each process now have to call gras_{init,exit}. One day, their log
2539 settings will be separated
2540 - Continue the code factorisation between SG, RL and common in Transport.
2542 2004-07-04 Martin Quinson
2544 - Redistribution between SG and RL.
2545 We wanna have to accept in SG, so move accepted related parts of RL in
2546 the common part. (more precisely, the dynar of all known sockets is no
2547 more a static in transport.c, but part of the process_data)
2549 [gras_stub_generator]
2550 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2552 2004-07-03 Martin Quinson
2553 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2554 them all up in one shot)
2556 - Fix the ugly case of reference to dynamic array.
2557 - New (semi-public) function gras_datadesc_size to allow the messaging
2558 layer to malloc the needed space for the buffer.
2560 - gras_socket_close now expect the socket to close (and not its address to
2561 put NULL in it after it). This is because the socket passed to handlers
2562 is one of their argument (=> not writable).
2564 - propagate the interface cleanup from last week in datadesc, ie remove a
2565 superfluous level of indirection. User pass adress of variable
2566 containing data (both when sending and receiving), and not of a variable
2567 being a pointer to the data. Let's say that I like it better ;)
2568 The price for that is constructs like "int msg=*(int*)payload" in
2569 handlers, but it's a fine price, IMHO.
2571 - Let it work in RL (yuhu)
2573 2004-06-21 Martin Quinson
2575 - porting SG plugin and SG select to new standards (works almost).
2576 - plug memleaks and fix bugs around.
2579 - cleanup the prototype of data recv and force users to specify when they
2580 want to handle references to objects. Test case working even for cycles.
2581 - plug memleaks. Valgrind is perfectly ok with this.
2583 2004-06-12 Martin Quinson
2585 - cleanup the separation between plugin and main code in plugin creation
2587 2004-06-11 Martin Quinson
2589 - Reput hook for raw sockets, needed for BW experiments
2590 - kill a few lines of dead code
2591 [Data description] Interface cleanup
2592 - gras_datadesc_by_name returns the searched type or NULL.
2593 That way, no variable is needed to use a type desc once, which makes
2595 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2596 two parameters were strings (field name, type name), leading to
2598 [Dicos] Interface cleanup
2599 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2600 This is consistant with the dynar API.
2602 2004-04-21 Martin Quinson
2604 - Porting to new standards.
2606 - interface cleanup.
2607 There is no bag anymore, no need to take extra provision to mask the
2608 pointers behind "ID".
2609 Better splitup of functions between files create/exchange/convert.
2610 This is still a bit artificial since convert and receive are so
2611 interleaved, but anyway.
2613 - add a queued message list to procdata (the ones not matching criteria
2615 - factorize some more code between SG and RL wrt procdata
2617 - use gras_exit in example to track memleaks
2618 - get rid of gs_example now that GS is properly integrated into gras
2619 - update run_test to integrate the latest tests (datadesc)
2621 - rename WARNINGn macros to WARNn since it prooved error-prone
2623 2004-04-19 Martin Quinson
2625 - register init/exit functions within gras module mechanism
2626 - send/receive function.
2627 Convertion is not implemented, but short-cutted if not needed.
2628 struct/array elements are sent one by one (instead of block-wise), but
2629 nobody really cares (yet). Get a prototype before optimizing.
2630 - tests (using a file socket) for DD send/receive on:
2631 - base types: int, float
2632 - array: fixed size, string (ie ref to dynamic string)
2633 - structure: homogeneous, heterogeneous
2634 - chained list, graph with cycle
2635 Believe it or not, valgrind is not too unhappy with the results. The
2636 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2640 - Bugfix when using a filename explicitely (instead of '-')
2642 2004-04-09 Martin Quinson
2644 - factorize more code between RL and SG in socket creation
2645 - Complete the implementation and tests of:
2647 o file (only in RL, and mainly for debugging)
2649 I lost 3 days to design a portable address resolver, and then decided
2650 that the prototype mainly have to run on my box.
2651 Addressing portability too early may be like optimizing too early :-/
2653 - use gras_init in the Tests instead of the crappy parse_log_opt
2654 (the latter function is removed)
2655 [Conditional execution]
2656 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2657 [Code reorganisation]
2658 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2659 Build examples against the RL library, since there is no way to disable
2660 its creation for now.
2662 For information, the beginning of coding on GRAS was back in june
2663 2003. I guess that every line has been rewritten at least twice since