1 SimGrid (3.10-rc1) unstable; urgency=low
4 * Reintegrate Java to the main archive as desynchronizing these
5 package is not acceptable anymore (Java is now considered stable)
6 * Add explicit synchronization facilities through semaphores
7 * Bug fix: Task.setDataSize() only changed the C world, not the value
8 cached in the Java world
11 * Dramatically change the way files are handled. API and internals changed, but
12 this part of MSG was not considered as production grade either.
13 * Add explicit synchronization facilities through semaphores
14 * Add a new function MSG_host_get_process_list()
15 * Preliminary DVFS support (see examples/msg/energy/ for details)
18 * SMPI is now included directly in the libsimgrid as the windows
19 linker doesn't force us on splitting it anymore.
20 * Improvements of the SMPI replay tool:
21 - Most of the collective communications are now rooted in the same process as
22 in the original application.
23 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
24 used until now). Multiple data types can now be used in a trace.
25 - The replay tool now supports traces produce either by TAU or a modified
27 - Bug Fix: the compute part of the reduce action is now taken into account.
28 * smpirun generates the host file if needed (with given host count and platform)
29 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
30 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
31 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
32 --cfg=smpi/coll_selector:(mpich/ompi)
33 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
35 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
36 happening inside a collective SMPI call.
37 * Fix the behavior of complex data types handling.
38 * Make MPI_Wtime another synchronization point to take computations into account.
39 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
40 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
41 * Add all missing Fortran bindings, SMPI should work with Fortran 90
42 (no privatization of global variables yet)
45 * Verification of liveness properties is now available for SMPI applications
46 (in addition to MSG applications)
47 * Bugged examples using SMPI in examples/smpi/mc/
48 * Add --cfg=model-check/visited option. Allows the verification of infinite
49 programs. Detection of loops in the execution thanks to the system state
50 comparison and reduction of the state space to explore. Can be combined with
51 DPOR for safety properties.
54 * Allow to change SimGrid configuration (see --help) within the code
55 thanks to SD_config() as it can be done in MSG.
56 * Add a new function SD_task_set_amount() upon user request.
59 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
60 * Remove rule based routing (no more PCRE dependency)
61 * Add a limiter_link option to cluster tag, to specify a maximum reachable
62 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
63 * Add a loopback_bw and loopback_lat options to cluster tag.
64 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
65 routing. Now peers are encapsulated in an AS and have their own private
66 router but this is transparent.
69 * Our own implementation of getline is renamed xbt_getline, and gets
70 used even if the OS provide a getline(). This should reduce the
71 configuration complexity by using the same code on all platforms.
72 * New type: xbt_cfg_elm_boolean.
73 * Allow to use yes/no for boolean configuration options in the command line.
74 * Allow to disable SimGrid cleanups at exit from command line option.
75 There are situations where one may want a simulation to end with an exit.
76 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
77 annoying when scripting around the simulator. Adding a
78 --cfg=clean_atexit:no allows to circumvent this issue.
79 * New command line option --version, to get SimGrid version information.
80 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
83 -- Mon Oct 7 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
85 SimGrid (3.9) stable; urgency=low
87 The Grasgory release: GRAS is really dead now.
89 * Complete overhaul of the internal host structures scheme.
92 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
93 was considered as experimental and badly maintained since then.
94 * Keeping it was thus a trap to our potential users, that could take
95 it instead of MSG or SMPI by mistake despite is pity state.
96 * GRAS seems to have very few users (if any), and no one volunteered
97 to maintain it further. It also induces a lot of XBT code (for
98 portability sake), that must be maintained too.
99 * For all these reasons, we killed GRAS. If someone wants to revive it
100 in the future, don't cry, our git history still remembers of GRAS.
103 * Major overhaul. Merge our documentation again as time proved that
104 splitting it was really not helping our users.
105 * Further improve the developer documentation to help newcomers
106 hacking on SimGrid itself. The user documentation (and in
107 particular, the beginner documentation) is still in a sorry state.
110 * Now works on Windows too!
111 * Much more extensive test suite, from MPICH
114 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
115 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
116 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
117 * Bug fix: let task be scheduled when the last dependency to be solved is
118 a control dependency.
119 * Remove SD_load_environment_script function.
120 Use the C sg_platf function if you want to declare a platform
124 * New function: MSG_process_get_number()
125 * Old function documented: MSG_config()
126 * Remove MSG_load_platform_script function
127 Use the C sg_platf function if you want to declare a platform
131 * Change the default value of the TCP_gamma constant (maximal size of TCP
132 congestion window) to a more realistic 4MiB value. If you notice changes in
133 your simulation results, you can fall back to the previous 20k tiny window
134 by adding --cfg=network/TCP_gamma:20000 on command line.
135 * (Hopefully) fix a bug wrt periodic availability and state traces
136 * Bug fix: use default values at start when first event in availability/state
137 trace is not at time 0.
140 * remove the "new_" part of function name sg_platf_new_trace_connect
141 (resulting in sg_platf_trace_connect), since it does not create
145 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
146 use case where it's really mandatory, and maintaining it was a pain in
148 * New: xbt_fifo_search(), search an item with a user-provided
149 comparison function instead of dumb pointer comparison.
152 * Fix the lua deployment:
153 Use `simgrid.init_application()` before deployment instead of
154 `simgrid.msg_register_application()` after.
157 * Transfer the tracing files into the corresponding modules.
159 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
161 SimGrid (3.8.1) stable; urgency=low
163 The "we are told that some people want to also *install* the simgrid
166 * Add missing file "tesh.1" to the archive.
168 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
170 SimGrid (3.8) stable; urgency=low
172 The Psssshiiiit release: SimGrid jumps into the Cloud.
175 * Add an experimental interface to manipulate VMs. They are mainly
176 process groups with very few intrinsic semantic, but they should
177 allow you to build the semantic you want easily.
178 * New function: MSG_host_set_property_value()
179 * New function: MSG_process_on_exit(). To clean memory in all cases.
180 * Bug fixes that made the host (and link) failures unusable.
181 * Add a way to auto-restart process when the host in which they are
182 executing comes back (ON_FAILURE="RESTART" on deployment file,
183 MSG_process_auto_restart_set).
184 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
185 please stop using the old ones, they are DEPRECATED.
187 * Deprecate functions MSG_global_init() / MSG_global_init_args()
188 Please use MSG_init() instead. (reducing the amount of entry
189 points in the library helps us).
190 * Make it impossible to link against the wrong version of the lib
191 * Deprecate MSG_clean(). No need to call it anymore.
192 * Function MSG_get_host_number() is not deprecated anymore.
195 * Split the doc into a user guide and a reference guide.
196 * Start a developper guide to help people hacking on SimGrid.
199 * Enable tracing by default. This modules rocks you should use it.
200 * Remove option custom_flags. Now use environment variables CFLAGS
202 * Use default cmake things to detect lua instead of home grown ones.
203 * New option "enable_mallocators" to disable mallocators, for debugging
204 purpose ("on" by default).
207 * Bug fixes around the resource failures: don't let the processes
208 survive the host they are running onto.
209 * Add an interface to auto-restart processes when the host in which they are
210 executing comes back.
211 * Ensures that SIMIX_clean is called automatically. It's not part of
212 the public interface anymore (bindings should be updated).
215 * Bug fix for when SD_Simulate is called with a positive value: be careful
216 when comparing doubles. Sometimes they are different for non significant
218 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
219 parallel task whose initial work is distributed among host according
220 to the Amdahl's law. Such tasks are created with a parameter alpha
221 that corresponds to the non-parallelizable part of the computation.
222 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
223 between two sets of workstations assuming a 1D block distribution (each
224 workstation owns a similar share of data) on both sides.
226 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
227 Data redistribution will be automatically scheduled once parent and child
228 are both scheduled. The filling of computation_amount and
229 communication_amount structures is now done seamlessly thanks to the chosen
231 * New function SD_workstation_dump to display various information
232 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
233 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
234 bandwidth on the route onto which the task is scheduled and the amount of
236 To divide the nominal bandwidth by 2, the rate then has to be :
237 rate = bandwidth/(2*amount)
238 * Compute tasks that have failed can now be rescheduled and executed again
239 (from their beginning)
240 * Increasing source code coverage (src/simdag is now covered at 95.8%
244 * Re-implement time-independent trace replay using SMPI (at the
245 smpi_smp_* level) instead of MSG. This should replace
246 examples/msg/actions/actions.c
247 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
248 hindexed and structs)
249 * Implement the exchange of non-contiguous data.
250 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
251 * Correct behavior of smpi/sender_gap and set its default value to 0
252 * Add option to asynchronously send small messages to allow better
253 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
254 specifies the size in bytes under which messages will be asynchronously sent.
255 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
256 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
259 * New C interface to define a platform: XML is now optional.
260 For more info, please check include/simgrid/platf.h.
261 * New interface to define random platforms from the C:
262 For more info, please check include/simgrid/platf_generator.h and
263 examples/msg/masterslave/masterslave_platfgen.c
264 * Export a sg_cmdline dynar containing all the arguments we got from
268 * Two new tracing options for adding comments to trace file so you
269 can track your experiments (see --help-tracing for details).
270 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
271 * Adding the SimGrid version that generated the trace file as a comment.
272 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
273 * Fix to avoid key clashes on Paje links
274 * Other minor fixes related to the Paje specification
277 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
278 and renamed to xbt_str_hash() and xbt_str_hash_ext().
279 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
280 * Greatly improve the robustness of mmalloc to user errors (such as
281 using an area after freeing it, or freeing it twice)
283 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
285 SimGrid (3.7.1) stable; urgency=low
288 * Restore the prototype of MSG_process_create_with_environment() to
289 the pre-3.7 situation by removing the kill_time argument.
290 * Add a MSG_process_set_kill_time() function instead.
293 * Fix weird behaviors when dealing with parallel tasks.
296 * Simgrid is now built as a dll.
297 * Simgrid-java now works on Windows.
298 * Simgrid-Java is now included into Windows package.
301 * First pre-build package for MacOSX.
304 * Fix compilation when using MSG_USE_DEPRECATED.
305 * Fix some compilation issues on Macs and Windows.
306 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
307 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
309 We discovered that the Lua console is broken, but we are missing the
310 manpower to fix it right now. The problem existed in 3.7 too, so we
311 are not blocking the release for that. Sorry if you depended on this
312 feature, any help would be really welcome.
314 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
316 SimGrid (3.7) stable; urgency=low
318 The "spring cleanups (before the next Big Project kicks in)" release.
321 * We can specify the SMPI latency/bandwidth factor with command line
322 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
323 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
324 You can also use the "config tag" from platform file by adding this line
325 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
326 (see "example/platforms/tag_config.xml" to use "config tag").
327 Note that the command line supersedes the platform file configuration.
328 * Change the correction factors used in LMM model, according to
329 the latest experiments described in INRIA RR-7821.
330 Accuracy should be improved this way.
331 * Use the partial invalidation optimization by default for the
332 network too. Should produce the exact same results, only faster.
333 * Major cleanup in surf to merge models and split some optimization
334 mechanisms from the core of the models. As a result you can now
335 specify which model to use (e.g., --cfg=network/model:LV08
336 --cfg=cpu/model:Cas01) and which optimization mode to use
337 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
338 Incompatible combinations should err at initialization. See
339 --help-models for the list of all models and optimization modes.
340 * The CLM03 workstation model was dropped for simplicity because it
341 used the deprecated CM02 network model. Use default instead.
342 * Rename the TCP_gamma configuration option to network/TCP_gamma
343 * Rename the coordinates configuration option to
344 network/coordinates, and document it
345 * Use now crosstraffic keyword instead of the terribly misleading
346 fullduplex keyword. It is activated by default now in the current
347 default model, use --cfg=network/crosstraffic:0 to turn it off.
348 * Ongoing refactoring the model parsing to make XML files optional
349 See include/simgrid/platf.h for details (still to be completed)
352 * Major overhaul of the documentation. Almost instructive now :/
353 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
354 functions and friends. This interface was considered as
355 deprecated since over 2 years, it's time to inform our users that it is.
356 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
357 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
358 use these (crappy) functions in your code.
359 These functions will be removed soon. Stop using them now.
360 * Deprecate MSG_get_host_{table,number}
361 Implement MSG_hosts_as_dynar() instead.
362 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
363 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
366 * Stabilize the parallel execution mode of user contexts
367 * Introduce configuration variables to control parallel execution:
368 - contexts/synchro: Synchronization mode to use when running
369 contexts in parallel (either futex, posix or busy_wait)
370 - contexts/parallel_threshold: Minimal number of user contexts
371 that must be part of a scheduling round to switch to parallel
372 execution mode (raw contexts only)
373 * Fix bugs that prevented to use suspend/resume along with
374 synchronization structures.
375 * Fix bugs in process termination that lead to invalid memory access
376 in very specific conditions.
379 * Introduce a parallel mode for the models (controlled by surf/nthreads
380 configuration item). In our tests, running the models in parallel
381 never lead to any speedups because they are so fast that the gain
382 of computing each model in parallel does not amortizes the
383 synchronization costs, even when ultra fast futexes are used.
384 This is released anyway because YMMV.
387 * Performance boost by using a swag internally to compute the set of
388 tasks that are finished and should constitute the return value of
392 * Enable it by default now that it is considered rather stable.
395 * Documentation of the tracing functions.
396 * Performance gains when tracing categorized/uncategorized resource
397 utilization by avoiding calls to get route when updating resource
398 variables. LMM constraints are being used instead.
399 * API changed to set task categories. Use MSG_task_set_category instead
400 of TRACE_msg_set_task_category, and SD_task_set_category instead
401 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
402 * Bugfix for graphicator, routes not correctly obtained, memory leaks
403 * Examples for link user variables added (see at examples/msg/tracing/)
404 * Deprecated function TRACE_msg_set_process_category completely removed
405 * Trace header updated according to the latest Paje file format
406 * Tracing network lazy updates, no longer obligate users to use full updates
407 * --cfg=tracing/platform:1 also registers power/bandwidth variables
408 * Experimental: let user code declare/set/push/pop application states for hosts
409 * API changed to allow the manual creation of graph configuration files
410 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
413 * Improve the API of Lua MSG bindings, using the Lua spirit.
414 * Each simulated process now lives in its own Lua world (globals are
415 automatically duplicated). It helps writing simulators. Will allow
416 to run Splay programs within SimGrid in the future.
417 * Add a Chord example in Lua, equivalent to the MSG one.
420 * Start the implementation of a solution to express temporal
421 properties, not only local assertions. This is still an
422 experimental work in progress, stay clear from it to be safe.
426 - Add new runtime parameters --help-logs and --help-log-categories
427 to display informations about supported logging parameters and
429 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
430 don't exists anymore.
431 * Mallocators: allow value NULL for the reset function.
433 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
434 create homogeneous dictionaries, where all the elements share the
435 same free function. Non homogeneous dictionaries will be
436 deprecated in the next release.
437 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
438 - Multi-level dictionaries are deprecated.
440 - new function xbt_dynar_search_or_negative() that is useful when
441 you have less than 2 million elements in your dynar and don't
442 want of the extra complexity of catching exceptions when the
443 element is not found.
445 - Make xbt_os_thread module (for thread portability) public.
446 Documentation is still to come, sorry.
448 - Cleanups and simplifications to make it maintainable again.
449 - Exotic features (such as memalign and valloc) were removed.
450 - The metadata were extended and improved so that the
451 model-checker becomes able to explore and inspect the heaps.
452 - This may induce a performance drop when enable_model-checking is
453 ON in cmake (even if it's not used in the simulation), but it is
454 necessary at this point to get MC working.
456 Turn model-checking OFF if simulation performance matters to you.
457 Not enabling it at runtime is not enough, disable it in cmake.
459 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
461 SimGrid (3.6.2) stable; urgency=low
463 The "Not coding new stuff allows to polish old things" release.
466 * New bindings to the NS3 packet level simulator (experimental)
467 * Use the raw (efficient) execution contexts instead of the sysv
468 (portable) ones when possible.
469 * libpcre is now mandatory in any cases since not using it led to
470 severe performance loss and possibly other issues
471 * Update the XML platforms:
472 - G5K: include the latest machine in Nancy
473 - GridPP and LCG: new platforms
474 * Documentation was partially updated, at least (more to come)
476 Bug fixes, cosmetics and small improvements
477 * Free terminated processes before the end of the simulation to avoid
478 exhausting the memory of users having very dynamic amount of
480 * Bug fix and cosmetics about canceling non-running tasks
481 * Bug fix about the dot loader's issues when using libcgraph
484 * Create an installer for windows with nsis (amd64 and win32)
485 - Add an hello world project to illustrate simgrid project creation.
486 - Embed libpcre into the Simgrid installer to avoid
487 its compilation burden
488 * The raw execution contexts should work on Apple now
489 * Port to Windows 64 bits
490 - Sysv contexts now have an implementation for this arch
491 - GRAS communication features now support this arch
492 * Drop support for borland compiler on windows
493 - this code was not maintained, and we kinda depend on gcc nowadays
494 * Fix portability issues on kfreebsd/gnu: build error about semaphores
495 * Fix portability issue on unstable ubuntu: linker became picky on
498 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
500 SimGrid (3.6.1) stable; urgency=low
502 The "Oops, we broke Macs too" release
505 * Fixed contexts detection so that raw ones are used when possible
506 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
507 leads to a strange error, with user code segfaulting sometimes when
508 the generated code is not perfectly aligned (which is not
509 controllable from the user side, depends on the amount of code)
512 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
514 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
516 SimGrid (3.6) stable; urgency=medium
518 The Summer Release, also known as the "OMG! They Killed Kenny!" version
521 * Bindings now constitute their own package, separated from the main one.
522 Rationale: reduce our maintenance nightmare by reducing the module coupling
523 They will soon be released on their own on gforge.
525 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
526 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
528 GRAS: It is not considered as stable anymore, but experimental. Sorry.
529 * It's not quite deprecated for now because we have no replacement,
530 but it may soon become the case.
533 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
534 * Fortran: New user-level cache variable to store the rank of the running
535 process. This improves performance by an order of magnitude.
536 * C: New coccinelle script to automatically locate and modify global and
537 local static variables.
538 * Improved SMPI network model with a sender-side gap to account for multiple
542 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
543 only say if a communication is finished, no matter whether it succeeded or
544 failed. You can call MSG_comm_get_status() to know the status of a finished
546 * New function MSG_task_dsend() to send a task and detach it. When a
547 communication is detached, you are never notified of its success or failure
548 and the memory is released automatically once it is finished. This function
549 is useful when you don't care about the end nor the success of a
551 * Change the prototypes of action replay. Sorry for inconvenience,
552 but this is really more efficient this way (and to adapt your code,
553 you just have to fix the initialization, that shouldn't be too long)
554 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
555 ever managed to do anything useful with it.
556 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
557 takes a dynar of communications. It returns immediately and gives the
558 index of a finished communication (if any).
559 * New example: a basic implementation of the Chord P2P algorithm.
562 * New model for multi-core CPUs. You can now use the core attribute to
563 precise the number of cores of a host. This is a basic model. Every
564 process running on the host receives at most the power provided in
565 the DTD (throughput<=power). Total throughput of process cannot exceed
567 * New peer tag. This peer tag creates a tiny AS comprising a host and a
568 router linked by an up-link and a down-link (possibly asymmetrical).
569 This kind of pattern allows to easily build last-mile model style platforms.
570 Aggregating such patterns in a rule-based AS is thus the technique of
571 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
572 * New model for Vivaldi routing. We transformed the Vivaldi network model
573 into a Vivaldi routing model (based on the rule-based model). This allows to
574 combine Vivaldi based latencies with last-mile platforms.
577 * Added a check for NaN of IEEE754 infinite in the double entries of
579 * Introduce a new context factory "raw", highly inspirated from the
580 ucontext factory, but using manually crafted functions in assembly to
581 do the work in an efficient manner.
582 * Allow to change the used context factory at run time, not only at
583 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
584 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
585 contexts at runtime (only with raw contexts or ucontexts).
586 * Completely rewrote this module to allow parallel execution of user
587 processes. Use --cfg=contexts/nthreads:N to execute user processes
588 with N parallel threads (the default is 1, meaning no parallelism).
589 * Allow to decide dynamically between sequential and parallel modes.
590 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
591 processes in parallel only when their number is greater than or equal to P
593 * Added a check for NaN of IEEE754 infinite in the double entries of
597 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
598 won't output the state of processes when interrupted with Ctrl-C
599 * Add a new function xbt_dynar_to_array that transforms a dynar into a
600 NULL-terminated array. This may solve backward compatibility issues
601 due to the change to return type of SD_simulate. See also:
602 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
603 * Add new macros with variable number of arguments.
604 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
605 - in xbt/asserts.h: xbt_assert
606 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
607 - in xbt/ex.h: THROWF and RETHROWF.
608 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
610 * Change xbt_die() to accept a format string with arguments, just like printf.
611 * New data structure: xbt_lib_t, like a dict but more general and with better
615 * New configuration options
616 Options triva/categorized and triva/uncategorized can be used to generate
617 graph configuration files for Triva visualization tool.
618 * Configuration option tracing/platform is renamed to tracing/categorized
619 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
620 * New TRACE_user_link_variable interface:
621 User provides the name of the link and the tracing variable to attach to it
622 * the declaration of tracing categories must be done after the environment
624 * simpler tracing interface, just one way to declare categories
625 TRACE_category or TRACE_category_with_color, it is up to you
626 * links in the trace file are again identified by their names
627 * trace contains the full platform hierarchy exactly as declared using the ASes
628 * Options tracing/msg/[task|process]:1 groups the process by hosts
629 for both cases, tasks and processes must have names that are unique during the simulation
630 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
631 * The experimental option tracing/msg/volume is deprecated
632 its functionality may be reincorporated if needed
634 The tracing generates a trace file with unordered timestamped events,
635 because of the way the core simulator (surf) works. A script available
636 at the tools directory (fix-paje-trace.sh) can be used to put the events
637 in order. We have changed the tracing so it can generate ordered timestamped
638 events in the final trace, but depending on the simulator (and how much time
639 is simulated) that can lead to a huge memory utilization. It is deactivated
640 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
643 * Define a SIMGRID_VERSION macro in simgrid_config.h.
644 - We are trying hard to keep the API stable, but it may happen that
645 some things change (we're a research project after all, not a
646 nuclear plant operating system). If such things should happen, you
647 could rely on that macro to adapt.
648 - current value: 30600 for 3.06.00, aka 3.6
649 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
650 a number that can be compared with SIMGRID_VERSION.
651 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
652 assertions and verbose/debug logging events are disabled at compile time.
654 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
656 SimGrid (3.5) stable; urgency=medium
659 * New feature to any SimGrid-based simulator: Model-Checking
660 Check SIN#1 for more details.
663 * New Model SMPI (three-interval linear regression for correction factors)
664 See RR-7426, available at http://hal.inria.fr/inria-00527150
665 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
666 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
667 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
668 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
669 * See also src/smpi/README
674 - Tracing API changes: TRACE_start and TRACE_end should not be called
675 by user-code. They are automatically called by simulators created
676 with SimDAG, MSG and SMPI if the toolkit is compiled with
677 tracing_enabled=ON. Categories declaration and utilization remain the
678 same for MSG and SimDag.
679 - A function was added to the tracing API to declare categories with
681 - TRACE_category_with_color (char *category, char *color)
682 where color must be in the following format
683 "%f %f %f", red, green, blue
684 and red, green, blue are float values in the interval [0, 1]
685 - User can specify NULL as color parameter, or continue calling
687 On that case, the tracing system will define random colors
688 - The following command-line options are supported:
689 --cfg=tracing/filename:msg.trace
690 --cfg=tracing:1 (activate tracing, needed to use others)
691 --cfg=tracing/platform:1 (categorized resource use)
692 --cfg=tracing/uncategorized:1 (uncategorized resource use)
693 --cfg=tracing/msg/task:1 (task creation)
694 --cfg=tracing/msg/process:1 (process creation, migration)
695 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
696 --cfg=tracing/smpi:1 (SMPI interface tracing)
697 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
698 - examples of examples/msg/tracing updated
700 - DAXLoader and DOTLoader functions can generate tasks with categories
701 - A new function to attribute a category to SD tasks:
702 TRACE_sd_set_task_category (SD_task_t task, char *category)
703 * Tracing the MPI interface implemented by SMPI
704 - Collective operations are traced with states
705 - Point-to-Point operations are traced with states/links
706 - Tracing activated by a parameter "-trace filename" passed
707 to smpirun during execution (considering that simgrid
708 is compiled with tracing enabled)
709 - To run the simulation with gdb, the simulator
710 accepts --cfg=tracing/smpi:1 to trace SMPI
711 - tesh files to check if smpi tracing is ok
712 - See examples/smpi/NAS/DT-trace
713 * GTNetS tracing re-worked
714 - adaptation to the tracing system of GTNets to cope
715 with modifications regarding the fullduplex mode
716 - new tesh files to check if gtnets tracing is ok
719 * Asynchronous communications through the functions:
720 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
721 * New function: MSG_load_platform_script()
722 to make possible using a lua script instead of XML files to set up platforms
723 * New function: MSG_set_function
724 to associate functions to processes, used when bypassing the parser
725 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
727 Platforms: Add some more examples in examples/platforms
728 * Grid'5000: see www.grid5000.fr
729 * *_30000_hosts.xml: various huge files [mainly scalability testing]
732 * Change the XML format. This is a very important modification. SimGrid
733 3.5 introduces a new hierarchical format based on the notion of
734 Autonomous Systems. Compatibility with old format is ensured
735 through the perl script provided in the install bin directory
736 bin/simgrid_update_xml.
737 It is now possible to build platforms with specific routing
738 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
739 connect several platforms together. We will try to provide soon
740 set of realistic platforms exploiting these properties (have a
741 look at examples/platforms/ for the moment).
742 * Take the opportunity of the XML format change to be a good XML citizen:
743 rename link:ctn to link_ctn and similar changes (also dealt with by
745 * Add a new routing scheme (rule-based) using regular expressions. It
746 enables to have an extremely low memory footprint when the
747 underlying routing is simple and can be compactly described. You
748 need to have libpcre4-dev (perl regular expressions) installed if
749 you want to use this routing scheme.
750 * Revive the cluster TAG and allow to easily and efficiently (both in
751 term of memory and speed) connect clusters together. Have a look
752 at teshsuite/simdag/platforms/ to see how this can be done. With
753 this tag, you can create clusters with thousands of tasks at no
754 cost (have a look at examples/platforms/).
755 Note: clusters are implemented as ASes, so there is no need for an
756 enclosing AS tag if you have only one cluster in your platform.
757 * Add new generic functions in the public interface that allows the user
758 to call SURF 'create_resource' methods from your code (same
759 functionality as the XML bypass mechanism but with a much lighter
761 * Add a new model (enabled through command line --cfg=network/model:SMPI)
762 that uses a piecewise linear approximation to produce better
763 results when exchanging small messages.
764 * Add a new parameter to handle correctly full duplex link and account
765 for interferences between uplink and downlink communications
766 (activate with --cfg=fullduplex:1).
769 * Rename the SD_READY (all dependencies are satisfied and task is
770 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
771 dependencies are satisfied) state.
772 This prevents a confusion between the notion of "ready to schedule"
773 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
774 simulated" (SD_RUNNABLE) used by the simulation kernel.
775 * Change the way a task is considered as ready. Instead of removing
776 dependencies when a task is done, a counter is decreased. This way,
777 it is always possible to reach ancestors thanks to the
778 SD_taks_get_parents function (even after the end of the simulation.)
779 * Change the return type of SD_Simulate from (SD_task_t*) into
780 xbt_dynar_t. This function was in handling a dynar internally and
781 converted it into a NULL terminated array for historical reasons.
782 * New function SD_dotload(char*) to load a DAG described in dot
783 format. This loader and the corresponding examples require the
784 installation of the graphviz library.
785 * Fix a bug in the management of tasks of size 0 in the surf network
786 models. This problem was only visible with SIMDAG and you should
787 thus disregard results produced with earlier versions if you
788 relied on this feature (some tasks were blocked because of this).
789 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
790 even though your applications were doing only point-to-point
791 communications and sequential computations. Now you can really use any
792 model you want (of course, if you create real parallel tasks, which are
793 not implemented in most models beside ptaskL07, this will abort).
794 * Add an example that schedules a DAX on an heterogeneous platform
795 using a Min-Min strategy.
796 * New function SD_workstation_get_current_task() that returns the kind
797 of task currently running on a workstation in the sequential access
799 * Raise some warnings when unexecuted tasks remains at the end of the
800 simulation. This is usually caused by cycles in the DAG.
803 * New function: SIMIX_process_set_function() called by MSG_set_function
804 * Change the underlying waiting queue in semaphores so that a process
805 can wait on several of them simultaneously (as in waitany).
806 * Fix the way to handle tokens in semaphores so that all access patterns
807 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
808 * kill the dirty pimple SIMIX_message_sizes_output()
809 Please use (proper) visualization instead
812 * New data container: setset (set of sets of elements)
813 * New module: mmalloc (mapped malloc, allowing to have several
814 independent segments of malloc)
815 * New function: xbt_dict_cursor_set_data()
816 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
817 * New function: xbt_dynar_is_empty()
818 * New function: xbt_fifo_get_last_item()
819 * Fix xbt_dynar_shrink(): use the right element size.
820 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
821 uninitialized areas during expand.
822 * Fix semaphores: previous implementation was severely broken.
823 * Use library init/fini functions for our initialization.
824 - you can use logs and other feature as soon as you want in your
825 code (even before the xbt_init / MSG_init)
826 - xbt_exit is now a no-op and produce a warning when used.
829 * Port GRAS to new SIMIX mechanisms. This allows gras users to
830 benefit from the latest improvement to the simulation kernel.
831 * Kill measurement sockets for now. If you rely on them, sorry. This
832 release is not for you. This feature will be reintroduced in the
833 future, but we cannot delay the release any further.
834 * New function: gras_msgtype_get_name().
835 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
836 * Fix (at last) the pmm example: it should not randomly fail anymore.
838 Build chain: bug fixes and overall polishing
839 * Cmake is now stable enough. Hence, we killed the autotools.
840 * Port to windows ( TM :)
841 * Fix the 'make install' target.
842 No need to use 'make install-simgrid' anymore
843 * Introduce a 'make dist' target compiling a *source* archive
844 'make package' compiles a binary archive
845 * Compile java files only on need
846 * Add --cd and --setenv command line options to tesh
847 * Out of source builds are not fully supported yet, but we are close
848 * Enable supernovae and optimization flags by default for our users
851 * Add layer to set up environment directly from lua, without XML.
852 * The effect of gras_stub_generator can be achieved through
853 lua too (check examples/gras/console/ping_generator.lua)
855 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
857 SimGrid (3.4.1) stable; urgency=low
859 The "Polishing easter eggs is probably a good idea" release.
860 This is a bug fixes release only.
864 * Fix a bug preventing the tasks from begin garbage collected.
867 * Fix a bug occuring when a host involved in a communication fails.
868 This was not detected properly by the other peer involved in the
869 communication. Now, it's reported as a network error.
872 * Warn the user about loop dependencies in data flow of DAX files
873 * Obey the control-flow dependencies of DAX files
876 * Add option "enable_smpi" allowing to not compile SMPI.
877 Probably useful for the (Mac) users experiencing a build error here
878 * Improve the detection of lua5.1 and ruby1.8
880 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
882 SimGrid (3.4) stable; urgency=low
884 The "Easter in Cargese" release. Also known as (major changes):
886 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
887 ~> bindings were greatly improved
888 ~> new tracing infrastructure for better visualization introduced
890 * the "Welcome to configury modernity" release.
891 ~> we switched from autotools to cmake, and improved our cdash
894 A more detailled list of changes follow (full detail in svn log).
896 Java Bindings: Various Cleanups
897 * (install java-gcj-compat-dev on debian-like to use them)
898 * Remove put/get: no need to export deprecated interface in Java
899 Use send/receive instead.
900 * Cleanup the examples and add a README per directory
901 * Remove example autoDestination (that's the only way to go now)
902 * Remove example explicitDestination (was a plain copy of basic)
903 * Make JniException a runtime exception, so that there is no need to
904 declare the fact that you may encounter such a beast. I guess that
905 nobody will ever want to survive such error.
906 * Create specific errors for each MSG case of failure:
907 host failure, transfer failure, timeout, task cancelled
908 * Cleanup the exceptions that may get thrown by each function
909 * Other internal cleanups in Java bindings. Performance still bad :/
910 Ruby and Lua Bindings: create them
911 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
912 * That's new and great, you should try them out.
913 Same functionalities than Java bindings, only even less polished
915 * Kill the useless "rate" argument of SD_task_get_execution_time()
916 Everyone used to provide -1 as a value, it was not used, and the
917 semantic of a possible use wasn't even clear.
918 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
919 either as comm costs or compute costs to mean that there is no
920 such thing for that specific task.
921 * Add a SD_task_set_name() function
922 * Fix SD_task_unschedule() on typed tasks
923 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
924 * In DAX loader, accept useless 'level' attributes to <job> since
925 LIGO DAGs have them (seem to be to ease graphical representation).
927 * Add an example masterslave_mailbox.c using send/receive and not
928 the deprecated put/get interface.
929 * Kill the MSG_paje_output() function. It's a noop since 2 years.
930 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
932 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
933 (declare MSG_USE_DEPRECATED to still have the old name)
934 * Add a MSG_task_set_data() function
935 * About trace replay (see examples/msg/actions):
937 - Allow to work with splitted trace files for each process
938 Give the specific trace file as argument of each process,
939 and call MSG_action_trace_run(NULL)
940 You can still have one merged file for all processes.
941 - Fix implementation of collective operations
942 * Allow task_execute() on 0-sized tasks (closes #10063)
944 * This is the first release of SimGrid where SMPI is not considered
945 beta anymore (even if some corners should still be improved)
946 * Port over the new SIMIX_network submodule (internal refactoring)
947 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
948 * Implement more missing elements of the standard:
950 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
952 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
953 MPI_Scan() MPI_Get_processor_name()
954 - Added implementation of missing case for Alltoall (warning: it's
955 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
956 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
957 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
958 SMPI_MPI_Allgatherv()
960 - MPI_Waitsome() was broken
961 - Allow relative includes in smpicc
962 - Command line cfg argument 'reference_speed' was ignored...
963 - Some functions did not properly lead to auto-benching of user code
964 - smpicc passes -O2 by default (just like openmpi one)
966 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
967 * Bug fixes about timeouts during communications
968 * add SIMIX_message_sizes_output() as a pimple to write to file the
969 amount of messages per size. Use gnuplot to get histogram.
970 Pimple because that's the only user-visible function of simix,
971 defined directly in xbt.h (irk, sorry)
973 - Add a SIMIX_sem_get_capacity() function
974 - Fix interactions with processe resume/suspende
975 - release_forever() was stupidly broken
976 - Fix SIMIX_display_process_status() for processes in a semaphore
977 - Make SIMIX_sem_block_onto() user-visible
978 * Refactoring context stuff:
979 - Use pseudo-OOP for better modularity
980 - reimplement SIMIX_process_kill() without process_schedule() so
981 that the latter can take as invariant that it is called from
983 - Merge context_start into context_new for sake of simplicity
985 * Add a Vivaldi network model, coded live during SUD'10 ;)
986 * Rename configuration variables to start a hierarchy:
987 o cpu_model -> cpu/model
988 o network_model -> network/model
989 o workstation_model -> workstation/model
990 * New configuration variables:
991 o network/bandwidth_factor: correction to bandwith
992 o network/latency_factor: correction to latency
993 o netwotk/weight_S: correction to the weight of competing streams
994 * Add a long description to the models, that users can see with such
995 argument on the command line: --cfg=cpu/model:help
996 * --help-models display the long description of all known models
998 * config: add the ability to set a default value after registration
999 Does not override any previously set value (e.g. from cmd line)
1000 * dict: allow to have integer key and data.
1001 When so, you need to use the following functions
1002 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1003 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1004 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1005 In contrary to regular dicts, the key is not malloced before copy.
1006 Mixing scalar and regular elements in the same dict is not tested
1008 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1009 Tracing for Visualization:
1010 * SimGrid is now instrumented in order to generate a trace file for
1011 visualization analysis: to use it, need to compile SimGrid with the
1012 "tracing" option enabled, and instrument the program using SimGrid with
1013 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1014 (among other functions).
1015 * The instrumentation only traces the platform utilization for now
1016 * Documentation to use the tracing functions and how to analyze the
1017 traces with the Triva tool is written.
1018 * More information about: SimGrid FAQ (in the section Tracing Simulations
1021 * We moved to cmake as default build system. Autotools support will
1022 be dropped soon. Check the FAQ for more info about how to use it.
1023 * Greatly improved our cdash/ctest interactions
1024 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1025 * Added memory checking tests with valgrind; lot of memleak fixing.
1026 This may be the first release of simgrid with so few memory issues
1027 * Added code coverage tests.
1028 Our coverage is still improvable, but at least we see it on cdash.
1030 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1032 SimGrid (3.3.4) stable; urgency=low
1034 The "Desktop Grid needs love too" release (also called Xmas release).
1036 Models improvements:
1037 * Major speedup in the maxmin system solving by using lazy evaluation
1038 Instead of solving completely the maxmin system at each iteration,
1039 only invalidate (and recompute) the modified parts.
1040 This new feature is enabled in default models but you can try to
1041 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1042 * Cas01 IMproved as default CPU model
1043 This CPU model is the same Cas01 model, but it uses the
1044 maxmin-selective-update flag and a heap structure to manage
1045 actions on SURF kernel.
1046 It reduces the complexity to find the next action to finish and,
1047 consequently, it's faster than the old Cas01.
1048 This is the new default CPU model (Cas01).
1049 * Rename the old Cas01 model to Cas01_fullupdate
1050 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1051 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1052 * CpuTI (CPU Trace Integration)
1053 A new CPU model whose objective is simulate faster when using
1054 availability trace files.
1055 Instead of using a full featured, over engineered maxmin system for
1056 CPU modeling, this model does the pre-integration of traces files
1057 to calculate the amount of CPU power available, and so, executes
1058 faster than the old CPU models.
1059 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1060 * Use LV08 as default network model since it gives better accuracy
1061 for small messages and shouldn't change things for big ones.
1062 Use --cfg=network_model:CM02 to get the previous behavior.
1065 ******************************************
1066 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1067 ******************************************
1068 * The new CPU model may changes simulations!
1069 The point is that events occurring at the exact same timestamp
1070 are not scheduled in the same order with the old and new
1071 version. This may be enough to completely change the execution
1072 of simulations in some cases.
1073 * The new network model will change simulations!
1074 This new model is more realistic than the previous one, so you
1075 should consider redoing your old experiments with this model.
1076 Sorry for the inconvenience.
1079 * Introduce the supernovae compilation mode
1080 When compiled that way, the whole SimGrid (or almost) is put in a
1081 single compilation unit and compiled in one shoot.
1082 This is to help gcc which has difficulties to inline stuff from one
1084 The speedup seem to be above 15%, althrough more tests are needed on
1085 amd64 to confirm that gain.
1088 * Port of MSG's mailbox on top of SIMIX network
1089 The put/get mechanism was greatly simplified on the way.
1092 * New SIMIX network module. Provides:
1093 - Mailbox: rendez-vous mecanism to find with who you want to speak
1094 - Synchronous send/recv: easier and hopefully faster since the
1095 logic is handled in the maestro process directly now
1096 - Asynchronous send/recv: you dreamt of it? It's here now
1097 Too bad that nobody cared enough to propagate the change to MSG.
1098 * Add semaphores as SIMIX synchronization mechanism.
1101 * new function SD_daxload(char*) to load a DAX file
1102 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1103 * Introduce typed tasks. Specify its kind and cost at creation.
1104 At scheduling, just give where it should be placed, and the cost
1105 for each involved resource is automatically computed.
1106 Existing constructors so far (more to come of course):
1107 - SD_task_create_comm_e2e() for end-to-end communication
1108 - SD_task_create_comp_seq() for sequential computation
1109 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1110 * new function SD_task_dump() for debuging display
1111 * new function SD_task_dotty(task,FILE*) writing to file the info
1112 about the task in dotty format
1113 * SD_task_dependency_exists() can now cope with having one of its
1114 arguments NULL. If so, it tests whether the other argument has any
1116 * Add getters on list of preceding/following tasks:
1117 SD_task_get_parents(task) and SD_task_get_children(task)
1118 * Add getters on amount of workstations and list:
1119 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1120 * Add getter on task kind: SD_task_get_kind(task)
1121 * Update the start_time and finish_time of tasks on completion/failure
1122 * Bugfix: Remove task from state swags when destroyed
1125 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1128 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1129 dynar with one separate thread per value of the dynar.
1130 * Change the prototype of xbt_thread_create(), sorry.
1131 Added a boolean parameter indicating whether we want to join this
1132 thread (used in SG only for now)
1133 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1136 * GTNetS wrappers should now be usable again (and betterly tested too)
1137 * Fix a major regression from 3.2 where the timeout provided to
1138 MSG_task_put_with_timeout() was used as absolute time before which
1139 the comm should be done.
1140 * Start to fix the <cluster> tag.
1141 - Internal links should be good now (beside of the loopback, which
1142 use the private link instead)
1143 - paths to the external world is still rather broken
1144 - the <route:multi> tag is just broken. Actually that's brain-dead.
1145 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1146 to make it less stupid
1147 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1148 * Fix a source-level compatibility glitch from 3.2: after defining
1149 MSG_USE_DEPRECATED, you can use the old name
1150 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1151 * Allow to compile from the SVN with automake 1.11
1152 * Fix some problems when using the "start_time" tag in deployment XMLs.
1153 * Fix #8569: XBT/synchro.h has redundant declarations
1154 * Fix #8563: MSG return values and exceptions
1155 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1156 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1157 Thanks to Robson Peixoto.
1158 * Fix a vicious bug in dictionaries inducing that some elements were
1159 not freed on xbt_dict_free()
1161 Portability report of this version:
1162 * Main portability targets:
1163 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1164 on (amd64/i386/ia64)
1165 - mac leopard on i386
1166 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1167 but nothing critical.
1168 * Other platforms: windows, AIX and others were not tested for this release
1170 Timing report of this version:
1171 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1172 scenario parameters). From 8h to a few seconds in desktop grid settings.
1173 * Supernovae brings about 25% speedup on i386.
1175 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1177 SimGrid (3.3.3) stable; urgency=low
1179 The "Need for Speed" release.
1181 The timings done to validate the 3.3.2 were faulty.
1182 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1184 The problem was a conversion from a manually handled vector to
1185 xbt_dynar_t on the critical path.
1186 xbt_dynar_foreach calls functions, inducing stack management crap.
1188 We inlined these functions and xbt_dynar_foreach is now breath taking.
1189 We also inlined xbt_swag_belong on the way.
1191 Here are some approximate speedup measurements (on master/slaves
1192 simulations lasting between 10s and 20s each):
1193 3.3.1 -> 3.3.2: about same performance
1194 3.3.2 -> 3.3.3: 40% speedup
1195 3.3.1 -> 3.3.3: 40% speedup
1196 3.3.1 with inline patch -> 3.3.3: 30% speedup
1198 Our reading is that the refactoring which occurred in 3.3.2 made us
1199 suffer much more from the xbt_dynar_foreach low performance, but
1200 once we solved this, this refactoring proved to be very performance
1201 effective. From the 40% speedup, somehow, 10% are due to the
1202 inlining and 30% to the refactoring.
1204 That's a pitty that gcc cannot inline functions placed in other files
1205 alone. We have to choose between:
1206 - break the encapsulation (by putting private data structures and
1207 accessors in headers files to help gcc)
1208 - live with low performance
1209 - switch to a decent compiler such as icc (not quite possible).
1211 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1213 SimGrid (3.3.2) stable; urgency=low
1215 The "Simplicity does not preceed complexity, but follows it" release.
1217 The main contributors of this release were (lexical order):
1218 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1221 * Extract the routing logic into its own object.
1222 (was dupplicated in network.c and workstation_LV07.c;
1223 Allows to implement other ways of storing that info)
1224 => kill now useless network_card concept
1225 - Use dynar to represent routes (instead of void** + int*)
1226 - kill link_set (use surf_network_model->resource_set instead)
1227 - Add a command-line option to choose the routing schema to use
1228 - Add three new models:
1229 * Floyd (shortest path computed at initialization)
1230 * Dijikstra (shortest path recomputed all the time)
1231 * Cached Dijikstra (shortest path computed on need)
1232 All these models where contributed by Silas De Munck, and are
1233 described in his ICCS09 paper.
1235 * Simplify model declaration
1236 (less redirections, less function to write when defining a model)
1237 - Factorize stuff between models:
1240 surf_model_resource_set(model)
1241 surf_model_resource_by_name(model, name)
1242 - Unify the types of models in s_surf_model_t (using an union)
1243 - Embeed fields of common_public directly into s_surf_model_t
1244 - Rename model methods:
1245 action_free ~> action_unref
1246 action_change_state ~> action_state_set
1247 action_get_state ~> action_state_get
1248 - Change model methods into functions :
1249 (model)->common_public->action_use ~> surf_action_ref
1251 * Implement a generic resource; use it as ancestor to specific ones
1252 (allows to kill duplicated code in models)
1253 Drawback: timer command don't need no name nor properties;
1254 workstation_CLM03 don't need no properties
1255 (but I guess we can live with those few bytes wasted)
1257 * Improve the action object model
1258 - implement a constructor avoiding dupplicated code about field
1259 initialization in generic_action part.
1261 * Kill the SDP model: it has an external dependency, is deprecated
1262 in flavor of modern lmm models, and didn't compile since a while
1265 * Relocation of the context module from XBT to SIMIX.
1266 (the context were decoupled from the simix processes, duplicating a lot of code)
1267 => a lot of code was factorized
1268 - less overhead is introduced during scheduling
1269 - simpler API for the context factory
1270 - the logic for process creation,destruction and manipulation was simplified
1271 * Simplification of the s_smx_process_t data structure.
1272 => accesing the simix level data associated to a process is faster now,
1273 and the code is a lot more readable.
1276 * Implement some more MPI primitives:
1277 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1278 -implementation: Bcast: flat or 2-ary tree (default),
1279 Barrier: 4-ary tree,
1281 Allreduce: Reduce then Bcast
1282 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1283 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1284 Alltoallv: flat tree, like ompi
1286 * Add support for optimized collectives (Bcast is now binomial by default)
1287 * Port smpirun and smpicc to OS X
1290 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1291 Shout out if you used it.
1294 * Display the list of still queued messages in SG mode when existing
1298 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1299 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1300 * Add xbt_str_from_file(FILE*)
1301 * Add xbt_dict_get_key achieving a linear reverse search
1302 * Remove the context module
1304 Portability report of this version:
1305 * Main portability targets:
1306 - Linux(debian)/x86/context
1307 - Linux(debian)/x86/pthreads
1308 - Linux(debian)/amd64/context
1309 - Linux(debian)/amd64/pthreads
1310 On these, we still have the eratic breakages of gras/pmm and
1311 amok/saturate_sg reported in previous version. We still think
1312 that the tests are the cause of the fault, not the tested code.
1314 - Mac OSX Leopard/x86/context
1315 Still false negative in tesh autotesting.
1316 Smpi still fails, but this time because readlink does not accept -f
1317 Everything seems to work properly beside of that.
1320 - AIX version 5.3 (only tested contexts this time)
1321 Smpi still fails there because mktemp is not installed.
1322 Everything seems to work properly beside of that.
1324 I managed to compile it for the first time, but several breakages.
1325 Won't delay the release for this exotic platform.
1327 * Windows: it's still lagging behind. If you want to help, please
1330 Timing report of this version:
1331 This version seem to be more than 5% faster than 3.3.1 (on linux
1332 64bits with contextes). The gain is less than expected, we are
1333 investigating this for next release.
1335 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1337 SimGrid (3.3.1) stable; urgency=low
1340 * Implement a --cfg-help to show existing configuration variables
1341 * Build chain do not require doxygen in maintainer mode
1344 * fix a bug on struct sizeof computation, which prevented the
1345 exchange of arrays of structs in some conditions
1346 - added a regression test about this in datadesc_usage
1347 * Allow the exchange of 0-long dynamic vectors.
1348 - for that, use -1 as indicator of dynamic size instead of 0
1349 - This implied to change any size from unsigned long to long,
1350 reducing a bit communication abilities, but I guess that with
1351 64bits being quite common, this is more than enough.
1352 - This also induce a protocol change, thus bumping network protocol
1353 version from 0 to 1 (if we have external users, we have to get
1354 clean on that point too ;)
1355 - added two regression tests about this in datadesc_usage
1356 * Be more verbose when propagating local exceptions
1357 This helps debugging.
1358 * Display the status of simulated processes when receiving SIGINT in
1362 * Allow to control the simulation from a trace file.
1363 New functions MSG_action_register() and MSG_action_trace_run()
1364 The first one allows to associate a function execution to each
1365 kind of action while the second one parses a trace file and
1366 triggers the corresponding actions within the system.
1367 For now, only a toy example is provided in examples/msg/actions
1368 * Add an exemple of process migration in examples/msg/migration
1369 * Fix a bug in task exchange which broke MSG_task_get_sender()
1370 Add a teshsuite regression test for that.
1371 [Bug: if MSG_task_get_sender() is called after sender exit,
1373 * Fix a bug which prevented suspend/resume to work properly
1374 * Display the status of simulated processes when receiving SIGINT
1375 This fixes a regression of v3.3. due to the introduction of SIMIX
1376 * Bug fixing in failure management:
1377 - trace could not start by a failure at time 0
1378 - failure during communications were not working
1381 * Add SIMIX_process_set_name() to change the name of the current
1382 process in the log messages.
1383 * Store smx_hosts in a dict since we only retrieve them by name
1384 * Move the configuration infrastructure to surf
1387 * Move the configuration infrastructure to surf
1390 * Massive internal cleanups:
1391 - Store internal structures on processes instead of hosts (allows
1392 to have more than one process per host, in addition of being more
1394 - Cleanup the initialization/finalization process
1395 - Kill a whole bunch of unneeded synchronization:
1396 processes run in exclusive manner within the simulator
1397 - Move queues from global tables to process data fields
1399 - now accept -platform and -hostfile arguments
1400 - Pass the right rank value to processes according to the hostfile
1401 * Compile the examples by default, and use them as regression tests
1402 * Implement MPI_Wtime()
1403 * Change the reference speed to a command line option
1406 * TCP_gamma can now be specified as command line option using
1407 --cfg=TCP_gamma:10000000.0
1408 * Change the --surf-path cmd line option into --cfg=path:
1411 * Also include strbuff from xbt.h public header
1412 * xbt_ex_display(): do not free the exception after displaying
1413 This allows to do more with the given exception afterward.
1414 Users should call xbt_ex_free() themselves.
1418 Portability report of this version:
1419 * Main portability targets:
1420 - Linux(debian)/x86/context
1421 - Linux(debian)/x86/pthreads
1422 - Linux(debian)/amd64/context
1423 - Linux(debian)/amd64/pthreads
1424 These targets fail about 1/10 of times on gras/pmm, but we believe
1425 that this is because of the test, not because of simgrid.
1426 amok/saturate_sg fails even more rarely, and the test may not be
1429 - Mac OSX Leopard/x86/context
1430 The test suite still spits tons of errors because some obscure
1431 force prevents us from removing the temporary directories
1432 arguing that they still contain some metadata I've never heard of.
1433 Smpi fails because seq is not installed.
1434 Everything seems to work properly beside of that.
1437 - AIX version 5.3 (both contexts and pthread)
1438 Smpi still fails there because mktemp is not installed.
1439 XML inclusions seems rosty on AIX.
1441 * Windows: it's still lagging behind. If you want to help, please
1444 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1446 SimGrid (3.3) stable; urgency=high
1450 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1451 [Malek Cherier & Mt]
1453 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1456 Until now, GRAS were using MSG as an interface to SURF. It was
1457 quite difficult because both interface have several differences
1458 (MSG channels vs GRAS sockets were the most notable point).
1460 This also opens the gate to SMPI (which should occur soon) and speed
1461 up simulations by to 40% (even if it were not the main goal).
1463 **************************************
1464 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1465 **************************************
1466 The point is that events occuring at the exact same timestamp are
1467 not scheduled in the same order with the old and new version. This
1468 may be enough to completely change the execution of simulations in
1469 some cases. Sorry for the inconvenience.
1471 * Cleanup and upgrade the XML format to push further scalability
1472 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1474 * Improve the testing infrastructure with tesh. Now a very large part of
1475 the code is tested not only by being run but also by checking that the
1476 output match an expected output [Mt].
1478 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1479 is really less memory-demanding, which should allow you to use
1480 larger files in SimGrid [AL].
1482 * Inform valgrind about our contextes, so that it becomes usable
1483 with the default (and more effecient) version of SimGrid
1484 [contributed by Sékou Diakite, many thanks]
1487 * Introduce a listener thread in charge of receiving incoming
1488 messages from the network. It allows to overlap communication and
1489 computation but most notably, it removes some stupid deadlocks due
1490 to the fact that so far, a process could not send and receive at
1491 the same time. This made most non trivial communication schema
1493 * Convert the PIDs from long int to int to match the MSG ones (and
1494 linux ones too) [Mt]
1495 * New function: gras_agent_spawn() to launch a new process on
1496 current host. Only working in simulation for now. [Mt]
1497 * New function: gras_os_hostport() returning a constant form (ie,
1498 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1501 * Make the backtrace of exceptions more human readable [Mt]
1502 * New module: xbt/str [Mt]
1503 a ton of string utility functions (split, join, printf to a newly
1504 allocated buffer, trim, etc)
1505 * New module: xbt/hash [Mt]
1506 SHA1 hashing algorithm (more to come if needed)
1507 * New module: xbt/synchro [Mt]
1508 synchronization tools (mutex and conditions) working the same way
1509 in simulation and in real life (mainly useful for GRAS, but not
1511 * New module: xbt/queue [Mt]
1512 classical producer/consumer synchronization scheme
1513 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1514 (using the classical functions will get serialized) [Mt]
1515 * Make dictionary internal table dynamic. No need to specify its size
1516 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1518 * Make sure the log channels are organized as a tree under windows
1519 (because of ANSI C compatibility issue, any channel were child of
1520 root directly) [Mt].
1523 * Cleaned many thing in surf and fixed a few bugs [AL].
1524 * Add a nice command line configuration mechanism to compose models [AL].
1525 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1526 the previous one (KCCFLN05). It relies on something that looks like
1527 a max-min sharing mechanism but cannot be written as such. A new solver
1528 was thus designed [AL].
1529 * Add a new solver to lmm. Based on Lagrange optimization and
1530 gradient-based descent, it enables to efficiently maximise systems s.a
1532 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1534 This solver enables to propose two new network models for TCP Reno and
1535 TCP Vegas based on Low's work. These models still need to be fully
1536 tested though [Pedro Velho].
1539 * Bug fix in SD_simulate. Now the time bound given as argument is
1541 * Use the new parallel task model (ptask_L07) as default.
1542 * Use the SURF command line configuration mechanism.
1543 * 0-size tasks (for synchronization) should now work.
1545 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1547 SimGrid (3.2) stable; urgency=high
1551 We still experience issues on this platform, but we believe that at
1552 least MSG is usable.
1554 GRAS API BREAKAGE (for simplification purpose, sorry):
1555 * the gras_msgtype_by_name is not used anymore. Instead of
1556 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1557 you can write (and must)
1558 gras_msg_send(toserver, "request", &request);
1559 - If you still want to pass a gras_msgtype_t to the function (to cache
1560 the type and avoid the lookup time), use the gras_msg_send_() variant.
1561 - Impacted functions:
1562 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1563 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1564 * The callbacks are now expected to return 0 when everything went well
1565 (just like the main() function)
1567 GRAS new features and improvements:
1568 * New module mecanism where user code can use per process globals [Mt]
1569 This is similar to gras_userdata_*() functions, but for libraries. It
1570 factorize some code developped over and over in the examples and AMOK.
1571 It has still to be documented and used (only amok/peermanagement is
1573 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1574 loss when they were small enough to fit into the buffer and sent quickly
1575 enough so that they can all get received in one shoot.
1576 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1577 if not found. Use the *_or_null() variant for the old semantic.
1578 * In gras_msg_handle, do not discard messages without callback.
1579 They are probably messages to be explicitly awaited later (ie, proofs of
1580 mis-synchronization in userland since they are sent before being awaited)
1581 No big deal usually.
1582 * gras_socket_meas_send/recv: semantic changed!
1583 The numerical arguments used to be (1) the total amount of data to send
1584 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1585 messages. This was need for the fool willing to send more than MAXINT
1586 bytes on quite fat pipes.
1589 * Do really rename the hostmanagement module to peermanagement. [Mt]
1590 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1591 but this is rather new and this was documented in the module
1592 documentation (poor excuses, I admit)
1593 * Bandwidth measurement semantic changed! This follows the changes to
1594 gras_socket_meas_send/recv explained above.
1597 * A sequential mode has been added to the workstations. When a workstation
1598 is in sequential mode, it can execute only one task, and the other tasks
1599 are waiting in a FIFO. [Christophe Thiery]
1602 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1603 model for SIMDAG. [Christophe Thiery]
1604 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1605 equal to 0, which caused some bad behaviors in
1606 saturated_constraint_set_update. I now use a threshold mechanism like in
1610 * When running manually src/testall, you select specific units [Mt]
1611 testall is the result of our cunit mecanism, and should replace all
1612 the scripty thingy around since bash don't run easily on billware.
1614 * A mallocator system has been added. [Christophe Thiery]
1615 Mallocators allow you to recycle your unused objects instead of freeing them
1616 and allocating new ones.
1618 Documentation update:
1619 * FAQ reworking + New FAQs:
1620 - "Valgrind spits tons of errors!" [Mt]
1621 - "How to repport bugs" [Mt]
1622 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1623 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1624 - Communication time measurement within MSG [AL]
1625 - I experience weird communication times when I change the latency [AL]
1626 * GRAS tutorial [Mt]
1628 - an introduction to the framework and to the used communication model
1629 - an initiatic tour introducing the most proheminent features:
1631 . Lesson 0: Installing GRAS
1632 . Lesson 1: Setting up your own project
1633 o Part 2: Message passing
1634 . Lesson 2: Exchanging simple messages
1635 . Lesson 3: Passing arguments to the processes (in SG)
1636 . Lesson 4: Attaching callbacks to messages
1637 . Lesson 5: Using globals in processes
1638 . Lesson 6: Logging informations properly
1639 . Lesson 7: Using internal timers
1640 . Lesson 8: Handling errors through exceptions
1641 . Lesson 9: Exchanging simple data
1642 . Lesson 10: Remote Procedure Calling (RPC)
1643 . Lesson 11: Explicitely waiting for messages
1644 . Recapping of message passing features in GRAS
1645 - A HOWTO section containing:
1646 o HOWTO design a GRAS application
1647 More are due, of course. They will come latter. In the meanwhile, you can
1648 check the examples which are still here.
1650 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1652 SimGrid (3.1) stable; urgency=high
1656 There was a stack corruption somewhere, visible only when optimizing
1657 with these versions. [Vince]
1660 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1661 module that enabled to play with Directed Acyclic Graphs. It is built
1662 directly on top of SURF and provides an API rather close to the old
1663 SG. Some old codes using SG are currently under rewrite to check that
1664 all needful functions are provided. [Christophe Thiery]
1667 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1668 extension of the classical CLM03 model that gracefully handles
1669 failures. This is now the default model for MSG and GRAS. It doesn't
1670 handle parallel tasks yet though. [AL]
1671 * Bug fix: Weights were not correctly set in the network part.
1672 WARNING: This may have resulted in incorrect results with simulations
1673 where there are more than one flow on a given link. [AL]
1676 * After a (long ?) discussion on simgrid-devel, we have decided that the
1677 convention we had on units was stupid. That is why it has been decided
1678 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1679 WARNING : This means that all previous platform files will not work as
1680 such with this version! A warning is issued to ask users to update
1682 A conversion script can be found in the contrib module of the CVS, under
1683 the name contrib/platform_generation/surfxml_update.pl [MQ]
1686 * Bug fix: Processes were started in reverse order, wrt deployment file.
1687 WARNING: if your code relies on this bug, please fix it. [AL]
1688 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1689 being executed on two different locations. [AL]
1690 * Bug fix: Failures are now better supported thanks to Derrick's tests
1691 (there was many failure situations I hadn't thought of and that weren't
1692 correctly handled). [AL]
1693 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1697 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1698 Exception occurring on server-side are propagated back to client (!).
1700 API CHANGE: the callback changed their prototype. Change:
1701 int my_handler(gras_socket_t expeditor, void *payload_data) {
1703 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1704 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1706 * New! function: gras_msg_handleall to deal with all messages arriving
1707 within a given period.
1708 * New! function: gras_socket_server_range to get a server socket in a
1709 range of port numbers (ease to avoid port number conflicts) [MQ]
1710 * New! gras processes display their backtrace when they get a SIGUSR1
1711 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1712 Sweet to debug RL processes [MQ]
1716 - Do not force experiment sizes to be expressed in kb, or it becomes
1717 impossible to measure the latency this way (needs one byte-long tests)
1718 WARNING: this changes the amok_bw_* function semantic. [MQ]
1719 - Implements the link saturation stuff. [MQ]
1720 * Peer management module:
1721 New! module factorizing code that we wrote over and over [MQ].
1724 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1725 - Test units are placed directly into the library code, they get extracted
1726 automatically and placed into the src/testall binary.
1727 - Convert most of the XBT tests to this system.
1728 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1729 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1730 live in the same C code. This required to change a little bit the API
1731 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1732 * New module: sparse graph structure with basic algorithms (this is work
1733 in progress and the API is not considered to be frozen yet). [AL]
1734 * Display more information on backtraces: source line & function names are
1735 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1736 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1737 * Reworked a little bit some #include statements to load only required
1738 headers. Some user code that relied on SimGrid to include stdlib or
1739 stdio may need to include it by themselves. [AL]
1740 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1742 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1743 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1744 but I guess that almost nobody used those parts. [MQ]
1746 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1748 SimGrid (3.0.1) stable; urgency=low
1751 * Unfortunately, I had missed 5 misnamed functions:
1752 xbt_fifo_item_t xbt_fifo_newitem(void);
1753 void xbt_fifo_freeitem(xbt_fifo_item_t);
1754 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1755 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1756 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1757 They're now deprecated. Please use their new versions:
1758 xbt_fifo_item_t xbt_fifo_new_item(void);
1759 void xbt_fifo_free_item(xbt_fifo_item_t);
1760 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1761 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1762 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1764 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1765 * Documentation: xbt_log module unmercifully reworked [MQ]
1766 * Bugfix: there was a problem with the ending of contexts with
1767 the pthread backend. It caused some weird deadlock or behavior
1768 depending on the pthread implementation. [AL]
1769 * Bugfix: get the exceptions raised in the simulator repport where
1770 and why they come from when they are not catched in time [AL, MQ]
1773 * Bugfix: Do repport the error when two non-connected hosts try to
1774 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1777 * Add additionnal checkings on communications. Assert that two
1778 communicating hosts are connected by a set of links... [AL]
1781 * Add additionnal checkings on channel values in communication [AL]
1782 * New: MSG_task_get_source to see on which host a task was generated [HC]
1783 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1784 the number of tasks waiting to be received on channel and sent
1786 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1787 waits for the first task coming from a given host.. [AL]
1789 GRAS new functionnalities: [MQ]
1790 * Enhance the parsing macro to allow the size of multidimentional objects
1791 to be given thru annotations.
1792 * New example (and documentation): Matrix Multiplication a la RPC
1793 (as when I was young!) and fix a bunch of bugs found on the way.
1795 GRAS performance improvements: [MQ]
1797 * Reduce the amount of cbps creation/destruction by making it static to
1798 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1801 * Change libdata to a set so that we can search for stuff by ID (and thus
1802 reduce the insane amount of dict lookups)
1805 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1806 RL communication on top of SG since it's so uneffective.
1807 It may also be used for inter-thread communication in RL, one day.
1808 * Use gras_datadesc_copy() to exchange messages on top of SG
1810 - improve message exchange performance on top of SG
1811 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1812 * Don't exchange on the network the size of the used part of buffer,
1813 instead, specify the possible buffer size to read().
1815 - reduces the amount of read/write calls (one pair per exchange)
1816 - reduces the amount of exchanged data (the size)
1817 - allows to retrieve all arrived data on receiver side, if we don't need
1818 it right now (subsequent read will peek the buffer)
1819 - allows the receiver to proceed with the begining of the stream before
1820 everything is arrived
1821 - make it possible to build an iov transport (using readv/writev)
1823 - take care of the data with non-stable storage (like stacked data),
1825 * If possible, TCP send uses vector I/O (when writev() is here)
1826 - Don't use it for receive since we send data sizes and data on the
1827 same stream, so we wouldn't be able to chain large amount of chunks
1828 before having to flush the stuff to read the size.
1829 * Rework the transport plugin mecanism to simplify it and reduce the
1830 amount of pointer dereferencement when searching for the right function
1833 * I guess that now, we do almost as few system calls as possible while
1834 doing as few data copy as possible.
1836 To improve it further, we could try to send all the sizes first and then
1837 all the data (to use iov on receiving size), but it's only a partial
1838 solution: when you have 2 dimensional data, the sizes of the second
1839 dimension is data of the first dimension, so you need 3 streams.
1841 I'm not sure the potential performance gains justify the coding burden.
1843 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1845 SimGrid (3.00) stable; urgency=high
1848 * New! Give the possibility to hijack the surf parser and thus bypass
1849 MSG_create_environment and MSG_launch_application. Have a look at
1850 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1852 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1854 SimGrid (2.96) unstable; urgency=low
1859 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1860 This deprecates the xbt_error_t mecanisms.
1861 It modifies (simplifies) all XBT and GRAS API.
1862 MSG API keeps unchanged (exceptions raised by XBT are catched from
1863 within MSG and masked with existing error handling facilities)
1866 * New! Add a FATPIPE model. [AL]
1867 * New! Add a parallel task model. [AL]
1868 * New! Add automatically a loopback interface (in the default
1869 network model) if none was precised.
1872 * Bugfix: MSG_process_resume now works with the current running process.
1874 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1875 * Modification of MSG_task_get_compute_duration. Once a task has been
1876 processed, the value returned by this function is now equal to 0. [AL]
1877 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1878 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1879 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1880 of a m_task. For now, MSG_task_cancel only works with computation
1882 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1883 of the processor (in Mflop/s) regardless of the current load on the
1885 * API Change: use proper naming convention for MSG_getClock and
1886 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1888 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1889 This function changes the priority of a computation task. This priority
1890 doesn't affect the transfer rate. A priority of 2 will make a task
1891 receive two times more cpu power than the other ones. This function
1892 has been added to suit the needs of Nguyen The Loc and hasn't been that
1893 much tested yet. So if it fails, please report it and send me your code.
1895 * API Change: removed all functions and types that were marked "deprecated"
1896 since many months. Renamed MSG_global_init_args to MSG_global_init.
1898 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1900 SimGrid (2.95) unstable; urgency=low
1903 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1904 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1905 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1908 * Bugfix: Allow absolute paths for platform description files [MQ]
1909 * Bugfix: do free the variables after use. Leads to drastic performance
1911 * Implement max_duration (ie, timeouts) on resources [AL]
1914 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1916 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1917 the needed support (timeouts) [AL]
1920 * Implement measurement sockets. You can now get the bandwidth between two
1921 hosts thanks to AMOK (see below). [MQ]
1922 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1923 dynar over the network (yeah) [MQ]
1924 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1925 * simple/static token-ring example. [Alexandre Colucci and MQ]
1926 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1927 in place before [MQ]
1930 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1931 type parsing. "short" and "long" modifiers were also ignored; other
1932 modifier (such as reference level) are still ignored. [MQ]
1933 * Embeed the buffer size within the buffer itself on SG. [MQ]
1934 That way, send() are atomic and cannot get intermixed anymore (at least
1935 the ones which are less than 100k; bigger messages still have the issue)
1936 * Array size pushed by the field, not by the field type (or each
1937 and every long int will push stuff to the cbps) [MQ]
1938 * use select() to sleep since it allows to portably sleep less than one
1941 GRAS (minor cleanups)
1942 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1945 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1946 * type category 'ignored' killed as it was never used and were difficult
1948 * whether a type can cycle or not is now a flag, leaving room for more
1949 flags (as "ignored", if we feel the need one day ;)
1950 * Rename raw sockets to measurement sockets since "raw" has another
1951 meaning in networking community.
1954 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1955 and offers features not belonging to GRAS but that most applications
1956 need. One day, it may be a set of plugins loadable at runtime.
1957 * New module: bandwidth
1958 bandwidth measurement between arbitrary nodes running this module. [MQ]
1960 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1962 SimGrid (2.94) unstable; urgency=low
1964 The first beta release of SimGrid 3 !
1968 * Update the main page and the FAQ. Adding references to gforge.
1971 * Add a gras_os_getpid function.
1974 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1975 * Extend the logs so that they also print PID, hostname, date, ... if
1977 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1978 and kill the old version which were in testsuite/
1979 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1983 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1984 for users. Instead, each programming environment (GRAS, MSG,...) use it
1985 when needed to provide such a feature to users.
1986 Don't shortcut the mecanism or you will also shortcut the virtualization
1987 you need on the simulator.
1991 * Cleanups in configury with regard to compile optimization/warning flags.
1992 Also add -fno-loop-optimize to any powerpc since it's the optimization
1993 killing gcc (< 3.4.0).
1994 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1995 needed by MSG examples complications
1996 * Borrow configury beautifications from PHP
1999 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2000 warning (thanks loris for stumbling into this one).
2001 * Bugfix: stop loading private headers (gras_config.h) from the public
2005 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2006 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2007 RL would imply a DNS resolver, which is *hard* to do in a portable way
2008 (and therefore delayed).
2009 * Implement a real timer mecanism and use it in timing macros. This allows
2010 to avoid rounding errors and get a 0.000005 sec precision in timing
2011 macros. While I was at it, various cleanups:
2012 - allow to declare more than one timed section per file (fix a stupid bug)
2013 - move some private declaration to the right place
2014 - merge conditional execution and timing macros into emulation module
2015 - document the module
2016 - make sure the module cleanups its mess on gras_exit
2017 * Documentation improvements:
2018 - (new) how to compile applications using GRAS
2019 - (new) emulation support (timing macros)
2021 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2023 SimGrid (2.93) unstable; urgency=low
2025 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2028 - Use Paje properly where used. Still to be sanitized properly.
2029 - Portability fix: Add an implementation of the contexts using pthread
2033 - Add xbt_procname(): returns the name of the current process.
2034 Use it to show the current process's name in all logging.
2036 - fix detection of older flex version and the reaction, since we do
2037 depend on modern ones (we use lex_destroy)
2038 - Better separation of SG and RL in the libs: remove all simulation code
2039 from libgras. As a result, this lib is now only 200k when stripped.
2040 Some of the xbt modules may also be duplicated (two sets and such) and
2041 should be cleaned/killed before SG3.
2042 - Insist on using xlC on AIX because of weird problems involving gcc there.
2043 - Cleanup the make remote stuff. This is now done by scripts
2044 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2045 mainly for our private use, but we're working on changing them to user
2048 - Bugfix: flush the socket on close only if there is some *output*.
2049 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2051 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2054 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2056 SimGrid (2.92) unstable; urgency=low
2058 Alpha 3 on the path to SimGrid 3
2062 - New! First try of benchmarking macros.
2063 - New! First try so that gras_stub_generator generate deployment and
2064 remote compilation helpers.
2066 - Bugfix: Initialization fix in msg_test.
2070 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2072 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2074 - API Change: gras_os_sleep to take the amount of seconds as a double.
2075 Accepting an int was error prone since it was the only location where
2076 seconds were coded as such. It leaded to damn rounding errors.
2077 - Bugfix: Hard to belive that timers ever worked before this.
2079 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2081 SimGrid (2.91) unstable; urgency=low
2083 Alpha 2 on the path to SimGrid 3
2087 - Bug fix in the lmm_solver.
2089 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2090 through the function MSG_paje_output.
2091 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2092 - It is possible to bound the rate of a communication in MSG with
2093 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2094 to write it in the changelog).
2095 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2100 - Various cleanups to the autotools stuff
2101 - Begin to move Gras examples to examples/gras/
2102 - Let make distcheck work again (yeah!)
2104 - documentation overhauled using doxygen.
2105 gtk-doc-tools is dead in SimGrid now.
2106 - Automatically extract all existing logging categories, and add the list
2107 to the documentation (long standing one, to say the less)
2109 - Cleanup the known architecture table. Reorder the entries to group what
2110 should be, and use a more consistent naming scheme.
2111 (some of the test dataset are still to be regenerated)
2112 - New! Allow library to register globals on each process just as userdata
2114 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2115 lookup time (for now).
2116 Use it in msg and trp.
2117 This cleans a lot the internals and helps enforcing privacy of the
2118 headers between the gras components.
2119 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2120 - Bugfix: gras_os_time was delirious in RL.
2121 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2123 - Reenable GRAS now that it works.
2125 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2127 SimGrid (2.90) unstable; urgency=low
2129 Alpha 1 on the path to SimGrid 3
2131 * It is a long time since the last release of SimGrid. I'm sorry about
2132 that but as I had told you, I was rewriting a lot of things. I apologize
2133 to those who had been reporting bugs to me and that I had not answered.
2134 If your bug is still in the new version, please tell me. Here is a
2135 summary of the main changes.
2137 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2138 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2139 portability, speed, and a lot more but you'll figure it out later.
2140 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2141 I wanted to release the new MSG as soon as possible and I have
2142 broken GRAS, which is the reason why, for now, only MSG is fully
2143 functional. A laconic description of these projects is available
2144 in the documentation.
2146 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2147 kernel optimized for our needs (called SURF but only the developers
2148 should use it). Hence, MSG is now roughly 30 times faster and I think
2149 that by rewriting a little bit MSG, I could event speed it up a little
2150 bit more. Beside the gain in speed, it is also much easier to encode a
2151 new platform model with SURF than it was with SG. More to come...
2153 * REVOLUTION 3: I have tried to change a little as possible the API of
2154 MSG but a few things really had to disappear. The main differences
2155 with the previous version are :
2156 1) no more m_links_t and the corresponding functions. Platforms are
2157 directly read from a XML description and cannot be hard-coded
2158 anymore. The same format is used for application deployment
2159 description. The new format is described in the documentation.
2160 Have a look in tools/platform_generation. There is a tiny script
2161 that converts from the old platform format to the new one. Concerning
2162 the application deployment format, parsing the old one is tricky.
2163 I think most of you should however be able to convert your files. If
2164 it is really an issue, I can write a C code that does the conversion.
2166 2) the toolbox tbx does not exist anymore. We now have a library
2167 with much more data-structures but without the hash-tables (we have
2168 dictionaries that are much faster).
2170 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2172 *****************************************************************************
2173 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2174 * don't feel like converting the oldies. (Mt) *
2175 *****************************************************************************
2178 Version 2.90: "the long awaited one"
2179 - Finished rewriting and debugging MSG. Rewrote the documentation.
2180 - disable GRAS for now since it needs to be ported to the newest SG
2183 - Finish the port to windows (using mingw32 for cross-compile)
2186 - Main loop and datastructures of SURF. A cpu resource object is
2187 functional. Surf can thus be used to create cpu's with variable
2188 performance on which you can execute some actions.
2190 2004-11-15 Martin Quinson
2191 - Port to ARM. Simply added the alignment and size descriptions. Should
2192 work, but the ARM machines are so slow that I didn't had the opportunity
2193 to 'make check' over there yet.
2195 2004-11-15 Arnaud Legrand
2196 - Trace manager now written. It uses a heap structure and is therefore
2197 expected to be efficient. It may however be speeded up (particularly
2198 when many events occur at the same date) by using red and black
2199 trees. One day maybe...
2200 - Max-min linear system solver written. It uses a sparse matrix
2201 structure taking advantage of its expected use. Most operations are
2202 O(1) and free/calloc are called as few as possible. The computation of
2203 the minimum could however be improved by using a red and black tree
2206 2004-11-03 Arnaud Legrand
2207 - Rename every gras_* function that was in xbt/ to its xbt_
2209 - Add a heap and a doubly-linked list to xbt
2210 - Added a dichotomy to the dictionaries. make check works as well before
2211 so I assume that the patch is correct. I do not know however if things
2212 run effectively faster than before now. :)
2214 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2215 SimGrid, and the version number is bumped to 2.x
2217 2004-10-29 Martin Quinson
2218 - Introduction of the remote errors.
2219 They are the result of a RMI/RPC on the remote machine.
2220 ErrCodes being scalar values, you can't get the host on which those
2221 errors did happen. Extending the error mechanism as in Gnome is possible.
2222 No idea yet whether it is a good idea.
2224 2004-10-28 Martin Quinson
2225 - Interface revolution: the Starred Structure Eradication.
2226 I used to do typedef struct {} toto_t; and then handle *toto_t.
2227 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2228 - ???_t is a valid type (builded with typedef)
2229 - s_toto_t is a structure (access to fields with .)
2230 - s_toto is a structure needing 'struct' keyword to be used
2231 - e_toto_t is an enum
2232 - toto_t is an 'object' (struct*)
2234 typedef struct s_toto {} s_toto_t, *toto_t;
2235 typedef enum {} e_toto_t;
2236 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2237 s_toto_t) is private.
2239 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2240 so that it can set the variable to NULL. It was so for dicts and sets,
2241 it changed for dynars.
2243 - Fix a bunch of memleaks in dict_remove
2244 - Fix a bug in sg/server_socket opening: it failed all the time.
2246 2004-10-07 Martin Quinson
2247 - Speed up dynar lookup operation a bit.
2249 gras_dynar_get is dead.
2251 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2252 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2253 (which gives you the address of the stored data).
2255 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2256 copy of the data using an affectation to do the job and not a memcpy.
2258 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2259 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2261 It does not really speedup the dynar test because they are
2262 setting elements all the time (and look them seldom). But the dict does
2263 far more lookup than setting.
2265 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2267 2004-10-05 Martin Quinson
2268 - Allow to (en/dis)able the cycle detection at run time.
2270 Whether we should check for cycle or not is now a property of each
2271 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2272 datadesc_cycle_unset allow to remove this property when previously set.
2274 Note that the cycle detection is off by default since it impacts the
2275 performance. Watch the data you feed GRAS with ;)
2277 This property is hereditary. Any element embedded in a structure having it
2278 set have it set for the time of this data exchange.
2280 You should set it both on sender and receiver side. If you don't set it on
2281 sender side, it will enter an endless loop. If you forget on receiver
2282 side, the cycles won't be recreated after communication.
2284 - Header reorganization.
2285 Kill gras_private.h, each submodule must load the headers it needs.
2287 2004-10-04 Martin Quinson
2288 - Interface revolution: do not try to survive to malloc failure.
2290 Now, gras_malloc and friends call gras_abort() on failure.
2291 As a conclusion, malloc_error is not a valid error anymore, and all
2292 functions for which it was the only gras_error_t return value are
2293 changed. They now return void, or there result directly.
2294 This simplify the API a lot.
2296 2004-09-29 Martin Quinson
2297 - Re-enable raw sockets.
2298 Created by gras_socket_{client,server}_ext;
2299 Used with gras_raw_{send,recv}
2302 It should allow to kill the last bits of gras first version soon.
2304 This is not completely satisfactory yet (duplicate code with
2305 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2308 - Simplify transport plugin (internal) interface by not passing any
2309 argument to _server and _client, but embedding them in the socket
2312 2004-09-28 Martin Quinson
2313 - Finish the port to AIX.
2314 autoconf was my problem (segfault within the malloc replacement
2315 function. No idea why)
2317 2004-09-16 Martin Quinson
2318 - Fix some size_t madness on 64bit architectures.
2320 2004-09-08 Martin Quinson
2321 - Reduce the number of system headers loaded, overload some more system
2322 calls (such as malloc to cast the result of the system one, and work
2324 - Fix and reintroduce the config support
2326 2004-09-07 Martin Quinson
2327 - Source code reorganization to allow Arnaud to surf all over there.
2328 - Allow to document the logging categories.
2329 - Remove all uppercase from logging categories and useless cleanup in names.
2331 2004-08-18 Martin Quinson
2332 Version 0.6.2 (protocol not changed; API changed)
2333 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2334 gras_error_t), and NULL when not found. Functions expecting a msgtype
2335 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2336 hopefully usefull message.
2337 - Portability to prehistoric sparcs again
2339 2004-08-17 Martin Quinson
2340 Version 0.6.1 (protocol not changed; ABI not changed)
2341 - prealloc some buffers to speed things up
2343 2004-08-11 Martin Quinson
2344 Version 0.6 (protocol not changed; ABI expended)
2345 - The parsing macro can deal with the references, provided that you add
2346 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2348 2004-08-09 Martin Quinson
2349 Version 0.5 (protocol not changed; ABI changed)
2350 - Allow to off turn the cycle detection code in data exchange at
2351 compilation time. It should be at run time, but I'm short of time (and
2352 the config stuff is still broken). That way, we keep dict out of the
2353 critical path, which is good because the performance is poor:
2354 - search not dichotomial yet
2355 - dynar give no way to access their content and memcpy everytime
2356 - In composed data description (struct, ref and so on), stop foolness of
2357 keeping the subtype's ID, but store the type itself. This keeps sets out
2358 of the critical path, which is good since they rely on dynar and
2359 dictionnaries. The only loose of that is that we cannot detect the
2360 redeclaration of a structure/union with another content (but I'm not sure
2361 the code detected well this error before anyway). We still can detect
2362 the redefinition discrepancy for the other types.
2363 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2364 it breaks the code because of type-punning used all over the place).
2365 This breaks on all non-gcc architectures (for now).
2367 All those changes (plus the buffer of last time) allow me to gain 2 order
2368 of magnitude on cruel tests consisting of 800000 array of integers on two
2369 level of a hierarchical structure (200 secondes -> 4 secondes)
2372 - the selector of reference must now return the type it points to, not
2373 the ID of this type.
2375 2004-08-06 Martin Quinson
2376 Version 0.4 (protocol changed; ABI not changed)
2377 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2378 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2379 - Implement a buffer transport, and use it by default (it relies on tcp in
2380 real life and on sg in simulation).
2381 That's a bit hackish since I had a new field to the structure to store
2382 its data without interfering with the subtype ones. Inheritance
2383 is tricky in C. And that's a kind of reverse inheritance with one class
2384 derivating two classes. Or maybe a game with java interfaces. Anyway,
2385 that's damn hard in C (at least).
2386 Moreover, I got tired while trying to ensure plugin separation and
2387 genericity in SG mode. MSG wants me to do weird things, so let's go for
2388 cruel hacks (temporarily of course ;).
2389 See comment in transport_private.h:71
2390 - do not include all the _interface headers in private but in the files
2391 which really need them (to cut the compilation time when they are
2394 2004-07-26 Martin Quinson
2395 Version 0.3 (protocol not changed; ABI changed)
2396 - Major overhault of the datadesc interface to simplify it:
2397 - shorted the function names:
2398 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2399 - add a trivial way to push/pop integers into the cbps without malloc.
2400 This allows to make really generic sub_type description, which simply
2401 pop their size of the stack.
2402 - add a function gras_datadesc_ref_pop_arr() which does what users want
2403 most of the time: Declare a dynamic array (which pops its size of the
2404 stack) and declare a reference to it. Poor name, but anyway.
2405 - kill the post-send callback, add a post-receive one
2407 2004-07-23 Martin Quinson
2408 Version 0.2 (protocol changed; ABI changed)
2409 - add some testing for cpbs in the test cases, and fix some more bugs.
2410 This invalidate again the little64 data file, since I cannot regenerate
2412 - remove an awfull optimization in the logging stuff, allowing me to:
2413 - understand it again
2414 - learn gcc how to check that the argument match the provided format
2415 - fix all errors revealed by gcc after that
2416 - internal keys of dict are not \0 terminated. Deal with it properly in
2417 loggings instead of segfaulting when the user want to see the logs :-/
2419 2004-07-22 Martin Quinson
2420 - Fix some stupid bug preventing cbps (callback postit) from working
2422 2004-07-21 Martin Quinson
2423 - Some documentation cleanups
2424 - remove the useless last argument of msgtype_declare
2425 - rename the Virtu functions to fit into the 'os' namespace
2426 - move headers src/include -> src/include/gras/ and stop fooling with
2427 gras -> . symbolic link
2428 - make distcheck is now successful
2430 2004-07-19 Martin Quinson
2432 - Build shared library also
2433 - Install html doc to the right location
2434 - stop removing maintainer files in make clean
2435 - build tests only on make check
2437 2004-07-13 Martin Quinson
2439 - No major issue in previous version => change versionning schema
2440 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2441 corresponding dataset.
2443 2004-07-11 Martin Quinson
2445 - Get it working with any kind of structure (we can compute the padding
2446 bytes remotely for all the architectures I have access to)
2447 - Implement the structure parsing macro (still not quite robust/complete)
2448 - Improvement to the remote testing toysuite
2450 2004-07-10 Martin Quinson
2451 [autoconf mechanism]
2452 - get ride of a bunch of deprecated macros
2453 - actually run the test for two-compliment, not only compile it :-/
2454 - test whether the structures get packed (and bail out if yes. Damn.
2455 Alignment is a serious matter)
2456 - test whether the structures get compacted (but respecting the alignment
2457 constraints of each types)
2458 - test whether the array fields of structures can straddle alignment boundaries
2460 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2461 (took me 2 hours to find that bug, looking at the wrong place)
2462 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2463 before sending/receiving and are used to compute the offsets of fields
2464 - Given that padding size depend even on compiler options, keep track of
2465 alignment and aligned_size only for the current architecture. Not a big
2466 deal since we send structure fields one after the other (seems
2468 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2469 program, let it work on linux/gcc/little32. portability todo.
2471 2004-07-08 Martin Quinson
2472 - import and improve remote compilation support from FAST
2473 - make sure make check works on half a dozen of machines out there
2475 2004-07-07 Martin Quinson
2476 Let's say it's version 0.0.3 ;)
2477 - Implement conversions (yuhu!)
2478 - Let it work on solaris (beside conversion, of course)
2479 - Stupid me, using rand() to generate the conversion datatests in not wise.
2481 2004-07-06 Martin Quinson
2482 - Let make dist work, since I'm gonna need it to compile on remote hosts
2483 - Let Tests/datadesc_usage write the architecture on which the file was
2484 generated as first byte.
2485 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2486 architecture descriptions.
2487 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2488 on those architectures.
2489 - Optimization: send/recv array of scalar in one shoot
2491 2004-07-05 Martin Quinson
2492 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2494 - Plug a whole bunch of memleaks
2495 - each process now have to call gras_{init,exit}. One day, their log
2496 settings will be separated
2497 - Continue the code factorisation between SG, RL and common in Transport.
2499 2004-07-04 Martin Quinson
2501 - Redistribution between SG and RL.
2502 We wanna have to accept in SG, so move accepted related parts of RL in
2503 the common part. (more precisely, the dynar of all known sockets is no
2504 more a static in transport.c, but part of the process_data)
2506 [gras_stub_generator]
2507 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2509 2004-07-03 Martin Quinson
2510 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2511 them all up in one shot)
2513 - Fix the ugly case of reference to dynamic array.
2514 - New (semi-public) function gras_datadesc_size to allow the messaging
2515 layer to malloc the needed space for the buffer.
2517 - gras_socket_close now expect the socket to close (and not its address to
2518 put NULL in it after it). This is because the socket passed to handlers
2519 is one of their argument (=> not writable).
2521 - propagate the interface cleanup from last week in datadesc, ie remove a
2522 superfluous level of indirection. User pass adress of variable
2523 containing data (both when sending and receiving), and not of a variable
2524 being a pointer to the data. Let's say that I like it better ;)
2525 The price for that is constructs like "int msg=*(int*)payload" in
2526 handlers, but it's a fine price, IMHO.
2528 - Let it work in RL (yuhu)
2530 2004-06-21 Martin Quinson
2532 - porting SG plugin and SG select to new standards (works almost).
2533 - plug memleaks and fix bugs around.
2536 - cleanup the prototype of data recv and force users to specify when they
2537 want to handle references to objects. Test case working even for cycles.
2538 - plug memleaks. Valgrind is perfectly ok with this.
2540 2004-06-12 Martin Quinson
2542 - cleanup the separation between plugin and main code in plugin creation
2544 2004-06-11 Martin Quinson
2546 - Reput hook for raw sockets, needed for BW experiments
2547 - kill a few lines of dead code
2548 [Data description] Interface cleanup
2549 - gras_datadesc_by_name returns the searched type or NULL.
2550 That way, no variable is needed to use a type desc once, which makes
2552 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2553 two parameters were strings (field name, type name), leading to
2555 [Dicos] Interface cleanup
2556 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2557 This is consistant with the dynar API.
2559 2004-04-21 Martin Quinson
2561 - Porting to new standards.
2563 - interface cleanup.
2564 There is no bag anymore, no need to take extra provision to mask the
2565 pointers behind "ID".
2566 Better splitup of functions between files create/exchange/convert.
2567 This is still a bit artificial since convert and receive are so
2568 interleaved, but anyway.
2570 - add a queued message list to procdata (the ones not matching criteria
2572 - factorize some more code between SG and RL wrt procdata
2574 - use gras_exit in example to track memleaks
2575 - get rid of gs_example now that GS is properly integrated into gras
2576 - update run_test to integrate the latest tests (datadesc)
2578 - rename WARNINGn macros to WARNn since it prooved error-prone
2580 2004-04-19 Martin Quinson
2582 - register init/exit functions within gras module mechanism
2583 - send/receive function.
2584 Convertion is not implemented, but short-cutted if not needed.
2585 struct/array elements are sent one by one (instead of block-wise), but
2586 nobody really cares (yet). Get a prototype before optimizing.
2587 - tests (using a file socket) for DD send/receive on:
2588 - base types: int, float
2589 - array: fixed size, string (ie ref to dynamic string)
2590 - structure: homogeneous, heterogeneous
2591 - chained list, graph with cycle
2592 Believe it or not, valgrind is not too unhappy with the results. The
2593 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2597 - Bugfix when using a filename explicitely (instead of '-')
2599 2004-04-09 Martin Quinson
2601 - factorize more code between RL and SG in socket creation
2602 - Complete the implementation and tests of:
2604 o file (only in RL, and mainly for debugging)
2606 I lost 3 days to design a portable address resolver, and then decided
2607 that the prototype mainly have to run on my box.
2608 Addressing portability too early may be like optimizing too early :-/
2610 - use gras_init in the Tests instead of the crappy parse_log_opt
2611 (the latter function is removed)
2612 [Conditional execution]
2613 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2614 [Code reorganisation]
2615 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2616 Build examples against the RL library, since there is no way to disable
2617 its creation for now.
2619 For information, the beginning of coding on GRAS was back in june
2620 2003. I guess that every line has been rewritten at least twice since