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 - Gatherv collective is now supported
29 - Simgrid (SMPI for now) can generate replay traces as well. Option -trace-ti
30 of smpirun outputs time independent traces for the current run. One file
31 is created per process. If too many processes are simulated, this behavior
32 can be changed to one file for all processes by using the
33 tracing/smpi/format/ti_one_file flag
34 * smpirun generates the host file if needed (with given host count and platform)
35 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
36 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
37 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
38 --cfg=smpi/coll_selector:(mpich/ompi)
39 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
41 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
42 happening inside a collective SMPI call.
43 * Fix the behavior of complex data types handling.
44 * Make MPI_Wtime another synchronization point to take computations into account.
45 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
46 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
47 * Add all missing Fortran bindings, SMPI should work with Fortran 90
48 (no privatization of global variables yet)
51 * Verification of liveness properties is now available for SMPI applications
52 (in addition to MSG applications)
53 * Bugged examples using SMPI in examples/smpi/mc/
54 * Add --cfg=model-check/visited option. Allows the verification of infinite
55 programs. Detection of loops in the execution thanks to the system state
56 comparison and reduction of the state space to explore. Can be combined with
57 DPOR for safety properties.
60 * Allow to change SimGrid configuration (see --help) within the code
61 thanks to SD_config() as it can be done in MSG.
62 * Add a new function SD_task_set_amount() upon user request.
65 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
66 * Remove rule based routing (no more PCRE dependency)
67 * Add a limiter_link option to cluster tag, to specify a maximum reachable
68 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
69 * Add a loopback_bw and loopback_lat options to cluster tag.
70 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
71 routing. Now peers are encapsulated in an AS and have their own private
72 router but this is transparent.
75 * Our own implementation of getline is renamed xbt_getline, and gets
76 used even if the OS provide a getline(). This should reduce the
77 configuration complexity by using the same code on all platforms.
78 * New type: xbt_cfg_elm_boolean.
79 * Allow to use yes/no for boolean configuration options in the command line.
80 * Allow to disable SimGrid cleanups at exit from command line option.
81 There are situations where one may want a simulation to end with an exit.
82 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
83 annoying when scripting around the simulator. Adding a
84 --cfg=clean_atexit:no allows to circumvent this issue.
87 * New command line option --version, to get SimGrid version information.
88 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
90 * Supernovae builds are deprecated, and expected to be removed in the next
93 -- Mon Oct 7 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
95 SimGrid (3.9) stable; urgency=low
97 The Grasgory release: GRAS is really dead now.
99 * Complete overhaul of the internal host structures scheme.
102 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
103 was considered as experimental and badly maintained since then.
104 * Keeping it was thus a trap to our potential users, that could take
105 it instead of MSG or SMPI by mistake despite is pity state.
106 * GRAS seems to have very few users (if any), and no one volunteered
107 to maintain it further. It also induces a lot of XBT code (for
108 portability sake), that must be maintained too.
109 * For all these reasons, we killed GRAS. If someone wants to revive it
110 in the future, don't cry, our git history still remembers of GRAS.
113 * Major overhaul. Merge our documentation again as time proved that
114 splitting it was really not helping our users.
115 * Further improve the developer documentation to help newcomers
116 hacking on SimGrid itself. The user documentation (and in
117 particular, the beginner documentation) is still in a sorry state.
120 * Now works on Windows too!
121 * Much more extensive test suite, from MPICH
124 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
125 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
126 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
127 * Bug fix: let task be scheduled when the last dependency to be solved is
128 a control dependency.
129 * Remove SD_load_environment_script function.
130 Use the C sg_platf function if you want to declare a platform
134 * New function: MSG_process_get_number()
135 * Old function documented: MSG_config()
136 * Remove MSG_load_platform_script function
137 Use the C sg_platf function if you want to declare a platform
141 * Change the default value of the TCP_gamma constant (maximal size of TCP
142 congestion window) to a more realistic 4MiB value. If you notice changes in
143 your simulation results, you can fall back to the previous 20k tiny window
144 by adding --cfg=network/TCP_gamma:20000 on command line.
145 * (Hopefully) fix a bug wrt periodic availability and state traces
146 * Bug fix: use default values at start when first event in availability/state
147 trace is not at time 0.
150 * remove the "new_" part of function name sg_platf_new_trace_connect
151 (resulting in sg_platf_trace_connect), since it does not create
155 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
156 use case where it's really mandatory, and maintaining it was a pain in
158 * New: xbt_fifo_search(), search an item with a user-provided
159 comparison function instead of dumb pointer comparison.
162 * Fix the lua deployment:
163 Use `simgrid.init_application()` before deployment instead of
164 `simgrid.msg_register_application()` after.
167 * Transfer the tracing files into the corresponding modules.
169 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
171 SimGrid (3.8.1) stable; urgency=low
173 The "we are told that some people want to also *install* the simgrid
176 * Add missing file "tesh.1" to the archive.
178 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
180 SimGrid (3.8) stable; urgency=low
182 The Psssshiiiit release: SimGrid jumps into the Cloud.
185 * Add an experimental interface to manipulate VMs. They are mainly
186 process groups with very few intrinsic semantic, but they should
187 allow you to build the semantic you want easily.
188 * New function: MSG_host_set_property_value()
189 * New function: MSG_process_on_exit(). To clean memory in all cases.
190 * Bug fixes that made the host (and link) failures unusable.
191 * Add a way to auto-restart process when the host in which they are
192 executing comes back (ON_FAILURE="RESTART" on deployment file,
193 MSG_process_auto_restart_set).
194 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
195 please stop using the old ones, they are DEPRECATED.
197 * Deprecate functions MSG_global_init() / MSG_global_init_args()
198 Please use MSG_init() instead. (reducing the amount of entry
199 points in the library helps us).
200 * Make it impossible to link against the wrong version of the lib
201 * Deprecate MSG_clean(). No need to call it anymore.
202 * Function MSG_get_host_number() is not deprecated anymore.
205 * Split the doc into a user guide and a reference guide.
206 * Start a developper guide to help people hacking on SimGrid.
209 * Enable tracing by default. This modules rocks you should use it.
210 * Remove option custom_flags. Now use environment variables CFLAGS
212 * Use default cmake things to detect lua instead of home grown ones.
213 * New option "enable_mallocators" to disable mallocators, for debugging
214 purpose ("on" by default).
217 * Bug fixes around the resource failures: don't let the processes
218 survive the host they are running onto.
219 * Add an interface to auto-restart processes when the host in which they are
220 executing comes back.
221 * Ensures that SIMIX_clean is called automatically. It's not part of
222 the public interface anymore (bindings should be updated).
225 * Bug fix for when SD_Simulate is called with a positive value: be careful
226 when comparing doubles. Sometimes they are different for non significant
228 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
229 parallel task whose initial work is distributed among host according
230 to the Amdahl's law. Such tasks are created with a parameter alpha
231 that corresponds to the non-parallelizable part of the computation.
232 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
233 between two sets of workstations assuming a 1D block distribution (each
234 workstation owns a similar share of data) on both sides.
236 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
237 Data redistribution will be automatically scheduled once parent and child
238 are both scheduled. The filling of computation_amount and
239 communication_amount structures is now done seamlessly thanks to the chosen
241 * New function SD_workstation_dump to display various information
242 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
243 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
244 bandwidth on the route onto which the task is scheduled and the amount of
246 To divide the nominal bandwidth by 2, the rate then has to be :
247 rate = bandwidth/(2*amount)
248 * Compute tasks that have failed can now be rescheduled and executed again
249 (from their beginning)
250 * Increasing source code coverage (src/simdag is now covered at 95.8%
254 * Re-implement time-independent trace replay using SMPI (at the
255 smpi_smp_* level) instead of MSG. This should replace
256 examples/msg/actions/actions.c
257 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
258 hindexed and structs)
259 * Implement the exchange of non-contiguous data.
260 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
261 * Correct behavior of smpi/sender_gap and set its default value to 0
262 * Add option to asynchronously send small messages to allow better
263 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
264 specifies the size in bytes under which messages will be asynchronously sent.
265 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
266 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
269 * New C interface to define a platform: XML is now optional.
270 For more info, please check include/simgrid/platf.h.
271 * New interface to define random platforms from the C:
272 For more info, please check include/simgrid/platf_generator.h and
273 examples/msg/masterslave/masterslave_platfgen.c
274 * Export a sg_cmdline dynar containing all the arguments we got from
278 * Two new tracing options for adding comments to trace file so you
279 can track your experiments (see --help-tracing for details).
280 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
281 * Adding the SimGrid version that generated the trace file as a comment.
282 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
283 * Fix to avoid key clashes on Paje links
284 * Other minor fixes related to the Paje specification
287 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
288 and renamed to xbt_str_hash() and xbt_str_hash_ext().
289 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
290 * Greatly improve the robustness of mmalloc to user errors (such as
291 using an area after freeing it, or freeing it twice)
293 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
295 SimGrid (3.7.1) stable; urgency=low
298 * Restore the prototype of MSG_process_create_with_environment() to
299 the pre-3.7 situation by removing the kill_time argument.
300 * Add a MSG_process_set_kill_time() function instead.
303 * Fix weird behaviors when dealing with parallel tasks.
306 * Simgrid is now built as a dll.
307 * Simgrid-java now works on Windows.
308 * Simgrid-Java is now included into Windows package.
311 * First pre-build package for MacOSX.
314 * Fix compilation when using MSG_USE_DEPRECATED.
315 * Fix some compilation issues on Macs and Windows.
316 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
317 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
319 We discovered that the Lua console is broken, but we are missing the
320 manpower to fix it right now. The problem existed in 3.7 too, so we
321 are not blocking the release for that. Sorry if you depended on this
322 feature, any help would be really welcome.
324 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
326 SimGrid (3.7) stable; urgency=low
328 The "spring cleanups (before the next Big Project kicks in)" release.
331 * We can specify the SMPI latency/bandwidth factor with command line
332 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
333 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
334 You can also use the "config tag" from platform file by adding this line
335 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
336 (see "example/platforms/tag_config.xml" to use "config tag").
337 Note that the command line supersedes the platform file configuration.
338 * Change the correction factors used in LMM model, according to
339 the latest experiments described in INRIA RR-7821.
340 Accuracy should be improved this way.
341 * Use the partial invalidation optimization by default for the
342 network too. Should produce the exact same results, only faster.
343 * Major cleanup in surf to merge models and split some optimization
344 mechanisms from the core of the models. As a result you can now
345 specify which model to use (e.g., --cfg=network/model:LV08
346 --cfg=cpu/model:Cas01) and which optimization mode to use
347 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
348 Incompatible combinations should err at initialization. See
349 --help-models for the list of all models and optimization modes.
350 * The CLM03 workstation model was dropped for simplicity because it
351 used the deprecated CM02 network model. Use default instead.
352 * Rename the TCP_gamma configuration option to network/TCP_gamma
353 * Rename the coordinates configuration option to
354 network/coordinates, and document it
355 * Use now crosstraffic keyword instead of the terribly misleading
356 fullduplex keyword. It is activated by default now in the current
357 default model, use --cfg=network/crosstraffic:0 to turn it off.
358 * Ongoing refactoring the model parsing to make XML files optional
359 See include/simgrid/platf.h for details (still to be completed)
362 * Major overhaul of the documentation. Almost instructive now :/
363 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
364 functions and friends. This interface was considered as
365 deprecated since over 2 years, it's time to inform our users that it is.
366 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
367 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
368 use these (crappy) functions in your code.
369 These functions will be removed soon. Stop using them now.
370 * Deprecate MSG_get_host_{table,number}
371 Implement MSG_hosts_as_dynar() instead.
372 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
373 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
376 * Stabilize the parallel execution mode of user contexts
377 * Introduce configuration variables to control parallel execution:
378 - contexts/synchro: Synchronization mode to use when running
379 contexts in parallel (either futex, posix or busy_wait)
380 - contexts/parallel_threshold: Minimal number of user contexts
381 that must be part of a scheduling round to switch to parallel
382 execution mode (raw contexts only)
383 * Fix bugs that prevented to use suspend/resume along with
384 synchronization structures.
385 * Fix bugs in process termination that lead to invalid memory access
386 in very specific conditions.
389 * Introduce a parallel mode for the models (controlled by surf/nthreads
390 configuration item). In our tests, running the models in parallel
391 never lead to any speedups because they are so fast that the gain
392 of computing each model in parallel does not amortizes the
393 synchronization costs, even when ultra fast futexes are used.
394 This is released anyway because YMMV.
397 * Performance boost by using a swag internally to compute the set of
398 tasks that are finished and should constitute the return value of
402 * Enable it by default now that it is considered rather stable.
405 * Documentation of the tracing functions.
406 * Performance gains when tracing categorized/uncategorized resource
407 utilization by avoiding calls to get route when updating resource
408 variables. LMM constraints are being used instead.
409 * API changed to set task categories. Use MSG_task_set_category instead
410 of TRACE_msg_set_task_category, and SD_task_set_category instead
411 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
412 * Bugfix for graphicator, routes not correctly obtained, memory leaks
413 * Examples for link user variables added (see at examples/msg/tracing/)
414 * Deprecated function TRACE_msg_set_process_category completely removed
415 * Trace header updated according to the latest Paje file format
416 * Tracing network lazy updates, no longer obligate users to use full updates
417 * --cfg=tracing/platform:1 also registers power/bandwidth variables
418 * Experimental: let user code declare/set/push/pop application states for hosts
419 * API changed to allow the manual creation of graph configuration files
420 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
423 * Improve the API of Lua MSG bindings, using the Lua spirit.
424 * Each simulated process now lives in its own Lua world (globals are
425 automatically duplicated). It helps writing simulators. Will allow
426 to run Splay programs within SimGrid in the future.
427 * Add a Chord example in Lua, equivalent to the MSG one.
430 * Start the implementation of a solution to express temporal
431 properties, not only local assertions. This is still an
432 experimental work in progress, stay clear from it to be safe.
436 - Add new runtime parameters --help-logs and --help-log-categories
437 to display informations about supported logging parameters and
439 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
440 don't exists anymore.
441 * Mallocators: allow value NULL for the reset function.
443 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
444 create homogeneous dictionaries, where all the elements share the
445 same free function. Non homogeneous dictionaries will be
446 deprecated in the next release.
447 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
448 - Multi-level dictionaries are deprecated.
450 - new function xbt_dynar_search_or_negative() that is useful when
451 you have less than 2 million elements in your dynar and don't
452 want of the extra complexity of catching exceptions when the
453 element is not found.
455 - Make xbt_os_thread module (for thread portability) public.
456 Documentation is still to come, sorry.
458 - Cleanups and simplifications to make it maintainable again.
459 - Exotic features (such as memalign and valloc) were removed.
460 - The metadata were extended and improved so that the
461 model-checker becomes able to explore and inspect the heaps.
462 - This may induce a performance drop when enable_model-checking is
463 ON in cmake (even if it's not used in the simulation), but it is
464 necessary at this point to get MC working.
466 Turn model-checking OFF if simulation performance matters to you.
467 Not enabling it at runtime is not enough, disable it in cmake.
469 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
471 SimGrid (3.6.2) stable; urgency=low
473 The "Not coding new stuff allows to polish old things" release.
476 * New bindings to the NS3 packet level simulator (experimental)
477 * Use the raw (efficient) execution contexts instead of the sysv
478 (portable) ones when possible.
479 * libpcre is now mandatory in any cases since not using it led to
480 severe performance loss and possibly other issues
481 * Update the XML platforms:
482 - G5K: include the latest machine in Nancy
483 - GridPP and LCG: new platforms
484 * Documentation was partially updated, at least (more to come)
486 Bug fixes, cosmetics and small improvements
487 * Free terminated processes before the end of the simulation to avoid
488 exhausting the memory of users having very dynamic amount of
490 * Bug fix and cosmetics about canceling non-running tasks
491 * Bug fix about the dot loader's issues when using libcgraph
494 * Create an installer for windows with nsis (amd64 and win32)
495 - Add an hello world project to illustrate simgrid project creation.
496 - Embed libpcre into the Simgrid installer to avoid
497 its compilation burden
498 * The raw execution contexts should work on Apple now
499 * Port to Windows 64 bits
500 - Sysv contexts now have an implementation for this arch
501 - GRAS communication features now support this arch
502 * Drop support for borland compiler on windows
503 - this code was not maintained, and we kinda depend on gcc nowadays
504 * Fix portability issues on kfreebsd/gnu: build error about semaphores
505 * Fix portability issue on unstable ubuntu: linker became picky on
508 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
510 SimGrid (3.6.1) stable; urgency=low
512 The "Oops, we broke Macs too" release
515 * Fixed contexts detection so that raw ones are used when possible
516 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
517 leads to a strange error, with user code segfaulting sometimes when
518 the generated code is not perfectly aligned (which is not
519 controllable from the user side, depends on the amount of code)
522 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
524 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
526 SimGrid (3.6) stable; urgency=medium
528 The Summer Release, also known as the "OMG! They Killed Kenny!" version
531 * Bindings now constitute their own package, separated from the main one.
532 Rationale: reduce our maintenance nightmare by reducing the module coupling
533 They will soon be released on their own on gforge.
535 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
536 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
538 GRAS: It is not considered as stable anymore, but experimental. Sorry.
539 * It's not quite deprecated for now because we have no replacement,
540 but it may soon become the case.
543 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
544 * Fortran: New user-level cache variable to store the rank of the running
545 process. This improves performance by an order of magnitude.
546 * C: New coccinelle script to automatically locate and modify global and
547 local static variables.
548 * Improved SMPI network model with a sender-side gap to account for multiple
552 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
553 only say if a communication is finished, no matter whether it succeeded or
554 failed. You can call MSG_comm_get_status() to know the status of a finished
556 * New function MSG_task_dsend() to send a task and detach it. When a
557 communication is detached, you are never notified of its success or failure
558 and the memory is released automatically once it is finished. This function
559 is useful when you don't care about the end nor the success of a
561 * Change the prototypes of action replay. Sorry for inconvenience,
562 but this is really more efficient this way (and to adapt your code,
563 you just have to fix the initialization, that shouldn't be too long)
564 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
565 ever managed to do anything useful with it.
566 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
567 takes a dynar of communications. It returns immediately and gives the
568 index of a finished communication (if any).
569 * New example: a basic implementation of the Chord P2P algorithm.
572 * New model for multi-core CPUs. You can now use the core attribute to
573 precise the number of cores of a host. This is a basic model. Every
574 process running on the host receives at most the power provided in
575 the DTD (throughput<=power). Total throughput of process cannot exceed
577 * New peer tag. This peer tag creates a tiny AS comprising a host and a
578 router linked by an up-link and a down-link (possibly asymmetrical).
579 This kind of pattern allows to easily build last-mile model style platforms.
580 Aggregating such patterns in a rule-based AS is thus the technique of
581 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
582 * New model for Vivaldi routing. We transformed the Vivaldi network model
583 into a Vivaldi routing model (based on the rule-based model). This allows to
584 combine Vivaldi based latencies with last-mile platforms.
587 * Added a check for NaN of IEEE754 infinite in the double entries of
589 * Introduce a new context factory "raw", highly inspirated from the
590 ucontext factory, but using manually crafted functions in assembly to
591 do the work in an efficient manner.
592 * Allow to change the used context factory at run time, not only at
593 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
594 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
595 contexts at runtime (only with raw contexts or ucontexts).
596 * Completely rewrote this module to allow parallel execution of user
597 processes. Use --cfg=contexts/nthreads:N to execute user processes
598 with N parallel threads (the default is 1, meaning no parallelism).
599 * Allow to decide dynamically between sequential and parallel modes.
600 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
601 processes in parallel only when their number is greater than or equal to P
603 * Added a check for NaN of IEEE754 infinite in the double entries of
607 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
608 won't output the state of processes when interrupted with Ctrl-C
609 * Add a new function xbt_dynar_to_array that transforms a dynar into a
610 NULL-terminated array. This may solve backward compatibility issues
611 due to the change to return type of SD_simulate. See also:
612 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
613 * Add new macros with variable number of arguments.
614 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
615 - in xbt/asserts.h: xbt_assert
616 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
617 - in xbt/ex.h: THROWF and RETHROWF.
618 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
620 * Change xbt_die() to accept a format string with arguments, just like printf.
621 * New data structure: xbt_lib_t, like a dict but more general and with better
625 * New configuration options
626 Options triva/categorized and triva/uncategorized can be used to generate
627 graph configuration files for Triva visualization tool.
628 * Configuration option tracing/platform is renamed to tracing/categorized
629 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
630 * New TRACE_user_link_variable interface:
631 User provides the name of the link and the tracing variable to attach to it
632 * the declaration of tracing categories must be done after the environment
634 * simpler tracing interface, just one way to declare categories
635 TRACE_category or TRACE_category_with_color, it is up to you
636 * links in the trace file are again identified by their names
637 * trace contains the full platform hierarchy exactly as declared using the ASes
638 * Options tracing/msg/[task|process]:1 groups the process by hosts
639 for both cases, tasks and processes must have names that are unique during the simulation
640 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
641 * The experimental option tracing/msg/volume is deprecated
642 its functionality may be reincorporated if needed
644 The tracing generates a trace file with unordered timestamped events,
645 because of the way the core simulator (surf) works. A script available
646 at the tools directory (fix-paje-trace.sh) can be used to put the events
647 in order. We have changed the tracing so it can generate ordered timestamped
648 events in the final trace, but depending on the simulator (and how much time
649 is simulated) that can lead to a huge memory utilization. It is deactivated
650 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
653 * Define a SIMGRID_VERSION macro in simgrid_config.h.
654 - We are trying hard to keep the API stable, but it may happen that
655 some things change (we're a research project after all, not a
656 nuclear plant operating system). If such things should happen, you
657 could rely on that macro to adapt.
658 - current value: 30600 for 3.06.00, aka 3.6
659 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
660 a number that can be compared with SIMGRID_VERSION.
661 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
662 assertions and verbose/debug logging events are disabled at compile time.
664 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
666 SimGrid (3.5) stable; urgency=medium
669 * New feature to any SimGrid-based simulator: Model-Checking
670 Check SIN#1 for more details.
673 * New Model SMPI (three-interval linear regression for correction factors)
674 See RR-7426, available at http://hal.inria.fr/inria-00527150
675 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
676 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
677 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
678 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
679 * See also src/smpi/README
684 - Tracing API changes: TRACE_start and TRACE_end should not be called
685 by user-code. They are automatically called by simulators created
686 with SimDAG, MSG and SMPI if the toolkit is compiled with
687 tracing_enabled=ON. Categories declaration and utilization remain the
688 same for MSG and SimDag.
689 - A function was added to the tracing API to declare categories with
691 - TRACE_category_with_color (char *category, char *color)
692 where color must be in the following format
693 "%f %f %f", red, green, blue
694 and red, green, blue are float values in the interval [0, 1]
695 - User can specify NULL as color parameter, or continue calling
697 On that case, the tracing system will define random colors
698 - The following command-line options are supported:
699 --cfg=tracing/filename:msg.trace
700 --cfg=tracing:1 (activate tracing, needed to use others)
701 --cfg=tracing/platform:1 (categorized resource use)
702 --cfg=tracing/uncategorized:1 (uncategorized resource use)
703 --cfg=tracing/msg/task:1 (task creation)
704 --cfg=tracing/msg/process:1 (process creation, migration)
705 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
706 --cfg=tracing/smpi:1 (SMPI interface tracing)
707 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
708 - examples of examples/msg/tracing updated
710 - DAXLoader and DOTLoader functions can generate tasks with categories
711 - A new function to attribute a category to SD tasks:
712 TRACE_sd_set_task_category (SD_task_t task, char *category)
713 * Tracing the MPI interface implemented by SMPI
714 - Collective operations are traced with states
715 - Point-to-Point operations are traced with states/links
716 - Tracing activated by a parameter "-trace filename" passed
717 to smpirun during execution (considering that simgrid
718 is compiled with tracing enabled)
719 - To run the simulation with gdb, the simulator
720 accepts --cfg=tracing/smpi:1 to trace SMPI
721 - tesh files to check if smpi tracing is ok
722 - See examples/smpi/NAS/DT-trace
723 * GTNetS tracing re-worked
724 - adaptation to the tracing system of GTNets to cope
725 with modifications regarding the fullduplex mode
726 - new tesh files to check if gtnets tracing is ok
729 * Asynchronous communications through the functions:
730 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
731 * New function: MSG_load_platform_script()
732 to make possible using a lua script instead of XML files to set up platforms
733 * New function: MSG_set_function
734 to associate functions to processes, used when bypassing the parser
735 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
737 Platforms: Add some more examples in examples/platforms
738 * Grid'5000: see www.grid5000.fr
739 * *_30000_hosts.xml: various huge files [mainly scalability testing]
742 * Change the XML format. This is a very important modification. SimGrid
743 3.5 introduces a new hierarchical format based on the notion of
744 Autonomous Systems. Compatibility with old format is ensured
745 through the perl script provided in the install bin directory
746 bin/simgrid_update_xml.
747 It is now possible to build platforms with specific routing
748 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
749 connect several platforms together. We will try to provide soon
750 set of realistic platforms exploiting these properties (have a
751 look at examples/platforms/ for the moment).
752 * Take the opportunity of the XML format change to be a good XML citizen:
753 rename link:ctn to link_ctn and similar changes (also dealt with by
755 * Add a new routing scheme (rule-based) using regular expressions. It
756 enables to have an extremely low memory footprint when the
757 underlying routing is simple and can be compactly described. You
758 need to have libpcre4-dev (perl regular expressions) installed if
759 you want to use this routing scheme.
760 * Revive the cluster TAG and allow to easily and efficiently (both in
761 term of memory and speed) connect clusters together. Have a look
762 at teshsuite/simdag/platforms/ to see how this can be done. With
763 this tag, you can create clusters with thousands of tasks at no
764 cost (have a look at examples/platforms/).
765 Note: clusters are implemented as ASes, so there is no need for an
766 enclosing AS tag if you have only one cluster in your platform.
767 * Add new generic functions in the public interface that allows the user
768 to call SURF 'create_resource' methods from your code (same
769 functionality as the XML bypass mechanism but with a much lighter
771 * Add a new model (enabled through command line --cfg=network/model:SMPI)
772 that uses a piecewise linear approximation to produce better
773 results when exchanging small messages.
774 * Add a new parameter to handle correctly full duplex link and account
775 for interferences between uplink and downlink communications
776 (activate with --cfg=fullduplex:1).
779 * Rename the SD_READY (all dependencies are satisfied and task is
780 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
781 dependencies are satisfied) state.
782 This prevents a confusion between the notion of "ready to schedule"
783 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
784 simulated" (SD_RUNNABLE) used by the simulation kernel.
785 * Change the way a task is considered as ready. Instead of removing
786 dependencies when a task is done, a counter is decreased. This way,
787 it is always possible to reach ancestors thanks to the
788 SD_taks_get_parents function (even after the end of the simulation.)
789 * Change the return type of SD_Simulate from (SD_task_t*) into
790 xbt_dynar_t. This function was in handling a dynar internally and
791 converted it into a NULL terminated array for historical reasons.
792 * New function SD_dotload(char*) to load a DAG described in dot
793 format. This loader and the corresponding examples require the
794 installation of the graphviz library.
795 * Fix a bug in the management of tasks of size 0 in the surf network
796 models. This problem was only visible with SIMDAG and you should
797 thus disregard results produced with earlier versions if you
798 relied on this feature (some tasks were blocked because of this).
799 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
800 even though your applications were doing only point-to-point
801 communications and sequential computations. Now you can really use any
802 model you want (of course, if you create real parallel tasks, which are
803 not implemented in most models beside ptaskL07, this will abort).
804 * Add an example that schedules a DAX on an heterogeneous platform
805 using a Min-Min strategy.
806 * New function SD_workstation_get_current_task() that returns the kind
807 of task currently running on a workstation in the sequential access
809 * Raise some warnings when unexecuted tasks remains at the end of the
810 simulation. This is usually caused by cycles in the DAG.
813 * New function: SIMIX_process_set_function() called by MSG_set_function
814 * Change the underlying waiting queue in semaphores so that a process
815 can wait on several of them simultaneously (as in waitany).
816 * Fix the way to handle tokens in semaphores so that all access patterns
817 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
818 * kill the dirty pimple SIMIX_message_sizes_output()
819 Please use (proper) visualization instead
822 * New data container: setset (set of sets of elements)
823 * New module: mmalloc (mapped malloc, allowing to have several
824 independent segments of malloc)
825 * New function: xbt_dict_cursor_set_data()
826 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
827 * New function: xbt_dynar_is_empty()
828 * New function: xbt_fifo_get_last_item()
829 * Fix xbt_dynar_shrink(): use the right element size.
830 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
831 uninitialized areas during expand.
832 * Fix semaphores: previous implementation was severely broken.
833 * Use library init/fini functions for our initialization.
834 - you can use logs and other feature as soon as you want in your
835 code (even before the xbt_init / MSG_init)
836 - xbt_exit is now a no-op and produce a warning when used.
839 * Port GRAS to new SIMIX mechanisms. This allows gras users to
840 benefit from the latest improvement to the simulation kernel.
841 * Kill measurement sockets for now. If you rely on them, sorry. This
842 release is not for you. This feature will be reintroduced in the
843 future, but we cannot delay the release any further.
844 * New function: gras_msgtype_get_name().
845 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
846 * Fix (at last) the pmm example: it should not randomly fail anymore.
848 Build chain: bug fixes and overall polishing
849 * Cmake is now stable enough. Hence, we killed the autotools.
850 * Port to windows ( TM :)
851 * Fix the 'make install' target.
852 No need to use 'make install-simgrid' anymore
853 * Introduce a 'make dist' target compiling a *source* archive
854 'make package' compiles a binary archive
855 * Compile java files only on need
856 * Add --cd and --setenv command line options to tesh
857 * Out of source builds are not fully supported yet, but we are close
858 * Enable supernovae and optimization flags by default for our users
861 * Add layer to set up environment directly from lua, without XML.
862 * The effect of gras_stub_generator can be achieved through
863 lua too (check examples/gras/console/ping_generator.lua)
865 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
867 SimGrid (3.4.1) stable; urgency=low
869 The "Polishing easter eggs is probably a good idea" release.
870 This is a bug fixes release only.
874 * Fix a bug preventing the tasks from begin garbage collected.
877 * Fix a bug occuring when a host involved in a communication fails.
878 This was not detected properly by the other peer involved in the
879 communication. Now, it's reported as a network error.
882 * Warn the user about loop dependencies in data flow of DAX files
883 * Obey the control-flow dependencies of DAX files
886 * Add option "enable_smpi" allowing to not compile SMPI.
887 Probably useful for the (Mac) users experiencing a build error here
888 * Improve the detection of lua5.1 and ruby1.8
890 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
892 SimGrid (3.4) stable; urgency=low
894 The "Easter in Cargese" release. Also known as (major changes):
896 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
897 ~> bindings were greatly improved
898 ~> new tracing infrastructure for better visualization introduced
900 * the "Welcome to configury modernity" release.
901 ~> we switched from autotools to cmake, and improved our cdash
904 A more detailled list of changes follow (full detail in svn log).
906 Java Bindings: Various Cleanups
907 * (install java-gcj-compat-dev on debian-like to use them)
908 * Remove put/get: no need to export deprecated interface in Java
909 Use send/receive instead.
910 * Cleanup the examples and add a README per directory
911 * Remove example autoDestination (that's the only way to go now)
912 * Remove example explicitDestination (was a plain copy of basic)
913 * Make JniException a runtime exception, so that there is no need to
914 declare the fact that you may encounter such a beast. I guess that
915 nobody will ever want to survive such error.
916 * Create specific errors for each MSG case of failure:
917 host failure, transfer failure, timeout, task cancelled
918 * Cleanup the exceptions that may get thrown by each function
919 * Other internal cleanups in Java bindings. Performance still bad :/
920 Ruby and Lua Bindings: create them
921 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
922 * That's new and great, you should try them out.
923 Same functionalities than Java bindings, only even less polished
925 * Kill the useless "rate" argument of SD_task_get_execution_time()
926 Everyone used to provide -1 as a value, it was not used, and the
927 semantic of a possible use wasn't even clear.
928 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
929 either as comm costs or compute costs to mean that there is no
930 such thing for that specific task.
931 * Add a SD_task_set_name() function
932 * Fix SD_task_unschedule() on typed tasks
933 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
934 * In DAX loader, accept useless 'level' attributes to <job> since
935 LIGO DAGs have them (seem to be to ease graphical representation).
937 * Add an example masterslave_mailbox.c using send/receive and not
938 the deprecated put/get interface.
939 * Kill the MSG_paje_output() function. It's a noop since 2 years.
940 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
942 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
943 (declare MSG_USE_DEPRECATED to still have the old name)
944 * Add a MSG_task_set_data() function
945 * About trace replay (see examples/msg/actions):
947 - Allow to work with splitted trace files for each process
948 Give the specific trace file as argument of each process,
949 and call MSG_action_trace_run(NULL)
950 You can still have one merged file for all processes.
951 - Fix implementation of collective operations
952 * Allow task_execute() on 0-sized tasks (closes #10063)
954 * This is the first release of SimGrid where SMPI is not considered
955 beta anymore (even if some corners should still be improved)
956 * Port over the new SIMIX_network submodule (internal refactoring)
957 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
958 * Implement more missing elements of the standard:
960 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
962 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
963 MPI_Scan() MPI_Get_processor_name()
964 - Added implementation of missing case for Alltoall (warning: it's
965 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
966 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
967 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
968 SMPI_MPI_Allgatherv()
970 - MPI_Waitsome() was broken
971 - Allow relative includes in smpicc
972 - Command line cfg argument 'reference_speed' was ignored...
973 - Some functions did not properly lead to auto-benching of user code
974 - smpicc passes -O2 by default (just like openmpi one)
976 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
977 * Bug fixes about timeouts during communications
978 * add SIMIX_message_sizes_output() as a pimple to write to file the
979 amount of messages per size. Use gnuplot to get histogram.
980 Pimple because that's the only user-visible function of simix,
981 defined directly in xbt.h (irk, sorry)
983 - Add a SIMIX_sem_get_capacity() function
984 - Fix interactions with processe resume/suspende
985 - release_forever() was stupidly broken
986 - Fix SIMIX_display_process_status() for processes in a semaphore
987 - Make SIMIX_sem_block_onto() user-visible
988 * Refactoring context stuff:
989 - Use pseudo-OOP for better modularity
990 - reimplement SIMIX_process_kill() without process_schedule() so
991 that the latter can take as invariant that it is called from
993 - Merge context_start into context_new for sake of simplicity
995 * Add a Vivaldi network model, coded live during SUD'10 ;)
996 * Rename configuration variables to start a hierarchy:
997 o cpu_model -> cpu/model
998 o network_model -> network/model
999 o workstation_model -> workstation/model
1000 * New configuration variables:
1001 o network/bandwidth_factor: correction to bandwith
1002 o network/latency_factor: correction to latency
1003 o netwotk/weight_S: correction to the weight of competing streams
1004 * Add a long description to the models, that users can see with such
1005 argument on the command line: --cfg=cpu/model:help
1006 * --help-models display the long description of all known models
1008 * config: add the ability to set a default value after registration
1009 Does not override any previously set value (e.g. from cmd line)
1010 * dict: allow to have integer key and data.
1011 When so, you need to use the following functions
1012 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1013 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1014 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1015 In contrary to regular dicts, the key is not malloced before copy.
1016 Mixing scalar and regular elements in the same dict is not tested
1018 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1019 Tracing for Visualization:
1020 * SimGrid is now instrumented in order to generate a trace file for
1021 visualization analysis: to use it, need to compile SimGrid with the
1022 "tracing" option enabled, and instrument the program using SimGrid with
1023 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1024 (among other functions).
1025 * The instrumentation only traces the platform utilization for now
1026 * Documentation to use the tracing functions and how to analyze the
1027 traces with the Triva tool is written.
1028 * More information about: SimGrid FAQ (in the section Tracing Simulations
1031 * We moved to cmake as default build system. Autotools support will
1032 be dropped soon. Check the FAQ for more info about how to use it.
1033 * Greatly improved our cdash/ctest interactions
1034 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1035 * Added memory checking tests with valgrind; lot of memleak fixing.
1036 This may be the first release of simgrid with so few memory issues
1037 * Added code coverage tests.
1038 Our coverage is still improvable, but at least we see it on cdash.
1040 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1042 SimGrid (3.3.4) stable; urgency=low
1044 The "Desktop Grid needs love too" release (also called Xmas release).
1046 Models improvements:
1047 * Major speedup in the maxmin system solving by using lazy evaluation
1048 Instead of solving completely the maxmin system at each iteration,
1049 only invalidate (and recompute) the modified parts.
1050 This new feature is enabled in default models but you can try to
1051 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1052 * Cas01 IMproved as default CPU model
1053 This CPU model is the same Cas01 model, but it uses the
1054 maxmin-selective-update flag and a heap structure to manage
1055 actions on SURF kernel.
1056 It reduces the complexity to find the next action to finish and,
1057 consequently, it's faster than the old Cas01.
1058 This is the new default CPU model (Cas01).
1059 * Rename the old Cas01 model to Cas01_fullupdate
1060 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1061 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1062 * CpuTI (CPU Trace Integration)
1063 A new CPU model whose objective is simulate faster when using
1064 availability trace files.
1065 Instead of using a full featured, over engineered maxmin system for
1066 CPU modeling, this model does the pre-integration of traces files
1067 to calculate the amount of CPU power available, and so, executes
1068 faster than the old CPU models.
1069 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1070 * Use LV08 as default network model since it gives better accuracy
1071 for small messages and shouldn't change things for big ones.
1072 Use --cfg=network_model:CM02 to get the previous behavior.
1075 ******************************************
1076 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1077 ******************************************
1078 * The new CPU model may changes simulations!
1079 The point is that events occurring at the exact same timestamp
1080 are not scheduled in the same order with the old and new
1081 version. This may be enough to completely change the execution
1082 of simulations in some cases.
1083 * The new network model will change simulations!
1084 This new model is more realistic than the previous one, so you
1085 should consider redoing your old experiments with this model.
1086 Sorry for the inconvenience.
1089 * Introduce the supernovae compilation mode
1090 When compiled that way, the whole SimGrid (or almost) is put in a
1091 single compilation unit and compiled in one shoot.
1092 This is to help gcc which has difficulties to inline stuff from one
1094 The speedup seem to be above 15%, althrough more tests are needed on
1095 amd64 to confirm that gain.
1098 * Port of MSG's mailbox on top of SIMIX network
1099 The put/get mechanism was greatly simplified on the way.
1102 * New SIMIX network module. Provides:
1103 - Mailbox: rendez-vous mecanism to find with who you want to speak
1104 - Synchronous send/recv: easier and hopefully faster since the
1105 logic is handled in the maestro process directly now
1106 - Asynchronous send/recv: you dreamt of it? It's here now
1107 Too bad that nobody cared enough to propagate the change to MSG.
1108 * Add semaphores as SIMIX synchronization mechanism.
1111 * new function SD_daxload(char*) to load a DAX file
1112 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1113 * Introduce typed tasks. Specify its kind and cost at creation.
1114 At scheduling, just give where it should be placed, and the cost
1115 for each involved resource is automatically computed.
1116 Existing constructors so far (more to come of course):
1117 - SD_task_create_comm_e2e() for end-to-end communication
1118 - SD_task_create_comp_seq() for sequential computation
1119 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1120 * new function SD_task_dump() for debuging display
1121 * new function SD_task_dotty(task,FILE*) writing to file the info
1122 about the task in dotty format
1123 * SD_task_dependency_exists() can now cope with having one of its
1124 arguments NULL. If so, it tests whether the other argument has any
1126 * Add getters on list of preceding/following tasks:
1127 SD_task_get_parents(task) and SD_task_get_children(task)
1128 * Add getters on amount of workstations and list:
1129 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1130 * Add getter on task kind: SD_task_get_kind(task)
1131 * Update the start_time and finish_time of tasks on completion/failure
1132 * Bugfix: Remove task from state swags when destroyed
1135 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1138 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1139 dynar with one separate thread per value of the dynar.
1140 * Change the prototype of xbt_thread_create(), sorry.
1141 Added a boolean parameter indicating whether we want to join this
1142 thread (used in SG only for now)
1143 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1146 * GTNetS wrappers should now be usable again (and betterly tested too)
1147 * Fix a major regression from 3.2 where the timeout provided to
1148 MSG_task_put_with_timeout() was used as absolute time before which
1149 the comm should be done.
1150 * Start to fix the <cluster> tag.
1151 - Internal links should be good now (beside of the loopback, which
1152 use the private link instead)
1153 - paths to the external world is still rather broken
1154 - the <route:multi> tag is just broken. Actually that's brain-dead.
1155 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1156 to make it less stupid
1157 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1158 * Fix a source-level compatibility glitch from 3.2: after defining
1159 MSG_USE_DEPRECATED, you can use the old name
1160 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1161 * Allow to compile from the SVN with automake 1.11
1162 * Fix some problems when using the "start_time" tag in deployment XMLs.
1163 * Fix #8569: XBT/synchro.h has redundant declarations
1164 * Fix #8563: MSG return values and exceptions
1165 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1166 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1167 Thanks to Robson Peixoto.
1168 * Fix a vicious bug in dictionaries inducing that some elements were
1169 not freed on xbt_dict_free()
1171 Portability report of this version:
1172 * Main portability targets:
1173 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1174 on (amd64/i386/ia64)
1175 - mac leopard on i386
1176 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1177 but nothing critical.
1178 * Other platforms: windows, AIX and others were not tested for this release
1180 Timing report of this version:
1181 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1182 scenario parameters). From 8h to a few seconds in desktop grid settings.
1183 * Supernovae brings about 25% speedup on i386.
1185 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1187 SimGrid (3.3.3) stable; urgency=low
1189 The "Need for Speed" release.
1191 The timings done to validate the 3.3.2 were faulty.
1192 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1194 The problem was a conversion from a manually handled vector to
1195 xbt_dynar_t on the critical path.
1196 xbt_dynar_foreach calls functions, inducing stack management crap.
1198 We inlined these functions and xbt_dynar_foreach is now breath taking.
1199 We also inlined xbt_swag_belong on the way.
1201 Here are some approximate speedup measurements (on master/slaves
1202 simulations lasting between 10s and 20s each):
1203 3.3.1 -> 3.3.2: about same performance
1204 3.3.2 -> 3.3.3: 40% speedup
1205 3.3.1 -> 3.3.3: 40% speedup
1206 3.3.1 with inline patch -> 3.3.3: 30% speedup
1208 Our reading is that the refactoring which occurred in 3.3.2 made us
1209 suffer much more from the xbt_dynar_foreach low performance, but
1210 once we solved this, this refactoring proved to be very performance
1211 effective. From the 40% speedup, somehow, 10% are due to the
1212 inlining and 30% to the refactoring.
1214 That's a pitty that gcc cannot inline functions placed in other files
1215 alone. We have to choose between:
1216 - break the encapsulation (by putting private data structures and
1217 accessors in headers files to help gcc)
1218 - live with low performance
1219 - switch to a decent compiler such as icc (not quite possible).
1221 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1223 SimGrid (3.3.2) stable; urgency=low
1225 The "Simplicity does not preceed complexity, but follows it" release.
1227 The main contributors of this release were (lexical order):
1228 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1231 * Extract the routing logic into its own object.
1232 (was dupplicated in network.c and workstation_LV07.c;
1233 Allows to implement other ways of storing that info)
1234 => kill now useless network_card concept
1235 - Use dynar to represent routes (instead of void** + int*)
1236 - kill link_set (use surf_network_model->resource_set instead)
1237 - Add a command-line option to choose the routing schema to use
1238 - Add three new models:
1239 * Floyd (shortest path computed at initialization)
1240 * Dijikstra (shortest path recomputed all the time)
1241 * Cached Dijikstra (shortest path computed on need)
1242 All these models where contributed by Silas De Munck, and are
1243 described in his ICCS09 paper.
1245 * Simplify model declaration
1246 (less redirections, less function to write when defining a model)
1247 - Factorize stuff between models:
1250 surf_model_resource_set(model)
1251 surf_model_resource_by_name(model, name)
1252 - Unify the types of models in s_surf_model_t (using an union)
1253 - Embeed fields of common_public directly into s_surf_model_t
1254 - Rename model methods:
1255 action_free ~> action_unref
1256 action_change_state ~> action_state_set
1257 action_get_state ~> action_state_get
1258 - Change model methods into functions :
1259 (model)->common_public->action_use ~> surf_action_ref
1261 * Implement a generic resource; use it as ancestor to specific ones
1262 (allows to kill duplicated code in models)
1263 Drawback: timer command don't need no name nor properties;
1264 workstation_CLM03 don't need no properties
1265 (but I guess we can live with those few bytes wasted)
1267 * Improve the action object model
1268 - implement a constructor avoiding dupplicated code about field
1269 initialization in generic_action part.
1271 * Kill the SDP model: it has an external dependency, is deprecated
1272 in flavor of modern lmm models, and didn't compile since a while
1275 * Relocation of the context module from XBT to SIMIX.
1276 (the context were decoupled from the simix processes, duplicating a lot of code)
1277 => a lot of code was factorized
1278 - less overhead is introduced during scheduling
1279 - simpler API for the context factory
1280 - the logic for process creation,destruction and manipulation was simplified
1281 * Simplification of the s_smx_process_t data structure.
1282 => accesing the simix level data associated to a process is faster now,
1283 and the code is a lot more readable.
1286 * Implement some more MPI primitives:
1287 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1288 -implementation: Bcast: flat or 2-ary tree (default),
1289 Barrier: 4-ary tree,
1291 Allreduce: Reduce then Bcast
1292 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1293 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1294 Alltoallv: flat tree, like ompi
1296 * Add support for optimized collectives (Bcast is now binomial by default)
1297 * Port smpirun and smpicc to OS X
1300 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1301 Shout out if you used it.
1304 * Display the list of still queued messages in SG mode when existing
1308 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1309 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1310 * Add xbt_str_from_file(FILE*)
1311 * Add xbt_dict_get_key achieving a linear reverse search
1312 * Remove the context module
1314 Portability report of this version:
1315 * Main portability targets:
1316 - Linux(debian)/x86/context
1317 - Linux(debian)/x86/pthreads
1318 - Linux(debian)/amd64/context
1319 - Linux(debian)/amd64/pthreads
1320 On these, we still have the eratic breakages of gras/pmm and
1321 amok/saturate_sg reported in previous version. We still think
1322 that the tests are the cause of the fault, not the tested code.
1324 - Mac OSX Leopard/x86/context
1325 Still false negative in tesh autotesting.
1326 Smpi still fails, but this time because readlink does not accept -f
1327 Everything seems to work properly beside of that.
1330 - AIX version 5.3 (only tested contexts this time)
1331 Smpi still fails there because mktemp is not installed.
1332 Everything seems to work properly beside of that.
1334 I managed to compile it for the first time, but several breakages.
1335 Won't delay the release for this exotic platform.
1337 * Windows: it's still lagging behind. If you want to help, please
1340 Timing report of this version:
1341 This version seem to be more than 5% faster than 3.3.1 (on linux
1342 64bits with contextes). The gain is less than expected, we are
1343 investigating this for next release.
1345 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1347 SimGrid (3.3.1) stable; urgency=low
1350 * Implement a --cfg-help to show existing configuration variables
1351 * Build chain do not require doxygen in maintainer mode
1354 * fix a bug on struct sizeof computation, which prevented the
1355 exchange of arrays of structs in some conditions
1356 - added a regression test about this in datadesc_usage
1357 * Allow the exchange of 0-long dynamic vectors.
1358 - for that, use -1 as indicator of dynamic size instead of 0
1359 - This implied to change any size from unsigned long to long,
1360 reducing a bit communication abilities, but I guess that with
1361 64bits being quite common, this is more than enough.
1362 - This also induce a protocol change, thus bumping network protocol
1363 version from 0 to 1 (if we have external users, we have to get
1364 clean on that point too ;)
1365 - added two regression tests about this in datadesc_usage
1366 * Be more verbose when propagating local exceptions
1367 This helps debugging.
1368 * Display the status of simulated processes when receiving SIGINT in
1372 * Allow to control the simulation from a trace file.
1373 New functions MSG_action_register() and MSG_action_trace_run()
1374 The first one allows to associate a function execution to each
1375 kind of action while the second one parses a trace file and
1376 triggers the corresponding actions within the system.
1377 For now, only a toy example is provided in examples/msg/actions
1378 * Add an exemple of process migration in examples/msg/migration
1379 * Fix a bug in task exchange which broke MSG_task_get_sender()
1380 Add a teshsuite regression test for that.
1381 [Bug: if MSG_task_get_sender() is called after sender exit,
1383 * Fix a bug which prevented suspend/resume to work properly
1384 * Display the status of simulated processes when receiving SIGINT
1385 This fixes a regression of v3.3. due to the introduction of SIMIX
1386 * Bug fixing in failure management:
1387 - trace could not start by a failure at time 0
1388 - failure during communications were not working
1391 * Add SIMIX_process_set_name() to change the name of the current
1392 process in the log messages.
1393 * Store smx_hosts in a dict since we only retrieve them by name
1394 * Move the configuration infrastructure to surf
1397 * Move the configuration infrastructure to surf
1400 * Massive internal cleanups:
1401 - Store internal structures on processes instead of hosts (allows
1402 to have more than one process per host, in addition of being more
1404 - Cleanup the initialization/finalization process
1405 - Kill a whole bunch of unneeded synchronization:
1406 processes run in exclusive manner within the simulator
1407 - Move queues from global tables to process data fields
1409 - now accept -platform and -hostfile arguments
1410 - Pass the right rank value to processes according to the hostfile
1411 * Compile the examples by default, and use them as regression tests
1412 * Implement MPI_Wtime()
1413 * Change the reference speed to a command line option
1416 * TCP_gamma can now be specified as command line option using
1417 --cfg=TCP_gamma:10000000.0
1418 * Change the --surf-path cmd line option into --cfg=path:
1421 * Also include strbuff from xbt.h public header
1422 * xbt_ex_display(): do not free the exception after displaying
1423 This allows to do more with the given exception afterward.
1424 Users should call xbt_ex_free() themselves.
1428 Portability report of this version:
1429 * Main portability targets:
1430 - Linux(debian)/x86/context
1431 - Linux(debian)/x86/pthreads
1432 - Linux(debian)/amd64/context
1433 - Linux(debian)/amd64/pthreads
1434 These targets fail about 1/10 of times on gras/pmm, but we believe
1435 that this is because of the test, not because of simgrid.
1436 amok/saturate_sg fails even more rarely, and the test may not be
1439 - Mac OSX Leopard/x86/context
1440 The test suite still spits tons of errors because some obscure
1441 force prevents us from removing the temporary directories
1442 arguing that they still contain some metadata I've never heard of.
1443 Smpi fails because seq is not installed.
1444 Everything seems to work properly beside of that.
1447 - AIX version 5.3 (both contexts and pthread)
1448 Smpi still fails there because mktemp is not installed.
1449 XML inclusions seems rosty on AIX.
1451 * Windows: it's still lagging behind. If you want to help, please
1454 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1456 SimGrid (3.3) stable; urgency=high
1460 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1461 [Malek Cherier & Mt]
1463 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1466 Until now, GRAS were using MSG as an interface to SURF. It was
1467 quite difficult because both interface have several differences
1468 (MSG channels vs GRAS sockets were the most notable point).
1470 This also opens the gate to SMPI (which should occur soon) and speed
1471 up simulations by to 40% (even if it were not the main goal).
1473 **************************************
1474 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1475 **************************************
1476 The point is that events occuring at the exact same timestamp are
1477 not scheduled in the same order with the old and new version. This
1478 may be enough to completely change the execution of simulations in
1479 some cases. Sorry for the inconvenience.
1481 * Cleanup and upgrade the XML format to push further scalability
1482 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1484 * Improve the testing infrastructure with tesh. Now a very large part of
1485 the code is tested not only by being run but also by checking that the
1486 output match an expected output [Mt].
1488 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1489 is really less memory-demanding, which should allow you to use
1490 larger files in SimGrid [AL].
1492 * Inform valgrind about our contextes, so that it becomes usable
1493 with the default (and more effecient) version of SimGrid
1494 [contributed by Sékou Diakite, many thanks]
1497 * Introduce a listener thread in charge of receiving incoming
1498 messages from the network. It allows to overlap communication and
1499 computation but most notably, it removes some stupid deadlocks due
1500 to the fact that so far, a process could not send and receive at
1501 the same time. This made most non trivial communication schema
1503 * Convert the PIDs from long int to int to match the MSG ones (and
1504 linux ones too) [Mt]
1505 * New function: gras_agent_spawn() to launch a new process on
1506 current host. Only working in simulation for now. [Mt]
1507 * New function: gras_os_hostport() returning a constant form (ie,
1508 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1511 * Make the backtrace of exceptions more human readable [Mt]
1512 * New module: xbt/str [Mt]
1513 a ton of string utility functions (split, join, printf to a newly
1514 allocated buffer, trim, etc)
1515 * New module: xbt/hash [Mt]
1516 SHA1 hashing algorithm (more to come if needed)
1517 * New module: xbt/synchro [Mt]
1518 synchronization tools (mutex and conditions) working the same way
1519 in simulation and in real life (mainly useful for GRAS, but not
1521 * New module: xbt/queue [Mt]
1522 classical producer/consumer synchronization scheme
1523 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1524 (using the classical functions will get serialized) [Mt]
1525 * Make dictionary internal table dynamic. No need to specify its size
1526 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1528 * Make sure the log channels are organized as a tree under windows
1529 (because of ANSI C compatibility issue, any channel were child of
1530 root directly) [Mt].
1533 * Cleaned many thing in surf and fixed a few bugs [AL].
1534 * Add a nice command line configuration mechanism to compose models [AL].
1535 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1536 the previous one (KCCFLN05). It relies on something that looks like
1537 a max-min sharing mechanism but cannot be written as such. A new solver
1538 was thus designed [AL].
1539 * Add a new solver to lmm. Based on Lagrange optimization and
1540 gradient-based descent, it enables to efficiently maximise systems s.a
1542 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1544 This solver enables to propose two new network models for TCP Reno and
1545 TCP Vegas based on Low's work. These models still need to be fully
1546 tested though [Pedro Velho].
1549 * Bug fix in SD_simulate. Now the time bound given as argument is
1551 * Use the new parallel task model (ptask_L07) as default.
1552 * Use the SURF command line configuration mechanism.
1553 * 0-size tasks (for synchronization) should now work.
1555 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1557 SimGrid (3.2) stable; urgency=high
1561 We still experience issues on this platform, but we believe that at
1562 least MSG is usable.
1564 GRAS API BREAKAGE (for simplification purpose, sorry):
1565 * the gras_msgtype_by_name is not used anymore. Instead of
1566 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1567 you can write (and must)
1568 gras_msg_send(toserver, "request", &request);
1569 - If you still want to pass a gras_msgtype_t to the function (to cache
1570 the type and avoid the lookup time), use the gras_msg_send_() variant.
1571 - Impacted functions:
1572 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1573 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1574 * The callbacks are now expected to return 0 when everything went well
1575 (just like the main() function)
1577 GRAS new features and improvements:
1578 * New module mecanism where user code can use per process globals [Mt]
1579 This is similar to gras_userdata_*() functions, but for libraries. It
1580 factorize some code developped over and over in the examples and AMOK.
1581 It has still to be documented and used (only amok/peermanagement is
1583 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1584 loss when they were small enough to fit into the buffer and sent quickly
1585 enough so that they can all get received in one shoot.
1586 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1587 if not found. Use the *_or_null() variant for the old semantic.
1588 * In gras_msg_handle, do not discard messages without callback.
1589 They are probably messages to be explicitly awaited later (ie, proofs of
1590 mis-synchronization in userland since they are sent before being awaited)
1591 No big deal usually.
1592 * gras_socket_meas_send/recv: semantic changed!
1593 The numerical arguments used to be (1) the total amount of data to send
1594 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1595 messages. This was need for the fool willing to send more than MAXINT
1596 bytes on quite fat pipes.
1599 * Do really rename the hostmanagement module to peermanagement. [Mt]
1600 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1601 but this is rather new and this was documented in the module
1602 documentation (poor excuses, I admit)
1603 * Bandwidth measurement semantic changed! This follows the changes to
1604 gras_socket_meas_send/recv explained above.
1607 * A sequential mode has been added to the workstations. When a workstation
1608 is in sequential mode, it can execute only one task, and the other tasks
1609 are waiting in a FIFO. [Christophe Thiery]
1612 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1613 model for SIMDAG. [Christophe Thiery]
1614 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1615 equal to 0, which caused some bad behaviors in
1616 saturated_constraint_set_update. I now use a threshold mechanism like in
1620 * When running manually src/testall, you select specific units [Mt]
1621 testall is the result of our cunit mecanism, and should replace all
1622 the scripty thingy around since bash don't run easily on billware.
1624 * A mallocator system has been added. [Christophe Thiery]
1625 Mallocators allow you to recycle your unused objects instead of freeing them
1626 and allocating new ones.
1628 Documentation update:
1629 * FAQ reworking + New FAQs:
1630 - "Valgrind spits tons of errors!" [Mt]
1631 - "How to repport bugs" [Mt]
1632 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1633 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1634 - Communication time measurement within MSG [AL]
1635 - I experience weird communication times when I change the latency [AL]
1636 * GRAS tutorial [Mt]
1638 - an introduction to the framework and to the used communication model
1639 - an initiatic tour introducing the most proheminent features:
1641 . Lesson 0: Installing GRAS
1642 . Lesson 1: Setting up your own project
1643 o Part 2: Message passing
1644 . Lesson 2: Exchanging simple messages
1645 . Lesson 3: Passing arguments to the processes (in SG)
1646 . Lesson 4: Attaching callbacks to messages
1647 . Lesson 5: Using globals in processes
1648 . Lesson 6: Logging informations properly
1649 . Lesson 7: Using internal timers
1650 . Lesson 8: Handling errors through exceptions
1651 . Lesson 9: Exchanging simple data
1652 . Lesson 10: Remote Procedure Calling (RPC)
1653 . Lesson 11: Explicitely waiting for messages
1654 . Recapping of message passing features in GRAS
1655 - A HOWTO section containing:
1656 o HOWTO design a GRAS application
1657 More are due, of course. They will come latter. In the meanwhile, you can
1658 check the examples which are still here.
1660 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1662 SimGrid (3.1) stable; urgency=high
1666 There was a stack corruption somewhere, visible only when optimizing
1667 with these versions. [Vince]
1670 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1671 module that enabled to play with Directed Acyclic Graphs. It is built
1672 directly on top of SURF and provides an API rather close to the old
1673 SG. Some old codes using SG are currently under rewrite to check that
1674 all needful functions are provided. [Christophe Thiery]
1677 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1678 extension of the classical CLM03 model that gracefully handles
1679 failures. This is now the default model for MSG and GRAS. It doesn't
1680 handle parallel tasks yet though. [AL]
1681 * Bug fix: Weights were not correctly set in the network part.
1682 WARNING: This may have resulted in incorrect results with simulations
1683 where there are more than one flow on a given link. [AL]
1686 * After a (long ?) discussion on simgrid-devel, we have decided that the
1687 convention we had on units was stupid. That is why it has been decided
1688 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1689 WARNING : This means that all previous platform files will not work as
1690 such with this version! A warning is issued to ask users to update
1692 A conversion script can be found in the contrib module of the CVS, under
1693 the name contrib/platform_generation/surfxml_update.pl [MQ]
1696 * Bug fix: Processes were started in reverse order, wrt deployment file.
1697 WARNING: if your code relies on this bug, please fix it. [AL]
1698 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1699 being executed on two different locations. [AL]
1700 * Bug fix: Failures are now better supported thanks to Derrick's tests
1701 (there was many failure situations I hadn't thought of and that weren't
1702 correctly handled). [AL]
1703 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1707 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1708 Exception occurring on server-side are propagated back to client (!).
1710 API CHANGE: the callback changed their prototype. Change:
1711 int my_handler(gras_socket_t expeditor, void *payload_data) {
1713 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1714 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1716 * New! function: gras_msg_handleall to deal with all messages arriving
1717 within a given period.
1718 * New! function: gras_socket_server_range to get a server socket in a
1719 range of port numbers (ease to avoid port number conflicts) [MQ]
1720 * New! gras processes display their backtrace when they get a SIGUSR1
1721 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1722 Sweet to debug RL processes [MQ]
1726 - Do not force experiment sizes to be expressed in kb, or it becomes
1727 impossible to measure the latency this way (needs one byte-long tests)
1728 WARNING: this changes the amok_bw_* function semantic. [MQ]
1729 - Implements the link saturation stuff. [MQ]
1730 * Peer management module:
1731 New! module factorizing code that we wrote over and over [MQ].
1734 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1735 - Test units are placed directly into the library code, they get extracted
1736 automatically and placed into the src/testall binary.
1737 - Convert most of the XBT tests to this system.
1738 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1739 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1740 live in the same C code. This required to change a little bit the API
1741 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1742 * New module: sparse graph structure with basic algorithms (this is work
1743 in progress and the API is not considered to be frozen yet). [AL]
1744 * Display more information on backtraces: source line & function names are
1745 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1746 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1747 * Reworked a little bit some #include statements to load only required
1748 headers. Some user code that relied on SimGrid to include stdlib or
1749 stdio may need to include it by themselves. [AL]
1750 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1752 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1753 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1754 but I guess that almost nobody used those parts. [MQ]
1756 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1758 SimGrid (3.0.1) stable; urgency=low
1761 * Unfortunately, I had missed 5 misnamed functions:
1762 xbt_fifo_item_t xbt_fifo_newitem(void);
1763 void xbt_fifo_freeitem(xbt_fifo_item_t);
1764 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1765 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1766 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1767 They're now deprecated. Please use their new versions:
1768 xbt_fifo_item_t xbt_fifo_new_item(void);
1769 void xbt_fifo_free_item(xbt_fifo_item_t);
1770 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1771 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1772 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1774 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1775 * Documentation: xbt_log module unmercifully reworked [MQ]
1776 * Bugfix: there was a problem with the ending of contexts with
1777 the pthread backend. It caused some weird deadlock or behavior
1778 depending on the pthread implementation. [AL]
1779 * Bugfix: get the exceptions raised in the simulator repport where
1780 and why they come from when they are not catched in time [AL, MQ]
1783 * Bugfix: Do repport the error when two non-connected hosts try to
1784 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1787 * Add additionnal checkings on communications. Assert that two
1788 communicating hosts are connected by a set of links... [AL]
1791 * Add additionnal checkings on channel values in communication [AL]
1792 * New: MSG_task_get_source to see on which host a task was generated [HC]
1793 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1794 the number of tasks waiting to be received on channel and sent
1796 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1797 waits for the first task coming from a given host.. [AL]
1799 GRAS new functionnalities: [MQ]
1800 * Enhance the parsing macro to allow the size of multidimentional objects
1801 to be given thru annotations.
1802 * New example (and documentation): Matrix Multiplication a la RPC
1803 (as when I was young!) and fix a bunch of bugs found on the way.
1805 GRAS performance improvements: [MQ]
1807 * Reduce the amount of cbps creation/destruction by making it static to
1808 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1811 * Change libdata to a set so that we can search for stuff by ID (and thus
1812 reduce the insane amount of dict lookups)
1815 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1816 RL communication on top of SG since it's so uneffective.
1817 It may also be used for inter-thread communication in RL, one day.
1818 * Use gras_datadesc_copy() to exchange messages on top of SG
1820 - improve message exchange performance on top of SG
1821 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1822 * Don't exchange on the network the size of the used part of buffer,
1823 instead, specify the possible buffer size to read().
1825 - reduces the amount of read/write calls (one pair per exchange)
1826 - reduces the amount of exchanged data (the size)
1827 - allows to retrieve all arrived data on receiver side, if we don't need
1828 it right now (subsequent read will peek the buffer)
1829 - allows the receiver to proceed with the begining of the stream before
1830 everything is arrived
1831 - make it possible to build an iov transport (using readv/writev)
1833 - take care of the data with non-stable storage (like stacked data),
1835 * If possible, TCP send uses vector I/O (when writev() is here)
1836 - Don't use it for receive since we send data sizes and data on the
1837 same stream, so we wouldn't be able to chain large amount of chunks
1838 before having to flush the stuff to read the size.
1839 * Rework the transport plugin mecanism to simplify it and reduce the
1840 amount of pointer dereferencement when searching for the right function
1843 * I guess that now, we do almost as few system calls as possible while
1844 doing as few data copy as possible.
1846 To improve it further, we could try to send all the sizes first and then
1847 all the data (to use iov on receiving size), but it's only a partial
1848 solution: when you have 2 dimensional data, the sizes of the second
1849 dimension is data of the first dimension, so you need 3 streams.
1851 I'm not sure the potential performance gains justify the coding burden.
1853 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1855 SimGrid (3.00) stable; urgency=high
1858 * New! Give the possibility to hijack the surf parser and thus bypass
1859 MSG_create_environment and MSG_launch_application. Have a look at
1860 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1862 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1864 SimGrid (2.96) unstable; urgency=low
1869 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1870 This deprecates the xbt_error_t mecanisms.
1871 It modifies (simplifies) all XBT and GRAS API.
1872 MSG API keeps unchanged (exceptions raised by XBT are catched from
1873 within MSG and masked with existing error handling facilities)
1876 * New! Add a FATPIPE model. [AL]
1877 * New! Add a parallel task model. [AL]
1878 * New! Add automatically a loopback interface (in the default
1879 network model) if none was precised.
1882 * Bugfix: MSG_process_resume now works with the current running process.
1884 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1885 * Modification of MSG_task_get_compute_duration. Once a task has been
1886 processed, the value returned by this function is now equal to 0. [AL]
1887 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1888 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1889 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1890 of a m_task. For now, MSG_task_cancel only works with computation
1892 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1893 of the processor (in Mflop/s) regardless of the current load on the
1895 * API Change: use proper naming convention for MSG_getClock and
1896 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1898 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1899 This function changes the priority of a computation task. This priority
1900 doesn't affect the transfer rate. A priority of 2 will make a task
1901 receive two times more cpu power than the other ones. This function
1902 has been added to suit the needs of Nguyen The Loc and hasn't been that
1903 much tested yet. So if it fails, please report it and send me your code.
1905 * API Change: removed all functions and types that were marked "deprecated"
1906 since many months. Renamed MSG_global_init_args to MSG_global_init.
1908 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1910 SimGrid (2.95) unstable; urgency=low
1913 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1914 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1915 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1918 * Bugfix: Allow absolute paths for platform description files [MQ]
1919 * Bugfix: do free the variables after use. Leads to drastic performance
1921 * Implement max_duration (ie, timeouts) on resources [AL]
1924 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1926 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1927 the needed support (timeouts) [AL]
1930 * Implement measurement sockets. You can now get the bandwidth between two
1931 hosts thanks to AMOK (see below). [MQ]
1932 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1933 dynar over the network (yeah) [MQ]
1934 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1935 * simple/static token-ring example. [Alexandre Colucci and MQ]
1936 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1937 in place before [MQ]
1940 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1941 type parsing. "short" and "long" modifiers were also ignored; other
1942 modifier (such as reference level) are still ignored. [MQ]
1943 * Embeed the buffer size within the buffer itself on SG. [MQ]
1944 That way, send() are atomic and cannot get intermixed anymore (at least
1945 the ones which are less than 100k; bigger messages still have the issue)
1946 * Array size pushed by the field, not by the field type (or each
1947 and every long int will push stuff to the cbps) [MQ]
1948 * use select() to sleep since it allows to portably sleep less than one
1951 GRAS (minor cleanups)
1952 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1955 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1956 * type category 'ignored' killed as it was never used and were difficult
1958 * whether a type can cycle or not is now a flag, leaving room for more
1959 flags (as "ignored", if we feel the need one day ;)
1960 * Rename raw sockets to measurement sockets since "raw" has another
1961 meaning in networking community.
1964 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1965 and offers features not belonging to GRAS but that most applications
1966 need. One day, it may be a set of plugins loadable at runtime.
1967 * New module: bandwidth
1968 bandwidth measurement between arbitrary nodes running this module. [MQ]
1970 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1972 SimGrid (2.94) unstable; urgency=low
1974 The first beta release of SimGrid 3 !
1978 * Update the main page and the FAQ. Adding references to gforge.
1981 * Add a gras_os_getpid function.
1984 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1985 * Extend the logs so that they also print PID, hostname, date, ... if
1987 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1988 and kill the old version which were in testsuite/
1989 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1993 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1994 for users. Instead, each programming environment (GRAS, MSG,...) use it
1995 when needed to provide such a feature to users.
1996 Don't shortcut the mecanism or you will also shortcut the virtualization
1997 you need on the simulator.
2001 * Cleanups in configury with regard to compile optimization/warning flags.
2002 Also add -fno-loop-optimize to any powerpc since it's the optimization
2003 killing gcc (< 3.4.0).
2004 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2005 needed by MSG examples complications
2006 * Borrow configury beautifications from PHP
2009 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2010 warning (thanks loris for stumbling into this one).
2011 * Bugfix: stop loading private headers (gras_config.h) from the public
2015 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2016 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2017 RL would imply a DNS resolver, which is *hard* to do in a portable way
2018 (and therefore delayed).
2019 * Implement a real timer mecanism and use it in timing macros. This allows
2020 to avoid rounding errors and get a 0.000005 sec precision in timing
2021 macros. While I was at it, various cleanups:
2022 - allow to declare more than one timed section per file (fix a stupid bug)
2023 - move some private declaration to the right place
2024 - merge conditional execution and timing macros into emulation module
2025 - document the module
2026 - make sure the module cleanups its mess on gras_exit
2027 * Documentation improvements:
2028 - (new) how to compile applications using GRAS
2029 - (new) emulation support (timing macros)
2031 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2033 SimGrid (2.93) unstable; urgency=low
2035 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2038 - Use Paje properly where used. Still to be sanitized properly.
2039 - Portability fix: Add an implementation of the contexts using pthread
2043 - Add xbt_procname(): returns the name of the current process.
2044 Use it to show the current process's name in all logging.
2046 - fix detection of older flex version and the reaction, since we do
2047 depend on modern ones (we use lex_destroy)
2048 - Better separation of SG and RL in the libs: remove all simulation code
2049 from libgras. As a result, this lib is now only 200k when stripped.
2050 Some of the xbt modules may also be duplicated (two sets and such) and
2051 should be cleaned/killed before SG3.
2052 - Insist on using xlC on AIX because of weird problems involving gcc there.
2053 - Cleanup the make remote stuff. This is now done by scripts
2054 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2055 mainly for our private use, but we're working on changing them to user
2058 - Bugfix: flush the socket on close only if there is some *output*.
2059 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2061 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2064 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2066 SimGrid (2.92) unstable; urgency=low
2068 Alpha 3 on the path to SimGrid 3
2072 - New! First try of benchmarking macros.
2073 - New! First try so that gras_stub_generator generate deployment and
2074 remote compilation helpers.
2076 - Bugfix: Initialization fix in msg_test.
2080 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2082 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2084 - API Change: gras_os_sleep to take the amount of seconds as a double.
2085 Accepting an int was error prone since it was the only location where
2086 seconds were coded as such. It leaded to damn rounding errors.
2087 - Bugfix: Hard to belive that timers ever worked before this.
2089 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2091 SimGrid (2.91) unstable; urgency=low
2093 Alpha 2 on the path to SimGrid 3
2097 - Bug fix in the lmm_solver.
2099 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2100 through the function MSG_paje_output.
2101 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2102 - It is possible to bound the rate of a communication in MSG with
2103 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2104 to write it in the changelog).
2105 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2110 - Various cleanups to the autotools stuff
2111 - Begin to move Gras examples to examples/gras/
2112 - Let make distcheck work again (yeah!)
2114 - documentation overhauled using doxygen.
2115 gtk-doc-tools is dead in SimGrid now.
2116 - Automatically extract all existing logging categories, and add the list
2117 to the documentation (long standing one, to say the less)
2119 - Cleanup the known architecture table. Reorder the entries to group what
2120 should be, and use a more consistent naming scheme.
2121 (some of the test dataset are still to be regenerated)
2122 - New! Allow library to register globals on each process just as userdata
2124 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2125 lookup time (for now).
2126 Use it in msg and trp.
2127 This cleans a lot the internals and helps enforcing privacy of the
2128 headers between the gras components.
2129 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2130 - Bugfix: gras_os_time was delirious in RL.
2131 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2133 - Reenable GRAS now that it works.
2135 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2137 SimGrid (2.90) unstable; urgency=low
2139 Alpha 1 on the path to SimGrid 3
2141 * It is a long time since the last release of SimGrid. I'm sorry about
2142 that but as I had told you, I was rewriting a lot of things. I apologize
2143 to those who had been reporting bugs to me and that I had not answered.
2144 If your bug is still in the new version, please tell me. Here is a
2145 summary of the main changes.
2147 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2148 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2149 portability, speed, and a lot more but you'll figure it out later.
2150 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2151 I wanted to release the new MSG as soon as possible and I have
2152 broken GRAS, which is the reason why, for now, only MSG is fully
2153 functional. A laconic description of these projects is available
2154 in the documentation.
2156 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2157 kernel optimized for our needs (called SURF but only the developers
2158 should use it). Hence, MSG is now roughly 30 times faster and I think
2159 that by rewriting a little bit MSG, I could event speed it up a little
2160 bit more. Beside the gain in speed, it is also much easier to encode a
2161 new platform model with SURF than it was with SG. More to come...
2163 * REVOLUTION 3: I have tried to change a little as possible the API of
2164 MSG but a few things really had to disappear. The main differences
2165 with the previous version are :
2166 1) no more m_links_t and the corresponding functions. Platforms are
2167 directly read from a XML description and cannot be hard-coded
2168 anymore. The same format is used for application deployment
2169 description. The new format is described in the documentation.
2170 Have a look in tools/platform_generation. There is a tiny script
2171 that converts from the old platform format to the new one. Concerning
2172 the application deployment format, parsing the old one is tricky.
2173 I think most of you should however be able to convert your files. If
2174 it is really an issue, I can write a C code that does the conversion.
2176 2) the toolbox tbx does not exist anymore. We now have a library
2177 with much more data-structures but without the hash-tables (we have
2178 dictionaries that are much faster).
2180 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2182 *****************************************************************************
2183 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2184 * don't feel like converting the oldies. (Mt) *
2185 *****************************************************************************
2188 Version 2.90: "the long awaited one"
2189 - Finished rewriting and debugging MSG. Rewrote the documentation.
2190 - disable GRAS for now since it needs to be ported to the newest SG
2193 - Finish the port to windows (using mingw32 for cross-compile)
2196 - Main loop and datastructures of SURF. A cpu resource object is
2197 functional. Surf can thus be used to create cpu's with variable
2198 performance on which you can execute some actions.
2200 2004-11-15 Martin Quinson
2201 - Port to ARM. Simply added the alignment and size descriptions. Should
2202 work, but the ARM machines are so slow that I didn't had the opportunity
2203 to 'make check' over there yet.
2205 2004-11-15 Arnaud Legrand
2206 - Trace manager now written. It uses a heap structure and is therefore
2207 expected to be efficient. It may however be speeded up (particularly
2208 when many events occur at the same date) by using red and black
2209 trees. One day maybe...
2210 - Max-min linear system solver written. It uses a sparse matrix
2211 structure taking advantage of its expected use. Most operations are
2212 O(1) and free/calloc are called as few as possible. The computation of
2213 the minimum could however be improved by using a red and black tree
2216 2004-11-03 Arnaud Legrand
2217 - Rename every gras_* function that was in xbt/ to its xbt_
2219 - Add a heap and a doubly-linked list to xbt
2220 - Added a dichotomy to the dictionaries. make check works as well before
2221 so I assume that the patch is correct. I do not know however if things
2222 run effectively faster than before now. :)
2224 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2225 SimGrid, and the version number is bumped to 2.x
2227 2004-10-29 Martin Quinson
2228 - Introduction of the remote errors.
2229 They are the result of a RMI/RPC on the remote machine.
2230 ErrCodes being scalar values, you can't get the host on which those
2231 errors did happen. Extending the error mechanism as in Gnome is possible.
2232 No idea yet whether it is a good idea.
2234 2004-10-28 Martin Quinson
2235 - Interface revolution: the Starred Structure Eradication.
2236 I used to do typedef struct {} toto_t; and then handle *toto_t.
2237 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2238 - ???_t is a valid type (builded with typedef)
2239 - s_toto_t is a structure (access to fields with .)
2240 - s_toto is a structure needing 'struct' keyword to be used
2241 - e_toto_t is an enum
2242 - toto_t is an 'object' (struct*)
2244 typedef struct s_toto {} s_toto_t, *toto_t;
2245 typedef enum {} e_toto_t;
2246 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2247 s_toto_t) is private.
2249 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2250 so that it can set the variable to NULL. It was so for dicts and sets,
2251 it changed for dynars.
2253 - Fix a bunch of memleaks in dict_remove
2254 - Fix a bug in sg/server_socket opening: it failed all the time.
2256 2004-10-07 Martin Quinson
2257 - Speed up dynar lookup operation a bit.
2259 gras_dynar_get is dead.
2261 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2262 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2263 (which gives you the address of the stored data).
2265 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2266 copy of the data using an affectation to do the job and not a memcpy.
2268 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2269 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2271 It does not really speedup the dynar test because they are
2272 setting elements all the time (and look them seldom). But the dict does
2273 far more lookup than setting.
2275 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2277 2004-10-05 Martin Quinson
2278 - Allow to (en/dis)able the cycle detection at run time.
2280 Whether we should check for cycle or not is now a property of each
2281 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2282 datadesc_cycle_unset allow to remove this property when previously set.
2284 Note that the cycle detection is off by default since it impacts the
2285 performance. Watch the data you feed GRAS with ;)
2287 This property is hereditary. Any element embedded in a structure having it
2288 set have it set for the time of this data exchange.
2290 You should set it both on sender and receiver side. If you don't set it on
2291 sender side, it will enter an endless loop. If you forget on receiver
2292 side, the cycles won't be recreated after communication.
2294 - Header reorganization.
2295 Kill gras_private.h, each submodule must load the headers it needs.
2297 2004-10-04 Martin Quinson
2298 - Interface revolution: do not try to survive to malloc failure.
2300 Now, gras_malloc and friends call gras_abort() on failure.
2301 As a conclusion, malloc_error is not a valid error anymore, and all
2302 functions for which it was the only gras_error_t return value are
2303 changed. They now return void, or there result directly.
2304 This simplify the API a lot.
2306 2004-09-29 Martin Quinson
2307 - Re-enable raw sockets.
2308 Created by gras_socket_{client,server}_ext;
2309 Used with gras_raw_{send,recv}
2312 It should allow to kill the last bits of gras first version soon.
2314 This is not completely satisfactory yet (duplicate code with
2315 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2318 - Simplify transport plugin (internal) interface by not passing any
2319 argument to _server and _client, but embedding them in the socket
2322 2004-09-28 Martin Quinson
2323 - Finish the port to AIX.
2324 autoconf was my problem (segfault within the malloc replacement
2325 function. No idea why)
2327 2004-09-16 Martin Quinson
2328 - Fix some size_t madness on 64bit architectures.
2330 2004-09-08 Martin Quinson
2331 - Reduce the number of system headers loaded, overload some more system
2332 calls (such as malloc to cast the result of the system one, and work
2334 - Fix and reintroduce the config support
2336 2004-09-07 Martin Quinson
2337 - Source code reorganization to allow Arnaud to surf all over there.
2338 - Allow to document the logging categories.
2339 - Remove all uppercase from logging categories and useless cleanup in names.
2341 2004-08-18 Martin Quinson
2342 Version 0.6.2 (protocol not changed; API changed)
2343 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2344 gras_error_t), and NULL when not found. Functions expecting a msgtype
2345 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2346 hopefully usefull message.
2347 - Portability to prehistoric sparcs again
2349 2004-08-17 Martin Quinson
2350 Version 0.6.1 (protocol not changed; ABI not changed)
2351 - prealloc some buffers to speed things up
2353 2004-08-11 Martin Quinson
2354 Version 0.6 (protocol not changed; ABI expended)
2355 - The parsing macro can deal with the references, provided that you add
2356 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2358 2004-08-09 Martin Quinson
2359 Version 0.5 (protocol not changed; ABI changed)
2360 - Allow to off turn the cycle detection code in data exchange at
2361 compilation time. It should be at run time, but I'm short of time (and
2362 the config stuff is still broken). That way, we keep dict out of the
2363 critical path, which is good because the performance is poor:
2364 - search not dichotomial yet
2365 - dynar give no way to access their content and memcpy everytime
2366 - In composed data description (struct, ref and so on), stop foolness of
2367 keeping the subtype's ID, but store the type itself. This keeps sets out
2368 of the critical path, which is good since they rely on dynar and
2369 dictionnaries. The only loose of that is that we cannot detect the
2370 redeclaration of a structure/union with another content (but I'm not sure
2371 the code detected well this error before anyway). We still can detect
2372 the redefinition discrepancy for the other types.
2373 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2374 it breaks the code because of type-punning used all over the place).
2375 This breaks on all non-gcc architectures (for now).
2377 All those changes (plus the buffer of last time) allow me to gain 2 order
2378 of magnitude on cruel tests consisting of 800000 array of integers on two
2379 level of a hierarchical structure (200 secondes -> 4 secondes)
2382 - the selector of reference must now return the type it points to, not
2383 the ID of this type.
2385 2004-08-06 Martin Quinson
2386 Version 0.4 (protocol changed; ABI not changed)
2387 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2388 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2389 - Implement a buffer transport, and use it by default (it relies on tcp in
2390 real life and on sg in simulation).
2391 That's a bit hackish since I had a new field to the structure to store
2392 its data without interfering with the subtype ones. Inheritance
2393 is tricky in C. And that's a kind of reverse inheritance with one class
2394 derivating two classes. Or maybe a game with java interfaces. Anyway,
2395 that's damn hard in C (at least).
2396 Moreover, I got tired while trying to ensure plugin separation and
2397 genericity in SG mode. MSG wants me to do weird things, so let's go for
2398 cruel hacks (temporarily of course ;).
2399 See comment in transport_private.h:71
2400 - do not include all the _interface headers in private but in the files
2401 which really need them (to cut the compilation time when they are
2404 2004-07-26 Martin Quinson
2405 Version 0.3 (protocol not changed; ABI changed)
2406 - Major overhault of the datadesc interface to simplify it:
2407 - shorted the function names:
2408 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2409 - add a trivial way to push/pop integers into the cbps without malloc.
2410 This allows to make really generic sub_type description, which simply
2411 pop their size of the stack.
2412 - add a function gras_datadesc_ref_pop_arr() which does what users want
2413 most of the time: Declare a dynamic array (which pops its size of the
2414 stack) and declare a reference to it. Poor name, but anyway.
2415 - kill the post-send callback, add a post-receive one
2417 2004-07-23 Martin Quinson
2418 Version 0.2 (protocol changed; ABI changed)
2419 - add some testing for cpbs in the test cases, and fix some more bugs.
2420 This invalidate again the little64 data file, since I cannot regenerate
2422 - remove an awfull optimization in the logging stuff, allowing me to:
2423 - understand it again
2424 - learn gcc how to check that the argument match the provided format
2425 - fix all errors revealed by gcc after that
2426 - internal keys of dict are not \0 terminated. Deal with it properly in
2427 loggings instead of segfaulting when the user want to see the logs :-/
2429 2004-07-22 Martin Quinson
2430 - Fix some stupid bug preventing cbps (callback postit) from working
2432 2004-07-21 Martin Quinson
2433 - Some documentation cleanups
2434 - remove the useless last argument of msgtype_declare
2435 - rename the Virtu functions to fit into the 'os' namespace
2436 - move headers src/include -> src/include/gras/ and stop fooling with
2437 gras -> . symbolic link
2438 - make distcheck is now successful
2440 2004-07-19 Martin Quinson
2442 - Build shared library also
2443 - Install html doc to the right location
2444 - stop removing maintainer files in make clean
2445 - build tests only on make check
2447 2004-07-13 Martin Quinson
2449 - No major issue in previous version => change versionning schema
2450 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2451 corresponding dataset.
2453 2004-07-11 Martin Quinson
2455 - Get it working with any kind of structure (we can compute the padding
2456 bytes remotely for all the architectures I have access to)
2457 - Implement the structure parsing macro (still not quite robust/complete)
2458 - Improvement to the remote testing toysuite
2460 2004-07-10 Martin Quinson
2461 [autoconf mechanism]
2462 - get ride of a bunch of deprecated macros
2463 - actually run the test for two-compliment, not only compile it :-/
2464 - test whether the structures get packed (and bail out if yes. Damn.
2465 Alignment is a serious matter)
2466 - test whether the structures get compacted (but respecting the alignment
2467 constraints of each types)
2468 - test whether the array fields of structures can straddle alignment boundaries
2470 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2471 (took me 2 hours to find that bug, looking at the wrong place)
2472 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2473 before sending/receiving and are used to compute the offsets of fields
2474 - Given that padding size depend even on compiler options, keep track of
2475 alignment and aligned_size only for the current architecture. Not a big
2476 deal since we send structure fields one after the other (seems
2478 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2479 program, let it work on linux/gcc/little32. portability todo.
2481 2004-07-08 Martin Quinson
2482 - import and improve remote compilation support from FAST
2483 - make sure make check works on half a dozen of machines out there
2485 2004-07-07 Martin Quinson
2486 Let's say it's version 0.0.3 ;)
2487 - Implement conversions (yuhu!)
2488 - Let it work on solaris (beside conversion, of course)
2489 - Stupid me, using rand() to generate the conversion datatests in not wise.
2491 2004-07-06 Martin Quinson
2492 - Let make dist work, since I'm gonna need it to compile on remote hosts
2493 - Let Tests/datadesc_usage write the architecture on which the file was
2494 generated as first byte.
2495 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2496 architecture descriptions.
2497 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2498 on those architectures.
2499 - Optimization: send/recv array of scalar in one shoot
2501 2004-07-05 Martin Quinson
2502 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2504 - Plug a whole bunch of memleaks
2505 - each process now have to call gras_{init,exit}. One day, their log
2506 settings will be separated
2507 - Continue the code factorisation between SG, RL and common in Transport.
2509 2004-07-04 Martin Quinson
2511 - Redistribution between SG and RL.
2512 We wanna have to accept in SG, so move accepted related parts of RL in
2513 the common part. (more precisely, the dynar of all known sockets is no
2514 more a static in transport.c, but part of the process_data)
2516 [gras_stub_generator]
2517 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2519 2004-07-03 Martin Quinson
2520 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2521 them all up in one shot)
2523 - Fix the ugly case of reference to dynamic array.
2524 - New (semi-public) function gras_datadesc_size to allow the messaging
2525 layer to malloc the needed space for the buffer.
2527 - gras_socket_close now expect the socket to close (and not its address to
2528 put NULL in it after it). This is because the socket passed to handlers
2529 is one of their argument (=> not writable).
2531 - propagate the interface cleanup from last week in datadesc, ie remove a
2532 superfluous level of indirection. User pass adress of variable
2533 containing data (both when sending and receiving), and not of a variable
2534 being a pointer to the data. Let's say that I like it better ;)
2535 The price for that is constructs like "int msg=*(int*)payload" in
2536 handlers, but it's a fine price, IMHO.
2538 - Let it work in RL (yuhu)
2540 2004-06-21 Martin Quinson
2542 - porting SG plugin and SG select to new standards (works almost).
2543 - plug memleaks and fix bugs around.
2546 - cleanup the prototype of data recv and force users to specify when they
2547 want to handle references to objects. Test case working even for cycles.
2548 - plug memleaks. Valgrind is perfectly ok with this.
2550 2004-06-12 Martin Quinson
2552 - cleanup the separation between plugin and main code in plugin creation
2554 2004-06-11 Martin Quinson
2556 - Reput hook for raw sockets, needed for BW experiments
2557 - kill a few lines of dead code
2558 [Data description] Interface cleanup
2559 - gras_datadesc_by_name returns the searched type or NULL.
2560 That way, no variable is needed to use a type desc once, which makes
2562 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2563 two parameters were strings (field name, type name), leading to
2565 [Dicos] Interface cleanup
2566 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2567 This is consistant with the dynar API.
2569 2004-04-21 Martin Quinson
2571 - Porting to new standards.
2573 - interface cleanup.
2574 There is no bag anymore, no need to take extra provision to mask the
2575 pointers behind "ID".
2576 Better splitup of functions between files create/exchange/convert.
2577 This is still a bit artificial since convert and receive are so
2578 interleaved, but anyway.
2580 - add a queued message list to procdata (the ones not matching criteria
2582 - factorize some more code between SG and RL wrt procdata
2584 - use gras_exit in example to track memleaks
2585 - get rid of gs_example now that GS is properly integrated into gras
2586 - update run_test to integrate the latest tests (datadesc)
2588 - rename WARNINGn macros to WARNn since it prooved error-prone
2590 2004-04-19 Martin Quinson
2592 - register init/exit functions within gras module mechanism
2593 - send/receive function.
2594 Convertion is not implemented, but short-cutted if not needed.
2595 struct/array elements are sent one by one (instead of block-wise), but
2596 nobody really cares (yet). Get a prototype before optimizing.
2597 - tests (using a file socket) for DD send/receive on:
2598 - base types: int, float
2599 - array: fixed size, string (ie ref to dynamic string)
2600 - structure: homogeneous, heterogeneous
2601 - chained list, graph with cycle
2602 Believe it or not, valgrind is not too unhappy with the results. The
2603 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2607 - Bugfix when using a filename explicitely (instead of '-')
2609 2004-04-09 Martin Quinson
2611 - factorize more code between RL and SG in socket creation
2612 - Complete the implementation and tests of:
2614 o file (only in RL, and mainly for debugging)
2616 I lost 3 days to design a portable address resolver, and then decided
2617 that the prototype mainly have to run on my box.
2618 Addressing portability too early may be like optimizing too early :-/
2620 - use gras_init in the Tests instead of the crappy parse_log_opt
2621 (the latter function is removed)
2622 [Conditional execution]
2623 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2624 [Code reorganisation]
2625 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2626 Build examples against the RL library, since there is no way to disable
2627 its creation for now.
2629 For information, the beginning of coding on GRAS was back in june
2630 2003. I guess that every line has been rewritten at least twice since