1 SimGrid (3.10) NOT RELEASED; urgency=low
4 * Reintegrate Java to the main archive as desynchronizing these
5 package is not acceptable anymore (Java is now considered stable)
6 * Bugfix: Task.setDataSize() only changed the C world, not the value
7 cached in the Java world
9 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
11 SimGrid (3.9) stable; urgency=low
13 The Grasgory release: GRAS is really dead now.
15 * Complete overhaul of the internal host structures scheme.
18 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
19 was considered as experimental and badly maintained since then.
20 * Keeping it was thus a trap to our potential users, that could take
21 it instead of MSG or SMPI by mistake despite is pity state.
22 * GRAS seems to have very few users (if any), and no one volunteered
23 to maintain it further. It also induces a lot of XBT code (for
24 portability sake), that must be maintained too.
25 * For all these reasons, we killed GRAS. If someone wants to revive it
26 in the future, don't cry, our git history still remembers of GRAS.
29 * Major overhaul. Merge our documentation again as time proved that
30 splitting it was really not helping our users.
31 * Further improve the developer documentation to help newcomers
32 hacking on SimGrid itself. The user documentation (and in
33 particular, the beginner documentation) is still in a sorry state.
36 * Now works on Windows too!
37 * Much more extensive test suite, from MPICH
40 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
41 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
42 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
43 * Bug fix: let task be scheduled when the last dependency to be solved is
45 * Remove SD_load_environment_script function.
46 Use the C sg_platf function if you want to declare a platform
50 * New function: MSG_process_get_number()
51 * Old function documented: MSG_config()
52 * Remove MSG_load_platform_script function
53 Use the C sg_platf function if you want to declare a platform
57 * Change the default value of the TCP_gamma constant (maximal size of TCP
58 congestion window) to a more realistic 4MiB value. If you notice changes in
59 your simulation results, you can fall back to the previous 20k tiny window
60 by adding --cfg=network/TCP_gamma:20000 on command line.
61 * (Hopefully) fix a bug wrt periodic availability and state traces
62 * Bug fix: use default values at start when first event in availability/state
63 trace is not at time 0.
66 * remove the "new_" part of function name sg_platf_new_trace_connect
67 (resulting in sg_platf_trace_connect), since it does not create
71 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
72 use case where it's really mandatory, and maintaining it was a pain in
74 * New: xbt_fifo_search(), search an item with a user-provided
75 comparison function instead of dumb pointer comparison.
78 * Fix the lua deployment:
79 Use `simgrid.init_application()` before deployment instead of
80 `simgrid.msg_register_application()` after.
83 * Transfer the tracing files into the corresponding modules.
85 -- Tue Jan 29 19:38:56 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
87 SimGrid (3.8.1) stable; urgency=low
89 The "we are told that some people want to also *install* the simgrid
92 * Add missing file "tesh.1" to the archive.
94 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
96 SimGrid (3.8) stable; urgency=low
98 The Psssshiiiit release: SimGrid jumps into the Cloud.
101 * Add an experimental interface to manipulate VMs. They are mainly
102 process groups with very few intrinsic semantic, but they should
103 allow you to build the semantic you want easily.
104 * New function: MSG_host_set_property_value()
105 * New function: MSG_process_on_exit(). To clean memory in all cases.
106 * Bug fixes that made the host (and link) failures unusable.
107 * Add a way to auto-restart process when the host in which they are
108 executing comes back (ON_FAILURE="RESTART" on deployment file,
109 MSG_process_auto_restart_set).
110 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
111 please stop using the old ones, they are DEPRECATED.
113 * Deprecate functions MSG_global_init() / MSG_global_init_args()
114 Please use MSG_init() instead. (reducing the amount of entry
115 points in the library helps us).
116 * Make it impossible to link against the wrong version of the lib
117 * Deprecate MSG_clean(). No need to call it anymore.
118 * Function MSG_get_host_number() is not deprecated anymore.
121 * Split the doc into a user guide and a reference guide.
122 * Start a developper guide to help people hacking on SimGrid.
125 * Enable tracing by default. This modules rocks you should use it.
126 * Remove option custom_flags. Now use environment variables CFLAGS
128 * Use default cmake things to detect lua instead of home grown ones.
129 * New option "enable_mallocators" to disable mallocators, for debugging
130 purpose ("on" by default).
133 * Bug fixes around the resource failures: don't let the processes
134 survive the host they are running onto.
135 * Add an interface to auto-restart processes when the host in which they are
136 executing comes back.
137 * Ensures that SIMIX_clean is called automatically. It's not part of
138 the public interface anymore (bindings should be updated).
141 * Bug fix for when SD_Simulate is called with a positive value: be careful
142 when comparing doubles. Sometimes they are different for non significant
144 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
145 parallel task whose initial work is distributed among host according
146 to the Amdahl's law. Such tasks are created with a parameter alpha
147 that corresponds to the non-parallelizable part of the computation.
148 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
149 between two sets of workstations assuming a 1D block distribution (each
150 workstation owns a similar share of data) on both sides.
152 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
153 Data redistribution will be automatically scheduled once parent and child
154 are both scheduled. The filling of computation_amount and
155 communication_amount structures is now done seamlessly thanks to the chosen
157 * New function SD_workstation_dump to display various information
158 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
159 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
160 bandwidth on the route onto which the task is scheduled and the amount of
162 To divide the nominal bandwidth by 2, the rate then has to be :
163 rate = bandwidth/(2*amount)
164 * Compute tasks that have failed can now be rescheduled and executed again
165 (from their beginning)
166 * Increasing source code coverage (src/simdag is now covered at 95.8%
170 * Re-implement time-independent trace replay using SMPI (at the
171 smpi_smp_* level) instead of MSG. This should replace
172 examples/msg/actions/actions.c
173 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
174 hindexed and structs)
175 * Implement the exchange of non-contiguous data.
176 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
177 * Correct behavior of smpi/sender_gap and set its default value to 0
178 * Add option to asynchronously send small messages to allow better
179 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
180 specifies the size in bytes under which messages will be asynchronously sent.
181 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
182 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
185 * New C interface to define a platform: XML is now optional.
186 For more info, please check include/simgrid/platf.h.
187 * New interface to define random platforms from the C:
188 For more info, please check include/simgrid/platf_generator.h and
189 examples/msg/masterslave/masterslave_platfgen.c
190 * Export a sg_cmdline dynar containing all the arguments we got from
194 * Two new tracing options for adding comments to trace file so you
195 can track your experiments (see --help-tracing for details).
196 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
197 * Adding the SimGrid version that generated the trace file as a comment.
198 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
199 * Fix to avoid key clashes on Paje links
200 * Other minor fixes related to the Paje specification
203 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
204 and renamed to xbt_str_hash() and xbt_str_hash_ext().
205 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
206 * Greatly improve the robustness of mmalloc to user errors (such as
207 using an area after freeing it, or freeing it twice)
209 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
211 SimGrid (3.7.1) stable; urgency=low
214 * Restore the prototype of MSG_process_create_with_environment() to
215 the pre-3.7 situation by removing the kill_time argument.
216 * Add a MSG_process_set_kill_time() function instead.
219 * Fix weird behaviors when dealing with parallel tasks.
222 * Simgrid is now built as a dll.
223 * Simgrid-java now works on Windows.
224 * Simgrid-Java is now included into Windows package.
227 * First pre-build package for MacOSX.
230 * Fix compilation when using MSG_USE_DEPRECATED.
231 * Fix some compilation issues on Macs and Windows.
232 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
233 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
235 We discovered that the Lua console is broken, but we are missing the
236 manpower to fix it right now. The problem existed in 3.7 too, so we
237 are not blocking the release for that. Sorry if you depended on this
238 feature, any help would be really welcome.
240 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
242 SimGrid (3.7) stable; urgency=low
244 The "spring cleanups (before the next Big Project kicks in)" release.
247 * We can specify the SMPI latency/bandwidth factor with command line
248 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
249 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
250 You can also use the "config tag" from platform file by adding this line
251 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
252 (see "example/platforms/tag_config.xml" to use "config tag").
253 Note that the command line supersedes the platform file configuration.
254 * Change the correction factors used in LMM model, according to
255 the latest experiments described in INRIA RR-7821.
256 Accuracy should be improved this way.
257 * Use the partial invalidation optimization by default for the
258 network too. Should produce the exact same results, only faster.
259 * Major cleanup in surf to merge models and split some optimization
260 mechanisms from the core of the models. As a result you can now
261 specify which model to use (e.g., --cfg=network/model:LV08
262 --cfg=cpu/model:Cas01) and which optimization mode to use
263 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
264 Incompatible combinations should err at initialization. See
265 --help-models for the list of all models and optimization modes.
266 * The CLM03 workstation model was dropped for simplicity because it
267 used the deprecated CM02 network model. Use default instead.
268 * Rename the TCP_gamma configuration option to network/TCP_gamma
269 * Rename the coordinates configuration option to
270 network/coordinates, and document it
271 * Use now crosstraffic keyword instead of the terribly misleading
272 fullduplex keyword. It is activated by default now in the current
273 default model, use --cfg=network/crosstraffic:0 to turn it off.
274 * Ongoing refactoring the model parsing to make XML files optional
275 See include/simgrid/platf.h for details (still to be completed)
278 * Major overhaul of the documentation. Almost instructive now :/
279 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
280 functions and friends. This interface was considered as
281 deprecated since over 2 years, it's time to inform our users that it is.
282 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
283 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
284 use these (crappy) functions in your code.
285 These functions will be removed soon. Stop using them now.
286 * Deprecate MSG_get_host_{table,number}
287 Implement MSG_hosts_as_dynar() instead.
288 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
289 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
292 * Stabilize the parallel execution mode of user contexts
293 * Introduce configuration variables to control parallel execution:
294 - contexts/synchro: Synchronization mode to use when running
295 contexts in parallel (either futex, posix or busy_wait)
296 - contexts/parallel_threshold: Minimal number of user contexts
297 that must be part of a scheduling round to switch to parallel
298 execution mode (raw contexts only)
299 * Fix bugs that prevented to use suspend/resume along with
300 synchronization structures.
301 * Fix bugs in process termination that lead to invalid memory access
302 in very specific conditions.
305 * Introduce a parallel mode for the models (controlled by surf/nthreads
306 configuration item). In our tests, running the models in parallel
307 never lead to any speedups because they are so fast that the gain
308 of computing each model in parallel does not amortizes the
309 synchronization costs, even when ultra fast futexes are used.
310 This is released anyway because YMMV.
313 * Performance boost by using a swag internally to compute the set of
314 tasks that are finished and should constitute the return value of
318 * Enable it by default now that it is considered rather stable.
321 * Documentation of the tracing functions.
322 * Performance gains when tracing categorized/uncategorized resource
323 utilization by avoiding calls to get route when updating resource
324 variables. LMM constraints are being used instead.
325 * API changed to set task categories. Use MSG_task_set_category instead
326 of TRACE_msg_set_task_category, and SD_task_set_category instead
327 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
328 * Bugfix for graphicator, routes not correctly obtained, memory leaks
329 * Examples for link user variables added (see at examples/msg/tracing/)
330 * Deprecated function TRACE_msg_set_process_category completely removed
331 * Trace header updated according to the latest Paje file format
332 * Tracing network lazy updates, no longer obligate users to use full updates
333 * --cfg=tracing/platform:1 also registers power/bandwidth variables
334 * Experimental: let user code declare/set/push/pop application states for hosts
335 * API changed to allow the manual creation of graph configuration files
336 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
339 * Improve the API of Lua MSG bindings, using the Lua spirit.
340 * Each simulated process now lives in its own Lua world (globals are
341 automatically duplicated). It helps writing simulators. Will allow
342 to run Splay programs within SimGrid in the future.
343 * Add a Chord example in Lua, equivalent to the MSG one.
346 * Start the implementation of a solution to express temporal
347 properties, not only local assertions. This is still an
348 experimental work in progress, stay clear from it to be safe.
352 - Add new runtime parameters --help-logs and --help-log-categories
353 to display informations about supported logging parameters and
355 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
356 don't exists anymore.
357 * Mallocators: allow value NULL for the reset function.
359 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
360 create homogeneous dictionaries, where all the elements share the
361 same free function. Non homogeneous dictionaries will be
362 deprecated in the next release.
363 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
364 - Multi-level dictionaries are deprecated.
366 - new function xbt_dynar_search_or_negative() that is useful when
367 you have less than 2 million elements in your dynar and don't
368 want of the extra complexity of catching exceptions when the
369 element is not found.
371 - Make xbt_os_thread module (for thread portability) public.
372 Documentation is still to come, sorry.
374 - Cleanups and simplifications to make it maintainable again.
375 - Exotic features (such as memalign and valloc) were removed.
376 - The metadata were extended and improved so that the
377 model-checker becomes able to explore and inspect the heaps.
378 - This may induce a performance drop when enable_model-checking is
379 ON in cmake (even if it's not used in the simulation), but it is
380 necessary at this point to get MC working.
382 Turn model-checking OFF if simulation performance matters to you.
383 Not enabling it at runtime is not enough, disable it in cmake.
385 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
387 SimGrid (3.6.2) stable; urgency=low
389 The "Not coding new stuff allows to polish old things" release.
392 * New bindings to the NS3 packet level simulator (experimental)
393 * Use the raw (efficient) execution contexts instead of the sysv
394 (portable) ones when possible.
395 * libpcre is now mandatory in any cases since not using it led to
396 severe performance loss and possibly other issues
397 * Update the XML platforms:
398 - G5K: include the latest machine in Nancy
399 - GridPP and LCG: new platforms
400 * Documentation was partially updated, at least (more to come)
402 Bug fixes, cosmetics and small improvements
403 * Free terminated processes before the end of the simulation to avoid
404 exhausting the memory of users having very dynamic amount of
406 * Bug fix and cosmetics about canceling non-running tasks
407 * Bug fix about the dot loader's issues when using libcgraph
410 * Create an installer for windows with nsis (amd64 and win32)
411 - Add an hello world project to illustrate simgrid project creation.
412 - Embed libpcre into the Simgrid installer to avoid
413 its compilation burden
414 * The raw execution contexts should work on Apple now
415 * Port to Windows 64 bits
416 - Sysv contexts now have an implementation for this arch
417 - GRAS communication features now support this arch
418 * Drop support for borland compiler on windows
419 - this code was not maintained, and we kinda depend on gcc nowadays
420 * Fix portability issues on kfreebsd/gnu: build error about semaphores
421 * Fix portability issue on unstable ubuntu: linker became picky on
424 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
426 SimGrid (3.6.1) stable; urgency=low
428 The "Oops, we broke Macs too" release
431 * Fixed contexts detection so that raw ones are used when possible
432 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
433 leads to a strange error, with user code segfaulting sometimes when
434 the generated code is not perfectly aligned (which is not
435 controllable from the user side, depends on the amount of code)
438 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
440 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
442 SimGrid (3.6) stable; urgency=medium
444 The Summer Release, also known as the "OMG! They Killed Kenny!" version
447 * Bindings now constitute their own package, separated from the main one.
448 Rationale: reduce our maintenance nightmare by reducing the module coupling
449 They will soon be released on their own on gforge.
451 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
452 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
454 GRAS: It is not considered as stable anymore, but experimental. Sorry.
455 * It's not quite deprecated for now because we have no replacement,
456 but it may soon become the case.
459 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
460 * Fortran: New user-level cache variable to store the rank of the running
461 process. This improves performance by an order of magnitude.
462 * C: New coccinelle script to automatically locate and modify global and
463 local static variables.
464 * Improved SMPI network model with a sender-side gap to account for multiple
468 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
469 only say if a communication is finished, no matter whether it succeeded or
470 failed. You can call MSG_comm_get_status() to know the status of a finished
472 * New function MSG_task_dsend() to send a task and detach it. When a
473 communication is detached, you are never notified of its success or failure
474 and the memory is released automatically once it is finished. This function
475 is useful when you don't care about the end nor the success of a
477 * Change the prototypes of action replay. Sorry for inconvenience,
478 but this is really more efficient this way (and to adapt your code,
479 you just have to fix the initialization, that shouldn't be too long)
480 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
481 ever managed to do anything useful with it.
482 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
483 takes a dynar of communications. It returns immediately and gives the
484 index of a finished communication (if any).
485 * New example: a basic implementation of the Chord P2P algorithm.
488 * New model for multi-core CPUs. You can now use the core attribute to
489 precise the number of cores of a host. This is a basic model. Every
490 process running on the host receives at most the power provided in
491 the DTD (throughput<=power). Total throughput of process cannot exceed
493 * New peer tag. This peer tag creates a tiny AS comprising a host and a
494 router linked by an up-link and a down-link (possibly asymmetrical).
495 This kind of pattern allows to easily build last-mile model style platforms.
496 Aggregating such patterns in a rule-based AS is thus the technique of
497 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
498 * New model for Vivaldi routing. We transformed the Vivaldi network model
499 into a Vivaldi routing model (based on the rule-based model). This allows to
500 combine Vivaldi based latencies with last-mile platforms.
503 * Added a check for NaN of IEEE754 infinite in the double entries of
505 * Introduce a new context factory "raw", highly inspirated from the
506 ucontext factory, but using manually crafted functions in assembly to
507 do the work in an efficient manner.
508 * Allow to change the used context factory at run time, not only at
509 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
510 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
511 contexts at runtime (only with raw contexts or ucontexts).
512 * Completely rewrote this module to allow parallel execution of user
513 processes. Use --cfg=contexts/nthreads:N to execute user processes
514 with N parallel threads (the default is 1, meaning no parallelism).
515 * Allow to decide dynamically between sequential and parallel modes.
516 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
517 processes in parallel only when their number is greater than or equal to P
519 * Added a check for NaN of IEEE754 infinite in the double entries of
523 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
524 won't output the state of processes when interrupted with Ctrl-C
525 * Add a new function xbt_dynar_to_array that transforms a dynar into a
526 NULL-terminated array. This may solve backward compatibility issues
527 due to the change to return type of SD_simulate. See also:
528 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
529 * Add new macros with variable number of arguments.
530 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
531 - in xbt/asserts.h: xbt_assert
532 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
533 - in xbt/ex.h: THROWF and RETHROWF.
534 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
536 * Change xbt_die() to accept a format string with arguments, just like printf.
537 * New data structure: xbt_lib_t, like a dict but more general and with better
541 * New configuration options
542 Options triva/categorized and triva/uncategorized can be used to generate
543 graph configuration files for Triva visualization tool.
544 * Configuration option tracing/platform is renamed to tracing/categorized
545 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
546 * New TRACE_user_link_variable interface:
547 User provides the name of the link and the tracing variable to attach to it
548 * the declaration of tracing categories must be done after the environment
550 * simpler tracing interface, just one way to declare categories
551 TRACE_category or TRACE_category_with_color, it is up to you
552 * links in the trace file are again identified by their names
553 * trace contains the full platform hierarchy exactly as declared using the ASes
554 * Options tracing/msg/[task|process]:1 groups the process by hosts
555 for both cases, tasks and processes must have names that are unique during the simulation
556 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
557 * The experimental option tracing/msg/volume is deprecated
558 its functionality may be reincorporated if needed
560 The tracing generates a trace file with unordered timestamped events,
561 because of the way the core simulator (surf) works. A script available
562 at the tools directory (fix-paje-trace.sh) can be used to put the events
563 in order. We have changed the tracing so it can generate ordered timestamped
564 events in the final trace, but depending on the simulator (and how much time
565 is simulated) that can lead to a huge memory utilization. It is deactivated
566 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
569 * Define a SIMGRID_VERSION macro in simgrid_config.h.
570 - We are trying hard to keep the API stable, but it may happen that
571 some things change (we're a research project after all, not a
572 nuclear plant operating system). If such things should happen, you
573 could rely on that macro to adapt.
574 - current value: 30600 for 3.06.00, aka 3.6
575 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
576 a number that can be compared with SIMGRID_VERSION.
577 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
578 assertions and verbose/debug logging events are disabled at compile time.
580 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
582 SimGrid (3.5) stable; urgency=medium
585 * New feature to any SimGrid-based simulator: Model-Checking
586 Check SIN#1 for more details.
589 * New Model SMPI (three-interval linear regression for correction factors)
590 See RR-7426, available at http://hal.inria.fr/inria-00527150
591 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
592 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
593 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
594 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
595 * See also src/smpi/README
600 - Tracing API changes: TRACE_start and TRACE_end should not be called
601 by user-code. They are automatically called by simulators created
602 with SimDAG, MSG and SMPI if the toolkit is compiled with
603 tracing_enabled=ON. Categories declaration and utilization remain the
604 same for MSG and SimDag.
605 - A function was added to the tracing API to declare categories with
607 - TRACE_category_with_color (char *category, char *color)
608 where color must be in the following format
609 "%f %f %f", red, green, blue
610 and red, green, blue are float values in the interval [0, 1]
611 - User can specify NULL as color parameter, or continue calling
613 On that case, the tracing system will define random colors
614 - The following command-line options are supported:
615 --cfg=tracing/filename:msg.trace
616 --cfg=tracing:1 (activate tracing, needed to use others)
617 --cfg=tracing/platform:1 (categorized resource use)
618 --cfg=tracing/uncategorized:1 (uncategorized resource use)
619 --cfg=tracing/msg/task:1 (task creation)
620 --cfg=tracing/msg/process:1 (process creation, migration)
621 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
622 --cfg=tracing/smpi:1 (SMPI interface tracing)
623 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
624 - examples of examples/msg/tracing updated
626 - DAXLoader and DOTLoader functions can generate tasks with categories
627 - A new function to attribute a category to SD tasks:
628 TRACE_sd_set_task_category (SD_task_t task, char *category)
629 * Tracing the MPI interface implemented by SMPI
630 - Collective operations are traced with states
631 - Point-to-Point operations are traced with states/links
632 - Tracing activated by a parameter "-trace filename" passed
633 to smpirun during execution (considering that simgrid
634 is compiled with tracing enabled)
635 - To run the simulation with gdb, the simulator
636 accepts --cfg=tracing/smpi:1 to trace SMPI
637 - tesh files to check if smpi tracing is ok
638 - See examples/smpi/NAS/DT-trace
639 * GTNetS tracing re-worked
640 - adaptation to the tracing system of GTNets to cope
641 with modifications regarding the fullduplex mode
642 - new tesh files to check if gtnets tracing is ok
645 * Asynchronous communications through the functions:
646 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
647 * New function: MSG_load_platform_script()
648 to make possible using a lua script instead of XML files to set up platforms
649 * New function: MSG_set_function
650 to associate functions to processes, used when bypassing the parser
651 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
653 Platforms: Add some more examples in examples/platforms
654 * Grid'5000: see www.grid5000.fr
655 * *_30000_hosts.xml: various huge files [mainly scalability testing]
658 * Change the XML format. This is a very important modification. SimGrid
659 3.5 introduces a new hierarchical format based on the notion of
660 Autonomous Systems. Compatibility with old format is ensured
661 through the perl script provided in the install bin directory
662 bin/simgrid_update_xml.
663 It is now possible to build platforms with specific routing
664 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
665 connect several platforms together. We will try to provide soon
666 set of realistic platforms exploiting these properties (have a
667 look at examples/platforms/ for the moment).
668 * Take the opportunity of the XML format change to be a good XML citizen:
669 rename link:ctn to link_ctn and similar changes (also dealt with by
671 * Add a new routing scheme (rule-based) using regular expressions. It
672 enables to have an extremely low memory footprint when the
673 underlying routing is simple and can be compactly described. You
674 need to have libpcre4-dev (perl regular expressions) installed if
675 you want to use this routing scheme.
676 * Revive the cluster TAG and allow to easily and efficiently (both in
677 term of memory and speed) connect clusters together. Have a look
678 at teshsuite/simdag/platforms/ to see how this can be done. With
679 this tag, you can create clusters with thousands of tasks at no
680 cost (have a look at examples/platforms/).
681 Note: clusters are implemented as ASes, so there is no need for an
682 enclosing AS tag if you have only one cluster in your platform.
683 * Add new generic functions in the public interface that allows the user
684 to call SURF 'create_resource' methods from your code (same
685 functionality as the XML bypass mechanism but with a much lighter
687 * Add a new model (enabled through command line --cfg=network/model:SMPI)
688 that uses a piecewise linear approximation to produce better
689 results when exchanging small messages.
690 * Add a new parameter to handle correctly full duplex link and account
691 for interferences between uplink and downlink communications
692 (activate with --cfg=fullduplex:1).
695 * Rename the SD_READY (all dependencies are satisfied and task is
696 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
697 dependencies are satisfied) state.
698 This prevents a confusion between the notion of "ready to schedule"
699 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
700 simulated" (SD_RUNNABLE) used by the simulation kernel.
701 * Change the way a task is considered as ready. Instead of removing
702 dependencies when a task is done, a counter is decreased. This way,
703 it is always possible to reach ancestors thanks to the
704 SD_taks_get_parents function (even after the end of the simulation.)
705 * Change the return type of SD_Simulate from (SD_task_t*) into
706 xbt_dynar_t. This function was in handling a dynar internally and
707 converted it into a NULL terminated array for historical reasons.
708 * New function SD_dotload(char*) to load a DAG described in dot
709 format. This loader and the corresponding examples require the
710 installation of the graphviz library.
711 * Fix a bug in the management of tasks of size 0 in the surf network
712 models. This problem was only visible with SIMDAG and you should
713 thus disregard results produced with earlier versions if you
714 relied on this feature (some tasks were blocked because of this).
715 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
716 even though your applications were doing only point-to-point
717 communications and sequential computations. Now you can really use any
718 model you want (of course, if you create real parallel tasks, which are
719 not implemented in most models beside ptaskL07, this will abort).
720 * Add an example that schedules a DAX on an heterogeneous platform
721 using a Min-Min strategy.
722 * New function SD_workstation_get_current_task() that returns the kind
723 of task currently running on a workstation in the sequential access
725 * Raise some warnings when unexecuted tasks remains at the end of the
726 simulation. This is usually caused by cycles in the DAG.
729 * New function: SIMIX_process_set_function() called by MSG_set_function
730 * Change the underlying waiting queue in semaphores so that a process
731 can wait on several of them simultaneously (as in waitany).
732 * Fix the way to handle tokens in semaphores so that all access patterns
733 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
734 * kill the dirty pimple SIMIX_message_sizes_output()
735 Please use (proper) visualization instead
738 * New data container: setset (set of sets of elements)
739 * New module: mmalloc (mapped malloc, allowing to have several
740 independent segments of malloc)
741 * New function: xbt_dict_cursor_set_data()
742 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
743 * New function: xbt_dynar_is_empty()
744 * New function: xbt_fifo_get_last_item()
745 * Fix xbt_dynar_shrink(): use the right element size.
746 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
747 uninitialized areas during expand.
748 * Fix semaphores: previous implementation was severely broken.
749 * Use library init/fini functions for our initialization.
750 - you can use logs and other feature as soon as you want in your
751 code (even before the xbt_init / MSG_init)
752 - xbt_exit is now a no-op and produce a warning when used.
755 * Port GRAS to new SIMIX mechanisms. This allows gras users to
756 benefit from the latest improvement to the simulation kernel.
757 * Kill measurement sockets for now. If you rely on them, sorry. This
758 release is not for you. This feature will be reintroduced in the
759 future, but we cannot delay the release any further.
760 * New function: gras_msgtype_get_name().
761 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
762 * Fix (at last) the pmm example: it should not randomly fail anymore.
764 Build chain: bug fixes and overall polishing
765 * Cmake is now stable enough. Hence, we killed the autotools.
766 * Port to windows ( TM :)
767 * Fix the 'make install' target.
768 No need to use 'make install-simgrid' anymore
769 * Introduce a 'make dist' target compiling a *source* archive
770 'make package' compiles a binary archive
771 * Compile java files only on need
772 * Add --cd and --setenv command line options to tesh
773 * Out of source builds are not fully supported yet, but we are close
774 * Enable supernovae and optimization flags by default for our users
777 * Add layer to set up environment directly from lua, without XML.
778 * The effect of gras_stub_generator can be achieved through
779 lua too (check examples/gras/console/ping_generator.lua)
781 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
783 SimGrid (3.4.1) stable; urgency=low
785 The "Polishing easter eggs is probably a good idea" release.
786 This is a bug fixes release only.
790 * Fix a bug preventing the tasks from begin garbage collected.
793 * Fix a bug occuring when a host involved in a communication fails.
794 This was not detected properly by the other peer involved in the
795 communication. Now, it's reported as a network error.
798 * Warn the user about loop dependencies in data flow of DAX files
799 * Obey the control-flow dependencies of DAX files
802 * Add option "enable_smpi" allowing to not compile SMPI.
803 Probably useful for the (Mac) users experiencing a build error here
804 * Improve the detection of lua5.1 and ruby1.8
806 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
808 SimGrid (3.4) stable; urgency=low
810 The "Easter in Cargese" release. Also known as (major changes):
812 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
813 ~> bindings were greatly improved
814 ~> new tracing infrastructure for better visualization introduced
816 * the "Welcome to configury modernity" release.
817 ~> we switched from autotools to cmake, and improved our cdash
820 A more detailled list of changes follow (full detail in svn log).
822 Java Bindings: Various Cleanups
823 * (install java-gcj-compat-dev on debian-like to use them)
824 * Remove put/get: no need to export deprecated interface in Java
825 Use send/receive instead.
826 * Cleanup the examples and add a README per directory
827 * Remove example autoDestination (that's the only way to go now)
828 * Remove example explicitDestination (was a plain copy of basic)
829 * Make JniException a runtime exception, so that there is no need to
830 declare the fact that you may encounter such a beast. I guess that
831 nobody will ever want to survive such error.
832 * Create specific errors for each MSG case of failure:
833 host failure, transfer failure, timeout, task cancelled
834 * Cleanup the exceptions that may get thrown by each function
835 * Other internal cleanups in Java bindings. Performance still bad :/
836 Ruby and Lua Bindings: create them
837 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
838 * That's new and great, you should try them out.
839 Same functionalities than Java bindings, only even less polished
841 * Kill the useless "rate" argument of SD_task_get_execution_time()
842 Everyone used to provide -1 as a value, it was not used, and the
843 semantic of a possible use wasn't even clear.
844 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
845 either as comm costs or compute costs to mean that there is no
846 such thing for that specific task.
847 * Add a SD_task_set_name() function
848 * Fix SD_task_unschedule() on typed tasks
849 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
850 * In DAX loader, accept useless 'level' attributes to <job> since
851 LIGO DAGs have them (seem to be to ease graphical representation).
853 * Add an example masterslave_mailbox.c using send/receive and not
854 the deprecated put/get interface.
855 * Kill the MSG_paje_output() function. It's a noop since 2 years.
856 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
858 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
859 (declare MSG_USE_DEPRECATED to still have the old name)
860 * Add a MSG_task_set_data() function
861 * About trace replay (see examples/msg/actions):
863 - Allow to work with splitted trace files for each process
864 Give the specific trace file as argument of each process,
865 and call MSG_action_trace_run(NULL)
866 You can still have one merged file for all processes.
867 - Fix implementation of collective operations
868 * Allow task_execute() on 0-sized tasks (closes #10063)
870 * This is the first release of SimGrid where SMPI is not considered
871 beta anymore (even if some corners should still be improved)
872 * Port over the new SIMIX_network submodule (internal refactoring)
873 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
874 * Implement more missing elements of the standard:
876 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
878 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
879 MPI_Scan() MPI_Get_processor_name()
880 - Added implementation of missing case for Alltoall (warning: it's
881 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
882 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
883 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
884 SMPI_MPI_Allgatherv()
886 - MPI_Waitsome() was broken
887 - Allow relative includes in smpicc
888 - Command line cfg argument 'reference_speed' was ignored...
889 - Some functions did not properly lead to auto-benching of user code
890 - smpicc passes -O2 by default (just like openmpi one)
892 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
893 * Bug fixes about timeouts during communications
894 * add SIMIX_message_sizes_output() as a pimple to write to file the
895 amount of messages per size. Use gnuplot to get histogram.
896 Pimple because that's the only user-visible function of simix,
897 defined directly in xbt.h (irk, sorry)
899 - Add a SIMIX_sem_get_capacity() function
900 - Fix interactions with processe resume/suspende
901 - release_forever() was stupidly broken
902 - Fix SIMIX_display_process_status() for processes in a semaphore
903 - Make SIMIX_sem_block_onto() user-visible
904 * Refactoring context stuff:
905 - Use pseudo-OOP for better modularity
906 - reimplement SIMIX_process_kill() without process_schedule() so
907 that the latter can take as invariant that it is called from
909 - Merge context_start into context_new for sake of simplicity
911 * Add a Vivaldi network model, coded live during SUD'10 ;)
912 * Rename configuration variables to start a hierarchy:
913 o cpu_model -> cpu/model
914 o network_model -> network/model
915 o workstation_model -> workstation/model
916 * New configuration variables:
917 o network/bandwidth_factor: correction to bandwith
918 o network/latency_factor: correction to latency
919 o netwotk/weight_S: correction to the weight of competing streams
920 * Add a long description to the models, that users can see with such
921 argument on the command line: --cfg=cpu/model:help
922 * --help-models display the long description of all known models
924 * config: add the ability to set a default value after registration
925 Does not override any previously set value (e.g. from cmd line)
926 * dict: allow to have integer key and data.
927 When so, you need to use the following functions
928 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
929 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
930 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
931 In contrary to regular dicts, the key is not malloced before copy.
932 Mixing scalar and regular elements in the same dict is not tested
934 * Allow to use xbt_dynar_shrink() to expend the dynar instead
935 Tracing for Visualization:
936 * SimGrid is now instrumented in order to generate a trace file for
937 visualization analysis: to use it, need to compile SimGrid with the
938 "tracing" option enabled, and instrument the program using SimGrid with
939 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
940 (among other functions).
941 * The instrumentation only traces the platform utilization for now
942 * Documentation to use the tracing functions and how to analyze the
943 traces with the Triva tool is written.
944 * More information about: SimGrid FAQ (in the section Tracing Simulations
947 * We moved to cmake as default build system. Autotools support will
948 be dropped soon. Check the FAQ for more info about how to use it.
949 * Greatly improved our cdash/ctest interactions
950 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
951 * Added memory checking tests with valgrind; lot of memleak fixing.
952 This may be the first release of simgrid with so few memory issues
953 * Added code coverage tests.
954 Our coverage is still improvable, but at least we see it on cdash.
956 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
958 SimGrid (3.3.4) stable; urgency=low
960 The "Desktop Grid needs love too" release (also called Xmas release).
963 * Major speedup in the maxmin system solving by using lazy evaluation
964 Instead of solving completely the maxmin system at each iteration,
965 only invalidate (and recompute) the modified parts.
966 This new feature is enabled in default models but you can try to
967 turn it on with "--cfg:maxmin-selective-update=1" for other models.
968 * Cas01 IMproved as default CPU model
969 This CPU model is the same Cas01 model, but it uses the
970 maxmin-selective-update flag and a heap structure to manage
971 actions on SURF kernel.
972 It reduces the complexity to find the next action to finish and,
973 consequently, it's faster than the old Cas01.
974 This is the new default CPU model (Cas01).
975 * Rename the old Cas01 model to Cas01_fullupdate
976 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
977 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
978 * CpuTI (CPU Trace Integration)
979 A new CPU model whose objective is simulate faster when using
980 availability trace files.
981 Instead of using a full featured, over engineered maxmin system for
982 CPU modeling, this model does the pre-integration of traces files
983 to calculate the amount of CPU power available, and so, executes
984 faster than the old CPU models.
985 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
986 * Use LV08 as default network model since it gives better accuracy
987 for small messages and shouldn't change things for big ones.
988 Use --cfg=network_model:CM02 to get the previous behavior.
991 ******************************************
992 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
993 ******************************************
994 * The new CPU model may changes simulations!
995 The point is that events occurring at the exact same timestamp
996 are not scheduled in the same order with the old and new
997 version. This may be enough to completely change the execution
998 of simulations in some cases.
999 * The new network model will change simulations!
1000 This new model is more realistic than the previous one, so you
1001 should consider redoing your old experiments with this model.
1002 Sorry for the inconvenience.
1005 * Introduce the supernovae compilation mode
1006 When compiled that way, the whole SimGrid (or almost) is put in a
1007 single compilation unit and compiled in one shoot.
1008 This is to help gcc which has difficulties to inline stuff from one
1010 The speedup seem to be above 15%, althrough more tests are needed on
1011 amd64 to confirm that gain.
1014 * Port of MSG's mailbox on top of SIMIX network
1015 The put/get mechanism was greatly simplified on the way.
1018 * New SIMIX network module. Provides:
1019 - Mailbox: rendez-vous mecanism to find with who you want to speak
1020 - Synchronous send/recv: easier and hopefully faster since the
1021 logic is handled in the maestro process directly now
1022 - Asynchronous send/recv: you dreamt of it? It's here now
1023 Too bad that nobody cared enough to propagate the change to MSG.
1024 * Add semaphores as SIMIX synchronization mechanism.
1027 * new function SD_daxload(char*) to load a DAX file
1028 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1029 * Introduce typed tasks. Specify its kind and cost at creation.
1030 At scheduling, just give where it should be placed, and the cost
1031 for each involved resource is automatically computed.
1032 Existing constructors so far (more to come of course):
1033 - SD_task_create_comm_e2e() for end-to-end communication
1034 - SD_task_create_comp_seq() for sequential computation
1035 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1036 * new function SD_task_dump() for debuging display
1037 * new function SD_task_dotty(task,FILE*) writing to file the info
1038 about the task in dotty format
1039 * SD_task_dependency_exists() can now cope with having one of its
1040 arguments NULL. If so, it tests whether the other argument has any
1042 * Add getters on list of preceding/following tasks:
1043 SD_task_get_parents(task) and SD_task_get_children(task)
1044 * Add getters on amount of workstations and list:
1045 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1046 * Add getter on task kind: SD_task_get_kind(task)
1047 * Update the start_time and finish_time of tasks on completion/failure
1048 * Bugfix: Remove task from state swags when destroyed
1051 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1054 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1055 dynar with one separate thread per value of the dynar.
1056 * Change the prototype of xbt_thread_create(), sorry.
1057 Added a boolean parameter indicating whether we want to join this
1058 thread (used in SG only for now)
1059 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1062 * GTNetS wrappers should now be usable again (and betterly tested too)
1063 * Fix a major regression from 3.2 where the timeout provided to
1064 MSG_task_put_with_timeout() was used as absolute time before which
1065 the comm should be done.
1066 * Start to fix the <cluster> tag.
1067 - Internal links should be good now (beside of the loopback, which
1068 use the private link instead)
1069 - paths to the external world is still rather broken
1070 - the <route:multi> tag is just broken. Actually that's brain-dead.
1071 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1072 to make it less stupid
1073 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1074 * Fix a source-level compatibility glitch from 3.2: after defining
1075 MSG_USE_DEPRECATED, you can use the old name
1076 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1077 * Allow to compile from the SVN with automake 1.11
1078 * Fix some problems when using the "start_time" tag in deployment XMLs.
1079 * Fix #8569: XBT/synchro.h has redundant declarations
1080 * Fix #8563: MSG return values and exceptions
1081 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1082 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1083 Thanks to Robson Peixoto.
1084 * Fix a vicious bug in dictionaries inducing that some elements were
1085 not freed on xbt_dict_free()
1087 Portability report of this version:
1088 * Main portability targets:
1089 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1090 on (amd64/i386/ia64)
1091 - mac leopard on i386
1092 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1093 but nothing critical.
1094 * Other platforms: windows, AIX and others were not tested for this release
1096 Timing report of this version:
1097 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1098 scenario parameters). From 8h to a few seconds in desktop grid settings.
1099 * Supernovae brings about 25% speedup on i386.
1101 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1103 SimGrid (3.3.3) stable; urgency=low
1105 The "Need for Speed" release.
1107 The timings done to validate the 3.3.2 were faulty.
1108 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1110 The problem was a conversion from a manually handled vector to
1111 xbt_dynar_t on the critical path.
1112 xbt_dynar_foreach calls functions, inducing stack management crap.
1114 We inlined these functions and xbt_dynar_foreach is now breath taking.
1115 We also inlined xbt_swag_belong on the way.
1117 Here are some approximate speedup measurements (on master/slaves
1118 simulations lasting between 10s and 20s each):
1119 3.3.1 -> 3.3.2: about same performance
1120 3.3.2 -> 3.3.3: 40% speedup
1121 3.3.1 -> 3.3.3: 40% speedup
1122 3.3.1 with inline patch -> 3.3.3: 30% speedup
1124 Our reading is that the refactoring which occurred in 3.3.2 made us
1125 suffer much more from the xbt_dynar_foreach low performance, but
1126 once we solved this, this refactoring proved to be very performance
1127 effective. From the 40% speedup, somehow, 10% are due to the
1128 inlining and 30% to the refactoring.
1130 That's a pitty that gcc cannot inline functions placed in other files
1131 alone. We have to choose between:
1132 - break the encapsulation (by putting private data structures and
1133 accessors in headers files to help gcc)
1134 - live with low performance
1135 - switch to a decent compiler such as icc (not quite possible).
1137 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1139 SimGrid (3.3.2) stable; urgency=low
1141 The "Simplicity does not preceed complexity, but follows it" release.
1143 The main contributors of this release were (lexical order):
1144 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1147 * Extract the routing logic into its own object.
1148 (was dupplicated in network.c and workstation_LV07.c;
1149 Allows to implement other ways of storing that info)
1150 => kill now useless network_card concept
1151 - Use dynar to represent routes (instead of void** + int*)
1152 - kill link_set (use surf_network_model->resource_set instead)
1153 - Add a command-line option to choose the routing schema to use
1154 - Add three new models:
1155 * Floyd (shortest path computed at initialization)
1156 * Dijikstra (shortest path recomputed all the time)
1157 * Cached Dijikstra (shortest path computed on need)
1158 All these models where contributed by Silas De Munck, and are
1159 described in his ICCS09 paper.
1161 * Simplify model declaration
1162 (less redirections, less function to write when defining a model)
1163 - Factorize stuff between models:
1166 surf_model_resource_set(model)
1167 surf_model_resource_by_name(model, name)
1168 - Unify the types of models in s_surf_model_t (using an union)
1169 - Embeed fields of common_public directly into s_surf_model_t
1170 - Rename model methods:
1171 action_free ~> action_unref
1172 action_change_state ~> action_state_set
1173 action_get_state ~> action_state_get
1174 - Change model methods into functions :
1175 (model)->common_public->action_use ~> surf_action_ref
1177 * Implement a generic resource; use it as ancestor to specific ones
1178 (allows to kill duplicated code in models)
1179 Drawback: timer command don't need no name nor properties;
1180 workstation_CLM03 don't need no properties
1181 (but I guess we can live with those few bytes wasted)
1183 * Improve the action object model
1184 - implement a constructor avoiding dupplicated code about field
1185 initialization in generic_action part.
1187 * Kill the SDP model: it has an external dependency, is deprecated
1188 in flavor of modern lmm models, and didn't compile since a while
1191 * Relocation of the context module from XBT to SIMIX.
1192 (the context were decoupled from the simix processes, duplicating a lot of code)
1193 => a lot of code was factorized
1194 - less overhead is introduced during scheduling
1195 - simpler API for the context factory
1196 - the logic for process creation,destruction and manipulation was simplified
1197 * Simplification of the s_smx_process_t data structure.
1198 => accesing the simix level data associated to a process is faster now,
1199 and the code is a lot more readable.
1202 * Implement some more MPI primitives:
1203 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1204 -implementation: Bcast: flat or 2-ary tree (default),
1205 Barrier: 4-ary tree,
1207 Allreduce: Reduce then Bcast
1208 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1209 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1210 Alltoallv: flat tree, like ompi
1212 * Add support for optimized collectives (Bcast is now binomial by default)
1213 * Port smpirun and smpicc to OS X
1216 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1217 Shout out if you used it.
1220 * Display the list of still queued messages in SG mode when existing
1224 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1225 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1226 * Add xbt_str_from_file(FILE*)
1227 * Add xbt_dict_get_key achieving a linear reverse search
1228 * Remove the context module
1230 Portability report of this version:
1231 * Main portability targets:
1232 - Linux(debian)/x86/context
1233 - Linux(debian)/x86/pthreads
1234 - Linux(debian)/amd64/context
1235 - Linux(debian)/amd64/pthreads
1236 On these, we still have the eratic breakages of gras/pmm and
1237 amok/saturate_sg reported in previous version. We still think
1238 that the tests are the cause of the fault, not the tested code.
1240 - Mac OSX Leopard/x86/context
1241 Still false negative in tesh autotesting.
1242 Smpi still fails, but this time because readlink does not accept -f
1243 Everything seems to work properly beside of that.
1246 - AIX version 5.3 (only tested contexts this time)
1247 Smpi still fails there because mktemp is not installed.
1248 Everything seems to work properly beside of that.
1250 I managed to compile it for the first time, but several breakages.
1251 Won't delay the release for this exotic platform.
1253 * Windows: it's still lagging behind. If you want to help, please
1256 Timing report of this version:
1257 This version seem to be more than 5% faster than 3.3.1 (on linux
1258 64bits with contextes). The gain is less than expected, we are
1259 investigating this for next release.
1261 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1263 SimGrid (3.3.1) stable; urgency=low
1266 * Implement a --cfg-help to show existing configuration variables
1267 * Build chain do not require doxygen in maintainer mode
1270 * fix a bug on struct sizeof computation, which prevented the
1271 exchange of arrays of structs in some conditions
1272 - added a regression test about this in datadesc_usage
1273 * Allow the exchange of 0-long dynamic vectors.
1274 - for that, use -1 as indicator of dynamic size instead of 0
1275 - This implied to change any size from unsigned long to long,
1276 reducing a bit communication abilities, but I guess that with
1277 64bits being quite common, this is more than enough.
1278 - This also induce a protocol change, thus bumping network protocol
1279 version from 0 to 1 (if we have external users, we have to get
1280 clean on that point too ;)
1281 - added two regression tests about this in datadesc_usage
1282 * Be more verbose when propagating local exceptions
1283 This helps debugging.
1284 * Display the status of simulated processes when receiving SIGINT in
1288 * Allow to control the simulation from a trace file.
1289 New functions MSG_action_register() and MSG_action_trace_run()
1290 The first one allows to associate a function execution to each
1291 kind of action while the second one parses a trace file and
1292 triggers the corresponding actions within the system.
1293 For now, only a toy example is provided in examples/msg/actions
1294 * Add an exemple of process migration in examples/msg/migration
1295 * Fix a bug in task exchange which broke MSG_task_get_sender()
1296 Add a teshsuite regression test for that.
1297 [Bug: if MSG_task_get_sender() is called after sender exit,
1299 * Fix a bug which prevented suspend/resume to work properly
1300 * Display the status of simulated processes when receiving SIGINT
1301 This fixes a regression of v3.3. due to the introduction of SIMIX
1302 * Bug fixing in failure management:
1303 - trace could not start by a failure at time 0
1304 - failure during communications were not working
1307 * Add SIMIX_process_set_name() to change the name of the current
1308 process in the log messages.
1309 * Store smx_hosts in a dict since we only retrieve them by name
1310 * Move the configuration infrastructure to surf
1313 * Move the configuration infrastructure to surf
1316 * Massive internal cleanups:
1317 - Store internal structures on processes instead of hosts (allows
1318 to have more than one process per host, in addition of being more
1320 - Cleanup the initialization/finalization process
1321 - Kill a whole bunch of unneeded synchronization:
1322 processes run in exclusive manner within the simulator
1323 - Move queues from global tables to process data fields
1325 - now accept -platform and -hostfile arguments
1326 - Pass the right rank value to processes according to the hostfile
1327 * Compile the examples by default, and use them as regression tests
1328 * Implement MPI_Wtime()
1329 * Change the reference speed to a command line option
1332 * TCP_gamma can now be specified as command line option using
1333 --cfg=TCP_gamma:10000000.0
1334 * Change the --surf-path cmd line option into --cfg=path:
1337 * Also include strbuff from xbt.h public header
1338 * xbt_ex_display(): do not free the exception after displaying
1339 This allows to do more with the given exception afterward.
1340 Users should call xbt_ex_free() themselves.
1344 Portability report of this version:
1345 * Main portability targets:
1346 - Linux(debian)/x86/context
1347 - Linux(debian)/x86/pthreads
1348 - Linux(debian)/amd64/context
1349 - Linux(debian)/amd64/pthreads
1350 These targets fail about 1/10 of times on gras/pmm, but we believe
1351 that this is because of the test, not because of simgrid.
1352 amok/saturate_sg fails even more rarely, and the test may not be
1355 - Mac OSX Leopard/x86/context
1356 The test suite still spits tons of errors because some obscure
1357 force prevents us from removing the temporary directories
1358 arguing that they still contain some metadata I've never heard of.
1359 Smpi fails because seq is not installed.
1360 Everything seems to work properly beside of that.
1363 - AIX version 5.3 (both contexts and pthread)
1364 Smpi still fails there because mktemp is not installed.
1365 XML inclusions seems rosty on AIX.
1367 * Windows: it's still lagging behind. If you want to help, please
1370 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1372 SimGrid (3.3) stable; urgency=high
1376 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1377 [Malek Cherier & Mt]
1379 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1382 Until now, GRAS were using MSG as an interface to SURF. It was
1383 quite difficult because both interface have several differences
1384 (MSG channels vs GRAS sockets were the most notable point).
1386 This also opens the gate to SMPI (which should occur soon) and speed
1387 up simulations by to 40% (even if it were not the main goal).
1389 **************************************
1390 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1391 **************************************
1392 The point is that events occuring at the exact same timestamp are
1393 not scheduled in the same order with the old and new version. This
1394 may be enough to completely change the execution of simulations in
1395 some cases. Sorry for the inconvenience.
1397 * Cleanup and upgrade the XML format to push further scalability
1398 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1400 * Improve the testing infrastructure with tesh. Now a very large part of
1401 the code is tested not only by being run but also by checking that the
1402 output match an expected output [Mt].
1404 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1405 is really less memory-demanding, which should allow you to use
1406 larger files in SimGrid [AL].
1408 * Inform valgrind about our contextes, so that it becomes usable
1409 with the default (and more effecient) version of SimGrid
1410 [contributed by Sékou Diakite, many thanks]
1413 * Introduce a listener thread in charge of receiving incoming
1414 messages from the network. It allows to overlap communication and
1415 computation but most notably, it removes some stupid deadlocks due
1416 to the fact that so far, a process could not send and receive at
1417 the same time. This made most non trivial communication schema
1419 * Convert the PIDs from long int to int to match the MSG ones (and
1420 linux ones too) [Mt]
1421 * New function: gras_agent_spawn() to launch a new process on
1422 current host. Only working in simulation for now. [Mt]
1423 * New function: gras_os_hostport() returning a constant form (ie,
1424 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1427 * Make the backtrace of exceptions more human readable [Mt]
1428 * New module: xbt/str [Mt]
1429 a ton of string utility functions (split, join, printf to a newly
1430 allocated buffer, trim, etc)
1431 * New module: xbt/hash [Mt]
1432 SHA1 hashing algorithm (more to come if needed)
1433 * New module: xbt/synchro [Mt]
1434 synchronization tools (mutex and conditions) working the same way
1435 in simulation and in real life (mainly useful for GRAS, but not
1437 * New module: xbt/queue [Mt]
1438 classical producer/consumer synchronization scheme
1439 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1440 (using the classical functions will get serialized) [Mt]
1441 * Make dictionary internal table dynamic. No need to specify its size
1442 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1444 * Make sure the log channels are organized as a tree under windows
1445 (because of ANSI C compatibility issue, any channel were child of
1446 root directly) [Mt].
1449 * Cleaned many thing in surf and fixed a few bugs [AL].
1450 * Add a nice command line configuration mechanism to compose models [AL].
1451 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1452 the previous one (KCCFLN05). It relies on something that looks like
1453 a max-min sharing mechanism but cannot be written as such. A new solver
1454 was thus designed [AL].
1455 * Add a new solver to lmm. Based on Lagrange optimization and
1456 gradient-based descent, it enables to efficiently maximise systems s.a
1458 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1460 This solver enables to propose two new network models for TCP Reno and
1461 TCP Vegas based on Low's work. These models still need to be fully
1462 tested though [Pedro Velho].
1465 * Bug fix in SD_simulate. Now the time bound given as argument is
1467 * Use the new parallel task model (ptask_L07) as default.
1468 * Use the SURF command line configuration mechanism.
1469 * 0-size tasks (for synchronization) should now work.
1471 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1473 SimGrid (3.2) stable; urgency=high
1477 We still experience issues on this platform, but we believe that at
1478 least MSG is usable.
1480 GRAS API BREAKAGE (for simplification purpose, sorry):
1481 * the gras_msgtype_by_name is not used anymore. Instead of
1482 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1483 you can write (and must)
1484 gras_msg_send(toserver, "request", &request);
1485 - If you still want to pass a gras_msgtype_t to the function (to cache
1486 the type and avoid the lookup time), use the gras_msg_send_() variant.
1487 - Impacted functions:
1488 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1489 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1490 * The callbacks are now expected to return 0 when everything went well
1491 (just like the main() function)
1493 GRAS new features and improvements:
1494 * New module mecanism where user code can use per process globals [Mt]
1495 This is similar to gras_userdata_*() functions, but for libraries. It
1496 factorize some code developped over and over in the examples and AMOK.
1497 It has still to be documented and used (only amok/peermanagement is
1499 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1500 loss when they were small enough to fit into the buffer and sent quickly
1501 enough so that they can all get received in one shoot.
1502 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1503 if not found. Use the *_or_null() variant for the old semantic.
1504 * In gras_msg_handle, do not discard messages without callback.
1505 They are probably messages to be explicitly awaited later (ie, proofs of
1506 mis-synchronization in userland since they are sent before being awaited)
1507 No big deal usually.
1508 * gras_socket_meas_send/recv: semantic changed!
1509 The numerical arguments used to be (1) the total amount of data to send
1510 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1511 messages. This was need for the fool willing to send more than MAXINT
1512 bytes on quite fat pipes.
1515 * Do really rename the hostmanagement module to peermanagement. [Mt]
1516 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1517 but this is rather new and this was documented in the module
1518 documentation (poor excuses, I admit)
1519 * Bandwidth measurement semantic changed! This follows the changes to
1520 gras_socket_meas_send/recv explained above.
1523 * A sequential mode has been added to the workstations. When a workstation
1524 is in sequential mode, it can execute only one task, and the other tasks
1525 are waiting in a FIFO. [Christophe Thiery]
1528 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1529 model for SIMDAG. [Christophe Thiery]
1530 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1531 equal to 0, which caused some bad behaviors in
1532 saturated_constraint_set_update. I now use a threshold mechanism like in
1536 * When running manually src/testall, you select specific units [Mt]
1537 testall is the result of our cunit mecanism, and should replace all
1538 the scripty thingy around since bash don't run easily on billware.
1540 * A mallocator system has been added. [Christophe Thiery]
1541 Mallocators allow you to recycle your unused objects instead of freeing them
1542 and allocating new ones.
1544 Documentation update:
1545 * FAQ reworking + New FAQs:
1546 - "Valgrind spits tons of errors!" [Mt]
1547 - "How to repport bugs" [Mt]
1548 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1549 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1550 - Communication time measurement within MSG [AL]
1551 - I experience weird communication times when I change the latency [AL]
1552 * GRAS tutorial [Mt]
1554 - an introduction to the framework and to the used communication model
1555 - an initiatic tour introducing the most proheminent features:
1557 . Lesson 0: Installing GRAS
1558 . Lesson 1: Setting up your own project
1559 o Part 2: Message passing
1560 . Lesson 2: Exchanging simple messages
1561 . Lesson 3: Passing arguments to the processes (in SG)
1562 . Lesson 4: Attaching callbacks to messages
1563 . Lesson 5: Using globals in processes
1564 . Lesson 6: Logging informations properly
1565 . Lesson 7: Using internal timers
1566 . Lesson 8: Handling errors through exceptions
1567 . Lesson 9: Exchanging simple data
1568 . Lesson 10: Remote Procedure Calling (RPC)
1569 . Lesson 11: Explicitely waiting for messages
1570 . Recapping of message passing features in GRAS
1571 - A HOWTO section containing:
1572 o HOWTO design a GRAS application
1573 More are due, of course. They will come latter. In the meanwhile, you can
1574 check the examples which are still here.
1576 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1578 SimGrid (3.1) stable; urgency=high
1582 There was a stack corruption somewhere, visible only when optimizing
1583 with these versions. [Vince]
1586 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1587 module that enabled to play with Directed Acyclic Graphs. It is built
1588 directly on top of SURF and provides an API rather close to the old
1589 SG. Some old codes using SG are currently under rewrite to check that
1590 all needful functions are provided. [Christophe Thiery]
1593 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1594 extension of the classical CLM03 model that gracefully handles
1595 failures. This is now the default model for MSG and GRAS. It doesn't
1596 handle parallel tasks yet though. [AL]
1597 * Bug fix: Weights were not correctly set in the network part.
1598 WARNING: This may have resulted in incorrect results with simulations
1599 where there are more than one flow on a given link. [AL]
1602 * After a (long ?) discussion on simgrid-devel, we have decided that the
1603 convention we had on units was stupid. That is why it has been decided
1604 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1605 WARNING : This means that all previous platform files will not work as
1606 such with this version! A warning is issued to ask users to update
1608 A conversion script can be found in the contrib module of the CVS, under
1609 the name contrib/platform_generation/surfxml_update.pl [MQ]
1612 * Bug fix: Processes were started in reverse order, wrt deployment file.
1613 WARNING: if your code relies on this bug, please fix it. [AL]
1614 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1615 being executed on two different locations. [AL]
1616 * Bug fix: Failures are now better supported thanks to Derrick's tests
1617 (there was many failure situations I hadn't thought of and that weren't
1618 correctly handled). [AL]
1619 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1623 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1624 Exception occurring on server-side are propagated back to client (!).
1626 API CHANGE: the callback changed their prototype. Change:
1627 int my_handler(gras_socket_t expeditor, void *payload_data) {
1629 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1630 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1632 * New! function: gras_msg_handleall to deal with all messages arriving
1633 within a given period.
1634 * New! function: gras_socket_server_range to get a server socket in a
1635 range of port numbers (ease to avoid port number conflicts) [MQ]
1636 * New! gras processes display their backtrace when they get a SIGUSR1
1637 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1638 Sweet to debug RL processes [MQ]
1642 - Do not force experiment sizes to be expressed in kb, or it becomes
1643 impossible to measure the latency this way (needs one byte-long tests)
1644 WARNING: this changes the amok_bw_* function semantic. [MQ]
1645 - Implements the link saturation stuff. [MQ]
1646 * Peer management module:
1647 New! module factorizing code that we wrote over and over [MQ].
1650 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1651 - Test units are placed directly into the library code, they get extracted
1652 automatically and placed into the src/testall binary.
1653 - Convert most of the XBT tests to this system.
1654 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1655 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1656 live in the same C code. This required to change a little bit the API
1657 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1658 * New module: sparse graph structure with basic algorithms (this is work
1659 in progress and the API is not considered to be frozen yet). [AL]
1660 * Display more information on backtraces: source line & function names are
1661 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1662 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1663 * Reworked a little bit some #include statements to load only required
1664 headers. Some user code that relied on SimGrid to include stdlib or
1665 stdio may need to include it by themselves. [AL]
1666 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1668 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1669 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1670 but I guess that almost nobody used those parts. [MQ]
1672 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1674 SimGrid (3.0.1) stable; urgency=low
1677 * Unfortunately, I had missed 5 misnamed functions:
1678 xbt_fifo_item_t xbt_fifo_newitem(void);
1679 void xbt_fifo_freeitem(xbt_fifo_item_t);
1680 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1681 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1682 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1683 They're now deprecated. Please use their new versions:
1684 xbt_fifo_item_t xbt_fifo_new_item(void);
1685 void xbt_fifo_free_item(xbt_fifo_item_t);
1686 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1687 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1688 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1690 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1691 * Documentation: xbt_log module unmercifully reworked [MQ]
1692 * Bugfix: there was a problem with the ending of contexts with
1693 the pthread backend. It caused some weird deadlock or behavior
1694 depending on the pthread implementation. [AL]
1695 * Bugfix: get the exceptions raised in the simulator repport where
1696 and why they come from when they are not catched in time [AL, MQ]
1699 * Bugfix: Do repport the error when two non-connected hosts try to
1700 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1703 * Add additionnal checkings on communications. Assert that two
1704 communicating hosts are connected by a set of links... [AL]
1707 * Add additionnal checkings on channel values in communication [AL]
1708 * New: MSG_task_get_source to see on which host a task was generated [HC]
1709 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1710 the number of tasks waiting to be received on channel and sent
1712 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1713 waits for the first task coming from a given host.. [AL]
1715 GRAS new functionnalities: [MQ]
1716 * Enhance the parsing macro to allow the size of multidimentional objects
1717 to be given thru annotations.
1718 * New example (and documentation): Matrix Multiplication a la RPC
1719 (as when I was young!) and fix a bunch of bugs found on the way.
1721 GRAS performance improvements: [MQ]
1723 * Reduce the amount of cbps creation/destruction by making it static to
1724 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1727 * Change libdata to a set so that we can search for stuff by ID (and thus
1728 reduce the insane amount of dict lookups)
1731 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1732 RL communication on top of SG since it's so uneffective.
1733 It may also be used for inter-thread communication in RL, one day.
1734 * Use gras_datadesc_copy() to exchange messages on top of SG
1736 - improve message exchange performance on top of SG
1737 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1738 * Don't exchange on the network the size of the used part of buffer,
1739 instead, specify the possible buffer size to read().
1741 - reduces the amount of read/write calls (one pair per exchange)
1742 - reduces the amount of exchanged data (the size)
1743 - allows to retrieve all arrived data on receiver side, if we don't need
1744 it right now (subsequent read will peek the buffer)
1745 - allows the receiver to proceed with the begining of the stream before
1746 everything is arrived
1747 - make it possible to build an iov transport (using readv/writev)
1749 - take care of the data with non-stable storage (like stacked data),
1751 * If possible, TCP send uses vector I/O (when writev() is here)
1752 - Don't use it for receive since we send data sizes and data on the
1753 same stream, so we wouldn't be able to chain large amount of chunks
1754 before having to flush the stuff to read the size.
1755 * Rework the transport plugin mecanism to simplify it and reduce the
1756 amount of pointer dereferencement when searching for the right function
1759 * I guess that now, we do almost as few system calls as possible while
1760 doing as few data copy as possible.
1762 To improve it further, we could try to send all the sizes first and then
1763 all the data (to use iov on receiving size), but it's only a partial
1764 solution: when you have 2 dimensional data, the sizes of the second
1765 dimension is data of the first dimension, so you need 3 streams.
1767 I'm not sure the potential performance gains justify the coding burden.
1769 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1771 SimGrid (3.00) stable; urgency=high
1774 * New! Give the possibility to hijack the surf parser and thus bypass
1775 MSG_create_environment and MSG_launch_application. Have a look at
1776 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1778 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1780 SimGrid (2.96) unstable; urgency=low
1785 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1786 This deprecates the xbt_error_t mecanisms.
1787 It modifies (simplifies) all XBT and GRAS API.
1788 MSG API keeps unchanged (exceptions raised by XBT are catched from
1789 within MSG and masked with existing error handling facilities)
1792 * New! Add a FATPIPE model. [AL]
1793 * New! Add a parallel task model. [AL]
1794 * New! Add automatically a loopback interface (in the default
1795 network model) if none was precised.
1798 * Bugfix: MSG_process_resume now works with the current running process.
1800 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1801 * Modification of MSG_task_get_compute_duration. Once a task has been
1802 processed, the value returned by this function is now equal to 0. [AL]
1803 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1804 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1805 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1806 of a m_task. For now, MSG_task_cancel only works with computation
1808 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1809 of the processor (in Mflop/s) regardless of the current load on the
1811 * API Change: use proper naming convention for MSG_getClock and
1812 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1814 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1815 This function changes the priority of a computation task. This priority
1816 doesn't affect the transfer rate. A priority of 2 will make a task
1817 receive two times more cpu power than the other ones. This function
1818 has been added to suit the needs of Nguyen The Loc and hasn't been that
1819 much tested yet. So if it fails, please report it and send me your code.
1821 * API Change: removed all functions and types that were marked "deprecated"
1822 since many months. Renamed MSG_global_init_args to MSG_global_init.
1824 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1826 SimGrid (2.95) unstable; urgency=low
1829 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1830 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1831 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1834 * Bugfix: Allow absolute paths for platform description files [MQ]
1835 * Bugfix: do free the variables after use. Leads to drastic performance
1837 * Implement max_duration (ie, timeouts) on resources [AL]
1840 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1842 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1843 the needed support (timeouts) [AL]
1846 * Implement measurement sockets. You can now get the bandwidth between two
1847 hosts thanks to AMOK (see below). [MQ]
1848 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1849 dynar over the network (yeah) [MQ]
1850 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1851 * simple/static token-ring example. [Alexandre Colucci and MQ]
1852 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1853 in place before [MQ]
1856 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1857 type parsing. "short" and "long" modifiers were also ignored; other
1858 modifier (such as reference level) are still ignored. [MQ]
1859 * Embeed the buffer size within the buffer itself on SG. [MQ]
1860 That way, send() are atomic and cannot get intermixed anymore (at least
1861 the ones which are less than 100k; bigger messages still have the issue)
1862 * Array size pushed by the field, not by the field type (or each
1863 and every long int will push stuff to the cbps) [MQ]
1864 * use select() to sleep since it allows to portably sleep less than one
1867 GRAS (minor cleanups)
1868 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1871 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1872 * type category 'ignored' killed as it was never used and were difficult
1874 * whether a type can cycle or not is now a flag, leaving room for more
1875 flags (as "ignored", if we feel the need one day ;)
1876 * Rename raw sockets to measurement sockets since "raw" has another
1877 meaning in networking community.
1880 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1881 and offers features not belonging to GRAS but that most applications
1882 need. One day, it may be a set of plugins loadable at runtime.
1883 * New module: bandwidth
1884 bandwidth measurement between arbitrary nodes running this module. [MQ]
1886 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1888 SimGrid (2.94) unstable; urgency=low
1890 The first beta release of SimGrid 3 !
1894 * Update the main page and the FAQ. Adding references to gforge.
1897 * Add a gras_os_getpid function.
1900 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1901 * Extend the logs so that they also print PID, hostname, date, ... if
1903 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1904 and kill the old version which were in testsuite/
1905 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1909 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1910 for users. Instead, each programming environment (GRAS, MSG,...) use it
1911 when needed to provide such a feature to users.
1912 Don't shortcut the mecanism or you will also shortcut the virtualization
1913 you need on the simulator.
1917 * Cleanups in configury with regard to compile optimization/warning flags.
1918 Also add -fno-loop-optimize to any powerpc since it's the optimization
1919 killing gcc (< 3.4.0).
1920 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1921 needed by MSG examples complications
1922 * Borrow configury beautifications from PHP
1925 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1926 warning (thanks loris for stumbling into this one).
1927 * Bugfix: stop loading private headers (gras_config.h) from the public
1931 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1932 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1933 RL would imply a DNS resolver, which is *hard* to do in a portable way
1934 (and therefore delayed).
1935 * Implement a real timer mecanism and use it in timing macros. This allows
1936 to avoid rounding errors and get a 0.000005 sec precision in timing
1937 macros. While I was at it, various cleanups:
1938 - allow to declare more than one timed section per file (fix a stupid bug)
1939 - move some private declaration to the right place
1940 - merge conditional execution and timing macros into emulation module
1941 - document the module
1942 - make sure the module cleanups its mess on gras_exit
1943 * Documentation improvements:
1944 - (new) how to compile applications using GRAS
1945 - (new) emulation support (timing macros)
1947 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1949 SimGrid (2.93) unstable; urgency=low
1951 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1954 - Use Paje properly where used. Still to be sanitized properly.
1955 - Portability fix: Add an implementation of the contexts using pthread
1959 - Add xbt_procname(): returns the name of the current process.
1960 Use it to show the current process's name in all logging.
1962 - fix detection of older flex version and the reaction, since we do
1963 depend on modern ones (we use lex_destroy)
1964 - Better separation of SG and RL in the libs: remove all simulation code
1965 from libgras. As a result, this lib is now only 200k when stripped.
1966 Some of the xbt modules may also be duplicated (two sets and such) and
1967 should be cleaned/killed before SG3.
1968 - Insist on using xlC on AIX because of weird problems involving gcc there.
1969 - Cleanup the make remote stuff. This is now done by scripts
1970 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1971 mainly for our private use, but we're working on changing them to user
1974 - Bugfix: flush the socket on close only if there is some *output*.
1975 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1977 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1980 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1982 SimGrid (2.92) unstable; urgency=low
1984 Alpha 3 on the path to SimGrid 3
1988 - New! First try of benchmarking macros.
1989 - New! First try so that gras_stub_generator generate deployment and
1990 remote compilation helpers.
1992 - Bugfix: Initialization fix in msg_test.
1996 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1998 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2000 - API Change: gras_os_sleep to take the amount of seconds as a double.
2001 Accepting an int was error prone since it was the only location where
2002 seconds were coded as such. It leaded to damn rounding errors.
2003 - Bugfix: Hard to belive that timers ever worked before this.
2005 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2007 SimGrid (2.91) unstable; urgency=low
2009 Alpha 2 on the path to SimGrid 3
2013 - Bug fix in the lmm_solver.
2015 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2016 through the function MSG_paje_output.
2017 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2018 - It is possible to bound the rate of a communication in MSG with
2019 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2020 to write it in the changelog).
2021 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2026 - Various cleanups to the autotools stuff
2027 - Begin to move Gras examples to examples/gras/
2028 - Let make distcheck work again (yeah!)
2030 - documentation overhauled using doxygen.
2031 gtk-doc-tools is dead in SimGrid now.
2032 - Automatically extract all existing logging categories, and add the list
2033 to the documentation (long standing one, to say the less)
2035 - Cleanup the known architecture table. Reorder the entries to group what
2036 should be, and use a more consistent naming scheme.
2037 (some of the test dataset are still to be regenerated)
2038 - New! Allow library to register globals on each process just as userdata
2040 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2041 lookup time (for now).
2042 Use it in msg and trp.
2043 This cleans a lot the internals and helps enforcing privacy of the
2044 headers between the gras components.
2045 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2046 - Bugfix: gras_os_time was delirious in RL.
2047 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2049 - Reenable GRAS now that it works.
2051 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2053 SimGrid (2.90) unstable; urgency=low
2055 Alpha 1 on the path to SimGrid 3
2057 * It is a long time since the last release of SimGrid. I'm sorry about
2058 that but as I had told you, I was rewriting a lot of things. I apologize
2059 to those who had been reporting bugs to me and that I had not answered.
2060 If your bug is still in the new version, please tell me. Here is a
2061 summary of the main changes.
2063 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2064 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2065 portability, speed, and a lot more but you'll figure it out later.
2066 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2067 I wanted to release the new MSG as soon as possible and I have
2068 broken GRAS, which is the reason why, for now, only MSG is fully
2069 functional. A laconic description of these projects is available
2070 in the documentation.
2072 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2073 kernel optimized for our needs (called SURF but only the developers
2074 should use it). Hence, MSG is now roughly 30 times faster and I think
2075 that by rewriting a little bit MSG, I could event speed it up a little
2076 bit more. Beside the gain in speed, it is also much easier to encode a
2077 new platform model with SURF than it was with SG. More to come...
2079 * REVOLUTION 3: I have tried to change a little as possible the API of
2080 MSG but a few things really had to disappear. The main differences
2081 with the previous version are :
2082 1) no more m_links_t and the corresponding functions. Platforms are
2083 directly read from a XML description and cannot be hard-coded
2084 anymore. The same format is used for application deployment
2085 description. The new format is described in the documentation.
2086 Have a look in tools/platform_generation. There is a tiny script
2087 that converts from the old platform format to the new one. Concerning
2088 the application deployment format, parsing the old one is tricky.
2089 I think most of you should however be able to convert your files. If
2090 it is really an issue, I can write a C code that does the conversion.
2092 2) the toolbox tbx does not exist anymore. We now have a library
2093 with much more data-structures but without the hash-tables (we have
2094 dictionaries that are much faster).
2096 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2098 *****************************************************************************
2099 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2100 * don't feel like converting the oldies. (Mt) *
2101 *****************************************************************************
2104 Version 2.90: "the long awaited one"
2105 - Finished rewriting and debugging MSG. Rewrote the documentation.
2106 - disable GRAS for now since it needs to be ported to the newest SG
2109 - Finish the port to windows (using mingw32 for cross-compile)
2112 - Main loop and datastructures of SURF. A cpu resource object is
2113 functional. Surf can thus be used to create cpu's with variable
2114 performance on which you can execute some actions.
2116 2004-11-15 Martin Quinson
2117 - Port to ARM. Simply added the alignment and size descriptions. Should
2118 work, but the ARM machines are so slow that I didn't had the opportunity
2119 to 'make check' over there yet.
2121 2004-11-15 Arnaud Legrand
2122 - Trace manager now written. It uses a heap structure and is therefore
2123 expected to be efficient. It may however be speeded up (particularly
2124 when many events occur at the same date) by using red and black
2125 trees. One day maybe...
2126 - Max-min linear system solver written. It uses a sparse matrix
2127 structure taking advantage of its expected use. Most operations are
2128 O(1) and free/calloc are called as few as possible. The computation of
2129 the minimum could however be improved by using a red and black tree
2132 2004-11-03 Arnaud Legrand
2133 - Rename every gras_* function that was in xbt/ to its xbt_
2135 - Add a heap and a doubly-linked list to xbt
2136 - Added a dichotomy to the dictionaries. make check works as well before
2137 so I assume that the patch is correct. I do not know however if things
2138 run effectively faster than before now. :)
2140 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2141 SimGrid, and the version number is bumped to 2.x
2143 2004-10-29 Martin Quinson
2144 - Introduction of the remote errors.
2145 They are the result of a RMI/RPC on the remote machine.
2146 ErrCodes being scalar values, you can't get the host on which those
2147 errors did happen. Extending the error mechanism as in Gnome is possible.
2148 No idea yet whether it is a good idea.
2150 2004-10-28 Martin Quinson
2151 - Interface revolution: the Starred Structure Eradication.
2152 I used to do typedef struct {} toto_t; and then handle *toto_t.
2153 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2154 - ???_t is a valid type (builded with typedef)
2155 - s_toto_t is a structure (access to fields with .)
2156 - s_toto is a structure needing 'struct' keyword to be used
2157 - e_toto_t is an enum
2158 - toto_t is an 'object' (struct*)
2160 typedef struct s_toto {} s_toto_t, *toto_t;
2161 typedef enum {} e_toto_t;
2162 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2163 s_toto_t) is private.
2165 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2166 so that it can set the variable to NULL. It was so for dicts and sets,
2167 it changed for dynars.
2169 - Fix a bunch of memleaks in dict_remove
2170 - Fix a bug in sg/server_socket opening: it failed all the time.
2172 2004-10-07 Martin Quinson
2173 - Speed up dynar lookup operation a bit.
2175 gras_dynar_get is dead.
2177 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2178 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2179 (which gives you the address of the stored data).
2181 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2182 copy of the data using an affectation to do the job and not a memcpy.
2184 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2185 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2187 It does not really speedup the dynar test because they are
2188 setting elements all the time (and look them seldom). But the dict does
2189 far more lookup than setting.
2191 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2193 2004-10-05 Martin Quinson
2194 - Allow to (en/dis)able the cycle detection at run time.
2196 Whether we should check for cycle or not is now a property of each
2197 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2198 datadesc_cycle_unset allow to remove this property when previously set.
2200 Note that the cycle detection is off by default since it impacts the
2201 performance. Watch the data you feed GRAS with ;)
2203 This property is hereditary. Any element embedded in a structure having it
2204 set have it set for the time of this data exchange.
2206 You should set it both on sender and receiver side. If you don't set it on
2207 sender side, it will enter an endless loop. If you forget on receiver
2208 side, the cycles won't be recreated after communication.
2210 - Header reorganization.
2211 Kill gras_private.h, each submodule must load the headers it needs.
2213 2004-10-04 Martin Quinson
2214 - Interface revolution: do not try to survive to malloc failure.
2216 Now, gras_malloc and friends call gras_abort() on failure.
2217 As a conclusion, malloc_error is not a valid error anymore, and all
2218 functions for which it was the only gras_error_t return value are
2219 changed. They now return void, or there result directly.
2220 This simplify the API a lot.
2222 2004-09-29 Martin Quinson
2223 - Re-enable raw sockets.
2224 Created by gras_socket_{client,server}_ext;
2225 Used with gras_raw_{send,recv}
2228 It should allow to kill the last bits of gras first version soon.
2230 This is not completely satisfactory yet (duplicate code with
2231 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2234 - Simplify transport plugin (internal) interface by not passing any
2235 argument to _server and _client, but embedding them in the socket
2238 2004-09-28 Martin Quinson
2239 - Finish the port to AIX.
2240 autoconf was my problem (segfault within the malloc replacement
2241 function. No idea why)
2243 2004-09-16 Martin Quinson
2244 - Fix some size_t madness on 64bit architectures.
2246 2004-09-08 Martin Quinson
2247 - Reduce the number of system headers loaded, overload some more system
2248 calls (such as malloc to cast the result of the system one, and work
2250 - Fix and reintroduce the config support
2252 2004-09-07 Martin Quinson
2253 - Source code reorganization to allow Arnaud to surf all over there.
2254 - Allow to document the logging categories.
2255 - Remove all uppercase from logging categories and useless cleanup in names.
2257 2004-08-18 Martin Quinson
2258 Version 0.6.2 (protocol not changed; API changed)
2259 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2260 gras_error_t), and NULL when not found. Functions expecting a msgtype
2261 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2262 hopefully usefull message.
2263 - Portability to prehistoric sparcs again
2265 2004-08-17 Martin Quinson
2266 Version 0.6.1 (protocol not changed; ABI not changed)
2267 - prealloc some buffers to speed things up
2269 2004-08-11 Martin Quinson
2270 Version 0.6 (protocol not changed; ABI expended)
2271 - The parsing macro can deal with the references, provided that you add
2272 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2274 2004-08-09 Martin Quinson
2275 Version 0.5 (protocol not changed; ABI changed)
2276 - Allow to off turn the cycle detection code in data exchange at
2277 compilation time. It should be at run time, but I'm short of time (and
2278 the config stuff is still broken). That way, we keep dict out of the
2279 critical path, which is good because the performance is poor:
2280 - search not dichotomial yet
2281 - dynar give no way to access their content and memcpy everytime
2282 - In composed data description (struct, ref and so on), stop foolness of
2283 keeping the subtype's ID, but store the type itself. This keeps sets out
2284 of the critical path, which is good since they rely on dynar and
2285 dictionnaries. The only loose of that is that we cannot detect the
2286 redeclaration of a structure/union with another content (but I'm not sure
2287 the code detected well this error before anyway). We still can detect
2288 the redefinition discrepancy for the other types.
2289 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2290 it breaks the code because of type-punning used all over the place).
2291 This breaks on all non-gcc architectures (for now).
2293 All those changes (plus the buffer of last time) allow me to gain 2 order
2294 of magnitude on cruel tests consisting of 800000 array of integers on two
2295 level of a hierarchical structure (200 secondes -> 4 secondes)
2298 - the selector of reference must now return the type it points to, not
2299 the ID of this type.
2301 2004-08-06 Martin Quinson
2302 Version 0.4 (protocol changed; ABI not changed)
2303 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2304 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2305 - Implement a buffer transport, and use it by default (it relies on tcp in
2306 real life and on sg in simulation).
2307 That's a bit hackish since I had a new field to the structure to store
2308 its data without interfering with the subtype ones. Inheritance
2309 is tricky in C. And that's a kind of reverse inheritance with one class
2310 derivating two classes. Or maybe a game with java interfaces. Anyway,
2311 that's damn hard in C (at least).
2312 Moreover, I got tired while trying to ensure plugin separation and
2313 genericity in SG mode. MSG wants me to do weird things, so let's go for
2314 cruel hacks (temporarily of course ;).
2315 See comment in transport_private.h:71
2316 - do not include all the _interface headers in private but in the files
2317 which really need them (to cut the compilation time when they are
2320 2004-07-26 Martin Quinson
2321 Version 0.3 (protocol not changed; ABI changed)
2322 - Major overhault of the datadesc interface to simplify it:
2323 - shorted the function names:
2324 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2325 - add a trivial way to push/pop integers into the cbps without malloc.
2326 This allows to make really generic sub_type description, which simply
2327 pop their size of the stack.
2328 - add a function gras_datadesc_ref_pop_arr() which does what users want
2329 most of the time: Declare a dynamic array (which pops its size of the
2330 stack) and declare a reference to it. Poor name, but anyway.
2331 - kill the post-send callback, add a post-receive one
2333 2004-07-23 Martin Quinson
2334 Version 0.2 (protocol changed; ABI changed)
2335 - add some testing for cpbs in the test cases, and fix some more bugs.
2336 This invalidate again the little64 data file, since I cannot regenerate
2338 - remove an awfull optimization in the logging stuff, allowing me to:
2339 - understand it again
2340 - learn gcc how to check that the argument match the provided format
2341 - fix all errors revealed by gcc after that
2342 - internal keys of dict are not \0 terminated. Deal with it properly in
2343 loggings instead of segfaulting when the user want to see the logs :-/
2345 2004-07-22 Martin Quinson
2346 - Fix some stupid bug preventing cbps (callback postit) from working
2348 2004-07-21 Martin Quinson
2349 - Some documentation cleanups
2350 - remove the useless last argument of msgtype_declare
2351 - rename the Virtu functions to fit into the 'os' namespace
2352 - move headers src/include -> src/include/gras/ and stop fooling with
2353 gras -> . symbolic link
2354 - make distcheck is now successful
2356 2004-07-19 Martin Quinson
2358 - Build shared library also
2359 - Install html doc to the right location
2360 - stop removing maintainer files in make clean
2361 - build tests only on make check
2363 2004-07-13 Martin Quinson
2365 - No major issue in previous version => change versionning schema
2366 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2367 corresponding dataset.
2369 2004-07-11 Martin Quinson
2371 - Get it working with any kind of structure (we can compute the padding
2372 bytes remotely for all the architectures I have access to)
2373 - Implement the structure parsing macro (still not quite robust/complete)
2374 - Improvement to the remote testing toysuite
2376 2004-07-10 Martin Quinson
2377 [autoconf mechanism]
2378 - get ride of a bunch of deprecated macros
2379 - actually run the test for two-compliment, not only compile it :-/
2380 - test whether the structures get packed (and bail out if yes. Damn.
2381 Alignment is a serious matter)
2382 - test whether the structures get compacted (but respecting the alignment
2383 constraints of each types)
2384 - test whether the array fields of structures can straddle alignment boundaries
2386 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2387 (took me 2 hours to find that bug, looking at the wrong place)
2388 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2389 before sending/receiving and are used to compute the offsets of fields
2390 - Given that padding size depend even on compiler options, keep track of
2391 alignment and aligned_size only for the current architecture. Not a big
2392 deal since we send structure fields one after the other (seems
2394 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2395 program, let it work on linux/gcc/little32. portability todo.
2397 2004-07-08 Martin Quinson
2398 - import and improve remote compilation support from FAST
2399 - make sure make check works on half a dozen of machines out there
2401 2004-07-07 Martin Quinson
2402 Let's say it's version 0.0.3 ;)
2403 - Implement conversions (yuhu!)
2404 - Let it work on solaris (beside conversion, of course)
2405 - Stupid me, using rand() to generate the conversion datatests in not wise.
2407 2004-07-06 Martin Quinson
2408 - Let make dist work, since I'm gonna need it to compile on remote hosts
2409 - Let Tests/datadesc_usage write the architecture on which the file was
2410 generated as first byte.
2411 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2412 architecture descriptions.
2413 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2414 on those architectures.
2415 - Optimization: send/recv array of scalar in one shoot
2417 2004-07-05 Martin Quinson
2418 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2420 - Plug a whole bunch of memleaks
2421 - each process now have to call gras_{init,exit}. One day, their log
2422 settings will be separated
2423 - Continue the code factorisation between SG, RL and common in Transport.
2425 2004-07-04 Martin Quinson
2427 - Redistribution between SG and RL.
2428 We wanna have to accept in SG, so move accepted related parts of RL in
2429 the common part. (more precisely, the dynar of all known sockets is no
2430 more a static in transport.c, but part of the process_data)
2432 [gras_stub_generator]
2433 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2435 2004-07-03 Martin Quinson
2436 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2437 them all up in one shot)
2439 - Fix the ugly case of reference to dynamic array.
2440 - New (semi-public) function gras_datadesc_size to allow the messaging
2441 layer to malloc the needed space for the buffer.
2443 - gras_socket_close now expect the socket to close (and not its address to
2444 put NULL in it after it). This is because the socket passed to handlers
2445 is one of their argument (=> not writable).
2447 - propagate the interface cleanup from last week in datadesc, ie remove a
2448 superfluous level of indirection. User pass adress of variable
2449 containing data (both when sending and receiving), and not of a variable
2450 being a pointer to the data. Let's say that I like it better ;)
2451 The price for that is constructs like "int msg=*(int*)payload" in
2452 handlers, but it's a fine price, IMHO.
2454 - Let it work in RL (yuhu)
2456 2004-06-21 Martin Quinson
2458 - porting SG plugin and SG select to new standards (works almost).
2459 - plug memleaks and fix bugs around.
2462 - cleanup the prototype of data recv and force users to specify when they
2463 want to handle references to objects. Test case working even for cycles.
2464 - plug memleaks. Valgrind is perfectly ok with this.
2466 2004-06-12 Martin Quinson
2468 - cleanup the separation between plugin and main code in plugin creation
2470 2004-06-11 Martin Quinson
2472 - Reput hook for raw sockets, needed for BW experiments
2473 - kill a few lines of dead code
2474 [Data description] Interface cleanup
2475 - gras_datadesc_by_name returns the searched type or NULL.
2476 That way, no variable is needed to use a type desc once, which makes
2478 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2479 two parameters were strings (field name, type name), leading to
2481 [Dicos] Interface cleanup
2482 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2483 This is consistant with the dynar API.
2485 2004-04-21 Martin Quinson
2487 - Porting to new standards.
2489 - interface cleanup.
2490 There is no bag anymore, no need to take extra provision to mask the
2491 pointers behind "ID".
2492 Better splitup of functions between files create/exchange/convert.
2493 This is still a bit artificial since convert and receive are so
2494 interleaved, but anyway.
2496 - add a queued message list to procdata (the ones not matching criteria
2498 - factorize some more code between SG and RL wrt procdata
2500 - use gras_exit in example to track memleaks
2501 - get rid of gs_example now that GS is properly integrated into gras
2502 - update run_test to integrate the latest tests (datadesc)
2504 - rename WARNINGn macros to WARNn since it prooved error-prone
2506 2004-04-19 Martin Quinson
2508 - register init/exit functions within gras module mechanism
2509 - send/receive function.
2510 Convertion is not implemented, but short-cutted if not needed.
2511 struct/array elements are sent one by one (instead of block-wise), but
2512 nobody really cares (yet). Get a prototype before optimizing.
2513 - tests (using a file socket) for DD send/receive on:
2514 - base types: int, float
2515 - array: fixed size, string (ie ref to dynamic string)
2516 - structure: homogeneous, heterogeneous
2517 - chained list, graph with cycle
2518 Believe it or not, valgrind is not too unhappy with the results. The
2519 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2523 - Bugfix when using a filename explicitely (instead of '-')
2525 2004-04-09 Martin Quinson
2527 - factorize more code between RL and SG in socket creation
2528 - Complete the implementation and tests of:
2530 o file (only in RL, and mainly for debugging)
2532 I lost 3 days to design a portable address resolver, and then decided
2533 that the prototype mainly have to run on my box.
2534 Addressing portability too early may be like optimizing too early :-/
2536 - use gras_init in the Tests instead of the crappy parse_log_opt
2537 (the latter function is removed)
2538 [Conditional execution]
2539 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2540 [Code reorganisation]
2541 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2542 Build examples against the RL library, since there is no way to disable
2543 its creation for now.
2545 For information, the beginning of coding on GRAS was back in june
2546 2003. I guess that every line has been rewritten at least twice since