1 SimGrid (3.7) NOT RELEASED; urgency=low
3 The "sprint cleanups (before spring)" release.
6 * We can specify the smpi latency/bandwidth factor with command line
7 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
8 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
9 You can also use the "config tag" from platform file by adding this line
10 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
11 (see "example/platforms/tag_config.xml" to use "config tag").
12 Note that the command line supersedes the platform file configuration.
13 * Change the correction factors used in LMM model, according to
14 the latest experiments described in INRIA RR-7821.
15 Accuracy should be improved this way.
16 (TODO: this is not true in the code, yet)
17 * Use the partial invalidation optimization by default for the
18 network too. Should produce the exact same results, only faster.
19 * Major cleanup in surf to merge models and split some optimization
20 mechanisms from the core of the models. As a result you can now
21 specify which model to use (e.g., --cfg=network/model:LV08
22 --cfg=cpu/model:Cas01) and which optimization mode to use
23 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
24 Incompatible combinations should err at initialization. See
25 --help-models for the list of all models and optimization modes.
26 * The CLM03 workstation model was dropped for simplicity because it
27 used the deprecated CM02 network model. Use default instead.
28 * Rename the TCP_gamma configuration option to network/TCP_gamma
29 * Rename the coordinates configuration option to
30 network/coordinates, and document it
31 * Use now crosstraffic keyword instead of the terribly misleading
32 fullduplex keyword. It is activated by default now in the current
33 default model, use --cfg=network/crosstraffic:0 to turn it off.
34 * Experimental: new configuration option --cfg=surf/nthreads:N to run
35 analytical models in parallel with N threads (default is 1).
36 * Ongoing refactoring the model parsing to make XML files optional
37 (more work is needed here to fully achieve this goal)
40 * Deprecate the use of m_channel_t mecanism like MSG_task_{get,put}
41 functions and friends. This interface was considered as
42 deprecated since over 2 years, it's time to inform our users that it is.
43 Switch to MSG_task_{send,recv} instead, or compile Simgrid command line
44 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
45 use these (crappy) functions in your code.
46 These functions will be removed at some point. Stop using them now.
47 * Deprecate MSG_get_host_{table,number}
48 Implement MSG_hosts_as_dynar() instead.
49 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
52 * Stabilize the parallel execution mode of user contexts
53 * Introduce configuration variables to control parallel execution:
54 - contexts/synchro: Synchronization mode to use when running
55 contexts in parallel (either futex, posix or busy_wait)
56 - contexts/parallel_threshold: Minimal number of user contexts
57 to be run in parallel (raw contexts only)
60 * Introduce a parallel mode for the models (see surf/nthreads
61 configuration item). In our tests, running the models in parallel
62 never lead to any speedups because they are so fast that the gain
63 of computing each model in parallel does not amortizes the
64 synchronization costs, even when ultra fast futexes are used.
65 This is released anyway because YMMV.
68 * Performance boost by using a swag internally to compute the set of
69 tasks that are finished and should constitute the return value of
73 * Enable it by default now that it is considered rather stable.
76 * Documentation of the tracing functions.
77 * Performance gains when tracing categorized/uncategorized resource
78 utilization by avoiding calls to get route when updating resource
79 variables. LMM constraints are being used instead.
80 * API changed to set task categories. Use MSG_task_set_category instead
81 of TRACE_msg_set_task_category, and SD_task_set_category instead
82 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
83 * Bugfix for graphicator, routes not correctly obtained, memory leaks
84 * Examples for link user variables added (see at examples/msg/tracing/)
85 * Deprecated function TRACE_msg_set_process_category completely removed
86 * Trace header updated according to the latest Paje file format
87 * Tracing network lazy updates, no longer obligate users to use full updates
90 * Improve the API of Lua MSG bindings, using the Lua spirit.
91 * Each simulated process now lives in its own Lua world (globals are
92 automatically duplicated). It helps writing simulators. This is also the
93 first step towards running real distributed Lua programs with SimGrid.
94 * Add a Chord example in Lua, equivalent to the MSG one.
98 - Add new runtime parameters --help-logs and --help-log-categories
99 to display informations about supported logging parameters and
101 - Old deprecated parameters --{gras,surf,msg.simix,xbt}-log=...
102 don't exists anymore.
103 * Mallocators: allow value NULL for the reset function.
105 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
106 create homogeneous dictionaries, where all the elements share the
107 same free function. Non homogeneous dictionaries will be
108 deprecated in the next release.
109 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
110 - Multi-level dictionaries are deprecated.
112 - Cleanups and simplifications to make it maintainable again.
113 - Exotic features (such as memalign and valloc) were removed.
114 - The metadata were extended and improved so that the
115 model-checker becomes able to explore and inspect the heaps.
116 - This may induce a performance drop when enable_model-checking is
117 ON in cmake (even if it's not used in the simulation), but it is
118 necessary at this point to get MC working.
120 Turn model-checking OFF if simulation performance matters to you.
122 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
124 SimGrid (3.6.2) stable; urgency=low
126 The "Not coding new stuff allows to polish old things" release.
129 * New bindings to the NS3 packet level simulator (experimental)
130 * Use the raw (efficient) execution contexts instead of the sysv
131 (portable) ones when possible.
132 * libpcre is now mandatory in any cases since not using it led to
133 severe performance loss and possibly other issues
134 * Update the XML platforms:
135 - G5K: include the latest machine in Nancy
136 - GridPP and LCG: new platforms
137 * Documentation was partially updated, at least (more to come)
139 Bug fixes, cosmetics and small improvements
140 * Free terminated processes before the end of the simulation to avoid
141 exhausting the memory of users having very dynamic amount of
143 * Bug fix and cosmetics about canceling non-running tasks
144 * Bug fix about the dot loader's issues when using libcgraph
147 * Create an installer for windows with nsis (amd64 and win32)
148 - Add an hello world project to illustrate simgrid project creation.
149 - Embed libpcre into the Simgrid installer to avoid
150 its compilation burden
151 * The raw execution contexts should work on Apple now
152 * Port to Windows 64 bits
153 - Sysv contexts now have an implementation for this arch
154 - GRAS communication features now support this arch
155 * Drop support for borland compiler on windows
156 - this code was not maintained, and we kinda depend on gcc nowadays
157 * Fix portability issues on kfreebsd/gnu: build error about semaphores
158 * Fix portability issue on unstable ubuntu: linker became picky on
161 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
164 SimGrid (3.6.1) stable; urgency=low
166 The "Oops, we broke Macs too" release
169 * Fixed contexts detection so that raw ones are used when possible
170 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
171 leads to a strange error, with user code segfaulting sometimes when
172 the generated code is not perfectly aligned (which is not
173 controllable from the user side, depends on the amount of code)
176 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
178 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
180 SimGrid (3.6) stable; urgency=medium
182 The Summer Release, also known as the "OMG! They Killed Kenny!" version
185 * Bindings now constitute their own package, separated from the main one.
186 Rationale: reduce our maintenance nightmare by reducing the module coupling
187 They will soon be released on their own on gforge.
189 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
190 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
192 GRAS: It is not considered as stable anymore, but experimental. Sorry.
193 * It's not quite deprecated for now because we have no replacement,
194 but it may soon become the case.
197 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
198 * Fortran: New user-level cache variable to store the rank of the running
199 process. This improves performance by an order of magnitude.
200 * C: New coccinelle script to automatically locate and modify global and
201 local static variables.
202 * Improved SMPI network model with a sender-side gap to account for multiple
206 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
207 only say if a communication is finished, no matter whether it succeeded or
208 failed. You can call MSG_comm_get_status() to know the status of a finished
210 * New function MSG_task_dsend() to send a task and detach it. When a
211 communication is detached, you are never notified of its success or failure
212 and the memory is released automatically once it is finished. This function
213 is useful when you don't care about the end nor the success of a
215 * Change the prototypes of action replay. Sorry for inconvenience,
216 but this is really more efficient this way (and to adapt your code,
217 you just have to fix the initialization, that shouldn't be too long)
218 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
219 ever managed to do anything useful with it.
220 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
221 takes a dynar of communications. It returns immediately and gives the
222 index of a finished communication (if any).
223 * New example: a basic implementation of the Chord P2P algorithm.
226 * New model for multi-core CPUs. You can now use the core attribute to
227 precise the number of cores of a host. This is a basic model. Every
228 process running on the host receives at most the power provided in
229 the DTD (throughput<=power). Total throughput of process cannot exceed
231 * New peer tag. This peer tag creates a tiny AS comprising a host and a
232 router linked by an up-link and a down-link (possibly asymmetrical).
233 This kind of pattern allows to easily build last-mile model style platforms.
234 Aggregating such patterns in a rule-based AS is thus the technique of
235 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
236 * New model for Vivaldi routing. We transformed the Vivaldi network model
237 into a Vivaldi routing model (based on the rule-based model). This allows to
238 combine Vivaldi based latencies with last-mile platforms.
241 * Added a check for NaN of IEEE754 infinite in the double entries of
243 * Introduce a new context factory "raw", highly inspirated from the
244 ucontext factory, but using manually crafted functions in assembly to
245 do the work in an efficient manner.
246 * Allow to change the used context factory at run time, not only at
247 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
248 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
249 contexts at runtime (only with raw contexts or ucontexts).
250 * Completely rewrote this module to allow parallel execution of user
251 processes. Use --cfg=contexts/nthreads:N to execute user processes
252 with N parallel threads (the default is 1, meaning no parallelism).
253 * Allow to decide dynamically between sequential and parallel modes.
254 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
255 processes in parallel only when their number is greater than or equal to P
257 * Added a check for NaN of IEEE754 infinite in the double entries of
261 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
262 won't output the state of processes when interrupted with Ctrl-C
263 * Add a new function xbt_dynar_to_array that transforms a dynar into a
264 NULL-terminated array. This may solve backward compatibility issues
265 due to the change to return type of SD_simulate. See also:
266 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
267 * Add new macros with variable number of arguments.
268 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
269 - in xbt/asserts.h: xbt_assert
270 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
271 - in xbt/ex.h: THROWF and RETHROWF.
272 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
274 * Change xbt_die() to accept a format string with arguments, just like printf.
275 * New data structure: xbt_lib_t, like a dict but more general and with better
279 * New configuration options
280 Options triva/categorized and triva/uncategorized can be used to generate
281 graph configuration files for Triva visualization tool.
282 * Configuration option tracing/platform is renamed to tracing/categorized
283 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
284 * New TRACE_user_link_variable interface:
285 User provides the name of the link and the tracing variable to attach to it
286 * the declaration of tracing categories must be done after the environment
288 * simpler tracing interface, just one way to declare categories
289 TRACE_category or TRACE_category_with_color, it is up to you
290 * links in the trace file are again identified by their names
291 * trace contains the full platform hierarchy exactly as declared using the ASes
292 * Options tracing/msg/[task|process]:1 groups the process by hosts
293 for both cases, tasks and processes must have names that are unique during the simulation
294 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
295 * The experimental option tracing/msg/volume is deprecated
296 its functionality may be reincorporated if needed
298 The tracing generates a trace file with unordered timestamped events,
299 because of the way the core simulator (surf) works. A script available
300 at the tools directory (fix-paje-trace.sh) can be used to put the events
301 in order. We have changed the tracing so it can generate ordered timestamped
302 events in the final trace, but depending on the simulator (and how much time
303 is simulated) that can lead to a huge memory utilization. It is deactivated
304 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
307 * Define a SIMGRID_VERSION macro in simgrid_config.h.
308 - We are trying hard to keep the API stable, but it may happen that
309 some things change (we're a research project after all, not a
310 nuclear plant operating system). If such things should happen, you
311 could rely on that macro to adapt.
312 - current value: 30600 for 3.06.00, aka 3.6
313 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
314 a number that can be compared with SIMGRID_VERSION.
315 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
316 assertions and verbose/debug logging events are disabled at compile time.
318 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
320 SimGrid (3.5) stable; urgency=medium
323 * New feature to any SimGrid-based simulator: Model-Checking
324 Check SIN#1 for more details.
327 * New Model SMPI (three-interval linear regression for correction factors)
328 See RR-7426, available at http://hal.inria.fr/inria-00527150
329 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
330 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
331 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
332 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
333 * See also src/smpi/README
338 - Tracing API changes: TRACE_start and TRACE_end should not be called
339 by user-code. They are automatically called by simulators created
340 with SimDAG, MSG and SMPI if the toolkit is compiled with
341 tracing_enabled=ON. Categories declaration and utilization remain the
342 same for MSG and SimDag.
343 - A function was added to the tracing API to declare categories with
345 - TRACE_category_with_color (char *category, char *color)
346 where color must be in the following format
347 "%f %f %f", red, green, blue
348 and red, green, blue are float values in the interval [0, 1]
349 - User can specify NULL as color parameter, or continue calling
351 On that case, the tracing system will define random colors
352 - The following command-line options are supported:
353 --cfg=tracing/filename:msg.trace
354 --cfg=tracing:1 (activate tracing, needed to use others)
355 --cfg=tracing/platform:1 (categorized resource use)
356 --cfg=tracing/uncategorized:1 (uncategorized resource use)
357 --cfg=tracing/msg/task:1 (task creation)
358 --cfg=tracing/msg/process:1 (process creation, migration)
359 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
360 --cfg=tracing/smpi:1 (SMPI interface tracing)
361 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
362 - examples of examples/msg/tracing updated
364 - DAXLoader and DOTLoader functions can generate tasks with categories
365 - A new function to attribute a category to SD tasks:
366 TRACE_sd_set_task_category (SD_task_t task, char *category)
367 * Tracing the MPI interface implemented by SMPI
368 - Collective operations are traced with states
369 - Point-to-Point operations are traced with states/links
370 - Tracing activated by a parameter "-trace filename" passed
371 to smpirun during execution (considering that simgrid
372 is compiled with tracing enabled)
373 - To run the simulation with gdb, the simulator
374 accepts --cfg=tracing/smpi:1 to trace SMPI
375 - tesh files to check if smpi tracing is ok
376 - See examples/smpi/NAS/DT-trace
377 * GTNetS tracing re-worked
378 - adaptation to the tracing system of GTNets to cope
379 with modifications regarding the fullduplex mode
380 - new tesh files to check if gtnets tracing is ok
383 * Asynchronous communications through the functions:
384 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
385 * New function: MSG_load_platform_script()
386 to make possible using a lua script instead of XML files to set up platforms
387 * New function: MSG_set_function
388 to associate functions to processes, used when bypassing the parser
389 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
391 Platforms: Add some more examples in examples/platforms
392 * Grid'5000: see www.grid5000.fr
393 * *_30000_hosts.xml: various huge files [mainly scalability testing]
396 * Change the XML format. This is a very important modification. SimGrid
397 3.5 introduces a new hierarchical format based on the notion of
398 Autonomous Systems. Compatibility with old format is ensured
399 through the perl script provided in the install bin directory
400 bin/simgrid_update_xml.
401 It is now possible to build platforms with specific routing
402 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
403 connect several platforms together. We will try to provide soon
404 set of realistic platforms exploiting these properties (have a
405 look at examples/platforms/ for the moment).
406 * Take the opportunity of the XML format change to be a good XML citizen:
407 rename link:ctn to link_ctn and similar changes (also dealt with by
409 * Add a new routing scheme (rule-based) using regular expressions. It
410 enables to have an extremely low memory footprint when the
411 underlying routing is simple and can be compactly described. You
412 need to have libpcre4-dev (perl regular expressions) installed if
413 you want to use this routing scheme.
414 * Revive the cluster TAG and allow to easily and efficiently (both in
415 term of memory and speed) connect clusters together. Have a look
416 at teshsuite/simdag/platforms/ to see how this can be done. With
417 this tag, you can create clusters with thousands of tasks at no
418 cost (have a look at examples/platforms/).
419 Note: clusters are implemented as ASes, so there is no need for an
420 enclosing AS tag if you have only one cluster in your platform.
421 * Add new generic functions in the public interface that allows the user
422 to call SURF 'create_resource' methods from your code (same
423 functionality as the XML bypass mechanism but with a much lighter
425 * Add a new model (enabled through command line --cfg=network/model:SMPI)
426 that uses a piecewise linear approximation to produce better
427 results when exchanging small messages.
428 * Add a new parameter to handle correctly full duplex link and account
429 for interferences between uplink and downlink communications
430 (activate with --cfg=fullduplex:1).
433 * Rename the SD_READY (all dependencies are satisfied and task is
434 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
435 dependencies are satisfied) state.
436 This prevents a confusion between the notion of "ready to schedule"
437 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
438 simulated" (SD_RUNNABLE) used by the simulation kernel.
439 * Change the way a task is considered as ready. Instead of removing
440 dependencies when a task is done, a counter is decreased. This way,
441 it is always possible to reach ancestors thanks to the
442 SD_taks_get_parents function (even after the end of the simulation.)
443 * Change the return type of SD_Simulate from (SD_task_t*) into
444 xbt_dynar_t. This function was in handling a dynar internally and
445 converted it into a NULL terminated array for historical reasons.
446 * New function SD_dotload(char*) to load a DAG described in dot
447 format. This loader and the corresponding examples require the
448 installation of the graphviz library.
449 * Fix a bug in the management of tasks of size 0 in the surf network
450 models. This problem was only visible with SIMDAG and you should
451 thus disregard results produced with earlier versions if you
452 relied on this feature (some tasks were blocked because of this).
453 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
454 even though your applications were doing only point-to-point
455 communications and sequential computations. Now you can really use any
456 model you want (of course, if you create real parallel tasks, which are
457 not implemented in most models beside ptaskL07, this will abort).
458 * Add an example that schedules a DAX on an heterogeneous platform
459 using a Min-Min strategy.
460 * New function SD_workstation_get_current_task() that returns the kind
461 of task currently running on a workstation in the sequential access
463 * Raise some warnings when unexecuted tasks remains at the end of the
464 simulation. This is usually caused by cycles in the DAG.
467 * New function: SIMIX_process_set_function() called by MSG_set_function
468 * Change the underlying waiting queue in semaphores so that a process
469 can wait on several of them simultaneously (as in waitany).
470 * Fix the way to handle tokens in semaphores so that all access patterns
471 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
472 * kill the dirty pimple SIMIX_message_sizes_output()
473 Please use (proper) visualization instead
476 * New data container: setset (set of sets of elements)
477 * New module: mmalloc (mapped malloc, allowing to have several
478 independent segments of malloc)
479 * New function: xbt_dict_cursor_set_data()
480 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
481 * New function: xbt_dynar_is_empty()
482 * New function: xbt_fifo_get_last_item()
483 * Fix xbt_dynar_shrink(): use the right element size.
484 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
485 uninitialized areas during expand.
486 * Fix semaphores: previous implementation was severely broken.
487 * Use library init/fini functions for our initialization.
488 - you can use logs and other feature as soon as you want in your
489 code (even before the xbt_init / MSG_init)
490 - xbt_exit is now a no-op and produce a warning when used.
493 * Port GRAS to new SIMIX mechanisms. This allows gras users to
494 benefit from the latest improvement to the simulation kernel.
495 * Kill measurement sockets for now. If you rely on them, sorry. This
496 release is not for you. This feature will be reintroduced in the
497 future, but we cannot delay the release any further.
498 * New function: gras_msgtype_get_name().
499 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
500 * Fix (at last) the pmm example: it should not randomly fail anymore.
502 Build chain: bug fixes and overall polishing
503 * Cmake is now stable enough. Hence, we killed the autotools.
504 * Port to windows ( TM :)
505 * Fix the 'make install' target.
506 No need to use 'make install-simgrid' anymore
507 * Introduce a 'make dist' target compiling a *source* archive
508 'make package' compiles a binary archive
509 * Compile java files only on need
510 * Add --cd and --setenv command line options to tesh
511 * Out of source builds are not fully supported yet, but we are close
512 * Enable supernovae and optimization flags by default for our users
515 * Add layer to set up environment directly from lua, without XML.
516 * The effect of gras_stub_generator can be achieved through
517 lua too (check examples/gras/console/ping_generator.lua)
519 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
521 SimGrid (3.4.1) stable; urgency=low
523 The "Polishing easter eggs is probably a good idea" release.
524 This is a bug fixes release only.
528 * Fix a bug preventing the tasks from begin garbage collected.
531 * Fix a bug occuring when a host involved in a communication fails.
532 This was not detected properly by the other peer involved in the
533 communication. Now, it's reported as a network error.
536 * Warn the user about loop dependencies in data flow of DAX files
537 * Obey the control-flow dependencies of DAX files
540 * Add option "enable_smpi" allowing to not compile SMPI.
541 Probably useful for the (Mac) users experiencing a build error here
542 * Improve the detection of lua5.1 and ruby1.8
544 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
546 SimGrid (3.4) stable; urgency=low
548 The "Easter in Cargese" release. Also known as (major changes):
550 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
551 ~> bindings were greatly improved
552 ~> new tracing infrastructure for better visualization introduced
554 * the "Welcome to configury modernity" release.
555 ~> we switched from autotools to cmake, and improved our cdash
558 A more detailled list of changes follow (full detail in svn log).
560 Java Bindings: Various Cleanups
561 * (install java-gcj-compat-dev on debian-like to use them)
562 * Remove put/get: no need to export deprecated interface in Java
563 Use send/receive instead.
564 * Cleanup the examples and add a README per directory
565 * Remove example autoDestination (that's the only way to go now)
566 * Remove example explicitDestination (was a plain copy of basic)
567 * Make JniException a runtime exception, so that there is no need to
568 declare the fact that you may encounter such a beast. I guess that
569 nobody will ever want to survive such error.
570 * Create specific errors for each MSG case of failure:
571 host failure, transfer failure, timeout, task cancelled
572 * Cleanup the exceptions that may get thrown by each function
573 * Other internal cleanups in Java bindings. Performance still bad :/
574 Ruby and Lua Bindings: create them
575 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
576 * That's new and great, you should try them out.
577 Same functionalities than Java bindings, only even less polished
579 * Kill the useless "rate" argument of SD_task_get_execution_time()
580 Everyone used to provide -1 as a value, it was not used, and the
581 semantic of a possible use wasn't even clear.
582 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
583 either as comm costs or compute costs to mean that there is no
584 such thing for that specific task.
585 * Add a SD_task_set_name() function
586 * Fix SD_task_unschedule() on typed tasks
587 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
588 * In DAX loader, accept useless 'level' attributes to <job> since
589 LIGO DAGs have them (seem to be to ease graphical representation).
591 * Add an example masterslave_mailbox.c using send/receive and not
592 the deprecated put/get interface.
593 * Kill the MSG_paje_output() function. It's a noop since 2 years.
594 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
596 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
597 (declare MSG_USE_DEPRECATED to still have the old name)
598 * Add a MSG_task_set_data() function
599 * About trace replay (see examples/msg/actions):
601 - Allow to work with splitted trace files for each process
602 Give the specific trace file as argument of each process,
603 and call MSG_action_trace_run(NULL)
604 You can still have one merged file for all processes.
605 - Fix implementation of collective operations
606 * Allow task_execute() on 0-sized tasks (closes #10063)
608 * This is the first release of SimGrid where SMPI is not considered
609 beta anymore (even if some corners should still be improved)
610 * Port over the new SIMIX_network submodule (internal refactoring)
611 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
612 * Implement more missing elements of the standard:
614 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
616 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
617 MPI_Scan() MPI_Get_processor_name()
618 - Added implementation of missing case for Alltoall (warning: it's
619 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
620 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
621 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
622 SMPI_MPI_Allgatherv()
624 - MPI_Waitsome() was broken
625 - Allow relative includes in smpicc
626 - Command line cfg argument 'reference_speed' was ignored...
627 - Some functions did not properly lead to auto-benching of user code
628 - smpicc passes -O2 by default (just like openmpi one)
630 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
631 * Bug fixes about timeouts during communications
632 * add SIMIX_message_sizes_output() as a pimple to write to file the
633 amount of messages per size. Use gnuplot to get histogram.
634 Pimple because that's the only user-visible function of simix,
635 defined directly in xbt.h (irk, sorry)
637 - Add a SIMIX_sem_get_capacity() function
638 - Fix interactions with processe resume/suspende
639 - release_forever() was stupidly broken
640 - Fix SIMIX_display_process_status() for processes in a semaphore
641 - Make SIMIX_sem_block_onto() user-visible
642 * Refactoring context stuff:
643 - Use pseudo-OOP for better modularity
644 - reimplement SIMIX_process_kill() without process_schedule() so
645 that the latter can take as invariant that it is called from
647 - Merge context_start into context_new for sake of simplicity
649 * Add a Vivaldi network model, coded live during SUD'10 ;)
650 * Rename configuration variables to start a hierarchy:
651 o cpu_model -> cpu/model
652 o network_model -> network/model
653 o workstation_model -> workstation/model
654 * New configuration variables:
655 o network/bandwidth_factor: correction to bandwith
656 o network/latency_factor: correction to latency
657 o netwotk/weight_S: correction to the weight of competing streams
658 * Add a long description to the models, that users can see with such
659 argument on the command line: --cfg=cpu/model:help
660 * --help-models display the long description of all known models
662 * config: add the ability to set a default value after registration
663 Does not override any previously set value (e.g. from cmd line)
664 * dict: allow to have integer key and data.
665 When so, you need to use the following functions
666 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
667 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
668 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
669 In contrary to regular dicts, the key is not malloced before copy.
670 Mixing scalar and regular elements in the same dict is not tested
672 * Allow to use xbt_dynar_shrink() to expend the dynar instead
673 Tracing for Visualization:
674 * SimGrid is now instrumented in order to generate a trace file for
675 visualization analysis: to use it, need to compile SimGrid with the
676 "tracing" option enabled, and instrument the program using SimGrid with
677 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
678 (among other functions).
679 * The instrumentation only traces the platform utilization for now
680 * Documentation to use the tracing functions and how to analyze the
681 traces with the Triva tool is written.
682 * More information about: SimGrid FAQ (in the section Tracing Simulations
685 * We moved to cmake as default build system. Autotools support will
686 be dropped soon. Check the FAQ for more info about how to use it.
687 * Greatly improved our cdash/ctest interactions
688 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
689 * Added memory checking tests with valgrind; lot of memleak fixing.
690 This may be the first release of simgrid with so few memory issues
691 * Added code coverage tests.
692 Our coverage is still improvable, but at least we see it on cdash.
694 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
696 SimGrid (3.3.4) stable; urgency=low
698 The "Desktop Grid needs love too" release (also called Xmas release).
701 * Major speedup in the maxmin system solving by using lazy evaluation
702 Instead of solving completely the maxmin system at each iteration,
703 only invalidate (and recompute) the modified parts.
704 This new feature is enabled in default models but you can try to
705 turn it on with "--cfg:maxmin-selective-update=1" for other models.
706 * Cas01 IMproved as default CPU model
707 This CPU model is the same Cas01 model, but it uses the
708 maxmin-selective-update flag and a heap structure to manage
709 actions on SURF kernel.
710 It reduces the complexity to find the next action to finish and,
711 consequently, it's faster than the old Cas01.
712 This is the new default CPU model (Cas01).
713 * Rename the old Cas01 model to Cas01_fullupdate
714 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
715 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
716 * CpuTI (CPU Trace Integration)
717 A new CPU model whose objective is simulate faster when using
718 availability trace files.
719 Instead of using a full featured, over engineered maxmin system for
720 CPU modeling, this model does the pre-integration of traces files
721 to calculate the amount of CPU power available, and so, executes
722 faster than the old CPU models.
723 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
724 * Use LV08 as default network model since it gives better accuracy
725 for small messages and shouldn't change things for big ones.
726 Use --cfg=network_model:CM02 to get the previous behavior.
729 ******************************************
730 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
731 ******************************************
732 * The new CPU model may changes simulations!
733 The point is that events occurring at the exact same timestamp
734 are not scheduled in the same order with the old and new
735 version. This may be enough to completely change the execution
736 of simulations in some cases.
737 * The new network model will change simulations!
738 This new model is more realistic than the previous one, so you
739 should consider redoing your old experiments with this model.
740 Sorry for the inconvenience.
743 * Introduce the supernovae compilation mode
744 When compiled that way, the whole SimGrid (or almost) is put in a
745 single compilation unit and compiled in one shoot.
746 This is to help gcc which has difficulties to inline stuff from one
748 The speedup seem to be above 15%, althrough more tests are needed on
749 amd64 to confirm that gain.
752 * Port of MSG's mailbox on top of SIMIX network
753 The put/get mechanism was greatly simplified on the way.
756 * New SIMIX network module. Provides:
757 - Mailbox: rendez-vous mecanism to find with who you want to speak
758 - Synchronous send/recv: easier and hopefully faster since the
759 logic is handled in the maestro process directly now
760 - Asynchronous send/recv: you dreamt of it? It's here now
761 Too bad that nobody cared enough to propagate the change to MSG.
762 * Add semaphores as SIMIX synchronization mechanism.
765 * new function SD_daxload(char*) to load a DAX file
766 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
767 * Introduce typed tasks. Specify its kind and cost at creation.
768 At scheduling, just give where it should be placed, and the cost
769 for each involved resource is automatically computed.
770 Existing constructors so far (more to come of course):
771 - SD_task_create_comm_e2e() for end-to-end communication
772 - SD_task_create_comp_seq() for sequential computation
773 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
774 * new function SD_task_dump() for debuging display
775 * new function SD_task_dotty(task,FILE*) writing to file the info
776 about the task in dotty format
777 * SD_task_dependency_exists() can now cope with having one of its
778 arguments NULL. If so, it tests whether the other argument has any
780 * Add getters on list of preceding/following tasks:
781 SD_task_get_parents(task) and SD_task_get_children(task)
782 * Add getters on amount of workstations and list:
783 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
784 * Add getter on task kind: SD_task_get_kind(task)
785 * Update the start_time and finish_time of tasks on completion/failure
786 * Bugfix: Remove task from state swags when destroyed
789 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
792 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
793 dynar with one separate thread per value of the dynar.
794 * Change the prototype of xbt_thread_create(), sorry.
795 Added a boolean parameter indicating whether we want to join this
796 thread (used in SG only for now)
797 * Implement xbt_thread_join and xbt_thread_yield in SG also.
800 * GTNetS wrappers should now be usable again (and betterly tested too)
801 * Fix a major regression from 3.2 where the timeout provided to
802 MSG_task_put_with_timeout() was used as absolute time before which
803 the comm should be done.
804 * Start to fix the <cluster> tag.
805 - Internal links should be good now (beside of the loopback, which
806 use the private link instead)
807 - paths to the external world is still rather broken
808 - the <route:multi> tag is just broken. Actually that's brain-dead.
809 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
810 to make it less stupid
811 ** Check your platform with teshsuite/simdag/platforms/flatifier **
812 * Fix a source-level compatibility glitch from 3.2: after defining
813 MSG_USE_DEPRECATED, you can use the old name
814 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
815 * Allow to compile from the SVN with automake 1.11
816 * Fix some problems when using the "start_time" tag in deployment XMLs.
817 * Fix #8569: XBT/synchro.h has redundant declarations
818 * Fix #8563: MSG return values and exceptions
819 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
820 * Integrate patch #8636: Obey DESTDIR when installing documentation.
821 Thanks to Robson Peixoto.
822 * Fix a vicious bug in dictionaries inducing that some elements were
823 not freed on xbt_dict_free()
825 Portability report of this version:
826 * Main portability targets:
827 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
829 - mac leopard on i386
830 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
831 but nothing critical.
832 * Other platforms: windows, AIX and others were not tested for this release
834 Timing report of this version:
835 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
836 scenario parameters). From 8h to a few seconds in desktop grid settings.
837 * Supernovae brings about 25% speedup on i386.
839 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
841 SimGrid (3.3.3) stable; urgency=low
843 The "Need for Speed" release.
845 The timings done to validate the 3.3.2 were faulty.
846 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
848 The problem was a conversion from a manually handled vector to
849 xbt_dynar_t on the critical path.
850 xbt_dynar_foreach calls functions, inducing stack management crap.
852 We inlined these functions and xbt_dynar_foreach is now breath taking.
853 We also inlined xbt_swag_belong on the way.
855 Here are some approximate speedup measurements (on master/slaves
856 simulations lasting between 10s and 20s each):
857 3.3.1 -> 3.3.2: about same performance
858 3.3.2 -> 3.3.3: 40% speedup
859 3.3.1 -> 3.3.3: 40% speedup
860 3.3.1 with inline patch -> 3.3.3: 30% speedup
862 Our reading is that the refactoring which occurred in 3.3.2 made us
863 suffer much more from the xbt_dynar_foreach low performance, but
864 once we solved this, this refactoring proved to be very performance
865 effective. From the 40% speedup, somehow, 10% are due to the
866 inlining and 30% to the refactoring.
868 That's a pitty that gcc cannot inline functions placed in other files
869 alone. We have to choose between:
870 - break the encapsulation (by putting private data structures and
871 accessors in headers files to help gcc)
872 - live with low performance
873 - switch to a decent compiler such as icc (not quite possible).
875 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
877 SimGrid (3.3.2) stable; urgency=low
879 The "Simplicity does not preceed complexity, but follows it" release.
881 The main contributors of this release were (lexical order):
882 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
885 * Extract the routing logic into its own object.
886 (was dupplicated in network.c and workstation_LV07.c;
887 Allows to implement other ways of storing that info)
888 => kill now useless network_card concept
889 - Use dynar to represent routes (instead of void** + int*)
890 - kill link_set (use surf_network_model->resource_set instead)
891 - Add a command-line option to choose the routing schema to use
892 - Add three new models:
893 * Floyd (shortest path computed at initialization)
894 * Dijikstra (shortest path recomputed all the time)
895 * Cached Dijikstra (shortest path computed on need)
896 All these models where contributed by Silas De Munck, and are
897 described in his ICCS09 paper.
899 * Simplify model declaration
900 (less redirections, less function to write when defining a model)
901 - Factorize stuff between models:
904 surf_model_resource_set(model)
905 surf_model_resource_by_name(model, name)
906 - Unify the types of models in s_surf_model_t (using an union)
907 - Embeed fields of common_public directly into s_surf_model_t
908 - Rename model methods:
909 action_free ~> action_unref
910 action_change_state ~> action_state_set
911 action_get_state ~> action_state_get
912 - Change model methods into functions :
913 (model)->common_public->action_use ~> surf_action_ref
915 * Implement a generic resource; use it as ancestor to specific ones
916 (allows to kill duplicated code in models)
917 Drawback: timer command don't need no name nor properties;
918 workstation_CLM03 don't need no properties
919 (but I guess we can live with those few bytes wasted)
921 * Improve the action object model
922 - implement a constructor avoiding dupplicated code about field
923 initialization in generic_action part.
925 * Kill the SDP model: it has an external dependency, is deprecated
926 in flavor of modern lmm models, and didn't compile since a while
929 * Relocation of the context module from XBT to SIMIX.
930 (the context were decoupled from the simix processes, duplicating a lot of code)
931 => a lot of code was factorized
932 - less overhead is introduced during scheduling
933 - simpler API for the context factory
934 - the logic for process creation,destruction and manipulation was simplified
935 * Simplification of the s_smx_process_t data structure.
936 => accesing the simix level data associated to a process is faster now,
937 and the code is a lot more readable.
940 * Implement some more MPI primitives:
941 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
942 -implementation: Bcast: flat or 2-ary tree (default),
945 Allreduce: Reduce then Bcast
946 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
947 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
948 Alltoallv: flat tree, like ompi
950 * Add support for optimized collectives (Bcast is now binomial by default)
951 * Port smpirun and smpicc to OS X
954 * Kill SD_link_get_properties: hard to maintain and makes very little sense
955 Shout out if you used it.
958 * Display the list of still queued messages in SG mode when existing
962 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
963 * Add xbt_graph_node_get_outedges() [Silas De Munck]
964 * Add xbt_str_from_file(FILE*)
965 * Add xbt_dict_get_key achieving a linear reverse search
966 * Remove the context module
968 Portability report of this version:
969 * Main portability targets:
970 - Linux(debian)/x86/context
971 - Linux(debian)/x86/pthreads
972 - Linux(debian)/amd64/context
973 - Linux(debian)/amd64/pthreads
974 On these, we still have the eratic breakages of gras/pmm and
975 amok/saturate_sg reported in previous version. We still think
976 that the tests are the cause of the fault, not the tested code.
978 - Mac OSX Leopard/x86/context
979 Still false negative in tesh autotesting.
980 Smpi still fails, but this time because readlink does not accept -f
981 Everything seems to work properly beside of that.
984 - AIX version 5.3 (only tested contexts this time)
985 Smpi still fails there because mktemp is not installed.
986 Everything seems to work properly beside of that.
988 I managed to compile it for the first time, but several breakages.
989 Won't delay the release for this exotic platform.
991 * Windows: it's still lagging behind. If you want to help, please
994 Timing report of this version:
995 This version seem to be more than 5% faster than 3.3.1 (on linux
996 64bits with contextes). The gain is less than expected, we are
997 investigating this for next release.
999 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1001 SimGrid (3.3.1) stable; urgency=low
1004 * Implement a --cfg-help to show existing configuration variables
1005 * Build chain do not require doxygen in maintainer mode
1008 * fix a bug on struct sizeof computation, which prevented the
1009 exchange of arrays of structs in some conditions
1010 - added a regression test about this in datadesc_usage
1011 * Allow the exchange of 0-long dynamic vectors.
1012 - for that, use -1 as indicator of dynamic size instead of 0
1013 - This implied to change any size from unsigned long to long,
1014 reducing a bit communication abilities, but I guess that with
1015 64bits being quite common, this is more than enough.
1016 - This also induce a protocol change, thus bumping network protocol
1017 version from 0 to 1 (if we have external users, we have to get
1018 clean on that point too ;)
1019 - added two regression tests about this in datadesc_usage
1020 * Be more verbose when propagating local exceptions
1021 This helps debugging.
1022 * Display the status of simulated processes when receiving SIGINT in
1026 * Allow to control the simulation from a trace file.
1027 New functions MSG_action_register() and MSG_action_trace_run()
1028 The first one allows to associate a function execution to each
1029 kind of action while the second one parses a trace file and
1030 triggers the corresponding actions within the system.
1031 For now, only a toy example is provided in examples/msg/actions
1032 * Add an exemple of process migration in examples/msg/migration
1033 * Fix a bug in task exchange which broke MSG_task_get_sender()
1034 Add a teshsuite regression test for that.
1035 [Bug: if MSG_task_get_sender() is called after sender exit,
1037 * Fix a bug which prevented suspend/resume to work properly
1038 * Display the status of simulated processes when receiving SIGINT
1039 This fixes a regression of v3.3. due to the introduction of SIMIX
1040 * Bug fixing in failure management:
1041 - trace could not start by a failure at time 0
1042 - failure during communications were not working
1045 * Add SIMIX_process_set_name() to change the name of the current
1046 process in the log messages.
1047 * Store smx_hosts in a dict since we only retrieve them by name
1048 * Move the configuration infrastructure to surf
1051 * Move the configuration infrastructure to surf
1054 * Massive internal cleanups:
1055 - Store internal structures on processes instead of hosts (allows
1056 to have more than one process per host, in addition of being more
1058 - Cleanup the initialization/finalization process
1059 - Kill a whole bunch of unneeded synchronization:
1060 processes run in exclusive manner within the simulator
1061 - Move queues from global tables to process data fields
1063 - now accept -platform and -hostfile arguments
1064 - Pass the right rank value to processes according to the hostfile
1065 * Compile the examples by default, and use them as regression tests
1066 * Implement MPI_Wtime()
1067 * Change the reference speed to a command line option
1070 * TCP_gamma can now be specified as command line option using
1071 --cfg=TCP_gamma:10000000.0
1072 * Change the --surf-path cmd line option into --cfg=path:
1075 * Also include strbuff from xbt.h public header
1076 * xbt_ex_display(): do not free the exception after displaying
1077 This allows to do more with the given exception afterward.
1078 Users should call xbt_ex_free() themselves.
1082 Portability report of this version:
1083 * Main portability targets:
1084 - Linux(debian)/x86/context
1085 - Linux(debian)/x86/pthreads
1086 - Linux(debian)/amd64/context
1087 - Linux(debian)/amd64/pthreads
1088 These targets fail about 1/10 of times on gras/pmm, but we believe
1089 that this is because of the test, not because of simgrid.
1090 amok/saturate_sg fails even more rarely, and the test may not be
1093 - Mac OSX Leopard/x86/context
1094 The test suite still spits tons of errors because some obscure
1095 force prevents us from removing the temporary directories
1096 arguing that they still contain some metadata I've never heard of.
1097 Smpi fails because seq is not installed.
1098 Everything seems to work properly beside of that.
1101 - AIX version 5.3 (both contexts and pthread)
1102 Smpi still fails there because mktemp is not installed.
1103 XML inclusions seems rosty on AIX.
1105 * Windows: it's still lagging behind. If you want to help, please
1108 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1110 SimGrid (3.3) stable; urgency=high
1114 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1115 [Malek Cherier & Mt]
1117 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1120 Until now, GRAS were using MSG as an interface to SURF. It was
1121 quite difficult because both interface have several differences
1122 (MSG channels vs GRAS sockets were the most notable point).
1124 This also opens the gate to SMPI (which should occur soon) and speed
1125 up simulations by to 40% (even if it were not the main goal).
1127 **************************************
1128 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1129 **************************************
1130 The point is that events occuring at the exact same timestamp are
1131 not scheduled in the same order with the old and new version. This
1132 may be enough to completely change the execution of simulations in
1133 some cases. Sorry for the inconvenience.
1135 * Cleanup and upgrade the XML format to push further scalability
1136 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1138 * Improve the testing infrastructure with tesh. Now a very large part of
1139 the code is tested not only by being run but also by checking that the
1140 output match an expected output [Mt].
1142 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1143 is really less memory-demanding, which should allow you to use
1144 larger files in SimGrid [AL].
1146 * Inform valgrind about our contextes, so that it becomes usable
1147 with the default (and more effecient) version of SimGrid
1148 [contributed by Sékou Diakite, many thanks]
1151 * Introduce a listener thread in charge of receiving incomming
1152 messages from the network. It allows to overlap communication and
1153 computation but most notably, it removes some stupid deadlocks due
1154 to the fact that so far, a process could not send and receive at
1155 the same time. This made most non trivial communication schema
1157 * Convert the PIDs from long int to int to match the MSG ones (and
1158 linux ones too) [Mt]
1159 * New function: gras_agent_spawn() to launch a new process on
1160 current host. Only working in simulation for now. [Mt]
1161 * New function: gras_os_hostport() returning a constant form (ie,
1162 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1165 * Make the backtrace of exceptions more human readable [Mt]
1166 * New module: xbt/str [Mt]
1167 a ton of string utility functions (split, join, printf to a newly
1168 allocated buffer, trim, etc)
1169 * New module: xbt/hash [Mt]
1170 SHA1 hashing algorithm (more to come if needed)
1171 * New module: xbt/synchro [Mt]
1172 synchronization tools (mutex and conditions) working the same way
1173 in simulation and in real life (mainly useful for GRAS, but not
1175 * New module: xbt/queue [Mt]
1176 classical producer/consumer synchronization scheme
1177 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1178 (using the classical functions will get serialized) [Mt]
1179 * Make dictionary internal table dynamic. No need to specify its size
1180 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1182 * Make sure the log channels are organized as a tree under windows
1183 (because of ANSI C compatibility issue, any channel were child of
1184 root directly) [Mt].
1187 * Cleaned many thing in surf and fixed a few bugs [AL].
1188 * Add a nice command line configuration mechanism to compose models [AL].
1189 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1190 the previous one (KCCFLN05). It relies on something that looks like
1191 a max-min sharing mechanism but cannot be written as such. A new solver
1192 was thus designed [AL].
1193 * Add a new solver to lmm. Based on Lagrange optimization and
1194 gradient-based descent, it enables to efficiently maximise systems s.a
1196 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1198 This solver enables to propose two new network models for TCP Reno and
1199 TCP Vegas based on Low's work. These models still need to be fully
1200 tested though [Pedro Velho].
1203 * Bug fix in SD_simulate. Now the time bound given as argument is
1205 * Use the new parallel task model (ptask_L07) as default.
1206 * Use the SURF command line configuration mechanism.
1207 * 0-size tasks (for synchronization) should now work.
1209 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1211 SimGrid (3.2) stable; urgency=high
1215 We still experience issues on this platform, but we believe that at
1216 least MSG is usable.
1218 GRAS API BREAKAGE (for simplification purpose, sorry):
1219 * the gras_msgtype_by_name is not used anymore. Instead of
1220 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1221 you can write (and must)
1222 gras_msg_send(toserver, "request", &request);
1223 - If you still want to pass a gras_msgtype_t to the function (to cache
1224 the type and avoid the lookup time), use the gras_msg_send_() variant.
1225 - Impacted functions:
1226 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1227 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1228 * The callbacks are now expected to return 0 when everything went well
1229 (just like the main() function)
1231 GRAS new features and improvements:
1232 * New module mecanism where user code can use per process globals [Mt]
1233 This is similar to gras_userdata_*() functions, but for libraries. It
1234 factorize some code developped over and over in the examples and AMOK.
1235 It has still to be documented and used (only amok/peermanagement is
1237 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1238 loss when they were small enough to fit into the buffer and sent quickly
1239 enough so that they can all get received in one shoot.
1240 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1241 if not found. Use the *_or_null() variant for the old semantic.
1242 * In gras_msg_handle, do not discard messages without callback.
1243 They are probably messages to be explicitly awaited later (ie, proofs of
1244 mis-synchronization in userland since they are sent before being awaited)
1245 No big deal usually.
1246 * gras_socket_meas_send/recv: semantic changed!
1247 The numerical arguments used to be (1) the total amount of data to send
1248 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1249 messages. This was need for the fool willing to send more than MAXINT
1250 bytes on quite fat pipes.
1253 * Do really rename the hostmanagement module to peermanagement. [Mt]
1254 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1255 but this is rather new and this was documented in the module
1256 documentation (poor excuses, I admit)
1257 * Bandwidth measurement semantic changed! This follows the changes to
1258 gras_socket_meas_send/recv explained above.
1261 * A sequential mode has been added to the workstations. When a workstation
1262 is in sequential mode, it can execute only one task, and the other tasks
1263 are waiting in a FIFO. [Christophe Thiery]
1266 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1267 model for SIMDAG. [Christophe Thiery]
1268 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1269 equal to 0, which caused some bad behaviors in
1270 saturated_constraint_set_update. I now use a threshold mechanism like in
1274 * When running manually src/testall, you select specific units [Mt]
1275 testall is the result of our cunit mecanism, and should replace all
1276 the scripty thingy around since bash don't run easily on billware.
1278 * A mallocator system has been added. [Christophe Thiery]
1279 Mallocators allow you to recycle your unused objects instead of freeing them
1280 and allocating new ones.
1282 Documentation update:
1283 * FAQ reworking + New FAQs:
1284 - "Valgrind spits tons of errors!" [Mt]
1285 - "How to repport bugs" [Mt]
1286 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1287 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1288 - Communication time measurement within MSG [AL]
1289 - I experience weird communication times when I change the latency [AL]
1290 * GRAS tutorial [Mt]
1292 - an introduction to the framework and to the used communication model
1293 - an initiatic tour introducing the most proheminent features:
1295 . Lesson 0: Installing GRAS
1296 . Lesson 1: Setting up your own project
1297 o Part 2: Message passing
1298 . Lesson 2: Exchanging simple messages
1299 . Lesson 3: Passing arguments to the processes (in SG)
1300 . Lesson 4: Attaching callbacks to messages
1301 . Lesson 5: Using globals in processes
1302 . Lesson 6: Logging informations properly
1303 . Lesson 7: Using internal timers
1304 . Lesson 8: Handling errors through exceptions
1305 . Lesson 9: Exchanging simple data
1306 . Lesson 10: Remote Procedure Calling (RPC)
1307 . Lesson 11: Explicitely waiting for messages
1308 . Recapping of message passing features in GRAS
1309 - A HOWTO section containing:
1310 o HOWTO design a GRAS application
1311 More are due, of course. They will come latter. In the meanwhile, you can
1312 check the examples which are still here.
1314 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1316 SimGrid (3.1) stable; urgency=high
1320 There was a stack corruption somewhere, visible only when optimizing
1321 with these versions. [Vince]
1324 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1325 module that enabled to play with Directed Acyclic Graphs. It is built
1326 directly on top of SURF and provides an API rather close to the old
1327 SG. Some old codes using SG are currently under rewrite to check that
1328 all needful functions are provided. [Christophe Thiery]
1331 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1332 extension of the classical CLM03 model that gracefully handles
1333 failures. This is now the default model for MSG and GRAS. It doesn't
1334 handle parallel tasks yet though. [AL]
1335 * Bug fix: Weights were not correctly set in the network part.
1336 WARNING: This may have resulted in incorrect results with simulations
1337 where there are more than one flow on a given link. [AL]
1340 * After a (long ?) discussion on simgrid-devel, we have decided that the
1341 convention we had on units was stupid. That is why it has been decided
1342 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1343 WARNING : This means that all previous platform files will not work as
1344 such with this version! A warning is issued to ask users to update
1346 A conversion script can be found in the contrib module of the CVS, under
1347 the name contrib/platform_generation/surfxml_update.pl [MQ]
1350 * Bug fix: Processes were started in reverse order, wrt deployment file.
1351 WARNING: if your code relies on this bug, please fix it. [AL]
1352 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1353 being executed on two different locations. [AL]
1354 * Bug fix: Failures are now better supported thanks to Derrick's tests
1355 (there was many failure situations I hadn't thought of and that weren't
1356 correctly handled). [AL]
1357 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1361 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1362 Exception occurring on server-side are propagated back to client (!).
1364 API CHANGE: the callback changed their prototype. Change:
1365 int my_handler(gras_socket_t expeditor, void *payload_data) {
1367 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1368 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1370 * New! function: gras_msg_handleall to deal with all messages arriving
1371 within a given period.
1372 * New! function: gras_socket_server_range to get a server socket in a
1373 range of port numbers (ease to avoid port number conflicts) [MQ]
1374 * New! gras processes display their backtrace when they get a SIGUSR1
1375 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1376 Sweet to debug RL processes [MQ]
1380 - Do not force experiment sizes to be expressed in kb, or it becomes
1381 impossible to measure the latency this way (needs one byte-long tests)
1382 WARNING: this changes the amok_bw_* function semantic. [MQ]
1383 - Implements the link saturation stuff. [MQ]
1384 * Peer management module:
1385 New! module factorizing code that we wrote over and over [MQ].
1388 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1389 - Test units are placed directly into the library code, they get extracted
1390 automatically and placed into the src/testall binary.
1391 - Convert most of the XBT tests to this system.
1392 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1393 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1394 live in the same C code. This required to change a little bit the API
1395 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1396 * New module: sparse graph structure with basic algorithms (this is work
1397 in progress and the API is not considered to be frozen yet). [AL]
1398 * Display more information on backtraces: source line & function names are
1399 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1400 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1401 * Reworked a little bit some #include statements to load only required
1402 headers. Some user code that relied on SimGrid to include stdlib or
1403 stdio may need to include it by themselves. [AL]
1404 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1406 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1407 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1408 but I guess that almost nobody used those parts. [MQ]
1410 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1412 SimGrid (3.0.1) stable; urgency=low
1415 * Unfortunately, I had missed 5 misnamed functions:
1416 xbt_fifo_item_t xbt_fifo_newitem(void);
1417 void xbt_fifo_freeitem(xbt_fifo_item_t);
1418 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1419 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1420 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1421 They're now deprecated. Please use their new versions:
1422 xbt_fifo_item_t xbt_fifo_new_item(void);
1423 void xbt_fifo_free_item(xbt_fifo_item_t);
1424 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1425 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1426 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1428 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1429 * Documentation: xbt_log module unmercifully reworked [MQ]
1430 * Bugfix: there was a problem with the ending of contexts with
1431 the pthread backend. It caused some weird deadlock or behavior
1432 depending on the pthread implementation. [AL]
1433 * Bugfix: get the exceptions raised in the simulator repport where
1434 and why they come from when they are not catched in time [AL, MQ]
1437 * Bugfix: Do repport the error when two non-connected hosts try to
1438 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1441 * Add additionnal checkings on communications. Assert that two
1442 communicating hosts are connected by a set of links... [AL]
1445 * Add additionnal checkings on channel values in communication [AL]
1446 * New: MSG_task_get_source to see on which host a task was generated [HC]
1447 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1448 the number of tasks waiting to be received on channel and sent
1450 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1451 waits for the first task coming from a given host.. [AL]
1453 GRAS new functionnalities: [MQ]
1454 * Enhance the parsing macro to allow the size of multidimentional objects
1455 to be given thru annotations.
1456 * New example (and documentation): Matrix Multiplication a la RPC
1457 (as when I was young!) and fix a bunch of bugs found on the way.
1459 GRAS performance improvements: [MQ]
1461 * Reduce the amount of cbps creation/destruction by making it static to
1462 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1465 * Change libdata to a set so that we can search for stuff by ID (and thus
1466 reduce the insane amount of dict lookups)
1469 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1470 RL communication on top of SG since it's so uneffective.
1471 It may also be used for inter-thread communication in RL, one day.
1472 * Use gras_datadesc_copy() to exchange messages on top of SG
1474 - improve message exchange performance on top of SG
1475 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1476 * Don't exchange on the network the size of the used part of buffer,
1477 instead, specify the possible buffer size to read().
1479 - reduces the amount of read/write calls (one pair per exchange)
1480 - reduces the amount of exchanged data (the size)
1481 - allows to retrieve all arrived data on receiver side, if we don't need
1482 it right now (subsequent read will peek the buffer)
1483 - allows the receiver to proceed with the begining of the stream before
1484 everything is arrived
1485 - make it possible to build an iov transport (using readv/writev)
1487 - take care of the data with non-stable storage (like stacked data),
1489 * If possible, TCP send uses vector I/O (when writev() is here)
1490 - Don't use it for receive since we send data sizes and data on the
1491 same stream, so we wouldn't be able to chain large amount of chunks
1492 before having to flush the stuff to read the size.
1493 * Rework the transport plugin mecanism to simplify it and reduce the
1494 amount of pointer dereferencement when searching for the right function
1497 * I guess that now, we do almost as few system calls as possible while
1498 doing as few data copy as possible.
1500 To improve it further, we could try to send all the sizes first and then
1501 all the data (to use iov on receiving size), but it's only a partial
1502 solution: when you have 2 dimensional data, the sizes of the second
1503 dimension is data of the first dimension, so you need 3 streams.
1505 I'm not sure the potential performance gains justify the coding burden.
1507 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1509 SimGrid (3.00) stable; urgency=high
1512 * New! Give the possibility to hijack the surf parser and thus bypass
1513 MSG_create_environment and MSG_launch_application. Have a look at
1514 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1516 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1518 SimGrid (2.96) unstable; urgency=low
1523 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1524 This deprecates the xbt_error_t mecanisms.
1525 It modifies (simplifies) all XBT and GRAS API.
1526 MSG API keeps unchanged (exceptions raised by XBT are catched from
1527 within MSG and masked with existing error handling facilities)
1530 * New! Add a FATPIPE model. [AL]
1531 * New! Add a parallel task model. [AL]
1532 * New! Add automatically a loopback interface (in the default
1533 network model) if none was precised.
1536 * Bugfix: MSG_process_resume now works with the current running process.
1538 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1539 * Modification of MSG_task_get_compute_duration. Once a task has been
1540 processed, the value returned by this function is now equal to 0. [AL]
1541 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1542 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1543 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1544 of a m_task. For now, MSG_task_cancel only works with computation
1546 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1547 of the processor (in Mflop/s) regardless of the current load on the
1549 * API Change: use proper naming convention for MSG_getClock and
1550 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1552 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1553 This function changes the priority of a computation task. This priority
1554 doesn't affect the transfer rate. A priority of 2 will make a task
1555 receive two times more cpu power than the other ones. This function
1556 has been added to suit the needs of Nguyen The Loc and hasn't been that
1557 much tested yet. So if it fails, please report it and send me your code.
1559 * API Change: removed all functions and types that were marked "deprecated"
1560 since many months. Renamed MSG_global_init_args to MSG_global_init.
1562 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1564 SimGrid (2.95) unstable; urgency=low
1567 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1568 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1569 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1572 * Bugfix: Allow absolute paths for platform description files [MQ]
1573 * Bugfix: do free the variables after use. Leads to drastic performance
1575 * Implement max_duration (ie, timeouts) on resources [AL]
1578 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1580 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1581 the needed support (timeouts) [AL]
1584 * Implement measurement sockets. You can now get the bandwidth between two
1585 hosts thanks to AMOK (see below). [MQ]
1586 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1587 dynar over the network (yeah) [MQ]
1588 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1589 * simple/static token-ring example. [Alexandre Colucci and MQ]
1590 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1591 in place before [MQ]
1594 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1595 type parsing. "short" and "long" modifiers were also ignored; other
1596 modifier (such as reference level) are still ignored. [MQ]
1597 * Embeed the buffer size within the buffer itself on SG. [MQ]
1598 That way, send() are atomic and cannot get intermixed anymore (at least
1599 the ones which are less than 100k; bigger messages still have the issue)
1600 * Array size pushed by the field, not by the field type (or each
1601 and every long int will push stuff to the cbps) [MQ]
1602 * use select() to sleep since it allows to portably sleep less than one
1605 GRAS (minor cleanups)
1606 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1609 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1610 * type category 'ignored' killed as it was never used and were difficult
1612 * whether a type can cycle or not is now a flag, leaving room for more
1613 flags (as "ignored", if we feel the need one day ;)
1614 * Rename raw sockets to measurement sockets since "raw" has another
1615 meaning in networking community.
1618 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1619 and offers features not belonging to GRAS but that most applications
1620 need. One day, it may be a set of plugins loadable at runtime.
1621 * New module: bandwidth
1622 bandwidth measurement between arbitrary nodes running this module. [MQ]
1624 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1626 SimGrid (2.94) unstable; urgency=low
1628 The first beta release of SimGrid 3 !
1632 * Update the main page and the FAQ. Adding references to gforge.
1635 * Add a gras_os_getpid function.
1638 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1639 * Extend the logs so that they also print PID, hostname, date, ... if
1641 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1642 and kill the old version which were in testsuite/
1643 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1647 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1648 for users. Instead, each programming environment (GRAS, MSG,...) use it
1649 when needed to provide such a feature to users.
1650 Don't shortcut the mecanism or you will also shortcut the virtualization
1651 you need on the simulator.
1655 * Cleanups in configury with regard to compile optimization/warning flags.
1656 Also add -fno-loop-optimize to any powerpc since it's the optimization
1657 killing gcc (< 3.4.0).
1658 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1659 needed by MSG examples complications
1660 * Borrow configury beautifications from PHP
1663 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1664 warning (thanks loris for stumbling into this one).
1665 * Bugfix: stop loading private headers (gras_config.h) from the public
1669 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1670 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1671 RL would imply a DNS resolver, which is *hard* to do in a portable way
1672 (and therefore delayed).
1673 * Implement a real timer mecanism and use it in timing macros. This allows
1674 to avoid rounding errors and get a 0.000005 sec precision in timing
1675 macros. While I was at it, various cleanups:
1676 - allow to declare more than one timed section per file (fix a stupid bug)
1677 - move some private declaration to the right place
1678 - merge conditional execution and timing macros into emulation module
1679 - document the module
1680 - make sure the module cleanups its mess on gras_exit
1681 * Documentation improvements:
1682 - (new) how to compile applications using GRAS
1683 - (new) emulation support (timing macros)
1685 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1687 SimGrid (2.93) unstable; urgency=low
1689 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1692 - Use Paje properly where used. Still to be sanitized properly.
1693 - Portability fix: Add an implementation of the contexts using pthread
1697 - Add xbt_procname(): returns the name of the current process.
1698 Use it to show the current process's name in all logging.
1700 - fix detection of older flex version and the reaction, since we do
1701 depend on modern ones (we use lex_destroy)
1702 - Better separation of SG and RL in the libs: remove all simulation code
1703 from libgras. As a result, this lib is now only 200k when stripped.
1704 Some of the xbt modules may also be duplicated (two sets and such) and
1705 should be cleaned/killed before SG3.
1706 - Insist on using xlC on AIX because of weird problems involving gcc there.
1707 - Cleanup the make remote stuff. This is now done by scripts
1708 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1709 mainly for our private use, but we're working on changing them to user
1712 - Bugfix: flush the socket on close only if there is some *output*.
1713 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1715 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1718 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1720 SimGrid (2.92) unstable; urgency=low
1722 Alpha 3 on the path to SimGrid 3
1726 - New! First try of benchmarking macros.
1727 - New! First try so that gras_stub_generator generate deployment and
1728 remote compilation helpers.
1730 - Bugfix: Initialization fix in msg_test.
1734 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1736 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1738 - API Change: gras_os_sleep to take the amount of seconds as a double.
1739 Accepting an int was error prone since it was the only location where
1740 seconds were coded as such. It leaded to damn rounding errors.
1741 - Bugfix: Hard to belive that timers ever worked before this.
1743 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1745 SimGrid (2.91) unstable; urgency=low
1747 Alpha 2 on the path to SimGrid 3
1751 - Bug fix in the lmm_solver.
1753 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1754 through the function MSG_paje_output.
1755 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1756 - It is possible to bound the rate of a communication in MSG with
1757 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1758 to write it in the changelog).
1759 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1764 - Various cleanups to the autotools stuff
1765 - Begin to move Gras examples to examples/gras/
1766 - Let make distcheck work again (yeah!)
1768 - documentation overhauled using doxygen.
1769 gtk-doc-tools is dead in SimGrid now.
1770 - Automatically extract all existing logging categories, and add the list
1771 to the documentation (long standing one, to say the less)
1773 - Cleanup the known architecture table. Reorder the entries to group what
1774 should be, and use a more consistent naming scheme.
1775 (some of the test dataset are still to be regenerated)
1776 - New! Allow library to register globals on each process just as userdata
1778 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1779 lookup time (for now).
1780 Use it in msg and trp.
1781 This cleans a lot the internals and helps enforcing privacy of the
1782 headers between the gras components.
1783 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1784 - Bugfix: gras_os_time was delirious in RL.
1785 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1787 - Reenable GRAS now that it works.
1789 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1791 SimGrid (2.90) unstable; urgency=low
1793 Alpha 1 on the path to SimGrid 3
1795 * It is a long time since the last release of SimGrid. I'm sorry about
1796 that but as I had told you, I was rewriting a lot of things. I apologize
1797 to those who had been reporting bugs to me and that I had not answered.
1798 If your bug is still in the new version, please tell me. Here is a
1799 summary of the main changes.
1801 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1802 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1803 portability, speed, and a lot more but you'll figure it out later.
1804 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1805 I wanted to release the new MSG as soon as possible and I have
1806 broken GRAS, which is the reason why, for now, only MSG is fully
1807 functional. A laconic description of these projects is available
1808 in the documentation.
1810 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1811 kernel optimized for our needs (called SURF but only the developers
1812 should use it). Hence, MSG is now roughly 30 times faster and I think
1813 that by rewriting a little bit MSG, I could event speed it up a little
1814 bit more. Beside the gain in speed, it is also much easier to encode a
1815 new platform model with SURF than it was with SG. More to come...
1817 * REVOLUTION 3: I have tried to change a little as possible the API of
1818 MSG but a few things really had to disappear. The main differences
1819 with the previous version are :
1820 1) no more m_links_t and the corresponding functions. Platforms are
1821 directly read from a XML description and cannot be hard-coded
1822 anymore. The same format is used for application deployment
1823 description. The new format is described in the documentation.
1824 Have a look in tools/platform_generation. There is a tiny script
1825 that converts from the old platform format to the new one. Concerning
1826 the application deployment format, parsing the old one is tricky.
1827 I think most of you should however be able to convert your files. If
1828 it is really an issue, I can write a C code that does the conversion.
1830 2) the toolbox tbx does not exist anymore. We now have a library
1831 with much more data-structures but without the hash-tables (we have
1832 dictionaries that are much faster).
1834 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1836 *****************************************************************************
1837 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1838 * don't feel like converting the oldies. (Mt) *
1839 *****************************************************************************
1842 Version 2.90: "the long awaited one"
1843 - Finished rewriting and debugging MSG. Rewrote the documentation.
1844 - disable GRAS for now since it needs to be ported to the newest SG
1847 - Finish the port to windows (using mingw32 for cross-compile)
1850 - Main loop and datastructures of SURF. A cpu resource object is
1851 functional. Surf can thus be used to create cpu's with variable
1852 performance on which you can execute some actions.
1854 2004-11-15 Martin Quinson
1855 - Port to ARM. Simply added the alignment and size descriptions. Should
1856 work, but the ARM machines are so slow that I didn't had the opportunity
1857 to 'make check' over there yet.
1859 2004-11-15 Arnaud Legrand
1860 - Trace manager now written. It uses a heap structure and is therefore
1861 expected to be efficient. It may however be speeded up (particularly
1862 when many events occur at the same date) by using red and black
1863 trees. One day maybe...
1864 - Max-min linear system solver written. It uses a sparse matrix
1865 structure taking advantage of its expected use. Most operations are
1866 O(1) and free/calloc are called as few as possible. The computation of
1867 the minimum could however be improved by using a red and black tree
1870 2004-11-03 Arnaud Legrand
1871 - Rename every gras_* function that was in xbt/ to its xbt_
1873 - Add a heap and a doubly-linked list to xbt
1874 - Added a dichotomy to the dictionaries. make check works as well before
1875 so I assume that the patch is correct. I do not know however if things
1876 run effectively faster than before now. :)
1878 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1879 SimGrid, and the version number is bumped to 2.x
1881 2004-10-29 Martin Quinson
1882 - Introduction of the remote errors.
1883 They are the result of a RMI/RPC on the remote machine.
1884 ErrCodes being scalar values, you can't get the host on which those
1885 errors did happen. Extending the error mechanism as in Gnome is possible.
1886 No idea yet whether it is a good idea.
1888 2004-10-28 Martin Quinson
1889 - Interface revolution: the Starred Structure Eradication.
1890 I used to do typedef struct {} toto_t; and then handle *toto_t.
1891 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1892 - ???_t is a valid type (builded with typedef)
1893 - s_toto_t is a structure (access to fields with .)
1894 - s_toto is a structure needing 'struct' keyword to be used
1895 - e_toto_t is an enum
1896 - toto_t is an 'object' (struct*)
1898 typedef struct s_toto {} s_toto_t, *toto_t;
1899 typedef enum {} e_toto_t;
1900 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1901 s_toto_t) is private.
1903 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1904 so that it can set the variable to NULL. It was so for dicts and sets,
1905 it changed for dynars.
1907 - Fix a bunch of memleaks in dict_remove
1908 - Fix a bug in sg/server_socket opening: it failed all the time.
1910 2004-10-07 Martin Quinson
1911 - Speed up dynar lookup operation a bit.
1913 gras_dynar_get is dead.
1915 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1916 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1917 (which gives you the address of the stored data).
1919 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1920 copy of the data using an affectation to do the job and not a memcpy.
1922 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1923 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1925 It does not really speedup the dynar test because they are
1926 setting elements all the time (and look them seldom). But the dict does
1927 far more lookup than setting.
1929 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1931 2004-10-05 Martin Quinson
1932 - Allow to (en/dis)able the cycle detection at run time.
1934 Whether we should check for cycle or not is now a property of each
1935 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1936 datadesc_cycle_unset allow to remove this property when previously set.
1938 Note that the cycle detection is off by default since it impacts the
1939 performance. Watch the data you feed GRAS with ;)
1941 This property is hereditary. Any element embedded in a structure having it
1942 set have it set for the time of this data exchange.
1944 You should set it both on sender and receiver side. If you don't set it on
1945 sender side, it will enter an endless loop. If you forget on receiver
1946 side, the cycles won't be recreated after communication.
1948 - Header reorganization.
1949 Kill gras_private.h, each submodule must load the headers it needs.
1951 2004-10-04 Martin Quinson
1952 - Interface revolution: do not try to survive to malloc failure.
1954 Now, gras_malloc and friends call gras_abort() on failure.
1955 As a conclusion, malloc_error is not a valid error anymore, and all
1956 functions for which it was the only gras_error_t return value are
1957 changed. They now return void, or there result directly.
1958 This simplify the API a lot.
1960 2004-09-29 Martin Quinson
1961 - Re-enable raw sockets.
1962 Created by gras_socket_{client,server}_ext;
1963 Used with gras_raw_{send,recv}
1966 It should allow to kill the last bits of gras first version soon.
1968 This is not completely satisfactory yet (duplicate code with
1969 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1972 - Simplify transport plugin (internal) interface by not passing any
1973 argument to _server and _client, but embedding them in the socket
1976 2004-09-28 Martin Quinson
1977 - Finish the port to AIX.
1978 autoconf was my problem (segfault within the malloc replacement
1979 function. No idea why)
1981 2004-09-16 Martin Quinson
1982 - Fix some size_t madness on 64bit architectures.
1984 2004-09-08 Martin Quinson
1985 - Reduce the number of system headers loaded, overload some more system
1986 calls (such as malloc to cast the result of the system one, and work
1988 - Fix and reintroduce the config support
1990 2004-09-07 Martin Quinson
1991 - Source code reorganization to allow Arnaud to surf all over there.
1992 - Allow to document the logging categories.
1993 - Remove all uppercase from logging categories and useless cleanup in names.
1995 2004-08-18 Martin Quinson
1996 Version 0.6.2 (protocol not changed; API changed)
1997 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1998 gras_error_t), and NULL when not found. Functions expecting a msgtype
1999 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2000 hopefully usefull message.
2001 - Portability to prehistoric sparcs again
2003 2004-08-17 Martin Quinson
2004 Version 0.6.1 (protocol not changed; ABI not changed)
2005 - prealloc some buffers to speed things up
2007 2004-08-11 Martin Quinson
2008 Version 0.6 (protocol not changed; ABI expended)
2009 - The parsing macro can deal with the references, provided that you add
2010 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2012 2004-08-09 Martin Quinson
2013 Version 0.5 (protocol not changed; ABI changed)
2014 - Allow to off turn the cycle detection code in data exchange at
2015 compilation time. It should be at run time, but I'm short of time (and
2016 the config stuff is still broken). That way, we keep dict out of the
2017 critical path, which is good because the performance is poor:
2018 - search not dichotomial yet
2019 - dynar give no way to access their content and memcpy everytime
2020 - In composed data description (struct, ref and so on), stop foolness of
2021 keeping the subtype's ID, but store the type itself. This keeps sets out
2022 of the critical path, which is good since they rely on dynar and
2023 dictionnaries. The only loose of that is that we cannot detect the
2024 redeclaration of a structure/union with another content (but I'm not sure
2025 the code detected well this error before anyway). We still can detect
2026 the redefinition discrepancy for the other types.
2027 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2028 it breaks the code because of type-punning used all over the place).
2029 This breaks on all non-gcc architectures (for now).
2031 All those changes (plus the buffer of last time) allow me to gain 2 order
2032 of magnitude on cruel tests consisting of 800000 array of integers on two
2033 level of a hierarchical structure (200 secondes -> 4 secondes)
2036 - the selector of reference must now return the type it points to, not
2037 the ID of this type.
2039 2004-08-06 Martin Quinson
2040 Version 0.4 (protocol changed; ABI not changed)
2041 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2042 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2043 - Implement a buffer transport, and use it by default (it relies on tcp in
2044 real life and on sg in simulation).
2045 That's a bit hackish since I had a new field to the structure to store
2046 its data without interfering with the subtype ones. Inheritance
2047 is tricky in C. And that's a kind of reverse inheritance with one class
2048 derivating two classes. Or maybe a game with java interfaces. Anyway,
2049 that's damn hard in C (at least).
2050 Moreover, I got tired while trying to ensure plugin separation and
2051 genericity in SG mode. MSG wants me to do weird things, so let's go for
2052 cruel hacks (temporarily of course ;).
2053 See comment in transport_private.h:71
2054 - do not include all the _interface headers in private but in the files
2055 which really need them (to cut the compilation time when they are
2058 2004-07-26 Martin Quinson
2059 Version 0.3 (protocol not changed; ABI changed)
2060 - Major overhault of the datadesc interface to simplify it:
2061 - shorted the function names:
2062 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2063 - add a trivial way to push/pop integers into the cbps without malloc.
2064 This allows to make really generic sub_type description, which simply
2065 pop their size of the stack.
2066 - add a function gras_datadesc_ref_pop_arr() which does what users want
2067 most of the time: Declare a dynamic array (which pops its size of the
2068 stack) and declare a reference to it. Poor name, but anyway.
2069 - kill the post-send callback, add a post-receive one
2071 2004-07-23 Martin Quinson
2072 Version 0.2 (protocol changed; ABI changed)
2073 - add some testing for cpbs in the test cases, and fix some more bugs.
2074 This invalidate again the little64 data file, since I cannot regenerate
2076 - remove an awfull optimization in the logging stuff, allowing me to:
2077 - understand it again
2078 - learn gcc how to check that the argument match the provided format
2079 - fix all errors revealed by gcc after that
2080 - internal keys of dict are not \0 terminated. Deal with it properly in
2081 loggings instead of segfaulting when the user want to see the logs :-/
2083 2004-07-22 Martin Quinson
2084 - Fix some stupid bug preventing cbps (callback postit) from working
2086 2004-07-21 Martin Quinson
2087 - Some documentation cleanups
2088 - remove the useless last argument of msgtype_declare
2089 - rename the Virtu functions to fit into the 'os' namespace
2090 - move headers src/include -> src/include/gras/ and stop fooling with
2091 gras -> . symbolic link
2092 - make distcheck is now successful
2094 2004-07-19 Martin Quinson
2096 - Build shared library also
2097 - Install html doc to the right location
2098 - stop removing maintainer files in make clean
2099 - build tests only on make check
2101 2004-07-13 Martin Quinson
2103 - No major issue in previous version => change versionning schema
2104 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2105 corresponding dataset.
2107 2004-07-11 Martin Quinson
2109 - Get it working with any kind of structure (we can compute the padding
2110 bytes remotely for all the architectures I have access to)
2111 - Implement the structure parsing macro (still not quite robust/complete)
2112 - Improvement to the remote testing toysuite
2114 2004-07-10 Martin Quinson
2115 [autoconf mechanism]
2116 - get ride of a bunch of deprecated macros
2117 - actually run the test for two-compliment, not only compile it :-/
2118 - test whether the structures get packed (and bail out if yes. Damn.
2119 Alignment is a serious matter)
2120 - test whether the structures get compacted (but respecting the alignment
2121 constraints of each types)
2122 - test whether the array fields of structures can straddle alignment boundaries
2124 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2125 (took me 2 hours to find that bug, looking at the wrong place)
2126 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2127 before sending/receiving and are used to compute the offsets of fields
2128 - Given that padding size depend even on compiler options, keep track of
2129 alignment and aligned_size only for the current architecture. Not a big
2130 deal since we send structure fields one after the other (seems
2132 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2133 program, let it work on linux/gcc/little32. portability todo.
2135 2004-07-08 Martin Quinson
2136 - import and improve remote compilation support from FAST
2137 - make sure make check works on half a dozen of machines out there
2139 2004-07-07 Martin Quinson
2140 Let's say it's version 0.0.3 ;)
2141 - Implement conversions (yuhu!)
2142 - Let it work on solaris (beside conversion, of course)
2143 - Stupid me, using rand() to generate the conversion datatests in not wise.
2145 2004-07-06 Martin Quinson
2146 - Let make dist work, since I'm gonna need it to compile on remote hosts
2147 - Let Tests/datadesc_usage write the architecture on which the file was
2148 generated as first byte.
2149 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2150 architecture descriptions.
2151 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2152 on those architectures.
2153 - Optimization: send/recv array of scalar in one shoot
2155 2004-07-05 Martin Quinson
2156 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2158 - Plug a whole bunch of memleaks
2159 - each process now have to call gras_{init,exit}. One day, their log
2160 settings will be separated
2161 - Continue the code factorisation between SG, RL and common in Transport.
2163 2004-07-04 Martin Quinson
2165 - Redistribution between SG and RL.
2166 We wanna have to accept in SG, so move accepted related parts of RL in
2167 the common part. (more precisely, the dynar of all known sockets is no
2168 more a static in transport.c, but part of the process_data)
2170 [gras_stub_generator]
2171 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2173 2004-07-03 Martin Quinson
2174 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2175 them all up in one shot)
2177 - Fix the ugly case of reference to dynamic array.
2178 - New (semi-public) function gras_datadesc_size to allow the messaging
2179 layer to malloc the needed space for the buffer.
2181 - gras_socket_close now expect the socket to close (and not its address to
2182 put NULL in it after it). This is because the socket passed to handlers
2183 is one of their argument (=> not writable).
2185 - propagate the interface cleanup from last week in datadesc, ie remove a
2186 superfluous level of indirection. User pass adress of variable
2187 containing data (both when sending and receiving), and not of a variable
2188 being a pointer to the data. Let's say that I like it better ;)
2189 The price for that is constructs like "int msg=*(int*)payload" in
2190 handlers, but it's a fine price, IMHO.
2192 - Let it work in RL (yuhu)
2194 2004-06-21 Martin Quinson
2196 - porting SG plugin and SG select to new standards (works almost).
2197 - plug memleaks and fix bugs around.
2200 - cleanup the prototype of data recv and force users to specify when they
2201 want to handle references to objects. Test case working even for cycles.
2202 - plug memleaks. Valgrind is perfectly ok with this.
2204 2004-06-12 Martin Quinson
2206 - cleanup the separation between plugin and main code in plugin creation
2208 2004-06-11 Martin Quinson
2210 - Reput hook for raw sockets, needed for BW experiments
2211 - kill a few lines of dead code
2212 [Data description] Interface cleanup
2213 - gras_datadesc_by_name returns the searched type or NULL.
2214 That way, no variable is needed to use a type desc once, which makes
2216 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2217 two parameters were strings (field name, type name), leading to
2219 [Dicos] Interface cleanup
2220 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2221 This is consistant with the dynar API.
2223 2004-04-21 Martin Quinson
2225 - Porting to new standards.
2227 - interface cleanup.
2228 There is no bag anymore, no need to take extra provision to mask the
2229 pointers behind "ID".
2230 Better splitup of functions between files create/exchange/convert.
2231 This is still a bit artificial since convert and receive are so
2232 interleaved, but anyway.
2234 - add a queued message list to procdata (the ones not matching criteria
2236 - factorize some more code between SG and RL wrt procdata
2238 - use gras_exit in example to track memleaks
2239 - get rid of gs_example now that GS is properly integrated into gras
2240 - update run_test to integrate the lastest tests (datadesc)
2242 - rename WARNINGn macros to WARNn since it prooved error-prone
2244 2004-04-19 Martin Quinson
2246 - register init/exit functions within gras module mechanism
2247 - send/receive function.
2248 Convertion is not implemented, but short-cutted if not needed.
2249 struct/array elements are sent one by one (instead of block-wise), but
2250 nobody really cares (yet). Get a prototype before optimizing.
2251 - tests (using a file socket) for DD send/receive on:
2252 - base types: int, float
2253 - array: fixed size, string (ie ref to dynamic string)
2254 - structure: homogeneous, heterogeneous
2255 - chained list, graph with cycle
2256 Believe it or not, valgrind is not too unhappy with the results. The
2257 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2261 - Bugfix when using a filename explicitely (instead of '-')
2263 2004-04-09 Martin Quinson
2265 - factorize more code between RL and SG in socket creation
2266 - Complete the implementation and tests of:
2268 o file (only in RL, and mainly for debugging)
2270 I lost 3 days to design a portable address resolver, and then decided
2271 that the prototype mainly have to run on my box.
2272 Addressing portability too early may be like optimizing too early :-/
2274 - use gras_init in the Tests instead of the crappy parse_log_opt
2275 (the latter function is removed)
2276 [Conditional execution]
2277 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2278 [Code reorganisation]
2279 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2280 Build examples against the RL library, since there is no way to disable
2281 its creation for now.
2283 For information, the beginning of coding on GRAS was back in june
2284 2003. I guess that every line has been rewritten at least twice since