1 SimGrid (3.7.1) NOT RELEASED; urgency=low
4 * Fix weird behaviors when dealing with parallel tasks.
6 Bug fixes, cosmetics and small improvements
7 * Fix compilation when using MSG_USE_DEPRECATED.
10 * Fix compilation on Macs.
12 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
14 SimGrid (3.7) stable; urgency=low
16 The "spring cleanups (before the next Big Project kicks in)" release.
19 * We can specify the SMPI latency/bandwidth factor with command line
20 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
21 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
22 You can also use the "config tag" from platform file by adding this line
23 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
24 (see "example/platforms/tag_config.xml" to use "config tag").
25 Note that the command line supersedes the platform file configuration.
26 * Change the correction factors used in LMM model, according to
27 the latest experiments described in INRIA RR-7821.
28 Accuracy should be improved this way.
29 * Use the partial invalidation optimization by default for the
30 network too. Should produce the exact same results, only faster.
31 * Major cleanup in surf to merge models and split some optimization
32 mechanisms from the core of the models. As a result you can now
33 specify which model to use (e.g., --cfg=network/model:LV08
34 --cfg=cpu/model:Cas01) and which optimization mode to use
35 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
36 Incompatible combinations should err at initialization. See
37 --help-models for the list of all models and optimization modes.
38 * The CLM03 workstation model was dropped for simplicity because it
39 used the deprecated CM02 network model. Use default instead.
40 * Rename the TCP_gamma configuration option to network/TCP_gamma
41 * Rename the coordinates configuration option to
42 network/coordinates, and document it
43 * Use now crosstraffic keyword instead of the terribly misleading
44 fullduplex keyword. It is activated by default now in the current
45 default model, use --cfg=network/crosstraffic:0 to turn it off.
46 * Ongoing refactoring the model parsing to make XML files optional
47 See include/simgrid/platf.h for details (still to be completed)
50 * Major overhaul of the documentation. Almost instructive now :/
51 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
52 functions and friends. This interface was considered as
53 deprecated since over 2 years, it's time to inform our users that it is.
54 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
55 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
56 use these (crappy) functions in your code.
57 These functions will be removed soon. Stop using them now.
58 * Deprecate MSG_get_host_{table,number}
59 Implement MSG_hosts_as_dynar() instead.
60 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
61 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
64 * Stabilize the parallel execution mode of user contexts
65 * Introduce configuration variables to control parallel execution:
66 - contexts/synchro: Synchronization mode to use when running
67 contexts in parallel (either futex, posix or busy_wait)
68 - contexts/parallel_threshold: Minimal number of user contexts
69 that must be part of a scheduling round to switch to parallel
70 execution mode (raw contexts only)
71 * Fix bugs that prevented to use suspend/resume along with
72 synchronization structures.
73 * Fix bugs in process termination that lead to invalid memory access
74 in very specific conditions.
77 * Introduce a parallel mode for the models (controlled by surf/nthreads
78 configuration item). In our tests, running the models in parallel
79 never lead to any speedups because they are so fast that the gain
80 of computing each model in parallel does not amortizes the
81 synchronization costs, even when ultra fast futexes are used.
82 This is released anyway because YMMV.
85 * Performance boost by using a swag internally to compute the set of
86 tasks that are finished and should constitute the return value of
90 * Enable it by default now that it is considered rather stable.
93 * Documentation of the tracing functions.
94 * Performance gains when tracing categorized/uncategorized resource
95 utilization by avoiding calls to get route when updating resource
96 variables. LMM constraints are being used instead.
97 * API changed to set task categories. Use MSG_task_set_category instead
98 of TRACE_msg_set_task_category, and SD_task_set_category instead
99 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
100 * Bugfix for graphicator, routes not correctly obtained, memory leaks
101 * Examples for link user variables added (see at examples/msg/tracing/)
102 * Deprecated function TRACE_msg_set_process_category completely removed
103 * Trace header updated according to the latest Paje file format
104 * Tracing network lazy updates, no longer obligate users to use full updates
105 * --cfg=tracing/platform:1 also registers power/bandwidth variables
106 * Experimental: let user code declare/set/push/pop application states for hosts
107 * API changed to allow the manual creation of graph configuration files
108 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
111 * Improve the API of Lua MSG bindings, using the Lua spirit.
112 * Each simulated process now lives in its own Lua world (globals are
113 automatically duplicated). It helps writing simulators. Will allow
114 to run Splay programs within SimGrid in the future.
115 * Add a Chord example in Lua, equivalent to the MSG one.
118 * Start the implementation of a solution to express temporal
119 properties, not only local assertions. This is still an
120 experimental work in progress, stay clear from it to be safe.
124 - Add new runtime parameters --help-logs and --help-log-categories
125 to display informations about supported logging parameters and
127 - Old deprecated parameters --{gras,surf,msg.simix,xbt}-log=...
128 don't exists anymore.
129 * Mallocators: allow value NULL for the reset function.
131 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
132 create homogeneous dictionaries, where all the elements share the
133 same free function. Non homogeneous dictionaries will be
134 deprecated in the next release.
135 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
136 - Multi-level dictionaries are deprecated.
138 - new function xbt_dynar_search_or_negative() that is useful when
139 you have less than 2 million elements in your dynar and don't
140 want of the extra complexity of catching exceptions when the
141 element is not found.
143 - Make xbt_os_thread module (for thread portability) public.
144 Documentation is still to come, sorry.
146 - Cleanups and simplifications to make it maintainable again.
147 - Exotic features (such as memalign and valloc) were removed.
148 - The metadata were extended and improved so that the
149 model-checker becomes able to explore and inspect the heaps.
150 - This may induce a performance drop when enable_model-checking is
151 ON in cmake (even if it's not used in the simulation), but it is
152 necessary at this point to get MC working.
154 Turn model-checking OFF if simulation performance matters to you.
155 Not enabling it at runtime is not enough, disable it in cmake.
157 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
159 SimGrid (3.6.2) stable; urgency=low
161 The "Not coding new stuff allows to polish old things" release.
164 * New bindings to the NS3 packet level simulator (experimental)
165 * Use the raw (efficient) execution contexts instead of the sysv
166 (portable) ones when possible.
167 * libpcre is now mandatory in any cases since not using it led to
168 severe performance loss and possibly other issues
169 * Update the XML platforms:
170 - G5K: include the latest machine in Nancy
171 - GridPP and LCG: new platforms
172 * Documentation was partially updated, at least (more to come)
174 Bug fixes, cosmetics and small improvements
175 * Free terminated processes before the end of the simulation to avoid
176 exhausting the memory of users having very dynamic amount of
178 * Bug fix and cosmetics about canceling non-running tasks
179 * Bug fix about the dot loader's issues when using libcgraph
182 * Create an installer for windows with nsis (amd64 and win32)
183 - Add an hello world project to illustrate simgrid project creation.
184 - Embed libpcre into the Simgrid installer to avoid
185 its compilation burden
186 * The raw execution contexts should work on Apple now
187 * Port to Windows 64 bits
188 - Sysv contexts now have an implementation for this arch
189 - GRAS communication features now support this arch
190 * Drop support for borland compiler on windows
191 - this code was not maintained, and we kinda depend on gcc nowadays
192 * Fix portability issues on kfreebsd/gnu: build error about semaphores
193 * Fix portability issue on unstable ubuntu: linker became picky on
196 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
198 SimGrid (3.6.1) stable; urgency=low
200 The "Oops, we broke Macs too" release
203 * Fixed contexts detection so that raw ones are used when possible
204 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
205 leads to a strange error, with user code segfaulting sometimes when
206 the generated code is not perfectly aligned (which is not
207 controllable from the user side, depends on the amount of code)
210 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
212 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
214 SimGrid (3.6) stable; urgency=medium
216 The Summer Release, also known as the "OMG! They Killed Kenny!" version
219 * Bindings now constitute their own package, separated from the main one.
220 Rationale: reduce our maintenance nightmare by reducing the module coupling
221 They will soon be released on their own on gforge.
223 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
224 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
226 GRAS: It is not considered as stable anymore, but experimental. Sorry.
227 * It's not quite deprecated for now because we have no replacement,
228 but it may soon become the case.
231 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
232 * Fortran: New user-level cache variable to store the rank of the running
233 process. This improves performance by an order of magnitude.
234 * C: New coccinelle script to automatically locate and modify global and
235 local static variables.
236 * Improved SMPI network model with a sender-side gap to account for multiple
240 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
241 only say if a communication is finished, no matter whether it succeeded or
242 failed. You can call MSG_comm_get_status() to know the status of a finished
244 * New function MSG_task_dsend() to send a task and detach it. When a
245 communication is detached, you are never notified of its success or failure
246 and the memory is released automatically once it is finished. This function
247 is useful when you don't care about the end nor the success of a
249 * Change the prototypes of action replay. Sorry for inconvenience,
250 but this is really more efficient this way (and to adapt your code,
251 you just have to fix the initialization, that shouldn't be too long)
252 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
253 ever managed to do anything useful with it.
254 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
255 takes a dynar of communications. It returns immediately and gives the
256 index of a finished communication (if any).
257 * New example: a basic implementation of the Chord P2P algorithm.
260 * New model for multi-core CPUs. You can now use the core attribute to
261 precise the number of cores of a host. This is a basic model. Every
262 process running on the host receives at most the power provided in
263 the DTD (throughput<=power). Total throughput of process cannot exceed
265 * New peer tag. This peer tag creates a tiny AS comprising a host and a
266 router linked by an up-link and a down-link (possibly asymmetrical).
267 This kind of pattern allows to easily build last-mile model style platforms.
268 Aggregating such patterns in a rule-based AS is thus the technique of
269 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
270 * New model for Vivaldi routing. We transformed the Vivaldi network model
271 into a Vivaldi routing model (based on the rule-based model). This allows to
272 combine Vivaldi based latencies with last-mile platforms.
275 * Added a check for NaN of IEEE754 infinite in the double entries of
277 * Introduce a new context factory "raw", highly inspirated from the
278 ucontext factory, but using manually crafted functions in assembly to
279 do the work in an efficient manner.
280 * Allow to change the used context factory at run time, not only at
281 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
282 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
283 contexts at runtime (only with raw contexts or ucontexts).
284 * Completely rewrote this module to allow parallel execution of user
285 processes. Use --cfg=contexts/nthreads:N to execute user processes
286 with N parallel threads (the default is 1, meaning no parallelism).
287 * Allow to decide dynamically between sequential and parallel modes.
288 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
289 processes in parallel only when their number is greater than or equal to P
291 * Added a check for NaN of IEEE754 infinite in the double entries of
295 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
296 won't output the state of processes when interrupted with Ctrl-C
297 * Add a new function xbt_dynar_to_array that transforms a dynar into a
298 NULL-terminated array. This may solve backward compatibility issues
299 due to the change to return type of SD_simulate. See also:
300 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
301 * Add new macros with variable number of arguments.
302 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
303 - in xbt/asserts.h: xbt_assert
304 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
305 - in xbt/ex.h: THROWF and RETHROWF.
306 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
308 * Change xbt_die() to accept a format string with arguments, just like printf.
309 * New data structure: xbt_lib_t, like a dict but more general and with better
313 * New configuration options
314 Options triva/categorized and triva/uncategorized can be used to generate
315 graph configuration files for Triva visualization tool.
316 * Configuration option tracing/platform is renamed to tracing/categorized
317 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
318 * New TRACE_user_link_variable interface:
319 User provides the name of the link and the tracing variable to attach to it
320 * the declaration of tracing categories must be done after the environment
322 * simpler tracing interface, just one way to declare categories
323 TRACE_category or TRACE_category_with_color, it is up to you
324 * links in the trace file are again identified by their names
325 * trace contains the full platform hierarchy exactly as declared using the ASes
326 * Options tracing/msg/[task|process]:1 groups the process by hosts
327 for both cases, tasks and processes must have names that are unique during the simulation
328 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
329 * The experimental option tracing/msg/volume is deprecated
330 its functionality may be reincorporated if needed
332 The tracing generates a trace file with unordered timestamped events,
333 because of the way the core simulator (surf) works. A script available
334 at the tools directory (fix-paje-trace.sh) can be used to put the events
335 in order. We have changed the tracing so it can generate ordered timestamped
336 events in the final trace, but depending on the simulator (and how much time
337 is simulated) that can lead to a huge memory utilization. It is deactivated
338 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
341 * Define a SIMGRID_VERSION macro in simgrid_config.h.
342 - We are trying hard to keep the API stable, but it may happen that
343 some things change (we're a research project after all, not a
344 nuclear plant operating system). If such things should happen, you
345 could rely on that macro to adapt.
346 - current value: 30600 for 3.06.00, aka 3.6
347 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
348 a number that can be compared with SIMGRID_VERSION.
349 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
350 assertions and verbose/debug logging events are disabled at compile time.
352 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
354 SimGrid (3.5) stable; urgency=medium
357 * New feature to any SimGrid-based simulator: Model-Checking
358 Check SIN#1 for more details.
361 * New Model SMPI (three-interval linear regression for correction factors)
362 See RR-7426, available at http://hal.inria.fr/inria-00527150
363 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
364 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
365 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
366 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
367 * See also src/smpi/README
372 - Tracing API changes: TRACE_start and TRACE_end should not be called
373 by user-code. They are automatically called by simulators created
374 with SimDAG, MSG and SMPI if the toolkit is compiled with
375 tracing_enabled=ON. Categories declaration and utilization remain the
376 same for MSG and SimDag.
377 - A function was added to the tracing API to declare categories with
379 - TRACE_category_with_color (char *category, char *color)
380 where color must be in the following format
381 "%f %f %f", red, green, blue
382 and red, green, blue are float values in the interval [0, 1]
383 - User can specify NULL as color parameter, or continue calling
385 On that case, the tracing system will define random colors
386 - The following command-line options are supported:
387 --cfg=tracing/filename:msg.trace
388 --cfg=tracing:1 (activate tracing, needed to use others)
389 --cfg=tracing/platform:1 (categorized resource use)
390 --cfg=tracing/uncategorized:1 (uncategorized resource use)
391 --cfg=tracing/msg/task:1 (task creation)
392 --cfg=tracing/msg/process:1 (process creation, migration)
393 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
394 --cfg=tracing/smpi:1 (SMPI interface tracing)
395 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
396 - examples of examples/msg/tracing updated
398 - DAXLoader and DOTLoader functions can generate tasks with categories
399 - A new function to attribute a category to SD tasks:
400 TRACE_sd_set_task_category (SD_task_t task, char *category)
401 * Tracing the MPI interface implemented by SMPI
402 - Collective operations are traced with states
403 - Point-to-Point operations are traced with states/links
404 - Tracing activated by a parameter "-trace filename" passed
405 to smpirun during execution (considering that simgrid
406 is compiled with tracing enabled)
407 - To run the simulation with gdb, the simulator
408 accepts --cfg=tracing/smpi:1 to trace SMPI
409 - tesh files to check if smpi tracing is ok
410 - See examples/smpi/NAS/DT-trace
411 * GTNetS tracing re-worked
412 - adaptation to the tracing system of GTNets to cope
413 with modifications regarding the fullduplex mode
414 - new tesh files to check if gtnets tracing is ok
417 * Asynchronous communications through the functions:
418 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
419 * New function: MSG_load_platform_script()
420 to make possible using a lua script instead of XML files to set up platforms
421 * New function: MSG_set_function
422 to associate functions to processes, used when bypassing the parser
423 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
425 Platforms: Add some more examples in examples/platforms
426 * Grid'5000: see www.grid5000.fr
427 * *_30000_hosts.xml: various huge files [mainly scalability testing]
430 * Change the XML format. This is a very important modification. SimGrid
431 3.5 introduces a new hierarchical format based on the notion of
432 Autonomous Systems. Compatibility with old format is ensured
433 through the perl script provided in the install bin directory
434 bin/simgrid_update_xml.
435 It is now possible to build platforms with specific routing
436 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
437 connect several platforms together. We will try to provide soon
438 set of realistic platforms exploiting these properties (have a
439 look at examples/platforms/ for the moment).
440 * Take the opportunity of the XML format change to be a good XML citizen:
441 rename link:ctn to link_ctn and similar changes (also dealt with by
443 * Add a new routing scheme (rule-based) using regular expressions. It
444 enables to have an extremely low memory footprint when the
445 underlying routing is simple and can be compactly described. You
446 need to have libpcre4-dev (perl regular expressions) installed if
447 you want to use this routing scheme.
448 * Revive the cluster TAG and allow to easily and efficiently (both in
449 term of memory and speed) connect clusters together. Have a look
450 at teshsuite/simdag/platforms/ to see how this can be done. With
451 this tag, you can create clusters with thousands of tasks at no
452 cost (have a look at examples/platforms/).
453 Note: clusters are implemented as ASes, so there is no need for an
454 enclosing AS tag if you have only one cluster in your platform.
455 * Add new generic functions in the public interface that allows the user
456 to call SURF 'create_resource' methods from your code (same
457 functionality as the XML bypass mechanism but with a much lighter
459 * Add a new model (enabled through command line --cfg=network/model:SMPI)
460 that uses a piecewise linear approximation to produce better
461 results when exchanging small messages.
462 * Add a new parameter to handle correctly full duplex link and account
463 for interferences between uplink and downlink communications
464 (activate with --cfg=fullduplex:1).
467 * Rename the SD_READY (all dependencies are satisfied and task is
468 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
469 dependencies are satisfied) state.
470 This prevents a confusion between the notion of "ready to schedule"
471 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
472 simulated" (SD_RUNNABLE) used by the simulation kernel.
473 * Change the way a task is considered as ready. Instead of removing
474 dependencies when a task is done, a counter is decreased. This way,
475 it is always possible to reach ancestors thanks to the
476 SD_taks_get_parents function (even after the end of the simulation.)
477 * Change the return type of SD_Simulate from (SD_task_t*) into
478 xbt_dynar_t. This function was in handling a dynar internally and
479 converted it into a NULL terminated array for historical reasons.
480 * New function SD_dotload(char*) to load a DAG described in dot
481 format. This loader and the corresponding examples require the
482 installation of the graphviz library.
483 * Fix a bug in the management of tasks of size 0 in the surf network
484 models. This problem was only visible with SIMDAG and you should
485 thus disregard results produced with earlier versions if you
486 relied on this feature (some tasks were blocked because of this).
487 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
488 even though your applications were doing only point-to-point
489 communications and sequential computations. Now you can really use any
490 model you want (of course, if you create real parallel tasks, which are
491 not implemented in most models beside ptaskL07, this will abort).
492 * Add an example that schedules a DAX on an heterogeneous platform
493 using a Min-Min strategy.
494 * New function SD_workstation_get_current_task() that returns the kind
495 of task currently running on a workstation in the sequential access
497 * Raise some warnings when unexecuted tasks remains at the end of the
498 simulation. This is usually caused by cycles in the DAG.
501 * New function: SIMIX_process_set_function() called by MSG_set_function
502 * Change the underlying waiting queue in semaphores so that a process
503 can wait on several of them simultaneously (as in waitany).
504 * Fix the way to handle tokens in semaphores so that all access patterns
505 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
506 * kill the dirty pimple SIMIX_message_sizes_output()
507 Please use (proper) visualization instead
510 * New data container: setset (set of sets of elements)
511 * New module: mmalloc (mapped malloc, allowing to have several
512 independent segments of malloc)
513 * New function: xbt_dict_cursor_set_data()
514 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
515 * New function: xbt_dynar_is_empty()
516 * New function: xbt_fifo_get_last_item()
517 * Fix xbt_dynar_shrink(): use the right element size.
518 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
519 uninitialized areas during expand.
520 * Fix semaphores: previous implementation was severely broken.
521 * Use library init/fini functions for our initialization.
522 - you can use logs and other feature as soon as you want in your
523 code (even before the xbt_init / MSG_init)
524 - xbt_exit is now a no-op and produce a warning when used.
527 * Port GRAS to new SIMIX mechanisms. This allows gras users to
528 benefit from the latest improvement to the simulation kernel.
529 * Kill measurement sockets for now. If you rely on them, sorry. This
530 release is not for you. This feature will be reintroduced in the
531 future, but we cannot delay the release any further.
532 * New function: gras_msgtype_get_name().
533 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
534 * Fix (at last) the pmm example: it should not randomly fail anymore.
536 Build chain: bug fixes and overall polishing
537 * Cmake is now stable enough. Hence, we killed the autotools.
538 * Port to windows ( TM :)
539 * Fix the 'make install' target.
540 No need to use 'make install-simgrid' anymore
541 * Introduce a 'make dist' target compiling a *source* archive
542 'make package' compiles a binary archive
543 * Compile java files only on need
544 * Add --cd and --setenv command line options to tesh
545 * Out of source builds are not fully supported yet, but we are close
546 * Enable supernovae and optimization flags by default for our users
549 * Add layer to set up environment directly from lua, without XML.
550 * The effect of gras_stub_generator can be achieved through
551 lua too (check examples/gras/console/ping_generator.lua)
553 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
555 SimGrid (3.4.1) stable; urgency=low
557 The "Polishing easter eggs is probably a good idea" release.
558 This is a bug fixes release only.
562 * Fix a bug preventing the tasks from begin garbage collected.
565 * Fix a bug occuring when a host involved in a communication fails.
566 This was not detected properly by the other peer involved in the
567 communication. Now, it's reported as a network error.
570 * Warn the user about loop dependencies in data flow of DAX files
571 * Obey the control-flow dependencies of DAX files
574 * Add option "enable_smpi" allowing to not compile SMPI.
575 Probably useful for the (Mac) users experiencing a build error here
576 * Improve the detection of lua5.1 and ruby1.8
578 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
580 SimGrid (3.4) stable; urgency=low
582 The "Easter in Cargese" release. Also known as (major changes):
584 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
585 ~> bindings were greatly improved
586 ~> new tracing infrastructure for better visualization introduced
588 * the "Welcome to configury modernity" release.
589 ~> we switched from autotools to cmake, and improved our cdash
592 A more detailled list of changes follow (full detail in svn log).
594 Java Bindings: Various Cleanups
595 * (install java-gcj-compat-dev on debian-like to use them)
596 * Remove put/get: no need to export deprecated interface in Java
597 Use send/receive instead.
598 * Cleanup the examples and add a README per directory
599 * Remove example autoDestination (that's the only way to go now)
600 * Remove example explicitDestination (was a plain copy of basic)
601 * Make JniException a runtime exception, so that there is no need to
602 declare the fact that you may encounter such a beast. I guess that
603 nobody will ever want to survive such error.
604 * Create specific errors for each MSG case of failure:
605 host failure, transfer failure, timeout, task cancelled
606 * Cleanup the exceptions that may get thrown by each function
607 * Other internal cleanups in Java bindings. Performance still bad :/
608 Ruby and Lua Bindings: create them
609 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
610 * That's new and great, you should try them out.
611 Same functionalities than Java bindings, only even less polished
613 * Kill the useless "rate" argument of SD_task_get_execution_time()
614 Everyone used to provide -1 as a value, it was not used, and the
615 semantic of a possible use wasn't even clear.
616 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
617 either as comm costs or compute costs to mean that there is no
618 such thing for that specific task.
619 * Add a SD_task_set_name() function
620 * Fix SD_task_unschedule() on typed tasks
621 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
622 * In DAX loader, accept useless 'level' attributes to <job> since
623 LIGO DAGs have them (seem to be to ease graphical representation).
625 * Add an example masterslave_mailbox.c using send/receive and not
626 the deprecated put/get interface.
627 * Kill the MSG_paje_output() function. It's a noop since 2 years.
628 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
630 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
631 (declare MSG_USE_DEPRECATED to still have the old name)
632 * Add a MSG_task_set_data() function
633 * About trace replay (see examples/msg/actions):
635 - Allow to work with splitted trace files for each process
636 Give the specific trace file as argument of each process,
637 and call MSG_action_trace_run(NULL)
638 You can still have one merged file for all processes.
639 - Fix implementation of collective operations
640 * Allow task_execute() on 0-sized tasks (closes #10063)
642 * This is the first release of SimGrid where SMPI is not considered
643 beta anymore (even if some corners should still be improved)
644 * Port over the new SIMIX_network submodule (internal refactoring)
645 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
646 * Implement more missing elements of the standard:
648 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
650 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
651 MPI_Scan() MPI_Get_processor_name()
652 - Added implementation of missing case for Alltoall (warning: it's
653 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
654 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
655 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
656 SMPI_MPI_Allgatherv()
658 - MPI_Waitsome() was broken
659 - Allow relative includes in smpicc
660 - Command line cfg argument 'reference_speed' was ignored...
661 - Some functions did not properly lead to auto-benching of user code
662 - smpicc passes -O2 by default (just like openmpi one)
664 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
665 * Bug fixes about timeouts during communications
666 * add SIMIX_message_sizes_output() as a pimple to write to file the
667 amount of messages per size. Use gnuplot to get histogram.
668 Pimple because that's the only user-visible function of simix,
669 defined directly in xbt.h (irk, sorry)
671 - Add a SIMIX_sem_get_capacity() function
672 - Fix interactions with processe resume/suspende
673 - release_forever() was stupidly broken
674 - Fix SIMIX_display_process_status() for processes in a semaphore
675 - Make SIMIX_sem_block_onto() user-visible
676 * Refactoring context stuff:
677 - Use pseudo-OOP for better modularity
678 - reimplement SIMIX_process_kill() without process_schedule() so
679 that the latter can take as invariant that it is called from
681 - Merge context_start into context_new for sake of simplicity
683 * Add a Vivaldi network model, coded live during SUD'10 ;)
684 * Rename configuration variables to start a hierarchy:
685 o cpu_model -> cpu/model
686 o network_model -> network/model
687 o workstation_model -> workstation/model
688 * New configuration variables:
689 o network/bandwidth_factor: correction to bandwith
690 o network/latency_factor: correction to latency
691 o netwotk/weight_S: correction to the weight of competing streams
692 * Add a long description to the models, that users can see with such
693 argument on the command line: --cfg=cpu/model:help
694 * --help-models display the long description of all known models
696 * config: add the ability to set a default value after registration
697 Does not override any previously set value (e.g. from cmd line)
698 * dict: allow to have integer key and data.
699 When so, you need to use the following functions
700 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
701 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
702 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
703 In contrary to regular dicts, the key is not malloced before copy.
704 Mixing scalar and regular elements in the same dict is not tested
706 * Allow to use xbt_dynar_shrink() to expend the dynar instead
707 Tracing for Visualization:
708 * SimGrid is now instrumented in order to generate a trace file for
709 visualization analysis: to use it, need to compile SimGrid with the
710 "tracing" option enabled, and instrument the program using SimGrid with
711 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
712 (among other functions).
713 * The instrumentation only traces the platform utilization for now
714 * Documentation to use the tracing functions and how to analyze the
715 traces with the Triva tool is written.
716 * More information about: SimGrid FAQ (in the section Tracing Simulations
719 * We moved to cmake as default build system. Autotools support will
720 be dropped soon. Check the FAQ for more info about how to use it.
721 * Greatly improved our cdash/ctest interactions
722 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
723 * Added memory checking tests with valgrind; lot of memleak fixing.
724 This may be the first release of simgrid with so few memory issues
725 * Added code coverage tests.
726 Our coverage is still improvable, but at least we see it on cdash.
728 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
730 SimGrid (3.3.4) stable; urgency=low
732 The "Desktop Grid needs love too" release (also called Xmas release).
735 * Major speedup in the maxmin system solving by using lazy evaluation
736 Instead of solving completely the maxmin system at each iteration,
737 only invalidate (and recompute) the modified parts.
738 This new feature is enabled in default models but you can try to
739 turn it on with "--cfg:maxmin-selective-update=1" for other models.
740 * Cas01 IMproved as default CPU model
741 This CPU model is the same Cas01 model, but it uses the
742 maxmin-selective-update flag and a heap structure to manage
743 actions on SURF kernel.
744 It reduces the complexity to find the next action to finish and,
745 consequently, it's faster than the old Cas01.
746 This is the new default CPU model (Cas01).
747 * Rename the old Cas01 model to Cas01_fullupdate
748 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
749 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
750 * CpuTI (CPU Trace Integration)
751 A new CPU model whose objective is simulate faster when using
752 availability trace files.
753 Instead of using a full featured, over engineered maxmin system for
754 CPU modeling, this model does the pre-integration of traces files
755 to calculate the amount of CPU power available, and so, executes
756 faster than the old CPU models.
757 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
758 * Use LV08 as default network model since it gives better accuracy
759 for small messages and shouldn't change things for big ones.
760 Use --cfg=network_model:CM02 to get the previous behavior.
763 ******************************************
764 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
765 ******************************************
766 * The new CPU model may changes simulations!
767 The point is that events occurring at the exact same timestamp
768 are not scheduled in the same order with the old and new
769 version. This may be enough to completely change the execution
770 of simulations in some cases.
771 * The new network model will change simulations!
772 This new model is more realistic than the previous one, so you
773 should consider redoing your old experiments with this model.
774 Sorry for the inconvenience.
777 * Introduce the supernovae compilation mode
778 When compiled that way, the whole SimGrid (or almost) is put in a
779 single compilation unit and compiled in one shoot.
780 This is to help gcc which has difficulties to inline stuff from one
782 The speedup seem to be above 15%, althrough more tests are needed on
783 amd64 to confirm that gain.
786 * Port of MSG's mailbox on top of SIMIX network
787 The put/get mechanism was greatly simplified on the way.
790 * New SIMIX network module. Provides:
791 - Mailbox: rendez-vous mecanism to find with who you want to speak
792 - Synchronous send/recv: easier and hopefully faster since the
793 logic is handled in the maestro process directly now
794 - Asynchronous send/recv: you dreamt of it? It's here now
795 Too bad that nobody cared enough to propagate the change to MSG.
796 * Add semaphores as SIMIX synchronization mechanism.
799 * new function SD_daxload(char*) to load a DAX file
800 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
801 * Introduce typed tasks. Specify its kind and cost at creation.
802 At scheduling, just give where it should be placed, and the cost
803 for each involved resource is automatically computed.
804 Existing constructors so far (more to come of course):
805 - SD_task_create_comm_e2e() for end-to-end communication
806 - SD_task_create_comp_seq() for sequential computation
807 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
808 * new function SD_task_dump() for debuging display
809 * new function SD_task_dotty(task,FILE*) writing to file the info
810 about the task in dotty format
811 * SD_task_dependency_exists() can now cope with having one of its
812 arguments NULL. If so, it tests whether the other argument has any
814 * Add getters on list of preceding/following tasks:
815 SD_task_get_parents(task) and SD_task_get_children(task)
816 * Add getters on amount of workstations and list:
817 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
818 * Add getter on task kind: SD_task_get_kind(task)
819 * Update the start_time and finish_time of tasks on completion/failure
820 * Bugfix: Remove task from state swags when destroyed
823 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
826 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
827 dynar with one separate thread per value of the dynar.
828 * Change the prototype of xbt_thread_create(), sorry.
829 Added a boolean parameter indicating whether we want to join this
830 thread (used in SG only for now)
831 * Implement xbt_thread_join and xbt_thread_yield in SG also.
834 * GTNetS wrappers should now be usable again (and betterly tested too)
835 * Fix a major regression from 3.2 where the timeout provided to
836 MSG_task_put_with_timeout() was used as absolute time before which
837 the comm should be done.
838 * Start to fix the <cluster> tag.
839 - Internal links should be good now (beside of the loopback, which
840 use the private link instead)
841 - paths to the external world is still rather broken
842 - the <route:multi> tag is just broken. Actually that's brain-dead.
843 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
844 to make it less stupid
845 ** Check your platform with teshsuite/simdag/platforms/flatifier **
846 * Fix a source-level compatibility glitch from 3.2: after defining
847 MSG_USE_DEPRECATED, you can use the old name
848 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
849 * Allow to compile from the SVN with automake 1.11
850 * Fix some problems when using the "start_time" tag in deployment XMLs.
851 * Fix #8569: XBT/synchro.h has redundant declarations
852 * Fix #8563: MSG return values and exceptions
853 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
854 * Integrate patch #8636: Obey DESTDIR when installing documentation.
855 Thanks to Robson Peixoto.
856 * Fix a vicious bug in dictionaries inducing that some elements were
857 not freed on xbt_dict_free()
859 Portability report of this version:
860 * Main portability targets:
861 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
863 - mac leopard on i386
864 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
865 but nothing critical.
866 * Other platforms: windows, AIX and others were not tested for this release
868 Timing report of this version:
869 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
870 scenario parameters). From 8h to a few seconds in desktop grid settings.
871 * Supernovae brings about 25% speedup on i386.
873 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
875 SimGrid (3.3.3) stable; urgency=low
877 The "Need for Speed" release.
879 The timings done to validate the 3.3.2 were faulty.
880 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
882 The problem was a conversion from a manually handled vector to
883 xbt_dynar_t on the critical path.
884 xbt_dynar_foreach calls functions, inducing stack management crap.
886 We inlined these functions and xbt_dynar_foreach is now breath taking.
887 We also inlined xbt_swag_belong on the way.
889 Here are some approximate speedup measurements (on master/slaves
890 simulations lasting between 10s and 20s each):
891 3.3.1 -> 3.3.2: about same performance
892 3.3.2 -> 3.3.3: 40% speedup
893 3.3.1 -> 3.3.3: 40% speedup
894 3.3.1 with inline patch -> 3.3.3: 30% speedup
896 Our reading is that the refactoring which occurred in 3.3.2 made us
897 suffer much more from the xbt_dynar_foreach low performance, but
898 once we solved this, this refactoring proved to be very performance
899 effective. From the 40% speedup, somehow, 10% are due to the
900 inlining and 30% to the refactoring.
902 That's a pitty that gcc cannot inline functions placed in other files
903 alone. We have to choose between:
904 - break the encapsulation (by putting private data structures and
905 accessors in headers files to help gcc)
906 - live with low performance
907 - switch to a decent compiler such as icc (not quite possible).
909 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
911 SimGrid (3.3.2) stable; urgency=low
913 The "Simplicity does not preceed complexity, but follows it" release.
915 The main contributors of this release were (lexical order):
916 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
919 * Extract the routing logic into its own object.
920 (was dupplicated in network.c and workstation_LV07.c;
921 Allows to implement other ways of storing that info)
922 => kill now useless network_card concept
923 - Use dynar to represent routes (instead of void** + int*)
924 - kill link_set (use surf_network_model->resource_set instead)
925 - Add a command-line option to choose the routing schema to use
926 - Add three new models:
927 * Floyd (shortest path computed at initialization)
928 * Dijikstra (shortest path recomputed all the time)
929 * Cached Dijikstra (shortest path computed on need)
930 All these models where contributed by Silas De Munck, and are
931 described in his ICCS09 paper.
933 * Simplify model declaration
934 (less redirections, less function to write when defining a model)
935 - Factorize stuff between models:
938 surf_model_resource_set(model)
939 surf_model_resource_by_name(model, name)
940 - Unify the types of models in s_surf_model_t (using an union)
941 - Embeed fields of common_public directly into s_surf_model_t
942 - Rename model methods:
943 action_free ~> action_unref
944 action_change_state ~> action_state_set
945 action_get_state ~> action_state_get
946 - Change model methods into functions :
947 (model)->common_public->action_use ~> surf_action_ref
949 * Implement a generic resource; use it as ancestor to specific ones
950 (allows to kill duplicated code in models)
951 Drawback: timer command don't need no name nor properties;
952 workstation_CLM03 don't need no properties
953 (but I guess we can live with those few bytes wasted)
955 * Improve the action object model
956 - implement a constructor avoiding dupplicated code about field
957 initialization in generic_action part.
959 * Kill the SDP model: it has an external dependency, is deprecated
960 in flavor of modern lmm models, and didn't compile since a while
963 * Relocation of the context module from XBT to SIMIX.
964 (the context were decoupled from the simix processes, duplicating a lot of code)
965 => a lot of code was factorized
966 - less overhead is introduced during scheduling
967 - simpler API for the context factory
968 - the logic for process creation,destruction and manipulation was simplified
969 * Simplification of the s_smx_process_t data structure.
970 => accesing the simix level data associated to a process is faster now,
971 and the code is a lot more readable.
974 * Implement some more MPI primitives:
975 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
976 -implementation: Bcast: flat or 2-ary tree (default),
979 Allreduce: Reduce then Bcast
980 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
981 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
982 Alltoallv: flat tree, like ompi
984 * Add support for optimized collectives (Bcast is now binomial by default)
985 * Port smpirun and smpicc to OS X
988 * Kill SD_link_get_properties: hard to maintain and makes very little sense
989 Shout out if you used it.
992 * Display the list of still queued messages in SG mode when existing
996 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
997 * Add xbt_graph_node_get_outedges() [Silas De Munck]
998 * Add xbt_str_from_file(FILE*)
999 * Add xbt_dict_get_key achieving a linear reverse search
1000 * Remove the context module
1002 Portability report of this version:
1003 * Main portability targets:
1004 - Linux(debian)/x86/context
1005 - Linux(debian)/x86/pthreads
1006 - Linux(debian)/amd64/context
1007 - Linux(debian)/amd64/pthreads
1008 On these, we still have the eratic breakages of gras/pmm and
1009 amok/saturate_sg reported in previous version. We still think
1010 that the tests are the cause of the fault, not the tested code.
1012 - Mac OSX Leopard/x86/context
1013 Still false negative in tesh autotesting.
1014 Smpi still fails, but this time because readlink does not accept -f
1015 Everything seems to work properly beside of that.
1018 - AIX version 5.3 (only tested contexts this time)
1019 Smpi still fails there because mktemp is not installed.
1020 Everything seems to work properly beside of that.
1022 I managed to compile it for the first time, but several breakages.
1023 Won't delay the release for this exotic platform.
1025 * Windows: it's still lagging behind. If you want to help, please
1028 Timing report of this version:
1029 This version seem to be more than 5% faster than 3.3.1 (on linux
1030 64bits with contextes). The gain is less than expected, we are
1031 investigating this for next release.
1033 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1035 SimGrid (3.3.1) stable; urgency=low
1038 * Implement a --cfg-help to show existing configuration variables
1039 * Build chain do not require doxygen in maintainer mode
1042 * fix a bug on struct sizeof computation, which prevented the
1043 exchange of arrays of structs in some conditions
1044 - added a regression test about this in datadesc_usage
1045 * Allow the exchange of 0-long dynamic vectors.
1046 - for that, use -1 as indicator of dynamic size instead of 0
1047 - This implied to change any size from unsigned long to long,
1048 reducing a bit communication abilities, but I guess that with
1049 64bits being quite common, this is more than enough.
1050 - This also induce a protocol change, thus bumping network protocol
1051 version from 0 to 1 (if we have external users, we have to get
1052 clean on that point too ;)
1053 - added two regression tests about this in datadesc_usage
1054 * Be more verbose when propagating local exceptions
1055 This helps debugging.
1056 * Display the status of simulated processes when receiving SIGINT in
1060 * Allow to control the simulation from a trace file.
1061 New functions MSG_action_register() and MSG_action_trace_run()
1062 The first one allows to associate a function execution to each
1063 kind of action while the second one parses a trace file and
1064 triggers the corresponding actions within the system.
1065 For now, only a toy example is provided in examples/msg/actions
1066 * Add an exemple of process migration in examples/msg/migration
1067 * Fix a bug in task exchange which broke MSG_task_get_sender()
1068 Add a teshsuite regression test for that.
1069 [Bug: if MSG_task_get_sender() is called after sender exit,
1071 * Fix a bug which prevented suspend/resume to work properly
1072 * Display the status of simulated processes when receiving SIGINT
1073 This fixes a regression of v3.3. due to the introduction of SIMIX
1074 * Bug fixing in failure management:
1075 - trace could not start by a failure at time 0
1076 - failure during communications were not working
1079 * Add SIMIX_process_set_name() to change the name of the current
1080 process in the log messages.
1081 * Store smx_hosts in a dict since we only retrieve them by name
1082 * Move the configuration infrastructure to surf
1085 * Move the configuration infrastructure to surf
1088 * Massive internal cleanups:
1089 - Store internal structures on processes instead of hosts (allows
1090 to have more than one process per host, in addition of being more
1092 - Cleanup the initialization/finalization process
1093 - Kill a whole bunch of unneeded synchronization:
1094 processes run in exclusive manner within the simulator
1095 - Move queues from global tables to process data fields
1097 - now accept -platform and -hostfile arguments
1098 - Pass the right rank value to processes according to the hostfile
1099 * Compile the examples by default, and use them as regression tests
1100 * Implement MPI_Wtime()
1101 * Change the reference speed to a command line option
1104 * TCP_gamma can now be specified as command line option using
1105 --cfg=TCP_gamma:10000000.0
1106 * Change the --surf-path cmd line option into --cfg=path:
1109 * Also include strbuff from xbt.h public header
1110 * xbt_ex_display(): do not free the exception after displaying
1111 This allows to do more with the given exception afterward.
1112 Users should call xbt_ex_free() themselves.
1116 Portability report of this version:
1117 * Main portability targets:
1118 - Linux(debian)/x86/context
1119 - Linux(debian)/x86/pthreads
1120 - Linux(debian)/amd64/context
1121 - Linux(debian)/amd64/pthreads
1122 These targets fail about 1/10 of times on gras/pmm, but we believe
1123 that this is because of the test, not because of simgrid.
1124 amok/saturate_sg fails even more rarely, and the test may not be
1127 - Mac OSX Leopard/x86/context
1128 The test suite still spits tons of errors because some obscure
1129 force prevents us from removing the temporary directories
1130 arguing that they still contain some metadata I've never heard of.
1131 Smpi fails because seq is not installed.
1132 Everything seems to work properly beside of that.
1135 - AIX version 5.3 (both contexts and pthread)
1136 Smpi still fails there because mktemp is not installed.
1137 XML inclusions seems rosty on AIX.
1139 * Windows: it's still lagging behind. If you want to help, please
1142 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1144 SimGrid (3.3) stable; urgency=high
1148 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1149 [Malek Cherier & Mt]
1151 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1154 Until now, GRAS were using MSG as an interface to SURF. It was
1155 quite difficult because both interface have several differences
1156 (MSG channels vs GRAS sockets were the most notable point).
1158 This also opens the gate to SMPI (which should occur soon) and speed
1159 up simulations by to 40% (even if it were not the main goal).
1161 **************************************
1162 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1163 **************************************
1164 The point is that events occuring at the exact same timestamp are
1165 not scheduled in the same order with the old and new version. This
1166 may be enough to completely change the execution of simulations in
1167 some cases. Sorry for the inconvenience.
1169 * Cleanup and upgrade the XML format to push further scalability
1170 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1172 * Improve the testing infrastructure with tesh. Now a very large part of
1173 the code is tested not only by being run but also by checking that the
1174 output match an expected output [Mt].
1176 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1177 is really less memory-demanding, which should allow you to use
1178 larger files in SimGrid [AL].
1180 * Inform valgrind about our contextes, so that it becomes usable
1181 with the default (and more effecient) version of SimGrid
1182 [contributed by Sékou Diakite, many thanks]
1185 * Introduce a listener thread in charge of receiving incomming
1186 messages from the network. It allows to overlap communication and
1187 computation but most notably, it removes some stupid deadlocks due
1188 to the fact that so far, a process could not send and receive at
1189 the same time. This made most non trivial communication schema
1191 * Convert the PIDs from long int to int to match the MSG ones (and
1192 linux ones too) [Mt]
1193 * New function: gras_agent_spawn() to launch a new process on
1194 current host. Only working in simulation for now. [Mt]
1195 * New function: gras_os_hostport() returning a constant form (ie,
1196 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1199 * Make the backtrace of exceptions more human readable [Mt]
1200 * New module: xbt/str [Mt]
1201 a ton of string utility functions (split, join, printf to a newly
1202 allocated buffer, trim, etc)
1203 * New module: xbt/hash [Mt]
1204 SHA1 hashing algorithm (more to come if needed)
1205 * New module: xbt/synchro [Mt]
1206 synchronization tools (mutex and conditions) working the same way
1207 in simulation and in real life (mainly useful for GRAS, but not
1209 * New module: xbt/queue [Mt]
1210 classical producer/consumer synchronization scheme
1211 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1212 (using the classical functions will get serialized) [Mt]
1213 * Make dictionary internal table dynamic. No need to specify its size
1214 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1216 * Make sure the log channels are organized as a tree under windows
1217 (because of ANSI C compatibility issue, any channel were child of
1218 root directly) [Mt].
1221 * Cleaned many thing in surf and fixed a few bugs [AL].
1222 * Add a nice command line configuration mechanism to compose models [AL].
1223 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1224 the previous one (KCCFLN05). It relies on something that looks like
1225 a max-min sharing mechanism but cannot be written as such. A new solver
1226 was thus designed [AL].
1227 * Add a new solver to lmm. Based on Lagrange optimization and
1228 gradient-based descent, it enables to efficiently maximise systems s.a
1230 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1232 This solver enables to propose two new network models for TCP Reno and
1233 TCP Vegas based on Low's work. These models still need to be fully
1234 tested though [Pedro Velho].
1237 * Bug fix in SD_simulate. Now the time bound given as argument is
1239 * Use the new parallel task model (ptask_L07) as default.
1240 * Use the SURF command line configuration mechanism.
1241 * 0-size tasks (for synchronization) should now work.
1243 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1245 SimGrid (3.2) stable; urgency=high
1249 We still experience issues on this platform, but we believe that at
1250 least MSG is usable.
1252 GRAS API BREAKAGE (for simplification purpose, sorry):
1253 * the gras_msgtype_by_name is not used anymore. Instead of
1254 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1255 you can write (and must)
1256 gras_msg_send(toserver, "request", &request);
1257 - If you still want to pass a gras_msgtype_t to the function (to cache
1258 the type and avoid the lookup time), use the gras_msg_send_() variant.
1259 - Impacted functions:
1260 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1261 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1262 * The callbacks are now expected to return 0 when everything went well
1263 (just like the main() function)
1265 GRAS new features and improvements:
1266 * New module mecanism where user code can use per process globals [Mt]
1267 This is similar to gras_userdata_*() functions, but for libraries. It
1268 factorize some code developped over and over in the examples and AMOK.
1269 It has still to be documented and used (only amok/peermanagement is
1271 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1272 loss when they were small enough to fit into the buffer and sent quickly
1273 enough so that they can all get received in one shoot.
1274 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1275 if not found. Use the *_or_null() variant for the old semantic.
1276 * In gras_msg_handle, do not discard messages without callback.
1277 They are probably messages to be explicitly awaited later (ie, proofs of
1278 mis-synchronization in userland since they are sent before being awaited)
1279 No big deal usually.
1280 * gras_socket_meas_send/recv: semantic changed!
1281 The numerical arguments used to be (1) the total amount of data to send
1282 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1283 messages. This was need for the fool willing to send more than MAXINT
1284 bytes on quite fat pipes.
1287 * Do really rename the hostmanagement module to peermanagement. [Mt]
1288 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1289 but this is rather new and this was documented in the module
1290 documentation (poor excuses, I admit)
1291 * Bandwidth measurement semantic changed! This follows the changes to
1292 gras_socket_meas_send/recv explained above.
1295 * A sequential mode has been added to the workstations. When a workstation
1296 is in sequential mode, it can execute only one task, and the other tasks
1297 are waiting in a FIFO. [Christophe Thiery]
1300 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1301 model for SIMDAG. [Christophe Thiery]
1302 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1303 equal to 0, which caused some bad behaviors in
1304 saturated_constraint_set_update. I now use a threshold mechanism like in
1308 * When running manually src/testall, you select specific units [Mt]
1309 testall is the result of our cunit mecanism, and should replace all
1310 the scripty thingy around since bash don't run easily on billware.
1312 * A mallocator system has been added. [Christophe Thiery]
1313 Mallocators allow you to recycle your unused objects instead of freeing them
1314 and allocating new ones.
1316 Documentation update:
1317 * FAQ reworking + New FAQs:
1318 - "Valgrind spits tons of errors!" [Mt]
1319 - "How to repport bugs" [Mt]
1320 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1321 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1322 - Communication time measurement within MSG [AL]
1323 - I experience weird communication times when I change the latency [AL]
1324 * GRAS tutorial [Mt]
1326 - an introduction to the framework and to the used communication model
1327 - an initiatic tour introducing the most proheminent features:
1329 . Lesson 0: Installing GRAS
1330 . Lesson 1: Setting up your own project
1331 o Part 2: Message passing
1332 . Lesson 2: Exchanging simple messages
1333 . Lesson 3: Passing arguments to the processes (in SG)
1334 . Lesson 4: Attaching callbacks to messages
1335 . Lesson 5: Using globals in processes
1336 . Lesson 6: Logging informations properly
1337 . Lesson 7: Using internal timers
1338 . Lesson 8: Handling errors through exceptions
1339 . Lesson 9: Exchanging simple data
1340 . Lesson 10: Remote Procedure Calling (RPC)
1341 . Lesson 11: Explicitely waiting for messages
1342 . Recapping of message passing features in GRAS
1343 - A HOWTO section containing:
1344 o HOWTO design a GRAS application
1345 More are due, of course. They will come latter. In the meanwhile, you can
1346 check the examples which are still here.
1348 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1350 SimGrid (3.1) stable; urgency=high
1354 There was a stack corruption somewhere, visible only when optimizing
1355 with these versions. [Vince]
1358 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1359 module that enabled to play with Directed Acyclic Graphs. It is built
1360 directly on top of SURF and provides an API rather close to the old
1361 SG. Some old codes using SG are currently under rewrite to check that
1362 all needful functions are provided. [Christophe Thiery]
1365 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1366 extension of the classical CLM03 model that gracefully handles
1367 failures. This is now the default model for MSG and GRAS. It doesn't
1368 handle parallel tasks yet though. [AL]
1369 * Bug fix: Weights were not correctly set in the network part.
1370 WARNING: This may have resulted in incorrect results with simulations
1371 where there are more than one flow on a given link. [AL]
1374 * After a (long ?) discussion on simgrid-devel, we have decided that the
1375 convention we had on units was stupid. That is why it has been decided
1376 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1377 WARNING : This means that all previous platform files will not work as
1378 such with this version! A warning is issued to ask users to update
1380 A conversion script can be found in the contrib module of the CVS, under
1381 the name contrib/platform_generation/surfxml_update.pl [MQ]
1384 * Bug fix: Processes were started in reverse order, wrt deployment file.
1385 WARNING: if your code relies on this bug, please fix it. [AL]
1386 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1387 being executed on two different locations. [AL]
1388 * Bug fix: Failures are now better supported thanks to Derrick's tests
1389 (there was many failure situations I hadn't thought of and that weren't
1390 correctly handled). [AL]
1391 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1395 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1396 Exception occurring on server-side are propagated back to client (!).
1398 API CHANGE: the callback changed their prototype. Change:
1399 int my_handler(gras_socket_t expeditor, void *payload_data) {
1401 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1402 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1404 * New! function: gras_msg_handleall to deal with all messages arriving
1405 within a given period.
1406 * New! function: gras_socket_server_range to get a server socket in a
1407 range of port numbers (ease to avoid port number conflicts) [MQ]
1408 * New! gras processes display their backtrace when they get a SIGUSR1
1409 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1410 Sweet to debug RL processes [MQ]
1414 - Do not force experiment sizes to be expressed in kb, or it becomes
1415 impossible to measure the latency this way (needs one byte-long tests)
1416 WARNING: this changes the amok_bw_* function semantic. [MQ]
1417 - Implements the link saturation stuff. [MQ]
1418 * Peer management module:
1419 New! module factorizing code that we wrote over and over [MQ].
1422 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1423 - Test units are placed directly into the library code, they get extracted
1424 automatically and placed into the src/testall binary.
1425 - Convert most of the XBT tests to this system.
1426 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1427 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1428 live in the same C code. This required to change a little bit the API
1429 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1430 * New module: sparse graph structure with basic algorithms (this is work
1431 in progress and the API is not considered to be frozen yet). [AL]
1432 * Display more information on backtraces: source line & function names are
1433 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1434 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1435 * Reworked a little bit some #include statements to load only required
1436 headers. Some user code that relied on SimGrid to include stdlib or
1437 stdio may need to include it by themselves. [AL]
1438 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1440 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1441 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1442 but I guess that almost nobody used those parts. [MQ]
1444 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1446 SimGrid (3.0.1) stable; urgency=low
1449 * Unfortunately, I had missed 5 misnamed functions:
1450 xbt_fifo_item_t xbt_fifo_newitem(void);
1451 void xbt_fifo_freeitem(xbt_fifo_item_t);
1452 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1453 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1454 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1455 They're now deprecated. Please use their new versions:
1456 xbt_fifo_item_t xbt_fifo_new_item(void);
1457 void xbt_fifo_free_item(xbt_fifo_item_t);
1458 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1459 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1460 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1462 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1463 * Documentation: xbt_log module unmercifully reworked [MQ]
1464 * Bugfix: there was a problem with the ending of contexts with
1465 the pthread backend. It caused some weird deadlock or behavior
1466 depending on the pthread implementation. [AL]
1467 * Bugfix: get the exceptions raised in the simulator repport where
1468 and why they come from when they are not catched in time [AL, MQ]
1471 * Bugfix: Do repport the error when two non-connected hosts try to
1472 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1475 * Add additionnal checkings on communications. Assert that two
1476 communicating hosts are connected by a set of links... [AL]
1479 * Add additionnal checkings on channel values in communication [AL]
1480 * New: MSG_task_get_source to see on which host a task was generated [HC]
1481 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1482 the number of tasks waiting to be received on channel and sent
1484 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1485 waits for the first task coming from a given host.. [AL]
1487 GRAS new functionnalities: [MQ]
1488 * Enhance the parsing macro to allow the size of multidimentional objects
1489 to be given thru annotations.
1490 * New example (and documentation): Matrix Multiplication a la RPC
1491 (as when I was young!) and fix a bunch of bugs found on the way.
1493 GRAS performance improvements: [MQ]
1495 * Reduce the amount of cbps creation/destruction by making it static to
1496 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1499 * Change libdata to a set so that we can search for stuff by ID (and thus
1500 reduce the insane amount of dict lookups)
1503 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1504 RL communication on top of SG since it's so uneffective.
1505 It may also be used for inter-thread communication in RL, one day.
1506 * Use gras_datadesc_copy() to exchange messages on top of SG
1508 - improve message exchange performance on top of SG
1509 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1510 * Don't exchange on the network the size of the used part of buffer,
1511 instead, specify the possible buffer size to read().
1513 - reduces the amount of read/write calls (one pair per exchange)
1514 - reduces the amount of exchanged data (the size)
1515 - allows to retrieve all arrived data on receiver side, if we don't need
1516 it right now (subsequent read will peek the buffer)
1517 - allows the receiver to proceed with the begining of the stream before
1518 everything is arrived
1519 - make it possible to build an iov transport (using readv/writev)
1521 - take care of the data with non-stable storage (like stacked data),
1523 * If possible, TCP send uses vector I/O (when writev() is here)
1524 - Don't use it for receive since we send data sizes and data on the
1525 same stream, so we wouldn't be able to chain large amount of chunks
1526 before having to flush the stuff to read the size.
1527 * Rework the transport plugin mecanism to simplify it and reduce the
1528 amount of pointer dereferencement when searching for the right function
1531 * I guess that now, we do almost as few system calls as possible while
1532 doing as few data copy as possible.
1534 To improve it further, we could try to send all the sizes first and then
1535 all the data (to use iov on receiving size), but it's only a partial
1536 solution: when you have 2 dimensional data, the sizes of the second
1537 dimension is data of the first dimension, so you need 3 streams.
1539 I'm not sure the potential performance gains justify the coding burden.
1541 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1543 SimGrid (3.00) stable; urgency=high
1546 * New! Give the possibility to hijack the surf parser and thus bypass
1547 MSG_create_environment and MSG_launch_application. Have a look at
1548 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1550 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1552 SimGrid (2.96) unstable; urgency=low
1557 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1558 This deprecates the xbt_error_t mecanisms.
1559 It modifies (simplifies) all XBT and GRAS API.
1560 MSG API keeps unchanged (exceptions raised by XBT are catched from
1561 within MSG and masked with existing error handling facilities)
1564 * New! Add a FATPIPE model. [AL]
1565 * New! Add a parallel task model. [AL]
1566 * New! Add automatically a loopback interface (in the default
1567 network model) if none was precised.
1570 * Bugfix: MSG_process_resume now works with the current running process.
1572 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1573 * Modification of MSG_task_get_compute_duration. Once a task has been
1574 processed, the value returned by this function is now equal to 0. [AL]
1575 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1576 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1577 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1578 of a m_task. For now, MSG_task_cancel only works with computation
1580 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1581 of the processor (in Mflop/s) regardless of the current load on the
1583 * API Change: use proper naming convention for MSG_getClock and
1584 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1586 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1587 This function changes the priority of a computation task. This priority
1588 doesn't affect the transfer rate. A priority of 2 will make a task
1589 receive two times more cpu power than the other ones. This function
1590 has been added to suit the needs of Nguyen The Loc and hasn't been that
1591 much tested yet. So if it fails, please report it and send me your code.
1593 * API Change: removed all functions and types that were marked "deprecated"
1594 since many months. Renamed MSG_global_init_args to MSG_global_init.
1596 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1598 SimGrid (2.95) unstable; urgency=low
1601 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1602 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1603 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1606 * Bugfix: Allow absolute paths for platform description files [MQ]
1607 * Bugfix: do free the variables after use. Leads to drastic performance
1609 * Implement max_duration (ie, timeouts) on resources [AL]
1612 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1614 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1615 the needed support (timeouts) [AL]
1618 * Implement measurement sockets. You can now get the bandwidth between two
1619 hosts thanks to AMOK (see below). [MQ]
1620 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1621 dynar over the network (yeah) [MQ]
1622 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1623 * simple/static token-ring example. [Alexandre Colucci and MQ]
1624 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1625 in place before [MQ]
1628 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1629 type parsing. "short" and "long" modifiers were also ignored; other
1630 modifier (such as reference level) are still ignored. [MQ]
1631 * Embeed the buffer size within the buffer itself on SG. [MQ]
1632 That way, send() are atomic and cannot get intermixed anymore (at least
1633 the ones which are less than 100k; bigger messages still have the issue)
1634 * Array size pushed by the field, not by the field type (or each
1635 and every long int will push stuff to the cbps) [MQ]
1636 * use select() to sleep since it allows to portably sleep less than one
1639 GRAS (minor cleanups)
1640 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1643 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1644 * type category 'ignored' killed as it was never used and were difficult
1646 * whether a type can cycle or not is now a flag, leaving room for more
1647 flags (as "ignored", if we feel the need one day ;)
1648 * Rename raw sockets to measurement sockets since "raw" has another
1649 meaning in networking community.
1652 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1653 and offers features not belonging to GRAS but that most applications
1654 need. One day, it may be a set of plugins loadable at runtime.
1655 * New module: bandwidth
1656 bandwidth measurement between arbitrary nodes running this module. [MQ]
1658 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1660 SimGrid (2.94) unstable; urgency=low
1662 The first beta release of SimGrid 3 !
1666 * Update the main page and the FAQ. Adding references to gforge.
1669 * Add a gras_os_getpid function.
1672 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1673 * Extend the logs so that they also print PID, hostname, date, ... if
1675 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1676 and kill the old version which were in testsuite/
1677 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1681 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1682 for users. Instead, each programming environment (GRAS, MSG,...) use it
1683 when needed to provide such a feature to users.
1684 Don't shortcut the mecanism or you will also shortcut the virtualization
1685 you need on the simulator.
1689 * Cleanups in configury with regard to compile optimization/warning flags.
1690 Also add -fno-loop-optimize to any powerpc since it's the optimization
1691 killing gcc (< 3.4.0).
1692 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1693 needed by MSG examples complications
1694 * Borrow configury beautifications from PHP
1697 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1698 warning (thanks loris for stumbling into this one).
1699 * Bugfix: stop loading private headers (gras_config.h) from the public
1703 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1704 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1705 RL would imply a DNS resolver, which is *hard* to do in a portable way
1706 (and therefore delayed).
1707 * Implement a real timer mecanism and use it in timing macros. This allows
1708 to avoid rounding errors and get a 0.000005 sec precision in timing
1709 macros. While I was at it, various cleanups:
1710 - allow to declare more than one timed section per file (fix a stupid bug)
1711 - move some private declaration to the right place
1712 - merge conditional execution and timing macros into emulation module
1713 - document the module
1714 - make sure the module cleanups its mess on gras_exit
1715 * Documentation improvements:
1716 - (new) how to compile applications using GRAS
1717 - (new) emulation support (timing macros)
1719 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1721 SimGrid (2.93) unstable; urgency=low
1723 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1726 - Use Paje properly where used. Still to be sanitized properly.
1727 - Portability fix: Add an implementation of the contexts using pthread
1731 - Add xbt_procname(): returns the name of the current process.
1732 Use it to show the current process's name in all logging.
1734 - fix detection of older flex version and the reaction, since we do
1735 depend on modern ones (we use lex_destroy)
1736 - Better separation of SG and RL in the libs: remove all simulation code
1737 from libgras. As a result, this lib is now only 200k when stripped.
1738 Some of the xbt modules may also be duplicated (two sets and such) and
1739 should be cleaned/killed before SG3.
1740 - Insist on using xlC on AIX because of weird problems involving gcc there.
1741 - Cleanup the make remote stuff. This is now done by scripts
1742 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1743 mainly for our private use, but we're working on changing them to user
1746 - Bugfix: flush the socket on close only if there is some *output*.
1747 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1749 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1752 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1754 SimGrid (2.92) unstable; urgency=low
1756 Alpha 3 on the path to SimGrid 3
1760 - New! First try of benchmarking macros.
1761 - New! First try so that gras_stub_generator generate deployment and
1762 remote compilation helpers.
1764 - Bugfix: Initialization fix in msg_test.
1768 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1770 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1772 - API Change: gras_os_sleep to take the amount of seconds as a double.
1773 Accepting an int was error prone since it was the only location where
1774 seconds were coded as such. It leaded to damn rounding errors.
1775 - Bugfix: Hard to belive that timers ever worked before this.
1777 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1779 SimGrid (2.91) unstable; urgency=low
1781 Alpha 2 on the path to SimGrid 3
1785 - Bug fix in the lmm_solver.
1787 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1788 through the function MSG_paje_output.
1789 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1790 - It is possible to bound the rate of a communication in MSG with
1791 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1792 to write it in the changelog).
1793 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1798 - Various cleanups to the autotools stuff
1799 - Begin to move Gras examples to examples/gras/
1800 - Let make distcheck work again (yeah!)
1802 - documentation overhauled using doxygen.
1803 gtk-doc-tools is dead in SimGrid now.
1804 - Automatically extract all existing logging categories, and add the list
1805 to the documentation (long standing one, to say the less)
1807 - Cleanup the known architecture table. Reorder the entries to group what
1808 should be, and use a more consistent naming scheme.
1809 (some of the test dataset are still to be regenerated)
1810 - New! Allow library to register globals on each process just as userdata
1812 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1813 lookup time (for now).
1814 Use it in msg and trp.
1815 This cleans a lot the internals and helps enforcing privacy of the
1816 headers between the gras components.
1817 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1818 - Bugfix: gras_os_time was delirious in RL.
1819 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1821 - Reenable GRAS now that it works.
1823 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1825 SimGrid (2.90) unstable; urgency=low
1827 Alpha 1 on the path to SimGrid 3
1829 * It is a long time since the last release of SimGrid. I'm sorry about
1830 that but as I had told you, I was rewriting a lot of things. I apologize
1831 to those who had been reporting bugs to me and that I had not answered.
1832 If your bug is still in the new version, please tell me. Here is a
1833 summary of the main changes.
1835 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1836 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1837 portability, speed, and a lot more but you'll figure it out later.
1838 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1839 I wanted to release the new MSG as soon as possible and I have
1840 broken GRAS, which is the reason why, for now, only MSG is fully
1841 functional. A laconic description of these projects is available
1842 in the documentation.
1844 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1845 kernel optimized for our needs (called SURF but only the developers
1846 should use it). Hence, MSG is now roughly 30 times faster and I think
1847 that by rewriting a little bit MSG, I could event speed it up a little
1848 bit more. Beside the gain in speed, it is also much easier to encode a
1849 new platform model with SURF than it was with SG. More to come...
1851 * REVOLUTION 3: I have tried to change a little as possible the API of
1852 MSG but a few things really had to disappear. The main differences
1853 with the previous version are :
1854 1) no more m_links_t and the corresponding functions. Platforms are
1855 directly read from a XML description and cannot be hard-coded
1856 anymore. The same format is used for application deployment
1857 description. The new format is described in the documentation.
1858 Have a look in tools/platform_generation. There is a tiny script
1859 that converts from the old platform format to the new one. Concerning
1860 the application deployment format, parsing the old one is tricky.
1861 I think most of you should however be able to convert your files. If
1862 it is really an issue, I can write a C code that does the conversion.
1864 2) the toolbox tbx does not exist anymore. We now have a library
1865 with much more data-structures but without the hash-tables (we have
1866 dictionaries that are much faster).
1868 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1870 *****************************************************************************
1871 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1872 * don't feel like converting the oldies. (Mt) *
1873 *****************************************************************************
1876 Version 2.90: "the long awaited one"
1877 - Finished rewriting and debugging MSG. Rewrote the documentation.
1878 - disable GRAS for now since it needs to be ported to the newest SG
1881 - Finish the port to windows (using mingw32 for cross-compile)
1884 - Main loop and datastructures of SURF. A cpu resource object is
1885 functional. Surf can thus be used to create cpu's with variable
1886 performance on which you can execute some actions.
1888 2004-11-15 Martin Quinson
1889 - Port to ARM. Simply added the alignment and size descriptions. Should
1890 work, but the ARM machines are so slow that I didn't had the opportunity
1891 to 'make check' over there yet.
1893 2004-11-15 Arnaud Legrand
1894 - Trace manager now written. It uses a heap structure and is therefore
1895 expected to be efficient. It may however be speeded up (particularly
1896 when many events occur at the same date) by using red and black
1897 trees. One day maybe...
1898 - Max-min linear system solver written. It uses a sparse matrix
1899 structure taking advantage of its expected use. Most operations are
1900 O(1) and free/calloc are called as few as possible. The computation of
1901 the minimum could however be improved by using a red and black tree
1904 2004-11-03 Arnaud Legrand
1905 - Rename every gras_* function that was in xbt/ to its xbt_
1907 - Add a heap and a doubly-linked list to xbt
1908 - Added a dichotomy to the dictionaries. make check works as well before
1909 so I assume that the patch is correct. I do not know however if things
1910 run effectively faster than before now. :)
1912 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1913 SimGrid, and the version number is bumped to 2.x
1915 2004-10-29 Martin Quinson
1916 - Introduction of the remote errors.
1917 They are the result of a RMI/RPC on the remote machine.
1918 ErrCodes being scalar values, you can't get the host on which those
1919 errors did happen. Extending the error mechanism as in Gnome is possible.
1920 No idea yet whether it is a good idea.
1922 2004-10-28 Martin Quinson
1923 - Interface revolution: the Starred Structure Eradication.
1924 I used to do typedef struct {} toto_t; and then handle *toto_t.
1925 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1926 - ???_t is a valid type (builded with typedef)
1927 - s_toto_t is a structure (access to fields with .)
1928 - s_toto is a structure needing 'struct' keyword to be used
1929 - e_toto_t is an enum
1930 - toto_t is an 'object' (struct*)
1932 typedef struct s_toto {} s_toto_t, *toto_t;
1933 typedef enum {} e_toto_t;
1934 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1935 s_toto_t) is private.
1937 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1938 so that it can set the variable to NULL. It was so for dicts and sets,
1939 it changed for dynars.
1941 - Fix a bunch of memleaks in dict_remove
1942 - Fix a bug in sg/server_socket opening: it failed all the time.
1944 2004-10-07 Martin Quinson
1945 - Speed up dynar lookup operation a bit.
1947 gras_dynar_get is dead.
1949 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1950 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1951 (which gives you the address of the stored data).
1953 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1954 copy of the data using an affectation to do the job and not a memcpy.
1956 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1957 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1959 It does not really speedup the dynar test because they are
1960 setting elements all the time (and look them seldom). But the dict does
1961 far more lookup than setting.
1963 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1965 2004-10-05 Martin Quinson
1966 - Allow to (en/dis)able the cycle detection at run time.
1968 Whether we should check for cycle or not is now a property of each
1969 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1970 datadesc_cycle_unset allow to remove this property when previously set.
1972 Note that the cycle detection is off by default since it impacts the
1973 performance. Watch the data you feed GRAS with ;)
1975 This property is hereditary. Any element embedded in a structure having it
1976 set have it set for the time of this data exchange.
1978 You should set it both on sender and receiver side. If you don't set it on
1979 sender side, it will enter an endless loop. If you forget on receiver
1980 side, the cycles won't be recreated after communication.
1982 - Header reorganization.
1983 Kill gras_private.h, each submodule must load the headers it needs.
1985 2004-10-04 Martin Quinson
1986 - Interface revolution: do not try to survive to malloc failure.
1988 Now, gras_malloc and friends call gras_abort() on failure.
1989 As a conclusion, malloc_error is not a valid error anymore, and all
1990 functions for which it was the only gras_error_t return value are
1991 changed. They now return void, or there result directly.
1992 This simplify the API a lot.
1994 2004-09-29 Martin Quinson
1995 - Re-enable raw sockets.
1996 Created by gras_socket_{client,server}_ext;
1997 Used with gras_raw_{send,recv}
2000 It should allow to kill the last bits of gras first version soon.
2002 This is not completely satisfactory yet (duplicate code with
2003 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2006 - Simplify transport plugin (internal) interface by not passing any
2007 argument to _server and _client, but embedding them in the socket
2010 2004-09-28 Martin Quinson
2011 - Finish the port to AIX.
2012 autoconf was my problem (segfault within the malloc replacement
2013 function. No idea why)
2015 2004-09-16 Martin Quinson
2016 - Fix some size_t madness on 64bit architectures.
2018 2004-09-08 Martin Quinson
2019 - Reduce the number of system headers loaded, overload some more system
2020 calls (such as malloc to cast the result of the system one, and work
2022 - Fix and reintroduce the config support
2024 2004-09-07 Martin Quinson
2025 - Source code reorganization to allow Arnaud to surf all over there.
2026 - Allow to document the logging categories.
2027 - Remove all uppercase from logging categories and useless cleanup in names.
2029 2004-08-18 Martin Quinson
2030 Version 0.6.2 (protocol not changed; API changed)
2031 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2032 gras_error_t), and NULL when not found. Functions expecting a msgtype
2033 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2034 hopefully usefull message.
2035 - Portability to prehistoric sparcs again
2037 2004-08-17 Martin Quinson
2038 Version 0.6.1 (protocol not changed; ABI not changed)
2039 - prealloc some buffers to speed things up
2041 2004-08-11 Martin Quinson
2042 Version 0.6 (protocol not changed; ABI expended)
2043 - The parsing macro can deal with the references, provided that you add
2044 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2046 2004-08-09 Martin Quinson
2047 Version 0.5 (protocol not changed; ABI changed)
2048 - Allow to off turn the cycle detection code in data exchange at
2049 compilation time. It should be at run time, but I'm short of time (and
2050 the config stuff is still broken). That way, we keep dict out of the
2051 critical path, which is good because the performance is poor:
2052 - search not dichotomial yet
2053 - dynar give no way to access their content and memcpy everytime
2054 - In composed data description (struct, ref and so on), stop foolness of
2055 keeping the subtype's ID, but store the type itself. This keeps sets out
2056 of the critical path, which is good since they rely on dynar and
2057 dictionnaries. The only loose of that is that we cannot detect the
2058 redeclaration of a structure/union with another content (but I'm not sure
2059 the code detected well this error before anyway). We still can detect
2060 the redefinition discrepancy for the other types.
2061 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2062 it breaks the code because of type-punning used all over the place).
2063 This breaks on all non-gcc architectures (for now).
2065 All those changes (plus the buffer of last time) allow me to gain 2 order
2066 of magnitude on cruel tests consisting of 800000 array of integers on two
2067 level of a hierarchical structure (200 secondes -> 4 secondes)
2070 - the selector of reference must now return the type it points to, not
2071 the ID of this type.
2073 2004-08-06 Martin Quinson
2074 Version 0.4 (protocol changed; ABI not changed)
2075 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2076 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2077 - Implement a buffer transport, and use it by default (it relies on tcp in
2078 real life and on sg in simulation).
2079 That's a bit hackish since I had a new field to the structure to store
2080 its data without interfering with the subtype ones. Inheritance
2081 is tricky in C. And that's a kind of reverse inheritance with one class
2082 derivating two classes. Or maybe a game with java interfaces. Anyway,
2083 that's damn hard in C (at least).
2084 Moreover, I got tired while trying to ensure plugin separation and
2085 genericity in SG mode. MSG wants me to do weird things, so let's go for
2086 cruel hacks (temporarily of course ;).
2087 See comment in transport_private.h:71
2088 - do not include all the _interface headers in private but in the files
2089 which really need them (to cut the compilation time when they are
2092 2004-07-26 Martin Quinson
2093 Version 0.3 (protocol not changed; ABI changed)
2094 - Major overhault of the datadesc interface to simplify it:
2095 - shorted the function names:
2096 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2097 - add a trivial way to push/pop integers into the cbps without malloc.
2098 This allows to make really generic sub_type description, which simply
2099 pop their size of the stack.
2100 - add a function gras_datadesc_ref_pop_arr() which does what users want
2101 most of the time: Declare a dynamic array (which pops its size of the
2102 stack) and declare a reference to it. Poor name, but anyway.
2103 - kill the post-send callback, add a post-receive one
2105 2004-07-23 Martin Quinson
2106 Version 0.2 (protocol changed; ABI changed)
2107 - add some testing for cpbs in the test cases, and fix some more bugs.
2108 This invalidate again the little64 data file, since I cannot regenerate
2110 - remove an awfull optimization in the logging stuff, allowing me to:
2111 - understand it again
2112 - learn gcc how to check that the argument match the provided format
2113 - fix all errors revealed by gcc after that
2114 - internal keys of dict are not \0 terminated. Deal with it properly in
2115 loggings instead of segfaulting when the user want to see the logs :-/
2117 2004-07-22 Martin Quinson
2118 - Fix some stupid bug preventing cbps (callback postit) from working
2120 2004-07-21 Martin Quinson
2121 - Some documentation cleanups
2122 - remove the useless last argument of msgtype_declare
2123 - rename the Virtu functions to fit into the 'os' namespace
2124 - move headers src/include -> src/include/gras/ and stop fooling with
2125 gras -> . symbolic link
2126 - make distcheck is now successful
2128 2004-07-19 Martin Quinson
2130 - Build shared library also
2131 - Install html doc to the right location
2132 - stop removing maintainer files in make clean
2133 - build tests only on make check
2135 2004-07-13 Martin Quinson
2137 - No major issue in previous version => change versionning schema
2138 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2139 corresponding dataset.
2141 2004-07-11 Martin Quinson
2143 - Get it working with any kind of structure (we can compute the padding
2144 bytes remotely for all the architectures I have access to)
2145 - Implement the structure parsing macro (still not quite robust/complete)
2146 - Improvement to the remote testing toysuite
2148 2004-07-10 Martin Quinson
2149 [autoconf mechanism]
2150 - get ride of a bunch of deprecated macros
2151 - actually run the test for two-compliment, not only compile it :-/
2152 - test whether the structures get packed (and bail out if yes. Damn.
2153 Alignment is a serious matter)
2154 - test whether the structures get compacted (but respecting the alignment
2155 constraints of each types)
2156 - test whether the array fields of structures can straddle alignment boundaries
2158 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2159 (took me 2 hours to find that bug, looking at the wrong place)
2160 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2161 before sending/receiving and are used to compute the offsets of fields
2162 - Given that padding size depend even on compiler options, keep track of
2163 alignment and aligned_size only for the current architecture. Not a big
2164 deal since we send structure fields one after the other (seems
2166 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2167 program, let it work on linux/gcc/little32. portability todo.
2169 2004-07-08 Martin Quinson
2170 - import and improve remote compilation support from FAST
2171 - make sure make check works on half a dozen of machines out there
2173 2004-07-07 Martin Quinson
2174 Let's say it's version 0.0.3 ;)
2175 - Implement conversions (yuhu!)
2176 - Let it work on solaris (beside conversion, of course)
2177 - Stupid me, using rand() to generate the conversion datatests in not wise.
2179 2004-07-06 Martin Quinson
2180 - Let make dist work, since I'm gonna need it to compile on remote hosts
2181 - Let Tests/datadesc_usage write the architecture on which the file was
2182 generated as first byte.
2183 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2184 architecture descriptions.
2185 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2186 on those architectures.
2187 - Optimization: send/recv array of scalar in one shoot
2189 2004-07-05 Martin Quinson
2190 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2192 - Plug a whole bunch of memleaks
2193 - each process now have to call gras_{init,exit}. One day, their log
2194 settings will be separated
2195 - Continue the code factorisation between SG, RL and common in Transport.
2197 2004-07-04 Martin Quinson
2199 - Redistribution between SG and RL.
2200 We wanna have to accept in SG, so move accepted related parts of RL in
2201 the common part. (more precisely, the dynar of all known sockets is no
2202 more a static in transport.c, but part of the process_data)
2204 [gras_stub_generator]
2205 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2207 2004-07-03 Martin Quinson
2208 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2209 them all up in one shot)
2211 - Fix the ugly case of reference to dynamic array.
2212 - New (semi-public) function gras_datadesc_size to allow the messaging
2213 layer to malloc the needed space for the buffer.
2215 - gras_socket_close now expect the socket to close (and not its address to
2216 put NULL in it after it). This is because the socket passed to handlers
2217 is one of their argument (=> not writable).
2219 - propagate the interface cleanup from last week in datadesc, ie remove a
2220 superfluous level of indirection. User pass adress of variable
2221 containing data (both when sending and receiving), and not of a variable
2222 being a pointer to the data. Let's say that I like it better ;)
2223 The price for that is constructs like "int msg=*(int*)payload" in
2224 handlers, but it's a fine price, IMHO.
2226 - Let it work in RL (yuhu)
2228 2004-06-21 Martin Quinson
2230 - porting SG plugin and SG select to new standards (works almost).
2231 - plug memleaks and fix bugs around.
2234 - cleanup the prototype of data recv and force users to specify when they
2235 want to handle references to objects. Test case working even for cycles.
2236 - plug memleaks. Valgrind is perfectly ok with this.
2238 2004-06-12 Martin Quinson
2240 - cleanup the separation between plugin and main code in plugin creation
2242 2004-06-11 Martin Quinson
2244 - Reput hook for raw sockets, needed for BW experiments
2245 - kill a few lines of dead code
2246 [Data description] Interface cleanup
2247 - gras_datadesc_by_name returns the searched type or NULL.
2248 That way, no variable is needed to use a type desc once, which makes
2250 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2251 two parameters were strings (field name, type name), leading to
2253 [Dicos] Interface cleanup
2254 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2255 This is consistant with the dynar API.
2257 2004-04-21 Martin Quinson
2259 - Porting to new standards.
2261 - interface cleanup.
2262 There is no bag anymore, no need to take extra provision to mask the
2263 pointers behind "ID".
2264 Better splitup of functions between files create/exchange/convert.
2265 This is still a bit artificial since convert and receive are so
2266 interleaved, but anyway.
2268 - add a queued message list to procdata (the ones not matching criteria
2270 - factorize some more code between SG and RL wrt procdata
2272 - use gras_exit in example to track memleaks
2273 - get rid of gs_example now that GS is properly integrated into gras
2274 - update run_test to integrate the lastest tests (datadesc)
2276 - rename WARNINGn macros to WARNn since it prooved error-prone
2278 2004-04-19 Martin Quinson
2280 - register init/exit functions within gras module mechanism
2281 - send/receive function.
2282 Convertion is not implemented, but short-cutted if not needed.
2283 struct/array elements are sent one by one (instead of block-wise), but
2284 nobody really cares (yet). Get a prototype before optimizing.
2285 - tests (using a file socket) for DD send/receive on:
2286 - base types: int, float
2287 - array: fixed size, string (ie ref to dynamic string)
2288 - structure: homogeneous, heterogeneous
2289 - chained list, graph with cycle
2290 Believe it or not, valgrind is not too unhappy with the results. The
2291 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2295 - Bugfix when using a filename explicitely (instead of '-')
2297 2004-04-09 Martin Quinson
2299 - factorize more code between RL and SG in socket creation
2300 - Complete the implementation and tests of:
2302 o file (only in RL, and mainly for debugging)
2304 I lost 3 days to design a portable address resolver, and then decided
2305 that the prototype mainly have to run on my box.
2306 Addressing portability too early may be like optimizing too early :-/
2308 - use gras_init in the Tests instead of the crappy parse_log_opt
2309 (the latter function is removed)
2310 [Conditional execution]
2311 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2312 [Code reorganisation]
2313 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2314 Build examples against the RL library, since there is no way to disable
2315 its creation for now.
2317 For information, the beginning of coding on GRAS was back in june
2318 2003. I guess that every line has been rewritten at least twice since