1 SimGrid (3.7.1) NOT RELEASED; urgency=low
4 * Fix weird behaviors when dealing with parallel tasks.
7 * New type of typed tasks SD_TASK_COMP_PAR_AMDAHL that represents a
8 parallel task whose initial work is distributed among host according
9 to the Amdahl's law. Such tasks are created with a parameter alpha
10 that corresponds to the non-parallelizable part of the computation.
12 Before calling SD_task_schedulel or SD_task_schedulev, the
13 SD_task_distribute_comp_amdahl function has to be called with the
14 desired number of workstations. The filling of computation_amount and
15 communication_amount (actually empty) structures is then done
17 * New function SD_workstation_dump to display various information
19 Bug fixes, cosmetics and small improvements
20 * Fix compilation when using MSG_USE_DEPRECATED.
23 * Fix compilation on Macs.
25 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
27 SimGrid (3.7) stable; urgency=low
29 The "spring cleanups (before the next Big Project kicks in)" release.
32 * We can specify the SMPI latency/bandwidth factor with command line
33 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
34 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
35 You can also use the "config tag" from platform file by adding this line
36 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
37 (see "example/platforms/tag_config.xml" to use "config tag").
38 Note that the command line supersedes the platform file configuration.
39 * Change the correction factors used in LMM model, according to
40 the latest experiments described in INRIA RR-7821.
41 Accuracy should be improved this way.
42 * Use the partial invalidation optimization by default for the
43 network too. Should produce the exact same results, only faster.
44 * Major cleanup in surf to merge models and split some optimization
45 mechanisms from the core of the models. As a result you can now
46 specify which model to use (e.g., --cfg=network/model:LV08
47 --cfg=cpu/model:Cas01) and which optimization mode to use
48 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
49 Incompatible combinations should err at initialization. See
50 --help-models for the list of all models and optimization modes.
51 * The CLM03 workstation model was dropped for simplicity because it
52 used the deprecated CM02 network model. Use default instead.
53 * Rename the TCP_gamma configuration option to network/TCP_gamma
54 * Rename the coordinates configuration option to
55 network/coordinates, and document it
56 * Use now crosstraffic keyword instead of the terribly misleading
57 fullduplex keyword. It is activated by default now in the current
58 default model, use --cfg=network/crosstraffic:0 to turn it off.
59 * Ongoing refactoring the model parsing to make XML files optional
60 See include/simgrid/platf.h for details (still to be completed)
63 * Major overhaul of the documentation. Almost instructive now :/
64 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
65 functions and friends. This interface was considered as
66 deprecated since over 2 years, it's time to inform our users that it is.
67 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
68 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
69 use these (crappy) functions in your code.
70 These functions will be removed soon. Stop using them now.
71 * Deprecate MSG_get_host_{table,number}
72 Implement MSG_hosts_as_dynar() instead.
73 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
74 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
77 * Stabilize the parallel execution mode of user contexts
78 * Introduce configuration variables to control parallel execution:
79 - contexts/synchro: Synchronization mode to use when running
80 contexts in parallel (either futex, posix or busy_wait)
81 - contexts/parallel_threshold: Minimal number of user contexts
82 that must be part of a scheduling round to switch to parallel
83 execution mode (raw contexts only)
84 * Fix bugs that prevented to use suspend/resume along with
85 synchronization structures.
86 * Fix bugs in process termination that lead to invalid memory access
87 in very specific conditions.
90 * Introduce a parallel mode for the models (controlled by surf/nthreads
91 configuration item). In our tests, running the models in parallel
92 never lead to any speedups because they are so fast that the gain
93 of computing each model in parallel does not amortizes the
94 synchronization costs, even when ultra fast futexes are used.
95 This is released anyway because YMMV.
98 * Performance boost by using a swag internally to compute the set of
99 tasks that are finished and should constitute the return value of
103 * Enable it by default now that it is considered rather stable.
106 * Documentation of the tracing functions.
107 * Performance gains when tracing categorized/uncategorized resource
108 utilization by avoiding calls to get route when updating resource
109 variables. LMM constraints are being used instead.
110 * API changed to set task categories. Use MSG_task_set_category instead
111 of TRACE_msg_set_task_category, and SD_task_set_category instead
112 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
113 * Bugfix for graphicator, routes not correctly obtained, memory leaks
114 * Examples for link user variables added (see at examples/msg/tracing/)
115 * Deprecated function TRACE_msg_set_process_category completely removed
116 * Trace header updated according to the latest Paje file format
117 * Tracing network lazy updates, no longer obligate users to use full updates
118 * --cfg=tracing/platform:1 also registers power/bandwidth variables
119 * Experimental: let user code declare/set/push/pop application states for hosts
120 * API changed to allow the manual creation of graph configuration files
121 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
124 * Improve the API of Lua MSG bindings, using the Lua spirit.
125 * Each simulated process now lives in its own Lua world (globals are
126 automatically duplicated). It helps writing simulators. Will allow
127 to run Splay programs within SimGrid in the future.
128 * Add a Chord example in Lua, equivalent to the MSG one.
131 * Start the implementation of a solution to express temporal
132 properties, not only local assertions. This is still an
133 experimental work in progress, stay clear from it to be safe.
137 - Add new runtime parameters --help-logs and --help-log-categories
138 to display informations about supported logging parameters and
140 - Old deprecated parameters --{gras,surf,msg.simix,xbt}-log=...
141 don't exists anymore.
142 * Mallocators: allow value NULL for the reset function.
144 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
145 create homogeneous dictionaries, where all the elements share the
146 same free function. Non homogeneous dictionaries will be
147 deprecated in the next release.
148 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
149 - Multi-level dictionaries are deprecated.
151 - new function xbt_dynar_search_or_negative() that is useful when
152 you have less than 2 million elements in your dynar and don't
153 want of the extra complexity of catching exceptions when the
154 element is not found.
156 - Make xbt_os_thread module (for thread portability) public.
157 Documentation is still to come, sorry.
159 - Cleanups and simplifications to make it maintainable again.
160 - Exotic features (such as memalign and valloc) were removed.
161 - The metadata were extended and improved so that the
162 model-checker becomes able to explore and inspect the heaps.
163 - This may induce a performance drop when enable_model-checking is
164 ON in cmake (even if it's not used in the simulation), but it is
165 necessary at this point to get MC working.
167 Turn model-checking OFF if simulation performance matters to you.
168 Not enabling it at runtime is not enough, disable it in cmake.
170 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
172 SimGrid (3.6.2) stable; urgency=low
174 The "Not coding new stuff allows to polish old things" release.
177 * New bindings to the NS3 packet level simulator (experimental)
178 * Use the raw (efficient) execution contexts instead of the sysv
179 (portable) ones when possible.
180 * libpcre is now mandatory in any cases since not using it led to
181 severe performance loss and possibly other issues
182 * Update the XML platforms:
183 - G5K: include the latest machine in Nancy
184 - GridPP and LCG: new platforms
185 * Documentation was partially updated, at least (more to come)
187 Bug fixes, cosmetics and small improvements
188 * Free terminated processes before the end of the simulation to avoid
189 exhausting the memory of users having very dynamic amount of
191 * Bug fix and cosmetics about canceling non-running tasks
192 * Bug fix about the dot loader's issues when using libcgraph
195 * Create an installer for windows with nsis (amd64 and win32)
196 - Add an hello world project to illustrate simgrid project creation.
197 - Embed libpcre into the Simgrid installer to avoid
198 its compilation burden
199 * The raw execution contexts should work on Apple now
200 * Port to Windows 64 bits
201 - Sysv contexts now have an implementation for this arch
202 - GRAS communication features now support this arch
203 * Drop support for borland compiler on windows
204 - this code was not maintained, and we kinda depend on gcc nowadays
205 * Fix portability issues on kfreebsd/gnu: build error about semaphores
206 * Fix portability issue on unstable ubuntu: linker became picky on
209 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
211 SimGrid (3.6.1) stable; urgency=low
213 The "Oops, we broke Macs too" release
216 * Fixed contexts detection so that raw ones are used when possible
217 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
218 leads to a strange error, with user code segfaulting sometimes when
219 the generated code is not perfectly aligned (which is not
220 controllable from the user side, depends on the amount of code)
223 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
225 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
227 SimGrid (3.6) stable; urgency=medium
229 The Summer Release, also known as the "OMG! They Killed Kenny!" version
232 * Bindings now constitute their own package, separated from the main one.
233 Rationale: reduce our maintenance nightmare by reducing the module coupling
234 They will soon be released on their own on gforge.
236 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
237 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
239 GRAS: It is not considered as stable anymore, but experimental. Sorry.
240 * It's not quite deprecated for now because we have no replacement,
241 but it may soon become the case.
244 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
245 * Fortran: New user-level cache variable to store the rank of the running
246 process. This improves performance by an order of magnitude.
247 * C: New coccinelle script to automatically locate and modify global and
248 local static variables.
249 * Improved SMPI network model with a sender-side gap to account for multiple
253 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
254 only say if a communication is finished, no matter whether it succeeded or
255 failed. You can call MSG_comm_get_status() to know the status of a finished
257 * New function MSG_task_dsend() to send a task and detach it. When a
258 communication is detached, you are never notified of its success or failure
259 and the memory is released automatically once it is finished. This function
260 is useful when you don't care about the end nor the success of a
262 * Change the prototypes of action replay. Sorry for inconvenience,
263 but this is really more efficient this way (and to adapt your code,
264 you just have to fix the initialization, that shouldn't be too long)
265 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
266 ever managed to do anything useful with it.
267 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
268 takes a dynar of communications. It returns immediately and gives the
269 index of a finished communication (if any).
270 * New example: a basic implementation of the Chord P2P algorithm.
273 * New model for multi-core CPUs. You can now use the core attribute to
274 precise the number of cores of a host. This is a basic model. Every
275 process running on the host receives at most the power provided in
276 the DTD (throughput<=power). Total throughput of process cannot exceed
278 * New peer tag. This peer tag creates a tiny AS comprising a host and a
279 router linked by an up-link and a down-link (possibly asymmetrical).
280 This kind of pattern allows to easily build last-mile model style platforms.
281 Aggregating such patterns in a rule-based AS is thus the technique of
282 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
283 * New model for Vivaldi routing. We transformed the Vivaldi network model
284 into a Vivaldi routing model (based on the rule-based model). This allows to
285 combine Vivaldi based latencies with last-mile platforms.
288 * Added a check for NaN of IEEE754 infinite in the double entries of
290 * Introduce a new context factory "raw", highly inspirated from the
291 ucontext factory, but using manually crafted functions in assembly to
292 do the work in an efficient manner.
293 * Allow to change the used context factory at run time, not only at
294 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
295 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
296 contexts at runtime (only with raw contexts or ucontexts).
297 * Completely rewrote this module to allow parallel execution of user
298 processes. Use --cfg=contexts/nthreads:N to execute user processes
299 with N parallel threads (the default is 1, meaning no parallelism).
300 * Allow to decide dynamically between sequential and parallel modes.
301 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
302 processes in parallel only when their number is greater than or equal to P
304 * Added a check for NaN of IEEE754 infinite in the double entries of
308 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
309 won't output the state of processes when interrupted with Ctrl-C
310 * Add a new function xbt_dynar_to_array that transforms a dynar into a
311 NULL-terminated array. This may solve backward compatibility issues
312 due to the change to return type of SD_simulate. See also:
313 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
314 * Add new macros with variable number of arguments.
315 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
316 - in xbt/asserts.h: xbt_assert
317 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
318 - in xbt/ex.h: THROWF and RETHROWF.
319 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
321 * Change xbt_die() to accept a format string with arguments, just like printf.
322 * New data structure: xbt_lib_t, like a dict but more general and with better
326 * New configuration options
327 Options triva/categorized and triva/uncategorized can be used to generate
328 graph configuration files for Triva visualization tool.
329 * Configuration option tracing/platform is renamed to tracing/categorized
330 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
331 * New TRACE_user_link_variable interface:
332 User provides the name of the link and the tracing variable to attach to it
333 * the declaration of tracing categories must be done after the environment
335 * simpler tracing interface, just one way to declare categories
336 TRACE_category or TRACE_category_with_color, it is up to you
337 * links in the trace file are again identified by their names
338 * trace contains the full platform hierarchy exactly as declared using the ASes
339 * Options tracing/msg/[task|process]:1 groups the process by hosts
340 for both cases, tasks and processes must have names that are unique during the simulation
341 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
342 * The experimental option tracing/msg/volume is deprecated
343 its functionality may be reincorporated if needed
345 The tracing generates a trace file with unordered timestamped events,
346 because of the way the core simulator (surf) works. A script available
347 at the tools directory (fix-paje-trace.sh) can be used to put the events
348 in order. We have changed the tracing so it can generate ordered timestamped
349 events in the final trace, but depending on the simulator (and how much time
350 is simulated) that can lead to a huge memory utilization. It is deactivated
351 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
354 * Define a SIMGRID_VERSION macro in simgrid_config.h.
355 - We are trying hard to keep the API stable, but it may happen that
356 some things change (we're a research project after all, not a
357 nuclear plant operating system). If such things should happen, you
358 could rely on that macro to adapt.
359 - current value: 30600 for 3.06.00, aka 3.6
360 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
361 a number that can be compared with SIMGRID_VERSION.
362 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
363 assertions and verbose/debug logging events are disabled at compile time.
365 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
367 SimGrid (3.5) stable; urgency=medium
370 * New feature to any SimGrid-based simulator: Model-Checking
371 Check SIN#1 for more details.
374 * New Model SMPI (three-interval linear regression for correction factors)
375 See RR-7426, available at http://hal.inria.fr/inria-00527150
376 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
377 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
378 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
379 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
380 * See also src/smpi/README
385 - Tracing API changes: TRACE_start and TRACE_end should not be called
386 by user-code. They are automatically called by simulators created
387 with SimDAG, MSG and SMPI if the toolkit is compiled with
388 tracing_enabled=ON. Categories declaration and utilization remain the
389 same for MSG and SimDag.
390 - A function was added to the tracing API to declare categories with
392 - TRACE_category_with_color (char *category, char *color)
393 where color must be in the following format
394 "%f %f %f", red, green, blue
395 and red, green, blue are float values in the interval [0, 1]
396 - User can specify NULL as color parameter, or continue calling
398 On that case, the tracing system will define random colors
399 - The following command-line options are supported:
400 --cfg=tracing/filename:msg.trace
401 --cfg=tracing:1 (activate tracing, needed to use others)
402 --cfg=tracing/platform:1 (categorized resource use)
403 --cfg=tracing/uncategorized:1 (uncategorized resource use)
404 --cfg=tracing/msg/task:1 (task creation)
405 --cfg=tracing/msg/process:1 (process creation, migration)
406 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
407 --cfg=tracing/smpi:1 (SMPI interface tracing)
408 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
409 - examples of examples/msg/tracing updated
411 - DAXLoader and DOTLoader functions can generate tasks with categories
412 - A new function to attribute a category to SD tasks:
413 TRACE_sd_set_task_category (SD_task_t task, char *category)
414 * Tracing the MPI interface implemented by SMPI
415 - Collective operations are traced with states
416 - Point-to-Point operations are traced with states/links
417 - Tracing activated by a parameter "-trace filename" passed
418 to smpirun during execution (considering that simgrid
419 is compiled with tracing enabled)
420 - To run the simulation with gdb, the simulator
421 accepts --cfg=tracing/smpi:1 to trace SMPI
422 - tesh files to check if smpi tracing is ok
423 - See examples/smpi/NAS/DT-trace
424 * GTNetS tracing re-worked
425 - adaptation to the tracing system of GTNets to cope
426 with modifications regarding the fullduplex mode
427 - new tesh files to check if gtnets tracing is ok
430 * Asynchronous communications through the functions:
431 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
432 * New function: MSG_load_platform_script()
433 to make possible using a lua script instead of XML files to set up platforms
434 * New function: MSG_set_function
435 to associate functions to processes, used when bypassing the parser
436 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
438 Platforms: Add some more examples in examples/platforms
439 * Grid'5000: see www.grid5000.fr
440 * *_30000_hosts.xml: various huge files [mainly scalability testing]
443 * Change the XML format. This is a very important modification. SimGrid
444 3.5 introduces a new hierarchical format based on the notion of
445 Autonomous Systems. Compatibility with old format is ensured
446 through the perl script provided in the install bin directory
447 bin/simgrid_update_xml.
448 It is now possible to build platforms with specific routing
449 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
450 connect several platforms together. We will try to provide soon
451 set of realistic platforms exploiting these properties (have a
452 look at examples/platforms/ for the moment).
453 * Take the opportunity of the XML format change to be a good XML citizen:
454 rename link:ctn to link_ctn and similar changes (also dealt with by
456 * Add a new routing scheme (rule-based) using regular expressions. It
457 enables to have an extremely low memory footprint when the
458 underlying routing is simple and can be compactly described. You
459 need to have libpcre4-dev (perl regular expressions) installed if
460 you want to use this routing scheme.
461 * Revive the cluster TAG and allow to easily and efficiently (both in
462 term of memory and speed) connect clusters together. Have a look
463 at teshsuite/simdag/platforms/ to see how this can be done. With
464 this tag, you can create clusters with thousands of tasks at no
465 cost (have a look at examples/platforms/).
466 Note: clusters are implemented as ASes, so there is no need for an
467 enclosing AS tag if you have only one cluster in your platform.
468 * Add new generic functions in the public interface that allows the user
469 to call SURF 'create_resource' methods from your code (same
470 functionality as the XML bypass mechanism but with a much lighter
472 * Add a new model (enabled through command line --cfg=network/model:SMPI)
473 that uses a piecewise linear approximation to produce better
474 results when exchanging small messages.
475 * Add a new parameter to handle correctly full duplex link and account
476 for interferences between uplink and downlink communications
477 (activate with --cfg=fullduplex:1).
480 * Rename the SD_READY (all dependencies are satisfied and task is
481 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
482 dependencies are satisfied) state.
483 This prevents a confusion between the notion of "ready to schedule"
484 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
485 simulated" (SD_RUNNABLE) used by the simulation kernel.
486 * Change the way a task is considered as ready. Instead of removing
487 dependencies when a task is done, a counter is decreased. This way,
488 it is always possible to reach ancestors thanks to the
489 SD_taks_get_parents function (even after the end of the simulation.)
490 * Change the return type of SD_Simulate from (SD_task_t*) into
491 xbt_dynar_t. This function was in handling a dynar internally and
492 converted it into a NULL terminated array for historical reasons.
493 * New function SD_dotload(char*) to load a DAG described in dot
494 format. This loader and the corresponding examples require the
495 installation of the graphviz library.
496 * Fix a bug in the management of tasks of size 0 in the surf network
497 models. This problem was only visible with SIMDAG and you should
498 thus disregard results produced with earlier versions if you
499 relied on this feature (some tasks were blocked because of this).
500 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
501 even though your applications were doing only point-to-point
502 communications and sequential computations. Now you can really use any
503 model you want (of course, if you create real parallel tasks, which are
504 not implemented in most models beside ptaskL07, this will abort).
505 * Add an example that schedules a DAX on an heterogeneous platform
506 using a Min-Min strategy.
507 * New function SD_workstation_get_current_task() that returns the kind
508 of task currently running on a workstation in the sequential access
510 * Raise some warnings when unexecuted tasks remains at the end of the
511 simulation. This is usually caused by cycles in the DAG.
514 * New function: SIMIX_process_set_function() called by MSG_set_function
515 * Change the underlying waiting queue in semaphores so that a process
516 can wait on several of them simultaneously (as in waitany).
517 * Fix the way to handle tokens in semaphores so that all access patterns
518 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
519 * kill the dirty pimple SIMIX_message_sizes_output()
520 Please use (proper) visualization instead
523 * New data container: setset (set of sets of elements)
524 * New module: mmalloc (mapped malloc, allowing to have several
525 independent segments of malloc)
526 * New function: xbt_dict_cursor_set_data()
527 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
528 * New function: xbt_dynar_is_empty()
529 * New function: xbt_fifo_get_last_item()
530 * Fix xbt_dynar_shrink(): use the right element size.
531 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
532 uninitialized areas during expand.
533 * Fix semaphores: previous implementation was severely broken.
534 * Use library init/fini functions for our initialization.
535 - you can use logs and other feature as soon as you want in your
536 code (even before the xbt_init / MSG_init)
537 - xbt_exit is now a no-op and produce a warning when used.
540 * Port GRAS to new SIMIX mechanisms. This allows gras users to
541 benefit from the latest improvement to the simulation kernel.
542 * Kill measurement sockets for now. If you rely on them, sorry. This
543 release is not for you. This feature will be reintroduced in the
544 future, but we cannot delay the release any further.
545 * New function: gras_msgtype_get_name().
546 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
547 * Fix (at last) the pmm example: it should not randomly fail anymore.
549 Build chain: bug fixes and overall polishing
550 * Cmake is now stable enough. Hence, we killed the autotools.
551 * Port to windows ( TM :)
552 * Fix the 'make install' target.
553 No need to use 'make install-simgrid' anymore
554 * Introduce a 'make dist' target compiling a *source* archive
555 'make package' compiles a binary archive
556 * Compile java files only on need
557 * Add --cd and --setenv command line options to tesh
558 * Out of source builds are not fully supported yet, but we are close
559 * Enable supernovae and optimization flags by default for our users
562 * Add layer to set up environment directly from lua, without XML.
563 * The effect of gras_stub_generator can be achieved through
564 lua too (check examples/gras/console/ping_generator.lua)
566 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
568 SimGrid (3.4.1) stable; urgency=low
570 The "Polishing easter eggs is probably a good idea" release.
571 This is a bug fixes release only.
575 * Fix a bug preventing the tasks from begin garbage collected.
578 * Fix a bug occuring when a host involved in a communication fails.
579 This was not detected properly by the other peer involved in the
580 communication. Now, it's reported as a network error.
583 * Warn the user about loop dependencies in data flow of DAX files
584 * Obey the control-flow dependencies of DAX files
587 * Add option "enable_smpi" allowing to not compile SMPI.
588 Probably useful for the (Mac) users experiencing a build error here
589 * Improve the detection of lua5.1 and ruby1.8
591 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
593 SimGrid (3.4) stable; urgency=low
595 The "Easter in Cargese" release. Also known as (major changes):
597 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
598 ~> bindings were greatly improved
599 ~> new tracing infrastructure for better visualization introduced
601 * the "Welcome to configury modernity" release.
602 ~> we switched from autotools to cmake, and improved our cdash
605 A more detailled list of changes follow (full detail in svn log).
607 Java Bindings: Various Cleanups
608 * (install java-gcj-compat-dev on debian-like to use them)
609 * Remove put/get: no need to export deprecated interface in Java
610 Use send/receive instead.
611 * Cleanup the examples and add a README per directory
612 * Remove example autoDestination (that's the only way to go now)
613 * Remove example explicitDestination (was a plain copy of basic)
614 * Make JniException a runtime exception, so that there is no need to
615 declare the fact that you may encounter such a beast. I guess that
616 nobody will ever want to survive such error.
617 * Create specific errors for each MSG case of failure:
618 host failure, transfer failure, timeout, task cancelled
619 * Cleanup the exceptions that may get thrown by each function
620 * Other internal cleanups in Java bindings. Performance still bad :/
621 Ruby and Lua Bindings: create them
622 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
623 * That's new and great, you should try them out.
624 Same functionalities than Java bindings, only even less polished
626 * Kill the useless "rate" argument of SD_task_get_execution_time()
627 Everyone used to provide -1 as a value, it was not used, and the
628 semantic of a possible use wasn't even clear.
629 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
630 either as comm costs or compute costs to mean that there is no
631 such thing for that specific task.
632 * Add a SD_task_set_name() function
633 * Fix SD_task_unschedule() on typed tasks
634 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
635 * In DAX loader, accept useless 'level' attributes to <job> since
636 LIGO DAGs have them (seem to be to ease graphical representation).
638 * Add an example masterslave_mailbox.c using send/receive and not
639 the deprecated put/get interface.
640 * Kill the MSG_paje_output() function. It's a noop since 2 years.
641 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
643 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
644 (declare MSG_USE_DEPRECATED to still have the old name)
645 * Add a MSG_task_set_data() function
646 * About trace replay (see examples/msg/actions):
648 - Allow to work with splitted trace files for each process
649 Give the specific trace file as argument of each process,
650 and call MSG_action_trace_run(NULL)
651 You can still have one merged file for all processes.
652 - Fix implementation of collective operations
653 * Allow task_execute() on 0-sized tasks (closes #10063)
655 * This is the first release of SimGrid where SMPI is not considered
656 beta anymore (even if some corners should still be improved)
657 * Port over the new SIMIX_network submodule (internal refactoring)
658 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
659 * Implement more missing elements of the standard:
661 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
663 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
664 MPI_Scan() MPI_Get_processor_name()
665 - Added implementation of missing case for Alltoall (warning: it's
666 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
667 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
668 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
669 SMPI_MPI_Allgatherv()
671 - MPI_Waitsome() was broken
672 - Allow relative includes in smpicc
673 - Command line cfg argument 'reference_speed' was ignored...
674 - Some functions did not properly lead to auto-benching of user code
675 - smpicc passes -O2 by default (just like openmpi one)
677 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
678 * Bug fixes about timeouts during communications
679 * add SIMIX_message_sizes_output() as a pimple to write to file the
680 amount of messages per size. Use gnuplot to get histogram.
681 Pimple because that's the only user-visible function of simix,
682 defined directly in xbt.h (irk, sorry)
684 - Add a SIMIX_sem_get_capacity() function
685 - Fix interactions with processe resume/suspende
686 - release_forever() was stupidly broken
687 - Fix SIMIX_display_process_status() for processes in a semaphore
688 - Make SIMIX_sem_block_onto() user-visible
689 * Refactoring context stuff:
690 - Use pseudo-OOP for better modularity
691 - reimplement SIMIX_process_kill() without process_schedule() so
692 that the latter can take as invariant that it is called from
694 - Merge context_start into context_new for sake of simplicity
696 * Add a Vivaldi network model, coded live during SUD'10 ;)
697 * Rename configuration variables to start a hierarchy:
698 o cpu_model -> cpu/model
699 o network_model -> network/model
700 o workstation_model -> workstation/model
701 * New configuration variables:
702 o network/bandwidth_factor: correction to bandwith
703 o network/latency_factor: correction to latency
704 o netwotk/weight_S: correction to the weight of competing streams
705 * Add a long description to the models, that users can see with such
706 argument on the command line: --cfg=cpu/model:help
707 * --help-models display the long description of all known models
709 * config: add the ability to set a default value after registration
710 Does not override any previously set value (e.g. from cmd line)
711 * dict: allow to have integer key and data.
712 When so, you need to use the following functions
713 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
714 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
715 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
716 In contrary to regular dicts, the key is not malloced before copy.
717 Mixing scalar and regular elements in the same dict is not tested
719 * Allow to use xbt_dynar_shrink() to expend the dynar instead
720 Tracing for Visualization:
721 * SimGrid is now instrumented in order to generate a trace file for
722 visualization analysis: to use it, need to compile SimGrid with the
723 "tracing" option enabled, and instrument the program using SimGrid with
724 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
725 (among other functions).
726 * The instrumentation only traces the platform utilization for now
727 * Documentation to use the tracing functions and how to analyze the
728 traces with the Triva tool is written.
729 * More information about: SimGrid FAQ (in the section Tracing Simulations
732 * We moved to cmake as default build system. Autotools support will
733 be dropped soon. Check the FAQ for more info about how to use it.
734 * Greatly improved our cdash/ctest interactions
735 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
736 * Added memory checking tests with valgrind; lot of memleak fixing.
737 This may be the first release of simgrid with so few memory issues
738 * Added code coverage tests.
739 Our coverage is still improvable, but at least we see it on cdash.
741 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
743 SimGrid (3.3.4) stable; urgency=low
745 The "Desktop Grid needs love too" release (also called Xmas release).
748 * Major speedup in the maxmin system solving by using lazy evaluation
749 Instead of solving completely the maxmin system at each iteration,
750 only invalidate (and recompute) the modified parts.
751 This new feature is enabled in default models but you can try to
752 turn it on with "--cfg:maxmin-selective-update=1" for other models.
753 * Cas01 IMproved as default CPU model
754 This CPU model is the same Cas01 model, but it uses the
755 maxmin-selective-update flag and a heap structure to manage
756 actions on SURF kernel.
757 It reduces the complexity to find the next action to finish and,
758 consequently, it's faster than the old Cas01.
759 This is the new default CPU model (Cas01).
760 * Rename the old Cas01 model to Cas01_fullupdate
761 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
762 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
763 * CpuTI (CPU Trace Integration)
764 A new CPU model whose objective is simulate faster when using
765 availability trace files.
766 Instead of using a full featured, over engineered maxmin system for
767 CPU modeling, this model does the pre-integration of traces files
768 to calculate the amount of CPU power available, and so, executes
769 faster than the old CPU models.
770 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
771 * Use LV08 as default network model since it gives better accuracy
772 for small messages and shouldn't change things for big ones.
773 Use --cfg=network_model:CM02 to get the previous behavior.
776 ******************************************
777 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
778 ******************************************
779 * The new CPU model may changes simulations!
780 The point is that events occurring at the exact same timestamp
781 are not scheduled in the same order with the old and new
782 version. This may be enough to completely change the execution
783 of simulations in some cases.
784 * The new network model will change simulations!
785 This new model is more realistic than the previous one, so you
786 should consider redoing your old experiments with this model.
787 Sorry for the inconvenience.
790 * Introduce the supernovae compilation mode
791 When compiled that way, the whole SimGrid (or almost) is put in a
792 single compilation unit and compiled in one shoot.
793 This is to help gcc which has difficulties to inline stuff from one
795 The speedup seem to be above 15%, althrough more tests are needed on
796 amd64 to confirm that gain.
799 * Port of MSG's mailbox on top of SIMIX network
800 The put/get mechanism was greatly simplified on the way.
803 * New SIMIX network module. Provides:
804 - Mailbox: rendez-vous mecanism to find with who you want to speak
805 - Synchronous send/recv: easier and hopefully faster since the
806 logic is handled in the maestro process directly now
807 - Asynchronous send/recv: you dreamt of it? It's here now
808 Too bad that nobody cared enough to propagate the change to MSG.
809 * Add semaphores as SIMIX synchronization mechanism.
812 * new function SD_daxload(char*) to load a DAX file
813 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
814 * Introduce typed tasks. Specify its kind and cost at creation.
815 At scheduling, just give where it should be placed, and the cost
816 for each involved resource is automatically computed.
817 Existing constructors so far (more to come of course):
818 - SD_task_create_comm_e2e() for end-to-end communication
819 - SD_task_create_comp_seq() for sequential computation
820 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
821 * new function SD_task_dump() for debuging display
822 * new function SD_task_dotty(task,FILE*) writing to file the info
823 about the task in dotty format
824 * SD_task_dependency_exists() can now cope with having one of its
825 arguments NULL. If so, it tests whether the other argument has any
827 * Add getters on list of preceding/following tasks:
828 SD_task_get_parents(task) and SD_task_get_children(task)
829 * Add getters on amount of workstations and list:
830 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
831 * Add getter on task kind: SD_task_get_kind(task)
832 * Update the start_time and finish_time of tasks on completion/failure
833 * Bugfix: Remove task from state swags when destroyed
836 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
839 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
840 dynar with one separate thread per value of the dynar.
841 * Change the prototype of xbt_thread_create(), sorry.
842 Added a boolean parameter indicating whether we want to join this
843 thread (used in SG only for now)
844 * Implement xbt_thread_join and xbt_thread_yield in SG also.
847 * GTNetS wrappers should now be usable again (and betterly tested too)
848 * Fix a major regression from 3.2 where the timeout provided to
849 MSG_task_put_with_timeout() was used as absolute time before which
850 the comm should be done.
851 * Start to fix the <cluster> tag.
852 - Internal links should be good now (beside of the loopback, which
853 use the private link instead)
854 - paths to the external world is still rather broken
855 - the <route:multi> tag is just broken. Actually that's brain-dead.
856 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
857 to make it less stupid
858 ** Check your platform with teshsuite/simdag/platforms/flatifier **
859 * Fix a source-level compatibility glitch from 3.2: after defining
860 MSG_USE_DEPRECATED, you can use the old name
861 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
862 * Allow to compile from the SVN with automake 1.11
863 * Fix some problems when using the "start_time" tag in deployment XMLs.
864 * Fix #8569: XBT/synchro.h has redundant declarations
865 * Fix #8563: MSG return values and exceptions
866 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
867 * Integrate patch #8636: Obey DESTDIR when installing documentation.
868 Thanks to Robson Peixoto.
869 * Fix a vicious bug in dictionaries inducing that some elements were
870 not freed on xbt_dict_free()
872 Portability report of this version:
873 * Main portability targets:
874 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
876 - mac leopard on i386
877 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
878 but nothing critical.
879 * Other platforms: windows, AIX and others were not tested for this release
881 Timing report of this version:
882 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
883 scenario parameters). From 8h to a few seconds in desktop grid settings.
884 * Supernovae brings about 25% speedup on i386.
886 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
888 SimGrid (3.3.3) stable; urgency=low
890 The "Need for Speed" release.
892 The timings done to validate the 3.3.2 were faulty.
893 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
895 The problem was a conversion from a manually handled vector to
896 xbt_dynar_t on the critical path.
897 xbt_dynar_foreach calls functions, inducing stack management crap.
899 We inlined these functions and xbt_dynar_foreach is now breath taking.
900 We also inlined xbt_swag_belong on the way.
902 Here are some approximate speedup measurements (on master/slaves
903 simulations lasting between 10s and 20s each):
904 3.3.1 -> 3.3.2: about same performance
905 3.3.2 -> 3.3.3: 40% speedup
906 3.3.1 -> 3.3.3: 40% speedup
907 3.3.1 with inline patch -> 3.3.3: 30% speedup
909 Our reading is that the refactoring which occurred in 3.3.2 made us
910 suffer much more from the xbt_dynar_foreach low performance, but
911 once we solved this, this refactoring proved to be very performance
912 effective. From the 40% speedup, somehow, 10% are due to the
913 inlining and 30% to the refactoring.
915 That's a pitty that gcc cannot inline functions placed in other files
916 alone. We have to choose between:
917 - break the encapsulation (by putting private data structures and
918 accessors in headers files to help gcc)
919 - live with low performance
920 - switch to a decent compiler such as icc (not quite possible).
922 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
924 SimGrid (3.3.2) stable; urgency=low
926 The "Simplicity does not preceed complexity, but follows it" release.
928 The main contributors of this release were (lexical order):
929 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
932 * Extract the routing logic into its own object.
933 (was dupplicated in network.c and workstation_LV07.c;
934 Allows to implement other ways of storing that info)
935 => kill now useless network_card concept
936 - Use dynar to represent routes (instead of void** + int*)
937 - kill link_set (use surf_network_model->resource_set instead)
938 - Add a command-line option to choose the routing schema to use
939 - Add three new models:
940 * Floyd (shortest path computed at initialization)
941 * Dijikstra (shortest path recomputed all the time)
942 * Cached Dijikstra (shortest path computed on need)
943 All these models where contributed by Silas De Munck, and are
944 described in his ICCS09 paper.
946 * Simplify model declaration
947 (less redirections, less function to write when defining a model)
948 - Factorize stuff between models:
951 surf_model_resource_set(model)
952 surf_model_resource_by_name(model, name)
953 - Unify the types of models in s_surf_model_t (using an union)
954 - Embeed fields of common_public directly into s_surf_model_t
955 - Rename model methods:
956 action_free ~> action_unref
957 action_change_state ~> action_state_set
958 action_get_state ~> action_state_get
959 - Change model methods into functions :
960 (model)->common_public->action_use ~> surf_action_ref
962 * Implement a generic resource; use it as ancestor to specific ones
963 (allows to kill duplicated code in models)
964 Drawback: timer command don't need no name nor properties;
965 workstation_CLM03 don't need no properties
966 (but I guess we can live with those few bytes wasted)
968 * Improve the action object model
969 - implement a constructor avoiding dupplicated code about field
970 initialization in generic_action part.
972 * Kill the SDP model: it has an external dependency, is deprecated
973 in flavor of modern lmm models, and didn't compile since a while
976 * Relocation of the context module from XBT to SIMIX.
977 (the context were decoupled from the simix processes, duplicating a lot of code)
978 => a lot of code was factorized
979 - less overhead is introduced during scheduling
980 - simpler API for the context factory
981 - the logic for process creation,destruction and manipulation was simplified
982 * Simplification of the s_smx_process_t data structure.
983 => accesing the simix level data associated to a process is faster now,
984 and the code is a lot more readable.
987 * Implement some more MPI primitives:
988 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
989 -implementation: Bcast: flat or 2-ary tree (default),
992 Allreduce: Reduce then Bcast
993 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
994 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
995 Alltoallv: flat tree, like ompi
997 * Add support for optimized collectives (Bcast is now binomial by default)
998 * Port smpirun and smpicc to OS X
1001 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1002 Shout out if you used it.
1005 * Display the list of still queued messages in SG mode when existing
1009 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1010 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1011 * Add xbt_str_from_file(FILE*)
1012 * Add xbt_dict_get_key achieving a linear reverse search
1013 * Remove the context module
1015 Portability report of this version:
1016 * Main portability targets:
1017 - Linux(debian)/x86/context
1018 - Linux(debian)/x86/pthreads
1019 - Linux(debian)/amd64/context
1020 - Linux(debian)/amd64/pthreads
1021 On these, we still have the eratic breakages of gras/pmm and
1022 amok/saturate_sg reported in previous version. We still think
1023 that the tests are the cause of the fault, not the tested code.
1025 - Mac OSX Leopard/x86/context
1026 Still false negative in tesh autotesting.
1027 Smpi still fails, but this time because readlink does not accept -f
1028 Everything seems to work properly beside of that.
1031 - AIX version 5.3 (only tested contexts this time)
1032 Smpi still fails there because mktemp is not installed.
1033 Everything seems to work properly beside of that.
1035 I managed to compile it for the first time, but several breakages.
1036 Won't delay the release for this exotic platform.
1038 * Windows: it's still lagging behind. If you want to help, please
1041 Timing report of this version:
1042 This version seem to be more than 5% faster than 3.3.1 (on linux
1043 64bits with contextes). The gain is less than expected, we are
1044 investigating this for next release.
1046 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1048 SimGrid (3.3.1) stable; urgency=low
1051 * Implement a --cfg-help to show existing configuration variables
1052 * Build chain do not require doxygen in maintainer mode
1055 * fix a bug on struct sizeof computation, which prevented the
1056 exchange of arrays of structs in some conditions
1057 - added a regression test about this in datadesc_usage
1058 * Allow the exchange of 0-long dynamic vectors.
1059 - for that, use -1 as indicator of dynamic size instead of 0
1060 - This implied to change any size from unsigned long to long,
1061 reducing a bit communication abilities, but I guess that with
1062 64bits being quite common, this is more than enough.
1063 - This also induce a protocol change, thus bumping network protocol
1064 version from 0 to 1 (if we have external users, we have to get
1065 clean on that point too ;)
1066 - added two regression tests about this in datadesc_usage
1067 * Be more verbose when propagating local exceptions
1068 This helps debugging.
1069 * Display the status of simulated processes when receiving SIGINT in
1073 * Allow to control the simulation from a trace file.
1074 New functions MSG_action_register() and MSG_action_trace_run()
1075 The first one allows to associate a function execution to each
1076 kind of action while the second one parses a trace file and
1077 triggers the corresponding actions within the system.
1078 For now, only a toy example is provided in examples/msg/actions
1079 * Add an exemple of process migration in examples/msg/migration
1080 * Fix a bug in task exchange which broke MSG_task_get_sender()
1081 Add a teshsuite regression test for that.
1082 [Bug: if MSG_task_get_sender() is called after sender exit,
1084 * Fix a bug which prevented suspend/resume to work properly
1085 * Display the status of simulated processes when receiving SIGINT
1086 This fixes a regression of v3.3. due to the introduction of SIMIX
1087 * Bug fixing in failure management:
1088 - trace could not start by a failure at time 0
1089 - failure during communications were not working
1092 * Add SIMIX_process_set_name() to change the name of the current
1093 process in the log messages.
1094 * Store smx_hosts in a dict since we only retrieve them by name
1095 * Move the configuration infrastructure to surf
1098 * Move the configuration infrastructure to surf
1101 * Massive internal cleanups:
1102 - Store internal structures on processes instead of hosts (allows
1103 to have more than one process per host, in addition of being more
1105 - Cleanup the initialization/finalization process
1106 - Kill a whole bunch of unneeded synchronization:
1107 processes run in exclusive manner within the simulator
1108 - Move queues from global tables to process data fields
1110 - now accept -platform and -hostfile arguments
1111 - Pass the right rank value to processes according to the hostfile
1112 * Compile the examples by default, and use them as regression tests
1113 * Implement MPI_Wtime()
1114 * Change the reference speed to a command line option
1117 * TCP_gamma can now be specified as command line option using
1118 --cfg=TCP_gamma:10000000.0
1119 * Change the --surf-path cmd line option into --cfg=path:
1122 * Also include strbuff from xbt.h public header
1123 * xbt_ex_display(): do not free the exception after displaying
1124 This allows to do more with the given exception afterward.
1125 Users should call xbt_ex_free() themselves.
1129 Portability report of this version:
1130 * Main portability targets:
1131 - Linux(debian)/x86/context
1132 - Linux(debian)/x86/pthreads
1133 - Linux(debian)/amd64/context
1134 - Linux(debian)/amd64/pthreads
1135 These targets fail about 1/10 of times on gras/pmm, but we believe
1136 that this is because of the test, not because of simgrid.
1137 amok/saturate_sg fails even more rarely, and the test may not be
1140 - Mac OSX Leopard/x86/context
1141 The test suite still spits tons of errors because some obscure
1142 force prevents us from removing the temporary directories
1143 arguing that they still contain some metadata I've never heard of.
1144 Smpi fails because seq is not installed.
1145 Everything seems to work properly beside of that.
1148 - AIX version 5.3 (both contexts and pthread)
1149 Smpi still fails there because mktemp is not installed.
1150 XML inclusions seems rosty on AIX.
1152 * Windows: it's still lagging behind. If you want to help, please
1155 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1157 SimGrid (3.3) stable; urgency=high
1161 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1162 [Malek Cherier & Mt]
1164 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1167 Until now, GRAS were using MSG as an interface to SURF. It was
1168 quite difficult because both interface have several differences
1169 (MSG channels vs GRAS sockets were the most notable point).
1171 This also opens the gate to SMPI (which should occur soon) and speed
1172 up simulations by to 40% (even if it were not the main goal).
1174 **************************************
1175 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1176 **************************************
1177 The point is that events occuring at the exact same timestamp are
1178 not scheduled in the same order with the old and new version. This
1179 may be enough to completely change the execution of simulations in
1180 some cases. Sorry for the inconvenience.
1182 * Cleanup and upgrade the XML format to push further scalability
1183 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1185 * Improve the testing infrastructure with tesh. Now a very large part of
1186 the code is tested not only by being run but also by checking that the
1187 output match an expected output [Mt].
1189 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1190 is really less memory-demanding, which should allow you to use
1191 larger files in SimGrid [AL].
1193 * Inform valgrind about our contextes, so that it becomes usable
1194 with the default (and more effecient) version of SimGrid
1195 [contributed by Sékou Diakite, many thanks]
1198 * Introduce a listener thread in charge of receiving incomming
1199 messages from the network. It allows to overlap communication and
1200 computation but most notably, it removes some stupid deadlocks due
1201 to the fact that so far, a process could not send and receive at
1202 the same time. This made most non trivial communication schema
1204 * Convert the PIDs from long int to int to match the MSG ones (and
1205 linux ones too) [Mt]
1206 * New function: gras_agent_spawn() to launch a new process on
1207 current host. Only working in simulation for now. [Mt]
1208 * New function: gras_os_hostport() returning a constant form (ie,
1209 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1212 * Make the backtrace of exceptions more human readable [Mt]
1213 * New module: xbt/str [Mt]
1214 a ton of string utility functions (split, join, printf to a newly
1215 allocated buffer, trim, etc)
1216 * New module: xbt/hash [Mt]
1217 SHA1 hashing algorithm (more to come if needed)
1218 * New module: xbt/synchro [Mt]
1219 synchronization tools (mutex and conditions) working the same way
1220 in simulation and in real life (mainly useful for GRAS, but not
1222 * New module: xbt/queue [Mt]
1223 classical producer/consumer synchronization scheme
1224 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1225 (using the classical functions will get serialized) [Mt]
1226 * Make dictionary internal table dynamic. No need to specify its size
1227 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1229 * Make sure the log channels are organized as a tree under windows
1230 (because of ANSI C compatibility issue, any channel were child of
1231 root directly) [Mt].
1234 * Cleaned many thing in surf and fixed a few bugs [AL].
1235 * Add a nice command line configuration mechanism to compose models [AL].
1236 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1237 the previous one (KCCFLN05). It relies on something that looks like
1238 a max-min sharing mechanism but cannot be written as such. A new solver
1239 was thus designed [AL].
1240 * Add a new solver to lmm. Based on Lagrange optimization and
1241 gradient-based descent, it enables to efficiently maximise systems s.a
1243 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1245 This solver enables to propose two new network models for TCP Reno and
1246 TCP Vegas based on Low's work. These models still need to be fully
1247 tested though [Pedro Velho].
1250 * Bug fix in SD_simulate. Now the time bound given as argument is
1252 * Use the new parallel task model (ptask_L07) as default.
1253 * Use the SURF command line configuration mechanism.
1254 * 0-size tasks (for synchronization) should now work.
1256 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1258 SimGrid (3.2) stable; urgency=high
1262 We still experience issues on this platform, but we believe that at
1263 least MSG is usable.
1265 GRAS API BREAKAGE (for simplification purpose, sorry):
1266 * the gras_msgtype_by_name is not used anymore. Instead of
1267 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1268 you can write (and must)
1269 gras_msg_send(toserver, "request", &request);
1270 - If you still want to pass a gras_msgtype_t to the function (to cache
1271 the type and avoid the lookup time), use the gras_msg_send_() variant.
1272 - Impacted functions:
1273 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1274 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1275 * The callbacks are now expected to return 0 when everything went well
1276 (just like the main() function)
1278 GRAS new features and improvements:
1279 * New module mecanism where user code can use per process globals [Mt]
1280 This is similar to gras_userdata_*() functions, but for libraries. It
1281 factorize some code developped over and over in the examples and AMOK.
1282 It has still to be documented and used (only amok/peermanagement is
1284 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1285 loss when they were small enough to fit into the buffer and sent quickly
1286 enough so that they can all get received in one shoot.
1287 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1288 if not found. Use the *_or_null() variant for the old semantic.
1289 * In gras_msg_handle, do not discard messages without callback.
1290 They are probably messages to be explicitly awaited later (ie, proofs of
1291 mis-synchronization in userland since they are sent before being awaited)
1292 No big deal usually.
1293 * gras_socket_meas_send/recv: semantic changed!
1294 The numerical arguments used to be (1) the total amount of data to send
1295 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1296 messages. This was need for the fool willing to send more than MAXINT
1297 bytes on quite fat pipes.
1300 * Do really rename the hostmanagement module to peermanagement. [Mt]
1301 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1302 but this is rather new and this was documented in the module
1303 documentation (poor excuses, I admit)
1304 * Bandwidth measurement semantic changed! This follows the changes to
1305 gras_socket_meas_send/recv explained above.
1308 * A sequential mode has been added to the workstations. When a workstation
1309 is in sequential mode, it can execute only one task, and the other tasks
1310 are waiting in a FIFO. [Christophe Thiery]
1313 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1314 model for SIMDAG. [Christophe Thiery]
1315 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1316 equal to 0, which caused some bad behaviors in
1317 saturated_constraint_set_update. I now use a threshold mechanism like in
1321 * When running manually src/testall, you select specific units [Mt]
1322 testall is the result of our cunit mecanism, and should replace all
1323 the scripty thingy around since bash don't run easily on billware.
1325 * A mallocator system has been added. [Christophe Thiery]
1326 Mallocators allow you to recycle your unused objects instead of freeing them
1327 and allocating new ones.
1329 Documentation update:
1330 * FAQ reworking + New FAQs:
1331 - "Valgrind spits tons of errors!" [Mt]
1332 - "How to repport bugs" [Mt]
1333 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1334 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1335 - Communication time measurement within MSG [AL]
1336 - I experience weird communication times when I change the latency [AL]
1337 * GRAS tutorial [Mt]
1339 - an introduction to the framework and to the used communication model
1340 - an initiatic tour introducing the most proheminent features:
1342 . Lesson 0: Installing GRAS
1343 . Lesson 1: Setting up your own project
1344 o Part 2: Message passing
1345 . Lesson 2: Exchanging simple messages
1346 . Lesson 3: Passing arguments to the processes (in SG)
1347 . Lesson 4: Attaching callbacks to messages
1348 . Lesson 5: Using globals in processes
1349 . Lesson 6: Logging informations properly
1350 . Lesson 7: Using internal timers
1351 . Lesson 8: Handling errors through exceptions
1352 . Lesson 9: Exchanging simple data
1353 . Lesson 10: Remote Procedure Calling (RPC)
1354 . Lesson 11: Explicitely waiting for messages
1355 . Recapping of message passing features in GRAS
1356 - A HOWTO section containing:
1357 o HOWTO design a GRAS application
1358 More are due, of course. They will come latter. In the meanwhile, you can
1359 check the examples which are still here.
1361 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1363 SimGrid (3.1) stable; urgency=high
1367 There was a stack corruption somewhere, visible only when optimizing
1368 with these versions. [Vince]
1371 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1372 module that enabled to play with Directed Acyclic Graphs. It is built
1373 directly on top of SURF and provides an API rather close to the old
1374 SG. Some old codes using SG are currently under rewrite to check that
1375 all needful functions are provided. [Christophe Thiery]
1378 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1379 extension of the classical CLM03 model that gracefully handles
1380 failures. This is now the default model for MSG and GRAS. It doesn't
1381 handle parallel tasks yet though. [AL]
1382 * Bug fix: Weights were not correctly set in the network part.
1383 WARNING: This may have resulted in incorrect results with simulations
1384 where there are more than one flow on a given link. [AL]
1387 * After a (long ?) discussion on simgrid-devel, we have decided that the
1388 convention we had on units was stupid. That is why it has been decided
1389 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1390 WARNING : This means that all previous platform files will not work as
1391 such with this version! A warning is issued to ask users to update
1393 A conversion script can be found in the contrib module of the CVS, under
1394 the name contrib/platform_generation/surfxml_update.pl [MQ]
1397 * Bug fix: Processes were started in reverse order, wrt deployment file.
1398 WARNING: if your code relies on this bug, please fix it. [AL]
1399 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1400 being executed on two different locations. [AL]
1401 * Bug fix: Failures are now better supported thanks to Derrick's tests
1402 (there was many failure situations I hadn't thought of and that weren't
1403 correctly handled). [AL]
1404 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1408 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1409 Exception occurring on server-side are propagated back to client (!).
1411 API CHANGE: the callback changed their prototype. Change:
1412 int my_handler(gras_socket_t expeditor, void *payload_data) {
1414 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1415 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1417 * New! function: gras_msg_handleall to deal with all messages arriving
1418 within a given period.
1419 * New! function: gras_socket_server_range to get a server socket in a
1420 range of port numbers (ease to avoid port number conflicts) [MQ]
1421 * New! gras processes display their backtrace when they get a SIGUSR1
1422 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1423 Sweet to debug RL processes [MQ]
1427 - Do not force experiment sizes to be expressed in kb, or it becomes
1428 impossible to measure the latency this way (needs one byte-long tests)
1429 WARNING: this changes the amok_bw_* function semantic. [MQ]
1430 - Implements the link saturation stuff. [MQ]
1431 * Peer management module:
1432 New! module factorizing code that we wrote over and over [MQ].
1435 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1436 - Test units are placed directly into the library code, they get extracted
1437 automatically and placed into the src/testall binary.
1438 - Convert most of the XBT tests to this system.
1439 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1440 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1441 live in the same C code. This required to change a little bit the API
1442 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1443 * New module: sparse graph structure with basic algorithms (this is work
1444 in progress and the API is not considered to be frozen yet). [AL]
1445 * Display more information on backtraces: source line & function names are
1446 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1447 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1448 * Reworked a little bit some #include statements to load only required
1449 headers. Some user code that relied on SimGrid to include stdlib or
1450 stdio may need to include it by themselves. [AL]
1451 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1453 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1454 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1455 but I guess that almost nobody used those parts. [MQ]
1457 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1459 SimGrid (3.0.1) stable; urgency=low
1462 * Unfortunately, I had missed 5 misnamed functions:
1463 xbt_fifo_item_t xbt_fifo_newitem(void);
1464 void xbt_fifo_freeitem(xbt_fifo_item_t);
1465 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1466 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1467 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1468 They're now deprecated. Please use their new versions:
1469 xbt_fifo_item_t xbt_fifo_new_item(void);
1470 void xbt_fifo_free_item(xbt_fifo_item_t);
1471 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1472 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1473 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1475 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1476 * Documentation: xbt_log module unmercifully reworked [MQ]
1477 * Bugfix: there was a problem with the ending of contexts with
1478 the pthread backend. It caused some weird deadlock or behavior
1479 depending on the pthread implementation. [AL]
1480 * Bugfix: get the exceptions raised in the simulator repport where
1481 and why they come from when they are not catched in time [AL, MQ]
1484 * Bugfix: Do repport the error when two non-connected hosts try to
1485 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1488 * Add additionnal checkings on communications. Assert that two
1489 communicating hosts are connected by a set of links... [AL]
1492 * Add additionnal checkings on channel values in communication [AL]
1493 * New: MSG_task_get_source to see on which host a task was generated [HC]
1494 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1495 the number of tasks waiting to be received on channel and sent
1497 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1498 waits for the first task coming from a given host.. [AL]
1500 GRAS new functionnalities: [MQ]
1501 * Enhance the parsing macro to allow the size of multidimentional objects
1502 to be given thru annotations.
1503 * New example (and documentation): Matrix Multiplication a la RPC
1504 (as when I was young!) and fix a bunch of bugs found on the way.
1506 GRAS performance improvements: [MQ]
1508 * Reduce the amount of cbps creation/destruction by making it static to
1509 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1512 * Change libdata to a set so that we can search for stuff by ID (and thus
1513 reduce the insane amount of dict lookups)
1516 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1517 RL communication on top of SG since it's so uneffective.
1518 It may also be used for inter-thread communication in RL, one day.
1519 * Use gras_datadesc_copy() to exchange messages on top of SG
1521 - improve message exchange performance on top of SG
1522 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1523 * Don't exchange on the network the size of the used part of buffer,
1524 instead, specify the possible buffer size to read().
1526 - reduces the amount of read/write calls (one pair per exchange)
1527 - reduces the amount of exchanged data (the size)
1528 - allows to retrieve all arrived data on receiver side, if we don't need
1529 it right now (subsequent read will peek the buffer)
1530 - allows the receiver to proceed with the begining of the stream before
1531 everything is arrived
1532 - make it possible to build an iov transport (using readv/writev)
1534 - take care of the data with non-stable storage (like stacked data),
1536 * If possible, TCP send uses vector I/O (when writev() is here)
1537 - Don't use it for receive since we send data sizes and data on the
1538 same stream, so we wouldn't be able to chain large amount of chunks
1539 before having to flush the stuff to read the size.
1540 * Rework the transport plugin mecanism to simplify it and reduce the
1541 amount of pointer dereferencement when searching for the right function
1544 * I guess that now, we do almost as few system calls as possible while
1545 doing as few data copy as possible.
1547 To improve it further, we could try to send all the sizes first and then
1548 all the data (to use iov on receiving size), but it's only a partial
1549 solution: when you have 2 dimensional data, the sizes of the second
1550 dimension is data of the first dimension, so you need 3 streams.
1552 I'm not sure the potential performance gains justify the coding burden.
1554 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1556 SimGrid (3.00) stable; urgency=high
1559 * New! Give the possibility to hijack the surf parser and thus bypass
1560 MSG_create_environment and MSG_launch_application. Have a look at
1561 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1563 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1565 SimGrid (2.96) unstable; urgency=low
1570 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1571 This deprecates the xbt_error_t mecanisms.
1572 It modifies (simplifies) all XBT and GRAS API.
1573 MSG API keeps unchanged (exceptions raised by XBT are catched from
1574 within MSG and masked with existing error handling facilities)
1577 * New! Add a FATPIPE model. [AL]
1578 * New! Add a parallel task model. [AL]
1579 * New! Add automatically a loopback interface (in the default
1580 network model) if none was precised.
1583 * Bugfix: MSG_process_resume now works with the current running process.
1585 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1586 * Modification of MSG_task_get_compute_duration. Once a task has been
1587 processed, the value returned by this function is now equal to 0. [AL]
1588 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1589 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1590 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1591 of a m_task. For now, MSG_task_cancel only works with computation
1593 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1594 of the processor (in Mflop/s) regardless of the current load on the
1596 * API Change: use proper naming convention for MSG_getClock and
1597 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1599 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1600 This function changes the priority of a computation task. This priority
1601 doesn't affect the transfer rate. A priority of 2 will make a task
1602 receive two times more cpu power than the other ones. This function
1603 has been added to suit the needs of Nguyen The Loc and hasn't been that
1604 much tested yet. So if it fails, please report it and send me your code.
1606 * API Change: removed all functions and types that were marked "deprecated"
1607 since many months. Renamed MSG_global_init_args to MSG_global_init.
1609 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1611 SimGrid (2.95) unstable; urgency=low
1614 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1615 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1616 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1619 * Bugfix: Allow absolute paths for platform description files [MQ]
1620 * Bugfix: do free the variables after use. Leads to drastic performance
1622 * Implement max_duration (ie, timeouts) on resources [AL]
1625 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1627 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1628 the needed support (timeouts) [AL]
1631 * Implement measurement sockets. You can now get the bandwidth between two
1632 hosts thanks to AMOK (see below). [MQ]
1633 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1634 dynar over the network (yeah) [MQ]
1635 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1636 * simple/static token-ring example. [Alexandre Colucci and MQ]
1637 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1638 in place before [MQ]
1641 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1642 type parsing. "short" and "long" modifiers were also ignored; other
1643 modifier (such as reference level) are still ignored. [MQ]
1644 * Embeed the buffer size within the buffer itself on SG. [MQ]
1645 That way, send() are atomic and cannot get intermixed anymore (at least
1646 the ones which are less than 100k; bigger messages still have the issue)
1647 * Array size pushed by the field, not by the field type (or each
1648 and every long int will push stuff to the cbps) [MQ]
1649 * use select() to sleep since it allows to portably sleep less than one
1652 GRAS (minor cleanups)
1653 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1656 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1657 * type category 'ignored' killed as it was never used and were difficult
1659 * whether a type can cycle or not is now a flag, leaving room for more
1660 flags (as "ignored", if we feel the need one day ;)
1661 * Rename raw sockets to measurement sockets since "raw" has another
1662 meaning in networking community.
1665 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1666 and offers features not belonging to GRAS but that most applications
1667 need. One day, it may be a set of plugins loadable at runtime.
1668 * New module: bandwidth
1669 bandwidth measurement between arbitrary nodes running this module. [MQ]
1671 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1673 SimGrid (2.94) unstable; urgency=low
1675 The first beta release of SimGrid 3 !
1679 * Update the main page and the FAQ. Adding references to gforge.
1682 * Add a gras_os_getpid function.
1685 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1686 * Extend the logs so that they also print PID, hostname, date, ... if
1688 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1689 and kill the old version which were in testsuite/
1690 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1694 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1695 for users. Instead, each programming environment (GRAS, MSG,...) use it
1696 when needed to provide such a feature to users.
1697 Don't shortcut the mecanism or you will also shortcut the virtualization
1698 you need on the simulator.
1702 * Cleanups in configury with regard to compile optimization/warning flags.
1703 Also add -fno-loop-optimize to any powerpc since it's the optimization
1704 killing gcc (< 3.4.0).
1705 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1706 needed by MSG examples complications
1707 * Borrow configury beautifications from PHP
1710 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1711 warning (thanks loris for stumbling into this one).
1712 * Bugfix: stop loading private headers (gras_config.h) from the public
1716 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1717 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1718 RL would imply a DNS resolver, which is *hard* to do in a portable way
1719 (and therefore delayed).
1720 * Implement a real timer mecanism and use it in timing macros. This allows
1721 to avoid rounding errors and get a 0.000005 sec precision in timing
1722 macros. While I was at it, various cleanups:
1723 - allow to declare more than one timed section per file (fix a stupid bug)
1724 - move some private declaration to the right place
1725 - merge conditional execution and timing macros into emulation module
1726 - document the module
1727 - make sure the module cleanups its mess on gras_exit
1728 * Documentation improvements:
1729 - (new) how to compile applications using GRAS
1730 - (new) emulation support (timing macros)
1732 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1734 SimGrid (2.93) unstable; urgency=low
1736 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1739 - Use Paje properly where used. Still to be sanitized properly.
1740 - Portability fix: Add an implementation of the contexts using pthread
1744 - Add xbt_procname(): returns the name of the current process.
1745 Use it to show the current process's name in all logging.
1747 - fix detection of older flex version and the reaction, since we do
1748 depend on modern ones (we use lex_destroy)
1749 - Better separation of SG and RL in the libs: remove all simulation code
1750 from libgras. As a result, this lib is now only 200k when stripped.
1751 Some of the xbt modules may also be duplicated (two sets and such) and
1752 should be cleaned/killed before SG3.
1753 - Insist on using xlC on AIX because of weird problems involving gcc there.
1754 - Cleanup the make remote stuff. This is now done by scripts
1755 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1756 mainly for our private use, but we're working on changing them to user
1759 - Bugfix: flush the socket on close only if there is some *output*.
1760 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1762 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1765 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1767 SimGrid (2.92) unstable; urgency=low
1769 Alpha 3 on the path to SimGrid 3
1773 - New! First try of benchmarking macros.
1774 - New! First try so that gras_stub_generator generate deployment and
1775 remote compilation helpers.
1777 - Bugfix: Initialization fix in msg_test.
1781 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1783 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1785 - API Change: gras_os_sleep to take the amount of seconds as a double.
1786 Accepting an int was error prone since it was the only location where
1787 seconds were coded as such. It leaded to damn rounding errors.
1788 - Bugfix: Hard to belive that timers ever worked before this.
1790 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1792 SimGrid (2.91) unstable; urgency=low
1794 Alpha 2 on the path to SimGrid 3
1798 - Bug fix in the lmm_solver.
1800 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1801 through the function MSG_paje_output.
1802 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1803 - It is possible to bound the rate of a communication in MSG with
1804 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1805 to write it in the changelog).
1806 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1811 - Various cleanups to the autotools stuff
1812 - Begin to move Gras examples to examples/gras/
1813 - Let make distcheck work again (yeah!)
1815 - documentation overhauled using doxygen.
1816 gtk-doc-tools is dead in SimGrid now.
1817 - Automatically extract all existing logging categories, and add the list
1818 to the documentation (long standing one, to say the less)
1820 - Cleanup the known architecture table. Reorder the entries to group what
1821 should be, and use a more consistent naming scheme.
1822 (some of the test dataset are still to be regenerated)
1823 - New! Allow library to register globals on each process just as userdata
1825 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1826 lookup time (for now).
1827 Use it in msg and trp.
1828 This cleans a lot the internals and helps enforcing privacy of the
1829 headers between the gras components.
1830 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1831 - Bugfix: gras_os_time was delirious in RL.
1832 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1834 - Reenable GRAS now that it works.
1836 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1838 SimGrid (2.90) unstable; urgency=low
1840 Alpha 1 on the path to SimGrid 3
1842 * It is a long time since the last release of SimGrid. I'm sorry about
1843 that but as I had told you, I was rewriting a lot of things. I apologize
1844 to those who had been reporting bugs to me and that I had not answered.
1845 If your bug is still in the new version, please tell me. Here is a
1846 summary of the main changes.
1848 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1849 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1850 portability, speed, and a lot more but you'll figure it out later.
1851 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1852 I wanted to release the new MSG as soon as possible and I have
1853 broken GRAS, which is the reason why, for now, only MSG is fully
1854 functional. A laconic description of these projects is available
1855 in the documentation.
1857 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1858 kernel optimized for our needs (called SURF but only the developers
1859 should use it). Hence, MSG is now roughly 30 times faster and I think
1860 that by rewriting a little bit MSG, I could event speed it up a little
1861 bit more. Beside the gain in speed, it is also much easier to encode a
1862 new platform model with SURF than it was with SG. More to come...
1864 * REVOLUTION 3: I have tried to change a little as possible the API of
1865 MSG but a few things really had to disappear. The main differences
1866 with the previous version are :
1867 1) no more m_links_t and the corresponding functions. Platforms are
1868 directly read from a XML description and cannot be hard-coded
1869 anymore. The same format is used for application deployment
1870 description. The new format is described in the documentation.
1871 Have a look in tools/platform_generation. There is a tiny script
1872 that converts from the old platform format to the new one. Concerning
1873 the application deployment format, parsing the old one is tricky.
1874 I think most of you should however be able to convert your files. If
1875 it is really an issue, I can write a C code that does the conversion.
1877 2) the toolbox tbx does not exist anymore. We now have a library
1878 with much more data-structures but without the hash-tables (we have
1879 dictionaries that are much faster).
1881 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1883 *****************************************************************************
1884 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1885 * don't feel like converting the oldies. (Mt) *
1886 *****************************************************************************
1889 Version 2.90: "the long awaited one"
1890 - Finished rewriting and debugging MSG. Rewrote the documentation.
1891 - disable GRAS for now since it needs to be ported to the newest SG
1894 - Finish the port to windows (using mingw32 for cross-compile)
1897 - Main loop and datastructures of SURF. A cpu resource object is
1898 functional. Surf can thus be used to create cpu's with variable
1899 performance on which you can execute some actions.
1901 2004-11-15 Martin Quinson
1902 - Port to ARM. Simply added the alignment and size descriptions. Should
1903 work, but the ARM machines are so slow that I didn't had the opportunity
1904 to 'make check' over there yet.
1906 2004-11-15 Arnaud Legrand
1907 - Trace manager now written. It uses a heap structure and is therefore
1908 expected to be efficient. It may however be speeded up (particularly
1909 when many events occur at the same date) by using red and black
1910 trees. One day maybe...
1911 - Max-min linear system solver written. It uses a sparse matrix
1912 structure taking advantage of its expected use. Most operations are
1913 O(1) and free/calloc are called as few as possible. The computation of
1914 the minimum could however be improved by using a red and black tree
1917 2004-11-03 Arnaud Legrand
1918 - Rename every gras_* function that was in xbt/ to its xbt_
1920 - Add a heap and a doubly-linked list to xbt
1921 - Added a dichotomy to the dictionaries. make check works as well before
1922 so I assume that the patch is correct. I do not know however if things
1923 run effectively faster than before now. :)
1925 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1926 SimGrid, and the version number is bumped to 2.x
1928 2004-10-29 Martin Quinson
1929 - Introduction of the remote errors.
1930 They are the result of a RMI/RPC on the remote machine.
1931 ErrCodes being scalar values, you can't get the host on which those
1932 errors did happen. Extending the error mechanism as in Gnome is possible.
1933 No idea yet whether it is a good idea.
1935 2004-10-28 Martin Quinson
1936 - Interface revolution: the Starred Structure Eradication.
1937 I used to do typedef struct {} toto_t; and then handle *toto_t.
1938 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1939 - ???_t is a valid type (builded with typedef)
1940 - s_toto_t is a structure (access to fields with .)
1941 - s_toto is a structure needing 'struct' keyword to be used
1942 - e_toto_t is an enum
1943 - toto_t is an 'object' (struct*)
1945 typedef struct s_toto {} s_toto_t, *toto_t;
1946 typedef enum {} e_toto_t;
1947 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1948 s_toto_t) is private.
1950 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1951 so that it can set the variable to NULL. It was so for dicts and sets,
1952 it changed for dynars.
1954 - Fix a bunch of memleaks in dict_remove
1955 - Fix a bug in sg/server_socket opening: it failed all the time.
1957 2004-10-07 Martin Quinson
1958 - Speed up dynar lookup operation a bit.
1960 gras_dynar_get is dead.
1962 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1963 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1964 (which gives you the address of the stored data).
1966 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1967 copy of the data using an affectation to do the job and not a memcpy.
1969 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1970 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1972 It does not really speedup the dynar test because they are
1973 setting elements all the time (and look them seldom). But the dict does
1974 far more lookup than setting.
1976 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1978 2004-10-05 Martin Quinson
1979 - Allow to (en/dis)able the cycle detection at run time.
1981 Whether we should check for cycle or not is now a property of each
1982 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1983 datadesc_cycle_unset allow to remove this property when previously set.
1985 Note that the cycle detection is off by default since it impacts the
1986 performance. Watch the data you feed GRAS with ;)
1988 This property is hereditary. Any element embedded in a structure having it
1989 set have it set for the time of this data exchange.
1991 You should set it both on sender and receiver side. If you don't set it on
1992 sender side, it will enter an endless loop. If you forget on receiver
1993 side, the cycles won't be recreated after communication.
1995 - Header reorganization.
1996 Kill gras_private.h, each submodule must load the headers it needs.
1998 2004-10-04 Martin Quinson
1999 - Interface revolution: do not try to survive to malloc failure.
2001 Now, gras_malloc and friends call gras_abort() on failure.
2002 As a conclusion, malloc_error is not a valid error anymore, and all
2003 functions for which it was the only gras_error_t return value are
2004 changed. They now return void, or there result directly.
2005 This simplify the API a lot.
2007 2004-09-29 Martin Quinson
2008 - Re-enable raw sockets.
2009 Created by gras_socket_{client,server}_ext;
2010 Used with gras_raw_{send,recv}
2013 It should allow to kill the last bits of gras first version soon.
2015 This is not completely satisfactory yet (duplicate code with
2016 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2019 - Simplify transport plugin (internal) interface by not passing any
2020 argument to _server and _client, but embedding them in the socket
2023 2004-09-28 Martin Quinson
2024 - Finish the port to AIX.
2025 autoconf was my problem (segfault within the malloc replacement
2026 function. No idea why)
2028 2004-09-16 Martin Quinson
2029 - Fix some size_t madness on 64bit architectures.
2031 2004-09-08 Martin Quinson
2032 - Reduce the number of system headers loaded, overload some more system
2033 calls (such as malloc to cast the result of the system one, and work
2035 - Fix and reintroduce the config support
2037 2004-09-07 Martin Quinson
2038 - Source code reorganization to allow Arnaud to surf all over there.
2039 - Allow to document the logging categories.
2040 - Remove all uppercase from logging categories and useless cleanup in names.
2042 2004-08-18 Martin Quinson
2043 Version 0.6.2 (protocol not changed; API changed)
2044 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2045 gras_error_t), and NULL when not found. Functions expecting a msgtype
2046 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2047 hopefully usefull message.
2048 - Portability to prehistoric sparcs again
2050 2004-08-17 Martin Quinson
2051 Version 0.6.1 (protocol not changed; ABI not changed)
2052 - prealloc some buffers to speed things up
2054 2004-08-11 Martin Quinson
2055 Version 0.6 (protocol not changed; ABI expended)
2056 - The parsing macro can deal with the references, provided that you add
2057 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2059 2004-08-09 Martin Quinson
2060 Version 0.5 (protocol not changed; ABI changed)
2061 - Allow to off turn the cycle detection code in data exchange at
2062 compilation time. It should be at run time, but I'm short of time (and
2063 the config stuff is still broken). That way, we keep dict out of the
2064 critical path, which is good because the performance is poor:
2065 - search not dichotomial yet
2066 - dynar give no way to access their content and memcpy everytime
2067 - In composed data description (struct, ref and so on), stop foolness of
2068 keeping the subtype's ID, but store the type itself. This keeps sets out
2069 of the critical path, which is good since they rely on dynar and
2070 dictionnaries. The only loose of that is that we cannot detect the
2071 redeclaration of a structure/union with another content (but I'm not sure
2072 the code detected well this error before anyway). We still can detect
2073 the redefinition discrepancy for the other types.
2074 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2075 it breaks the code because of type-punning used all over the place).
2076 This breaks on all non-gcc architectures (for now).
2078 All those changes (plus the buffer of last time) allow me to gain 2 order
2079 of magnitude on cruel tests consisting of 800000 array of integers on two
2080 level of a hierarchical structure (200 secondes -> 4 secondes)
2083 - the selector of reference must now return the type it points to, not
2084 the ID of this type.
2086 2004-08-06 Martin Quinson
2087 Version 0.4 (protocol changed; ABI not changed)
2088 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2089 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2090 - Implement a buffer transport, and use it by default (it relies on tcp in
2091 real life and on sg in simulation).
2092 That's a bit hackish since I had a new field to the structure to store
2093 its data without interfering with the subtype ones. Inheritance
2094 is tricky in C. And that's a kind of reverse inheritance with one class
2095 derivating two classes. Or maybe a game with java interfaces. Anyway,
2096 that's damn hard in C (at least).
2097 Moreover, I got tired while trying to ensure plugin separation and
2098 genericity in SG mode. MSG wants me to do weird things, so let's go for
2099 cruel hacks (temporarily of course ;).
2100 See comment in transport_private.h:71
2101 - do not include all the _interface headers in private but in the files
2102 which really need them (to cut the compilation time when they are
2105 2004-07-26 Martin Quinson
2106 Version 0.3 (protocol not changed; ABI changed)
2107 - Major overhault of the datadesc interface to simplify it:
2108 - shorted the function names:
2109 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2110 - add a trivial way to push/pop integers into the cbps without malloc.
2111 This allows to make really generic sub_type description, which simply
2112 pop their size of the stack.
2113 - add a function gras_datadesc_ref_pop_arr() which does what users want
2114 most of the time: Declare a dynamic array (which pops its size of the
2115 stack) and declare a reference to it. Poor name, but anyway.
2116 - kill the post-send callback, add a post-receive one
2118 2004-07-23 Martin Quinson
2119 Version 0.2 (protocol changed; ABI changed)
2120 - add some testing for cpbs in the test cases, and fix some more bugs.
2121 This invalidate again the little64 data file, since I cannot regenerate
2123 - remove an awfull optimization in the logging stuff, allowing me to:
2124 - understand it again
2125 - learn gcc how to check that the argument match the provided format
2126 - fix all errors revealed by gcc after that
2127 - internal keys of dict are not \0 terminated. Deal with it properly in
2128 loggings instead of segfaulting when the user want to see the logs :-/
2130 2004-07-22 Martin Quinson
2131 - Fix some stupid bug preventing cbps (callback postit) from working
2133 2004-07-21 Martin Quinson
2134 - Some documentation cleanups
2135 - remove the useless last argument of msgtype_declare
2136 - rename the Virtu functions to fit into the 'os' namespace
2137 - move headers src/include -> src/include/gras/ and stop fooling with
2138 gras -> . symbolic link
2139 - make distcheck is now successful
2141 2004-07-19 Martin Quinson
2143 - Build shared library also
2144 - Install html doc to the right location
2145 - stop removing maintainer files in make clean
2146 - build tests only on make check
2148 2004-07-13 Martin Quinson
2150 - No major issue in previous version => change versionning schema
2151 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2152 corresponding dataset.
2154 2004-07-11 Martin Quinson
2156 - Get it working with any kind of structure (we can compute the padding
2157 bytes remotely for all the architectures I have access to)
2158 - Implement the structure parsing macro (still not quite robust/complete)
2159 - Improvement to the remote testing toysuite
2161 2004-07-10 Martin Quinson
2162 [autoconf mechanism]
2163 - get ride of a bunch of deprecated macros
2164 - actually run the test for two-compliment, not only compile it :-/
2165 - test whether the structures get packed (and bail out if yes. Damn.
2166 Alignment is a serious matter)
2167 - test whether the structures get compacted (but respecting the alignment
2168 constraints of each types)
2169 - test whether the array fields of structures can straddle alignment boundaries
2171 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2172 (took me 2 hours to find that bug, looking at the wrong place)
2173 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2174 before sending/receiving and are used to compute the offsets of fields
2175 - Given that padding size depend even on compiler options, keep track of
2176 alignment and aligned_size only for the current architecture. Not a big
2177 deal since we send structure fields one after the other (seems
2179 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2180 program, let it work on linux/gcc/little32. portability todo.
2182 2004-07-08 Martin Quinson
2183 - import and improve remote compilation support from FAST
2184 - make sure make check works on half a dozen of machines out there
2186 2004-07-07 Martin Quinson
2187 Let's say it's version 0.0.3 ;)
2188 - Implement conversions (yuhu!)
2189 - Let it work on solaris (beside conversion, of course)
2190 - Stupid me, using rand() to generate the conversion datatests in not wise.
2192 2004-07-06 Martin Quinson
2193 - Let make dist work, since I'm gonna need it to compile on remote hosts
2194 - Let Tests/datadesc_usage write the architecture on which the file was
2195 generated as first byte.
2196 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2197 architecture descriptions.
2198 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2199 on those architectures.
2200 - Optimization: send/recv array of scalar in one shoot
2202 2004-07-05 Martin Quinson
2203 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2205 - Plug a whole bunch of memleaks
2206 - each process now have to call gras_{init,exit}. One day, their log
2207 settings will be separated
2208 - Continue the code factorisation between SG, RL and common in Transport.
2210 2004-07-04 Martin Quinson
2212 - Redistribution between SG and RL.
2213 We wanna have to accept in SG, so move accepted related parts of RL in
2214 the common part. (more precisely, the dynar of all known sockets is no
2215 more a static in transport.c, but part of the process_data)
2217 [gras_stub_generator]
2218 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2220 2004-07-03 Martin Quinson
2221 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2222 them all up in one shot)
2224 - Fix the ugly case of reference to dynamic array.
2225 - New (semi-public) function gras_datadesc_size to allow the messaging
2226 layer to malloc the needed space for the buffer.
2228 - gras_socket_close now expect the socket to close (and not its address to
2229 put NULL in it after it). This is because the socket passed to handlers
2230 is one of their argument (=> not writable).
2232 - propagate the interface cleanup from last week in datadesc, ie remove a
2233 superfluous level of indirection. User pass adress of variable
2234 containing data (both when sending and receiving), and not of a variable
2235 being a pointer to the data. Let's say that I like it better ;)
2236 The price for that is constructs like "int msg=*(int*)payload" in
2237 handlers, but it's a fine price, IMHO.
2239 - Let it work in RL (yuhu)
2241 2004-06-21 Martin Quinson
2243 - porting SG plugin and SG select to new standards (works almost).
2244 - plug memleaks and fix bugs around.
2247 - cleanup the prototype of data recv and force users to specify when they
2248 want to handle references to objects. Test case working even for cycles.
2249 - plug memleaks. Valgrind is perfectly ok with this.
2251 2004-06-12 Martin Quinson
2253 - cleanup the separation between plugin and main code in plugin creation
2255 2004-06-11 Martin Quinson
2257 - Reput hook for raw sockets, needed for BW experiments
2258 - kill a few lines of dead code
2259 [Data description] Interface cleanup
2260 - gras_datadesc_by_name returns the searched type or NULL.
2261 That way, no variable is needed to use a type desc once, which makes
2263 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2264 two parameters were strings (field name, type name), leading to
2266 [Dicos] Interface cleanup
2267 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2268 This is consistant with the dynar API.
2270 2004-04-21 Martin Quinson
2272 - Porting to new standards.
2274 - interface cleanup.
2275 There is no bag anymore, no need to take extra provision to mask the
2276 pointers behind "ID".
2277 Better splitup of functions between files create/exchange/convert.
2278 This is still a bit artificial since convert and receive are so
2279 interleaved, but anyway.
2281 - add a queued message list to procdata (the ones not matching criteria
2283 - factorize some more code between SG and RL wrt procdata
2285 - use gras_exit in example to track memleaks
2286 - get rid of gs_example now that GS is properly integrated into gras
2287 - update run_test to integrate the lastest tests (datadesc)
2289 - rename WARNINGn macros to WARNn since it prooved error-prone
2291 2004-04-19 Martin Quinson
2293 - register init/exit functions within gras module mechanism
2294 - send/receive function.
2295 Convertion is not implemented, but short-cutted if not needed.
2296 struct/array elements are sent one by one (instead of block-wise), but
2297 nobody really cares (yet). Get a prototype before optimizing.
2298 - tests (using a file socket) for DD send/receive on:
2299 - base types: int, float
2300 - array: fixed size, string (ie ref to dynamic string)
2301 - structure: homogeneous, heterogeneous
2302 - chained list, graph with cycle
2303 Believe it or not, valgrind is not too unhappy with the results. The
2304 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2308 - Bugfix when using a filename explicitely (instead of '-')
2310 2004-04-09 Martin Quinson
2312 - factorize more code between RL and SG in socket creation
2313 - Complete the implementation and tests of:
2315 o file (only in RL, and mainly for debugging)
2317 I lost 3 days to design a portable address resolver, and then decided
2318 that the prototype mainly have to run on my box.
2319 Addressing portability too early may be like optimizing too early :-/
2321 - use gras_init in the Tests instead of the crappy parse_log_opt
2322 (the latter function is removed)
2323 [Conditional execution]
2324 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2325 [Code reorganisation]
2326 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2327 Build examples against the RL library, since there is no way to disable
2328 its creation for now.
2330 For information, the beginning of coding on GRAS was back in june
2331 2003. I guess that every line has been rewritten at least twice since