1 SimGrid (3.8) NOT RELEASED; urgency=low
4 * New type of typed tasks SD_TASK_COMP_PAR_AMDAHL that represents a
5 parallel task whose initial work is distributed among host according
6 to the Amdahl's law. Such tasks are created with a parameter alpha
7 that corresponds to the non-parallelizable part of the computation.
9 Before calling SD_task_schedulel or SD_task_schedulev, the
10 SD_task_distribute_comp_amdahl function has to be called with the
11 desired number of workstations. The filling of computation_amount and
12 communication_amount (actually empty) structures is then done
14 * New function SD_workstation_dump to display various information
16 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
18 SimGrid (3.7.1) NOT RELEASED; urgency=low
21 * Fix weird behaviors when dealing with parallel tasks.
23 Bug fixes, cosmetics and small improvements
24 * Fix compilation when using MSG_USE_DEPRECATED.
27 * Fix compilation on Macs.
29 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
31 SimGrid (3.7) stable; urgency=low
33 The "spring cleanups (before the next Big Project kicks in)" release.
36 * We can specify the SMPI latency/bandwidth factor with command line
37 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
38 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
39 You can also use the "config tag" from platform file by adding this line
40 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
41 (see "example/platforms/tag_config.xml" to use "config tag").
42 Note that the command line supersedes the platform file configuration.
43 * Change the correction factors used in LMM model, according to
44 the latest experiments described in INRIA RR-7821.
45 Accuracy should be improved this way.
46 * Use the partial invalidation optimization by default for the
47 network too. Should produce the exact same results, only faster.
48 * Major cleanup in surf to merge models and split some optimization
49 mechanisms from the core of the models. As a result you can now
50 specify which model to use (e.g., --cfg=network/model:LV08
51 --cfg=cpu/model:Cas01) and which optimization mode to use
52 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
53 Incompatible combinations should err at initialization. See
54 --help-models for the list of all models and optimization modes.
55 * The CLM03 workstation model was dropped for simplicity because it
56 used the deprecated CM02 network model. Use default instead.
57 * Rename the TCP_gamma configuration option to network/TCP_gamma
58 * Rename the coordinates configuration option to
59 network/coordinates, and document it
60 * Use now crosstraffic keyword instead of the terribly misleading
61 fullduplex keyword. It is activated by default now in the current
62 default model, use --cfg=network/crosstraffic:0 to turn it off.
63 * Ongoing refactoring the model parsing to make XML files optional
64 See include/simgrid/platf.h for details (still to be completed)
67 * Major overhaul of the documentation. Almost instructive now :/
68 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
69 functions and friends. This interface was considered as
70 deprecated since over 2 years, it's time to inform our users that it is.
71 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
72 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
73 use these (crappy) functions in your code.
74 These functions will be removed soon. Stop using them now.
75 * Deprecate MSG_get_host_{table,number}
76 Implement MSG_hosts_as_dynar() instead.
77 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
78 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
81 * Stabilize the parallel execution mode of user contexts
82 * Introduce configuration variables to control parallel execution:
83 - contexts/synchro: Synchronization mode to use when running
84 contexts in parallel (either futex, posix or busy_wait)
85 - contexts/parallel_threshold: Minimal number of user contexts
86 that must be part of a scheduling round to switch to parallel
87 execution mode (raw contexts only)
88 * Fix bugs that prevented to use suspend/resume along with
89 synchronization structures.
90 * Fix bugs in process termination that lead to invalid memory access
91 in very specific conditions.
94 * Introduce a parallel mode for the models (controlled by surf/nthreads
95 configuration item). In our tests, running the models in parallel
96 never lead to any speedups because they are so fast that the gain
97 of computing each model in parallel does not amortizes the
98 synchronization costs, even when ultra fast futexes are used.
99 This is released anyway because YMMV.
102 * Performance boost by using a swag internally to compute the set of
103 tasks that are finished and should constitute the return value of
107 * Enable it by default now that it is considered rather stable.
110 * Documentation of the tracing functions.
111 * Performance gains when tracing categorized/uncategorized resource
112 utilization by avoiding calls to get route when updating resource
113 variables. LMM constraints are being used instead.
114 * API changed to set task categories. Use MSG_task_set_category instead
115 of TRACE_msg_set_task_category, and SD_task_set_category instead
116 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
117 * Bugfix for graphicator, routes not correctly obtained, memory leaks
118 * Examples for link user variables added (see at examples/msg/tracing/)
119 * Deprecated function TRACE_msg_set_process_category completely removed
120 * Trace header updated according to the latest Paje file format
121 * Tracing network lazy updates, no longer obligate users to use full updates
122 * --cfg=tracing/platform:1 also registers power/bandwidth variables
123 * Experimental: let user code declare/set/push/pop application states for hosts
124 * API changed to allow the manual creation of graph configuration files
125 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
128 * Improve the API of Lua MSG bindings, using the Lua spirit.
129 * Each simulated process now lives in its own Lua world (globals are
130 automatically duplicated). It helps writing simulators. Will allow
131 to run Splay programs within SimGrid in the future.
132 * Add a Chord example in Lua, equivalent to the MSG one.
135 * Start the implementation of a solution to express temporal
136 properties, not only local assertions. This is still an
137 experimental work in progress, stay clear from it to be safe.
141 - Add new runtime parameters --help-logs and --help-log-categories
142 to display informations about supported logging parameters and
144 - Old deprecated parameters --{gras,surf,msg.simix,xbt}-log=...
145 don't exists anymore.
146 * Mallocators: allow value NULL for the reset function.
148 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
149 create homogeneous dictionaries, where all the elements share the
150 same free function. Non homogeneous dictionaries will be
151 deprecated in the next release.
152 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
153 - Multi-level dictionaries are deprecated.
155 - new function xbt_dynar_search_or_negative() that is useful when
156 you have less than 2 million elements in your dynar and don't
157 want of the extra complexity of catching exceptions when the
158 element is not found.
160 - Make xbt_os_thread module (for thread portability) public.
161 Documentation is still to come, sorry.
163 - Cleanups and simplifications to make it maintainable again.
164 - Exotic features (such as memalign and valloc) were removed.
165 - The metadata were extended and improved so that the
166 model-checker becomes able to explore and inspect the heaps.
167 - This may induce a performance drop when enable_model-checking is
168 ON in cmake (even if it's not used in the simulation), but it is
169 necessary at this point to get MC working.
171 Turn model-checking OFF if simulation performance matters to you.
172 Not enabling it at runtime is not enough, disable it in cmake.
174 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
176 SimGrid (3.6.2) stable; urgency=low
178 The "Not coding new stuff allows to polish old things" release.
181 * New bindings to the NS3 packet level simulator (experimental)
182 * Use the raw (efficient) execution contexts instead of the sysv
183 (portable) ones when possible.
184 * libpcre is now mandatory in any cases since not using it led to
185 severe performance loss and possibly other issues
186 * Update the XML platforms:
187 - G5K: include the latest machine in Nancy
188 - GridPP and LCG: new platforms
189 * Documentation was partially updated, at least (more to come)
191 Bug fixes, cosmetics and small improvements
192 * Free terminated processes before the end of the simulation to avoid
193 exhausting the memory of users having very dynamic amount of
195 * Bug fix and cosmetics about canceling non-running tasks
196 * Bug fix about the dot loader's issues when using libcgraph
199 * Create an installer for windows with nsis (amd64 and win32)
200 - Add an hello world project to illustrate simgrid project creation.
201 - Embed libpcre into the Simgrid installer to avoid
202 its compilation burden
203 * The raw execution contexts should work on Apple now
204 * Port to Windows 64 bits
205 - Sysv contexts now have an implementation for this arch
206 - GRAS communication features now support this arch
207 * Drop support for borland compiler on windows
208 - this code was not maintained, and we kinda depend on gcc nowadays
209 * Fix portability issues on kfreebsd/gnu: build error about semaphores
210 * Fix portability issue on unstable ubuntu: linker became picky on
213 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
215 SimGrid (3.6.1) stable; urgency=low
217 The "Oops, we broke Macs too" release
220 * Fixed contexts detection so that raw ones are used when possible
221 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
222 leads to a strange error, with user code segfaulting sometimes when
223 the generated code is not perfectly aligned (which is not
224 controllable from the user side, depends on the amount of code)
227 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
229 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
231 SimGrid (3.6) stable; urgency=medium
233 The Summer Release, also known as the "OMG! They Killed Kenny!" version
236 * Bindings now constitute their own package, separated from the main one.
237 Rationale: reduce our maintenance nightmare by reducing the module coupling
238 They will soon be released on their own on gforge.
240 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
241 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
243 GRAS: It is not considered as stable anymore, but experimental. Sorry.
244 * It's not quite deprecated for now because we have no replacement,
245 but it may soon become the case.
248 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
249 * Fortran: New user-level cache variable to store the rank of the running
250 process. This improves performance by an order of magnitude.
251 * C: New coccinelle script to automatically locate and modify global and
252 local static variables.
253 * Improved SMPI network model with a sender-side gap to account for multiple
257 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
258 only say if a communication is finished, no matter whether it succeeded or
259 failed. You can call MSG_comm_get_status() to know the status of a finished
261 * New function MSG_task_dsend() to send a task and detach it. When a
262 communication is detached, you are never notified of its success or failure
263 and the memory is released automatically once it is finished. This function
264 is useful when you don't care about the end nor the success of a
266 * Change the prototypes of action replay. Sorry for inconvenience,
267 but this is really more efficient this way (and to adapt your code,
268 you just have to fix the initialization, that shouldn't be too long)
269 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
270 ever managed to do anything useful with it.
271 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
272 takes a dynar of communications. It returns immediately and gives the
273 index of a finished communication (if any).
274 * New example: a basic implementation of the Chord P2P algorithm.
277 * New model for multi-core CPUs. You can now use the core attribute to
278 precise the number of cores of a host. This is a basic model. Every
279 process running on the host receives at most the power provided in
280 the DTD (throughput<=power). Total throughput of process cannot exceed
282 * New peer tag. This peer tag creates a tiny AS comprising a host and a
283 router linked by an up-link and a down-link (possibly asymmetrical).
284 This kind of pattern allows to easily build last-mile model style platforms.
285 Aggregating such patterns in a rule-based AS is thus the technique of
286 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
287 * New model for Vivaldi routing. We transformed the Vivaldi network model
288 into a Vivaldi routing model (based on the rule-based model). This allows to
289 combine Vivaldi based latencies with last-mile platforms.
292 * Added a check for NaN of IEEE754 infinite in the double entries of
294 * Introduce a new context factory "raw", highly inspirated from the
295 ucontext factory, but using manually crafted functions in assembly to
296 do the work in an efficient manner.
297 * Allow to change the used context factory at run time, not only at
298 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
299 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
300 contexts at runtime (only with raw contexts or ucontexts).
301 * Completely rewrote this module to allow parallel execution of user
302 processes. Use --cfg=contexts/nthreads:N to execute user processes
303 with N parallel threads (the default is 1, meaning no parallelism).
304 * Allow to decide dynamically between sequential and parallel modes.
305 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
306 processes in parallel only when their number is greater than or equal to P
308 * Added a check for NaN of IEEE754 infinite in the double entries of
312 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
313 won't output the state of processes when interrupted with Ctrl-C
314 * Add a new function xbt_dynar_to_array that transforms a dynar into a
315 NULL-terminated array. This may solve backward compatibility issues
316 due to the change to return type of SD_simulate. See also:
317 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
318 * Add new macros with variable number of arguments.
319 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
320 - in xbt/asserts.h: xbt_assert
321 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
322 - in xbt/ex.h: THROWF and RETHROWF.
323 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
325 * Change xbt_die() to accept a format string with arguments, just like printf.
326 * New data structure: xbt_lib_t, like a dict but more general and with better
330 * New configuration options
331 Options triva/categorized and triva/uncategorized can be used to generate
332 graph configuration files for Triva visualization tool.
333 * Configuration option tracing/platform is renamed to tracing/categorized
334 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
335 * New TRACE_user_link_variable interface:
336 User provides the name of the link and the tracing variable to attach to it
337 * the declaration of tracing categories must be done after the environment
339 * simpler tracing interface, just one way to declare categories
340 TRACE_category or TRACE_category_with_color, it is up to you
341 * links in the trace file are again identified by their names
342 * trace contains the full platform hierarchy exactly as declared using the ASes
343 * Options tracing/msg/[task|process]:1 groups the process by hosts
344 for both cases, tasks and processes must have names that are unique during the simulation
345 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
346 * The experimental option tracing/msg/volume is deprecated
347 its functionality may be reincorporated if needed
349 The tracing generates a trace file with unordered timestamped events,
350 because of the way the core simulator (surf) works. A script available
351 at the tools directory (fix-paje-trace.sh) can be used to put the events
352 in order. We have changed the tracing so it can generate ordered timestamped
353 events in the final trace, but depending on the simulator (and how much time
354 is simulated) that can lead to a huge memory utilization. It is deactivated
355 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
358 * Define a SIMGRID_VERSION macro in simgrid_config.h.
359 - We are trying hard to keep the API stable, but it may happen that
360 some things change (we're a research project after all, not a
361 nuclear plant operating system). If such things should happen, you
362 could rely on that macro to adapt.
363 - current value: 30600 for 3.06.00, aka 3.6
364 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
365 a number that can be compared with SIMGRID_VERSION.
366 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
367 assertions and verbose/debug logging events are disabled at compile time.
369 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
371 SimGrid (3.5) stable; urgency=medium
374 * New feature to any SimGrid-based simulator: Model-Checking
375 Check SIN#1 for more details.
378 * New Model SMPI (three-interval linear regression for correction factors)
379 See RR-7426, available at http://hal.inria.fr/inria-00527150
380 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
381 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
382 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
383 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
384 * See also src/smpi/README
389 - Tracing API changes: TRACE_start and TRACE_end should not be called
390 by user-code. They are automatically called by simulators created
391 with SimDAG, MSG and SMPI if the toolkit is compiled with
392 tracing_enabled=ON. Categories declaration and utilization remain the
393 same for MSG and SimDag.
394 - A function was added to the tracing API to declare categories with
396 - TRACE_category_with_color (char *category, char *color)
397 where color must be in the following format
398 "%f %f %f", red, green, blue
399 and red, green, blue are float values in the interval [0, 1]
400 - User can specify NULL as color parameter, or continue calling
402 On that case, the tracing system will define random colors
403 - The following command-line options are supported:
404 --cfg=tracing/filename:msg.trace
405 --cfg=tracing:1 (activate tracing, needed to use others)
406 --cfg=tracing/platform:1 (categorized resource use)
407 --cfg=tracing/uncategorized:1 (uncategorized resource use)
408 --cfg=tracing/msg/task:1 (task creation)
409 --cfg=tracing/msg/process:1 (process creation, migration)
410 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
411 --cfg=tracing/smpi:1 (SMPI interface tracing)
412 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
413 - examples of examples/msg/tracing updated
415 - DAXLoader and DOTLoader functions can generate tasks with categories
416 - A new function to attribute a category to SD tasks:
417 TRACE_sd_set_task_category (SD_task_t task, char *category)
418 * Tracing the MPI interface implemented by SMPI
419 - Collective operations are traced with states
420 - Point-to-Point operations are traced with states/links
421 - Tracing activated by a parameter "-trace filename" passed
422 to smpirun during execution (considering that simgrid
423 is compiled with tracing enabled)
424 - To run the simulation with gdb, the simulator
425 accepts --cfg=tracing/smpi:1 to trace SMPI
426 - tesh files to check if smpi tracing is ok
427 - See examples/smpi/NAS/DT-trace
428 * GTNetS tracing re-worked
429 - adaptation to the tracing system of GTNets to cope
430 with modifications regarding the fullduplex mode
431 - new tesh files to check if gtnets tracing is ok
434 * Asynchronous communications through the functions:
435 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
436 * New function: MSG_load_platform_script()
437 to make possible using a lua script instead of XML files to set up platforms
438 * New function: MSG_set_function
439 to associate functions to processes, used when bypassing the parser
440 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
442 Platforms: Add some more examples in examples/platforms
443 * Grid'5000: see www.grid5000.fr
444 * *_30000_hosts.xml: various huge files [mainly scalability testing]
447 * Change the XML format. This is a very important modification. SimGrid
448 3.5 introduces a new hierarchical format based on the notion of
449 Autonomous Systems. Compatibility with old format is ensured
450 through the perl script provided in the install bin directory
451 bin/simgrid_update_xml.
452 It is now possible to build platforms with specific routing
453 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
454 connect several platforms together. We will try to provide soon
455 set of realistic platforms exploiting these properties (have a
456 look at examples/platforms/ for the moment).
457 * Take the opportunity of the XML format change to be a good XML citizen:
458 rename link:ctn to link_ctn and similar changes (also dealt with by
460 * Add a new routing scheme (rule-based) using regular expressions. It
461 enables to have an extremely low memory footprint when the
462 underlying routing is simple and can be compactly described. You
463 need to have libpcre4-dev (perl regular expressions) installed if
464 you want to use this routing scheme.
465 * Revive the cluster TAG and allow to easily and efficiently (both in
466 term of memory and speed) connect clusters together. Have a look
467 at teshsuite/simdag/platforms/ to see how this can be done. With
468 this tag, you can create clusters with thousands of tasks at no
469 cost (have a look at examples/platforms/).
470 Note: clusters are implemented as ASes, so there is no need for an
471 enclosing AS tag if you have only one cluster in your platform.
472 * Add new generic functions in the public interface that allows the user
473 to call SURF 'create_resource' methods from your code (same
474 functionality as the XML bypass mechanism but with a much lighter
476 * Add a new model (enabled through command line --cfg=network/model:SMPI)
477 that uses a piecewise linear approximation to produce better
478 results when exchanging small messages.
479 * Add a new parameter to handle correctly full duplex link and account
480 for interferences between uplink and downlink communications
481 (activate with --cfg=fullduplex:1).
484 * Rename the SD_READY (all dependencies are satisfied and task is
485 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
486 dependencies are satisfied) state.
487 This prevents a confusion between the notion of "ready to schedule"
488 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
489 simulated" (SD_RUNNABLE) used by the simulation kernel.
490 * Change the way a task is considered as ready. Instead of removing
491 dependencies when a task is done, a counter is decreased. This way,
492 it is always possible to reach ancestors thanks to the
493 SD_taks_get_parents function (even after the end of the simulation.)
494 * Change the return type of SD_Simulate from (SD_task_t*) into
495 xbt_dynar_t. This function was in handling a dynar internally and
496 converted it into a NULL terminated array for historical reasons.
497 * New function SD_dotload(char*) to load a DAG described in dot
498 format. This loader and the corresponding examples require the
499 installation of the graphviz library.
500 * Fix a bug in the management of tasks of size 0 in the surf network
501 models. This problem was only visible with SIMDAG and you should
502 thus disregard results produced with earlier versions if you
503 relied on this feature (some tasks were blocked because of this).
504 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
505 even though your applications were doing only point-to-point
506 communications and sequential computations. Now you can really use any
507 model you want (of course, if you create real parallel tasks, which are
508 not implemented in most models beside ptaskL07, this will abort).
509 * Add an example that schedules a DAX on an heterogeneous platform
510 using a Min-Min strategy.
511 * New function SD_workstation_get_current_task() that returns the kind
512 of task currently running on a workstation in the sequential access
514 * Raise some warnings when unexecuted tasks remains at the end of the
515 simulation. This is usually caused by cycles in the DAG.
518 * New function: SIMIX_process_set_function() called by MSG_set_function
519 * Change the underlying waiting queue in semaphores so that a process
520 can wait on several of them simultaneously (as in waitany).
521 * Fix the way to handle tokens in semaphores so that all access patterns
522 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
523 * kill the dirty pimple SIMIX_message_sizes_output()
524 Please use (proper) visualization instead
527 * New data container: setset (set of sets of elements)
528 * New module: mmalloc (mapped malloc, allowing to have several
529 independent segments of malloc)
530 * New function: xbt_dict_cursor_set_data()
531 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
532 * New function: xbt_dynar_is_empty()
533 * New function: xbt_fifo_get_last_item()
534 * Fix xbt_dynar_shrink(): use the right element size.
535 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
536 uninitialized areas during expand.
537 * Fix semaphores: previous implementation was severely broken.
538 * Use library init/fini functions for our initialization.
539 - you can use logs and other feature as soon as you want in your
540 code (even before the xbt_init / MSG_init)
541 - xbt_exit is now a no-op and produce a warning when used.
544 * Port GRAS to new SIMIX mechanisms. This allows gras users to
545 benefit from the latest improvement to the simulation kernel.
546 * Kill measurement sockets for now. If you rely on them, sorry. This
547 release is not for you. This feature will be reintroduced in the
548 future, but we cannot delay the release any further.
549 * New function: gras_msgtype_get_name().
550 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
551 * Fix (at last) the pmm example: it should not randomly fail anymore.
553 Build chain: bug fixes and overall polishing
554 * Cmake is now stable enough. Hence, we killed the autotools.
555 * Port to windows ( TM :)
556 * Fix the 'make install' target.
557 No need to use 'make install-simgrid' anymore
558 * Introduce a 'make dist' target compiling a *source* archive
559 'make package' compiles a binary archive
560 * Compile java files only on need
561 * Add --cd and --setenv command line options to tesh
562 * Out of source builds are not fully supported yet, but we are close
563 * Enable supernovae and optimization flags by default for our users
566 * Add layer to set up environment directly from lua, without XML.
567 * The effect of gras_stub_generator can be achieved through
568 lua too (check examples/gras/console/ping_generator.lua)
570 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
572 SimGrid (3.4.1) stable; urgency=low
574 The "Polishing easter eggs is probably a good idea" release.
575 This is a bug fixes release only.
579 * Fix a bug preventing the tasks from begin garbage collected.
582 * Fix a bug occuring when a host involved in a communication fails.
583 This was not detected properly by the other peer involved in the
584 communication. Now, it's reported as a network error.
587 * Warn the user about loop dependencies in data flow of DAX files
588 * Obey the control-flow dependencies of DAX files
591 * Add option "enable_smpi" allowing to not compile SMPI.
592 Probably useful for the (Mac) users experiencing a build error here
593 * Improve the detection of lua5.1 and ruby1.8
595 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
597 SimGrid (3.4) stable; urgency=low
599 The "Easter in Cargese" release. Also known as (major changes):
601 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
602 ~> bindings were greatly improved
603 ~> new tracing infrastructure for better visualization introduced
605 * the "Welcome to configury modernity" release.
606 ~> we switched from autotools to cmake, and improved our cdash
609 A more detailled list of changes follow (full detail in svn log).
611 Java Bindings: Various Cleanups
612 * (install java-gcj-compat-dev on debian-like to use them)
613 * Remove put/get: no need to export deprecated interface in Java
614 Use send/receive instead.
615 * Cleanup the examples and add a README per directory
616 * Remove example autoDestination (that's the only way to go now)
617 * Remove example explicitDestination (was a plain copy of basic)
618 * Make JniException a runtime exception, so that there is no need to
619 declare the fact that you may encounter such a beast. I guess that
620 nobody will ever want to survive such error.
621 * Create specific errors for each MSG case of failure:
622 host failure, transfer failure, timeout, task cancelled
623 * Cleanup the exceptions that may get thrown by each function
624 * Other internal cleanups in Java bindings. Performance still bad :/
625 Ruby and Lua Bindings: create them
626 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
627 * That's new and great, you should try them out.
628 Same functionalities than Java bindings, only even less polished
630 * Kill the useless "rate" argument of SD_task_get_execution_time()
631 Everyone used to provide -1 as a value, it was not used, and the
632 semantic of a possible use wasn't even clear.
633 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
634 either as comm costs or compute costs to mean that there is no
635 such thing for that specific task.
636 * Add a SD_task_set_name() function
637 * Fix SD_task_unschedule() on typed tasks
638 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
639 * In DAX loader, accept useless 'level' attributes to <job> since
640 LIGO DAGs have them (seem to be to ease graphical representation).
642 * Add an example masterslave_mailbox.c using send/receive and not
643 the deprecated put/get interface.
644 * Kill the MSG_paje_output() function. It's a noop since 2 years.
645 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
647 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
648 (declare MSG_USE_DEPRECATED to still have the old name)
649 * Add a MSG_task_set_data() function
650 * About trace replay (see examples/msg/actions):
652 - Allow to work with splitted trace files for each process
653 Give the specific trace file as argument of each process,
654 and call MSG_action_trace_run(NULL)
655 You can still have one merged file for all processes.
656 - Fix implementation of collective operations
657 * Allow task_execute() on 0-sized tasks (closes #10063)
659 * This is the first release of SimGrid where SMPI is not considered
660 beta anymore (even if some corners should still be improved)
661 * Port over the new SIMIX_network submodule (internal refactoring)
662 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
663 * Implement more missing elements of the standard:
665 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
667 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
668 MPI_Scan() MPI_Get_processor_name()
669 - Added implementation of missing case for Alltoall (warning: it's
670 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
671 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
672 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
673 SMPI_MPI_Allgatherv()
675 - MPI_Waitsome() was broken
676 - Allow relative includes in smpicc
677 - Command line cfg argument 'reference_speed' was ignored...
678 - Some functions did not properly lead to auto-benching of user code
679 - smpicc passes -O2 by default (just like openmpi one)
681 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
682 * Bug fixes about timeouts during communications
683 * add SIMIX_message_sizes_output() as a pimple to write to file the
684 amount of messages per size. Use gnuplot to get histogram.
685 Pimple because that's the only user-visible function of simix,
686 defined directly in xbt.h (irk, sorry)
688 - Add a SIMIX_sem_get_capacity() function
689 - Fix interactions with processe resume/suspende
690 - release_forever() was stupidly broken
691 - Fix SIMIX_display_process_status() for processes in a semaphore
692 - Make SIMIX_sem_block_onto() user-visible
693 * Refactoring context stuff:
694 - Use pseudo-OOP for better modularity
695 - reimplement SIMIX_process_kill() without process_schedule() so
696 that the latter can take as invariant that it is called from
698 - Merge context_start into context_new for sake of simplicity
700 * Add a Vivaldi network model, coded live during SUD'10 ;)
701 * Rename configuration variables to start a hierarchy:
702 o cpu_model -> cpu/model
703 o network_model -> network/model
704 o workstation_model -> workstation/model
705 * New configuration variables:
706 o network/bandwidth_factor: correction to bandwith
707 o network/latency_factor: correction to latency
708 o netwotk/weight_S: correction to the weight of competing streams
709 * Add a long description to the models, that users can see with such
710 argument on the command line: --cfg=cpu/model:help
711 * --help-models display the long description of all known models
713 * config: add the ability to set a default value after registration
714 Does not override any previously set value (e.g. from cmd line)
715 * dict: allow to have integer key and data.
716 When so, you need to use the following functions
717 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
718 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
719 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
720 In contrary to regular dicts, the key is not malloced before copy.
721 Mixing scalar and regular elements in the same dict is not tested
723 * Allow to use xbt_dynar_shrink() to expend the dynar instead
724 Tracing for Visualization:
725 * SimGrid is now instrumented in order to generate a trace file for
726 visualization analysis: to use it, need to compile SimGrid with the
727 "tracing" option enabled, and instrument the program using SimGrid with
728 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
729 (among other functions).
730 * The instrumentation only traces the platform utilization for now
731 * Documentation to use the tracing functions and how to analyze the
732 traces with the Triva tool is written.
733 * More information about: SimGrid FAQ (in the section Tracing Simulations
736 * We moved to cmake as default build system. Autotools support will
737 be dropped soon. Check the FAQ for more info about how to use it.
738 * Greatly improved our cdash/ctest interactions
739 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
740 * Added memory checking tests with valgrind; lot of memleak fixing.
741 This may be the first release of simgrid with so few memory issues
742 * Added code coverage tests.
743 Our coverage is still improvable, but at least we see it on cdash.
745 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
747 SimGrid (3.3.4) stable; urgency=low
749 The "Desktop Grid needs love too" release (also called Xmas release).
752 * Major speedup in the maxmin system solving by using lazy evaluation
753 Instead of solving completely the maxmin system at each iteration,
754 only invalidate (and recompute) the modified parts.
755 This new feature is enabled in default models but you can try to
756 turn it on with "--cfg:maxmin-selective-update=1" for other models.
757 * Cas01 IMproved as default CPU model
758 This CPU model is the same Cas01 model, but it uses the
759 maxmin-selective-update flag and a heap structure to manage
760 actions on SURF kernel.
761 It reduces the complexity to find the next action to finish and,
762 consequently, it's faster than the old Cas01.
763 This is the new default CPU model (Cas01).
764 * Rename the old Cas01 model to Cas01_fullupdate
765 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
766 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
767 * CpuTI (CPU Trace Integration)
768 A new CPU model whose objective is simulate faster when using
769 availability trace files.
770 Instead of using a full featured, over engineered maxmin system for
771 CPU modeling, this model does the pre-integration of traces files
772 to calculate the amount of CPU power available, and so, executes
773 faster than the old CPU models.
774 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
775 * Use LV08 as default network model since it gives better accuracy
776 for small messages and shouldn't change things for big ones.
777 Use --cfg=network_model:CM02 to get the previous behavior.
780 ******************************************
781 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
782 ******************************************
783 * The new CPU model may changes simulations!
784 The point is that events occurring at the exact same timestamp
785 are not scheduled in the same order with the old and new
786 version. This may be enough to completely change the execution
787 of simulations in some cases.
788 * The new network model will change simulations!
789 This new model is more realistic than the previous one, so you
790 should consider redoing your old experiments with this model.
791 Sorry for the inconvenience.
794 * Introduce the supernovae compilation mode
795 When compiled that way, the whole SimGrid (or almost) is put in a
796 single compilation unit and compiled in one shoot.
797 This is to help gcc which has difficulties to inline stuff from one
799 The speedup seem to be above 15%, althrough more tests are needed on
800 amd64 to confirm that gain.
803 * Port of MSG's mailbox on top of SIMIX network
804 The put/get mechanism was greatly simplified on the way.
807 * New SIMIX network module. Provides:
808 - Mailbox: rendez-vous mecanism to find with who you want to speak
809 - Synchronous send/recv: easier and hopefully faster since the
810 logic is handled in the maestro process directly now
811 - Asynchronous send/recv: you dreamt of it? It's here now
812 Too bad that nobody cared enough to propagate the change to MSG.
813 * Add semaphores as SIMIX synchronization mechanism.
816 * new function SD_daxload(char*) to load a DAX file
817 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
818 * Introduce typed tasks. Specify its kind and cost at creation.
819 At scheduling, just give where it should be placed, and the cost
820 for each involved resource is automatically computed.
821 Existing constructors so far (more to come of course):
822 - SD_task_create_comm_e2e() for end-to-end communication
823 - SD_task_create_comp_seq() for sequential computation
824 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
825 * new function SD_task_dump() for debuging display
826 * new function SD_task_dotty(task,FILE*) writing to file the info
827 about the task in dotty format
828 * SD_task_dependency_exists() can now cope with having one of its
829 arguments NULL. If so, it tests whether the other argument has any
831 * Add getters on list of preceding/following tasks:
832 SD_task_get_parents(task) and SD_task_get_children(task)
833 * Add getters on amount of workstations and list:
834 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
835 * Add getter on task kind: SD_task_get_kind(task)
836 * Update the start_time and finish_time of tasks on completion/failure
837 * Bugfix: Remove task from state swags when destroyed
840 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
843 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
844 dynar with one separate thread per value of the dynar.
845 * Change the prototype of xbt_thread_create(), sorry.
846 Added a boolean parameter indicating whether we want to join this
847 thread (used in SG only for now)
848 * Implement xbt_thread_join and xbt_thread_yield in SG also.
851 * GTNetS wrappers should now be usable again (and betterly tested too)
852 * Fix a major regression from 3.2 where the timeout provided to
853 MSG_task_put_with_timeout() was used as absolute time before which
854 the comm should be done.
855 * Start to fix the <cluster> tag.
856 - Internal links should be good now (beside of the loopback, which
857 use the private link instead)
858 - paths to the external world is still rather broken
859 - the <route:multi> tag is just broken. Actually that's brain-dead.
860 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
861 to make it less stupid
862 ** Check your platform with teshsuite/simdag/platforms/flatifier **
863 * Fix a source-level compatibility glitch from 3.2: after defining
864 MSG_USE_DEPRECATED, you can use the old name
865 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
866 * Allow to compile from the SVN with automake 1.11
867 * Fix some problems when using the "start_time" tag in deployment XMLs.
868 * Fix #8569: XBT/synchro.h has redundant declarations
869 * Fix #8563: MSG return values and exceptions
870 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
871 * Integrate patch #8636: Obey DESTDIR when installing documentation.
872 Thanks to Robson Peixoto.
873 * Fix a vicious bug in dictionaries inducing that some elements were
874 not freed on xbt_dict_free()
876 Portability report of this version:
877 * Main portability targets:
878 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
880 - mac leopard on i386
881 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
882 but nothing critical.
883 * Other platforms: windows, AIX and others were not tested for this release
885 Timing report of this version:
886 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
887 scenario parameters). From 8h to a few seconds in desktop grid settings.
888 * Supernovae brings about 25% speedup on i386.
890 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
892 SimGrid (3.3.3) stable; urgency=low
894 The "Need for Speed" release.
896 The timings done to validate the 3.3.2 were faulty.
897 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
899 The problem was a conversion from a manually handled vector to
900 xbt_dynar_t on the critical path.
901 xbt_dynar_foreach calls functions, inducing stack management crap.
903 We inlined these functions and xbt_dynar_foreach is now breath taking.
904 We also inlined xbt_swag_belong on the way.
906 Here are some approximate speedup measurements (on master/slaves
907 simulations lasting between 10s and 20s each):
908 3.3.1 -> 3.3.2: about same performance
909 3.3.2 -> 3.3.3: 40% speedup
910 3.3.1 -> 3.3.3: 40% speedup
911 3.3.1 with inline patch -> 3.3.3: 30% speedup
913 Our reading is that the refactoring which occurred in 3.3.2 made us
914 suffer much more from the xbt_dynar_foreach low performance, but
915 once we solved this, this refactoring proved to be very performance
916 effective. From the 40% speedup, somehow, 10% are due to the
917 inlining and 30% to the refactoring.
919 That's a pitty that gcc cannot inline functions placed in other files
920 alone. We have to choose between:
921 - break the encapsulation (by putting private data structures and
922 accessors in headers files to help gcc)
923 - live with low performance
924 - switch to a decent compiler such as icc (not quite possible).
926 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
928 SimGrid (3.3.2) stable; urgency=low
930 The "Simplicity does not preceed complexity, but follows it" release.
932 The main contributors of this release were (lexical order):
933 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
936 * Extract the routing logic into its own object.
937 (was dupplicated in network.c and workstation_LV07.c;
938 Allows to implement other ways of storing that info)
939 => kill now useless network_card concept
940 - Use dynar to represent routes (instead of void** + int*)
941 - kill link_set (use surf_network_model->resource_set instead)
942 - Add a command-line option to choose the routing schema to use
943 - Add three new models:
944 * Floyd (shortest path computed at initialization)
945 * Dijikstra (shortest path recomputed all the time)
946 * Cached Dijikstra (shortest path computed on need)
947 All these models where contributed by Silas De Munck, and are
948 described in his ICCS09 paper.
950 * Simplify model declaration
951 (less redirections, less function to write when defining a model)
952 - Factorize stuff between models:
955 surf_model_resource_set(model)
956 surf_model_resource_by_name(model, name)
957 - Unify the types of models in s_surf_model_t (using an union)
958 - Embeed fields of common_public directly into s_surf_model_t
959 - Rename model methods:
960 action_free ~> action_unref
961 action_change_state ~> action_state_set
962 action_get_state ~> action_state_get
963 - Change model methods into functions :
964 (model)->common_public->action_use ~> surf_action_ref
966 * Implement a generic resource; use it as ancestor to specific ones
967 (allows to kill duplicated code in models)
968 Drawback: timer command don't need no name nor properties;
969 workstation_CLM03 don't need no properties
970 (but I guess we can live with those few bytes wasted)
972 * Improve the action object model
973 - implement a constructor avoiding dupplicated code about field
974 initialization in generic_action part.
976 * Kill the SDP model: it has an external dependency, is deprecated
977 in flavor of modern lmm models, and didn't compile since a while
980 * Relocation of the context module from XBT to SIMIX.
981 (the context were decoupled from the simix processes, duplicating a lot of code)
982 => a lot of code was factorized
983 - less overhead is introduced during scheduling
984 - simpler API for the context factory
985 - the logic for process creation,destruction and manipulation was simplified
986 * Simplification of the s_smx_process_t data structure.
987 => accesing the simix level data associated to a process is faster now,
988 and the code is a lot more readable.
991 * Implement some more MPI primitives:
992 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
993 -implementation: Bcast: flat or 2-ary tree (default),
996 Allreduce: Reduce then Bcast
997 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
998 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
999 Alltoallv: flat tree, like ompi
1001 * Add support for optimized collectives (Bcast is now binomial by default)
1002 * Port smpirun and smpicc to OS X
1005 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1006 Shout out if you used it.
1009 * Display the list of still queued messages in SG mode when existing
1013 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1014 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1015 * Add xbt_str_from_file(FILE*)
1016 * Add xbt_dict_get_key achieving a linear reverse search
1017 * Remove the context module
1019 Portability report of this version:
1020 * Main portability targets:
1021 - Linux(debian)/x86/context
1022 - Linux(debian)/x86/pthreads
1023 - Linux(debian)/amd64/context
1024 - Linux(debian)/amd64/pthreads
1025 On these, we still have the eratic breakages of gras/pmm and
1026 amok/saturate_sg reported in previous version. We still think
1027 that the tests are the cause of the fault, not the tested code.
1029 - Mac OSX Leopard/x86/context
1030 Still false negative in tesh autotesting.
1031 Smpi still fails, but this time because readlink does not accept -f
1032 Everything seems to work properly beside of that.
1035 - AIX version 5.3 (only tested contexts this time)
1036 Smpi still fails there because mktemp is not installed.
1037 Everything seems to work properly beside of that.
1039 I managed to compile it for the first time, but several breakages.
1040 Won't delay the release for this exotic platform.
1042 * Windows: it's still lagging behind. If you want to help, please
1045 Timing report of this version:
1046 This version seem to be more than 5% faster than 3.3.1 (on linux
1047 64bits with contextes). The gain is less than expected, we are
1048 investigating this for next release.
1050 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1052 SimGrid (3.3.1) stable; urgency=low
1055 * Implement a --cfg-help to show existing configuration variables
1056 * Build chain do not require doxygen in maintainer mode
1059 * fix a bug on struct sizeof computation, which prevented the
1060 exchange of arrays of structs in some conditions
1061 - added a regression test about this in datadesc_usage
1062 * Allow the exchange of 0-long dynamic vectors.
1063 - for that, use -1 as indicator of dynamic size instead of 0
1064 - This implied to change any size from unsigned long to long,
1065 reducing a bit communication abilities, but I guess that with
1066 64bits being quite common, this is more than enough.
1067 - This also induce a protocol change, thus bumping network protocol
1068 version from 0 to 1 (if we have external users, we have to get
1069 clean on that point too ;)
1070 - added two regression tests about this in datadesc_usage
1071 * Be more verbose when propagating local exceptions
1072 This helps debugging.
1073 * Display the status of simulated processes when receiving SIGINT in
1077 * Allow to control the simulation from a trace file.
1078 New functions MSG_action_register() and MSG_action_trace_run()
1079 The first one allows to associate a function execution to each
1080 kind of action while the second one parses a trace file and
1081 triggers the corresponding actions within the system.
1082 For now, only a toy example is provided in examples/msg/actions
1083 * Add an exemple of process migration in examples/msg/migration
1084 * Fix a bug in task exchange which broke MSG_task_get_sender()
1085 Add a teshsuite regression test for that.
1086 [Bug: if MSG_task_get_sender() is called after sender exit,
1088 * Fix a bug which prevented suspend/resume to work properly
1089 * Display the status of simulated processes when receiving SIGINT
1090 This fixes a regression of v3.3. due to the introduction of SIMIX
1091 * Bug fixing in failure management:
1092 - trace could not start by a failure at time 0
1093 - failure during communications were not working
1096 * Add SIMIX_process_set_name() to change the name of the current
1097 process in the log messages.
1098 * Store smx_hosts in a dict since we only retrieve them by name
1099 * Move the configuration infrastructure to surf
1102 * Move the configuration infrastructure to surf
1105 * Massive internal cleanups:
1106 - Store internal structures on processes instead of hosts (allows
1107 to have more than one process per host, in addition of being more
1109 - Cleanup the initialization/finalization process
1110 - Kill a whole bunch of unneeded synchronization:
1111 processes run in exclusive manner within the simulator
1112 - Move queues from global tables to process data fields
1114 - now accept -platform and -hostfile arguments
1115 - Pass the right rank value to processes according to the hostfile
1116 * Compile the examples by default, and use them as regression tests
1117 * Implement MPI_Wtime()
1118 * Change the reference speed to a command line option
1121 * TCP_gamma can now be specified as command line option using
1122 --cfg=TCP_gamma:10000000.0
1123 * Change the --surf-path cmd line option into --cfg=path:
1126 * Also include strbuff from xbt.h public header
1127 * xbt_ex_display(): do not free the exception after displaying
1128 This allows to do more with the given exception afterward.
1129 Users should call xbt_ex_free() themselves.
1133 Portability report of this version:
1134 * Main portability targets:
1135 - Linux(debian)/x86/context
1136 - Linux(debian)/x86/pthreads
1137 - Linux(debian)/amd64/context
1138 - Linux(debian)/amd64/pthreads
1139 These targets fail about 1/10 of times on gras/pmm, but we believe
1140 that this is because of the test, not because of simgrid.
1141 amok/saturate_sg fails even more rarely, and the test may not be
1144 - Mac OSX Leopard/x86/context
1145 The test suite still spits tons of errors because some obscure
1146 force prevents us from removing the temporary directories
1147 arguing that they still contain some metadata I've never heard of.
1148 Smpi fails because seq is not installed.
1149 Everything seems to work properly beside of that.
1152 - AIX version 5.3 (both contexts and pthread)
1153 Smpi still fails there because mktemp is not installed.
1154 XML inclusions seems rosty on AIX.
1156 * Windows: it's still lagging behind. If you want to help, please
1159 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1161 SimGrid (3.3) stable; urgency=high
1165 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1166 [Malek Cherier & Mt]
1168 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1171 Until now, GRAS were using MSG as an interface to SURF. It was
1172 quite difficult because both interface have several differences
1173 (MSG channels vs GRAS sockets were the most notable point).
1175 This also opens the gate to SMPI (which should occur soon) and speed
1176 up simulations by to 40% (even if it were not the main goal).
1178 **************************************
1179 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1180 **************************************
1181 The point is that events occuring at the exact same timestamp are
1182 not scheduled in the same order with the old and new version. This
1183 may be enough to completely change the execution of simulations in
1184 some cases. Sorry for the inconvenience.
1186 * Cleanup and upgrade the XML format to push further scalability
1187 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1189 * Improve the testing infrastructure with tesh. Now a very large part of
1190 the code is tested not only by being run but also by checking that the
1191 output match an expected output [Mt].
1193 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1194 is really less memory-demanding, which should allow you to use
1195 larger files in SimGrid [AL].
1197 * Inform valgrind about our contextes, so that it becomes usable
1198 with the default (and more effecient) version of SimGrid
1199 [contributed by Sékou Diakite, many thanks]
1202 * Introduce a listener thread in charge of receiving incomming
1203 messages from the network. It allows to overlap communication and
1204 computation but most notably, it removes some stupid deadlocks due
1205 to the fact that so far, a process could not send and receive at
1206 the same time. This made most non trivial communication schema
1208 * Convert the PIDs from long int to int to match the MSG ones (and
1209 linux ones too) [Mt]
1210 * New function: gras_agent_spawn() to launch a new process on
1211 current host. Only working in simulation for now. [Mt]
1212 * New function: gras_os_hostport() returning a constant form (ie,
1213 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1216 * Make the backtrace of exceptions more human readable [Mt]
1217 * New module: xbt/str [Mt]
1218 a ton of string utility functions (split, join, printf to a newly
1219 allocated buffer, trim, etc)
1220 * New module: xbt/hash [Mt]
1221 SHA1 hashing algorithm (more to come if needed)
1222 * New module: xbt/synchro [Mt]
1223 synchronization tools (mutex and conditions) working the same way
1224 in simulation and in real life (mainly useful for GRAS, but not
1226 * New module: xbt/queue [Mt]
1227 classical producer/consumer synchronization scheme
1228 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1229 (using the classical functions will get serialized) [Mt]
1230 * Make dictionary internal table dynamic. No need to specify its size
1231 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1233 * Make sure the log channels are organized as a tree under windows
1234 (because of ANSI C compatibility issue, any channel were child of
1235 root directly) [Mt].
1238 * Cleaned many thing in surf and fixed a few bugs [AL].
1239 * Add a nice command line configuration mechanism to compose models [AL].
1240 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1241 the previous one (KCCFLN05). It relies on something that looks like
1242 a max-min sharing mechanism but cannot be written as such. A new solver
1243 was thus designed [AL].
1244 * Add a new solver to lmm. Based on Lagrange optimization and
1245 gradient-based descent, it enables to efficiently maximise systems s.a
1247 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1249 This solver enables to propose two new network models for TCP Reno and
1250 TCP Vegas based on Low's work. These models still need to be fully
1251 tested though [Pedro Velho].
1254 * Bug fix in SD_simulate. Now the time bound given as argument is
1256 * Use the new parallel task model (ptask_L07) as default.
1257 * Use the SURF command line configuration mechanism.
1258 * 0-size tasks (for synchronization) should now work.
1260 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1262 SimGrid (3.2) stable; urgency=high
1266 We still experience issues on this platform, but we believe that at
1267 least MSG is usable.
1269 GRAS API BREAKAGE (for simplification purpose, sorry):
1270 * the gras_msgtype_by_name is not used anymore. Instead of
1271 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1272 you can write (and must)
1273 gras_msg_send(toserver, "request", &request);
1274 - If you still want to pass a gras_msgtype_t to the function (to cache
1275 the type and avoid the lookup time), use the gras_msg_send_() variant.
1276 - Impacted functions:
1277 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1278 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1279 * The callbacks are now expected to return 0 when everything went well
1280 (just like the main() function)
1282 GRAS new features and improvements:
1283 * New module mecanism where user code can use per process globals [Mt]
1284 This is similar to gras_userdata_*() functions, but for libraries. It
1285 factorize some code developped over and over in the examples and AMOK.
1286 It has still to be documented and used (only amok/peermanagement is
1288 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1289 loss when they were small enough to fit into the buffer and sent quickly
1290 enough so that they can all get received in one shoot.
1291 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1292 if not found. Use the *_or_null() variant for the old semantic.
1293 * In gras_msg_handle, do not discard messages without callback.
1294 They are probably messages to be explicitly awaited later (ie, proofs of
1295 mis-synchronization in userland since they are sent before being awaited)
1296 No big deal usually.
1297 * gras_socket_meas_send/recv: semantic changed!
1298 The numerical arguments used to be (1) the total amount of data to send
1299 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1300 messages. This was need for the fool willing to send more than MAXINT
1301 bytes on quite fat pipes.
1304 * Do really rename the hostmanagement module to peermanagement. [Mt]
1305 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1306 but this is rather new and this was documented in the module
1307 documentation (poor excuses, I admit)
1308 * Bandwidth measurement semantic changed! This follows the changes to
1309 gras_socket_meas_send/recv explained above.
1312 * A sequential mode has been added to the workstations. When a workstation
1313 is in sequential mode, it can execute only one task, and the other tasks
1314 are waiting in a FIFO. [Christophe Thiery]
1317 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1318 model for SIMDAG. [Christophe Thiery]
1319 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1320 equal to 0, which caused some bad behaviors in
1321 saturated_constraint_set_update. I now use a threshold mechanism like in
1325 * When running manually src/testall, you select specific units [Mt]
1326 testall is the result of our cunit mecanism, and should replace all
1327 the scripty thingy around since bash don't run easily on billware.
1329 * A mallocator system has been added. [Christophe Thiery]
1330 Mallocators allow you to recycle your unused objects instead of freeing them
1331 and allocating new ones.
1333 Documentation update:
1334 * FAQ reworking + New FAQs:
1335 - "Valgrind spits tons of errors!" [Mt]
1336 - "How to repport bugs" [Mt]
1337 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1338 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1339 - Communication time measurement within MSG [AL]
1340 - I experience weird communication times when I change the latency [AL]
1341 * GRAS tutorial [Mt]
1343 - an introduction to the framework and to the used communication model
1344 - an initiatic tour introducing the most proheminent features:
1346 . Lesson 0: Installing GRAS
1347 . Lesson 1: Setting up your own project
1348 o Part 2: Message passing
1349 . Lesson 2: Exchanging simple messages
1350 . Lesson 3: Passing arguments to the processes (in SG)
1351 . Lesson 4: Attaching callbacks to messages
1352 . Lesson 5: Using globals in processes
1353 . Lesson 6: Logging informations properly
1354 . Lesson 7: Using internal timers
1355 . Lesson 8: Handling errors through exceptions
1356 . Lesson 9: Exchanging simple data
1357 . Lesson 10: Remote Procedure Calling (RPC)
1358 . Lesson 11: Explicitely waiting for messages
1359 . Recapping of message passing features in GRAS
1360 - A HOWTO section containing:
1361 o HOWTO design a GRAS application
1362 More are due, of course. They will come latter. In the meanwhile, you can
1363 check the examples which are still here.
1365 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1367 SimGrid (3.1) stable; urgency=high
1371 There was a stack corruption somewhere, visible only when optimizing
1372 with these versions. [Vince]
1375 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1376 module that enabled to play with Directed Acyclic Graphs. It is built
1377 directly on top of SURF and provides an API rather close to the old
1378 SG. Some old codes using SG are currently under rewrite to check that
1379 all needful functions are provided. [Christophe Thiery]
1382 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1383 extension of the classical CLM03 model that gracefully handles
1384 failures. This is now the default model for MSG and GRAS. It doesn't
1385 handle parallel tasks yet though. [AL]
1386 * Bug fix: Weights were not correctly set in the network part.
1387 WARNING: This may have resulted in incorrect results with simulations
1388 where there are more than one flow on a given link. [AL]
1391 * After a (long ?) discussion on simgrid-devel, we have decided that the
1392 convention we had on units was stupid. That is why it has been decided
1393 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1394 WARNING : This means that all previous platform files will not work as
1395 such with this version! A warning is issued to ask users to update
1397 A conversion script can be found in the contrib module of the CVS, under
1398 the name contrib/platform_generation/surfxml_update.pl [MQ]
1401 * Bug fix: Processes were started in reverse order, wrt deployment file.
1402 WARNING: if your code relies on this bug, please fix it. [AL]
1403 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1404 being executed on two different locations. [AL]
1405 * Bug fix: Failures are now better supported thanks to Derrick's tests
1406 (there was many failure situations I hadn't thought of and that weren't
1407 correctly handled). [AL]
1408 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1412 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1413 Exception occurring on server-side are propagated back to client (!).
1415 API CHANGE: the callback changed their prototype. Change:
1416 int my_handler(gras_socket_t expeditor, void *payload_data) {
1418 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1419 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1421 * New! function: gras_msg_handleall to deal with all messages arriving
1422 within a given period.
1423 * New! function: gras_socket_server_range to get a server socket in a
1424 range of port numbers (ease to avoid port number conflicts) [MQ]
1425 * New! gras processes display their backtrace when they get a SIGUSR1
1426 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1427 Sweet to debug RL processes [MQ]
1431 - Do not force experiment sizes to be expressed in kb, or it becomes
1432 impossible to measure the latency this way (needs one byte-long tests)
1433 WARNING: this changes the amok_bw_* function semantic. [MQ]
1434 - Implements the link saturation stuff. [MQ]
1435 * Peer management module:
1436 New! module factorizing code that we wrote over and over [MQ].
1439 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1440 - Test units are placed directly into the library code, they get extracted
1441 automatically and placed into the src/testall binary.
1442 - Convert most of the XBT tests to this system.
1443 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1444 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1445 live in the same C code. This required to change a little bit the API
1446 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1447 * New module: sparse graph structure with basic algorithms (this is work
1448 in progress and the API is not considered to be frozen yet). [AL]
1449 * Display more information on backtraces: source line & function names are
1450 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1451 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1452 * Reworked a little bit some #include statements to load only required
1453 headers. Some user code that relied on SimGrid to include stdlib or
1454 stdio may need to include it by themselves. [AL]
1455 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1457 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1458 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1459 but I guess that almost nobody used those parts. [MQ]
1461 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1463 SimGrid (3.0.1) stable; urgency=low
1466 * Unfortunately, I had missed 5 misnamed functions:
1467 xbt_fifo_item_t xbt_fifo_newitem(void);
1468 void xbt_fifo_freeitem(xbt_fifo_item_t);
1469 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1470 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1471 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1472 They're now deprecated. Please use their new versions:
1473 xbt_fifo_item_t xbt_fifo_new_item(void);
1474 void xbt_fifo_free_item(xbt_fifo_item_t);
1475 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1476 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1477 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1479 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1480 * Documentation: xbt_log module unmercifully reworked [MQ]
1481 * Bugfix: there was a problem with the ending of contexts with
1482 the pthread backend. It caused some weird deadlock or behavior
1483 depending on the pthread implementation. [AL]
1484 * Bugfix: get the exceptions raised in the simulator repport where
1485 and why they come from when they are not catched in time [AL, MQ]
1488 * Bugfix: Do repport the error when two non-connected hosts try to
1489 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1492 * Add additionnal checkings on communications. Assert that two
1493 communicating hosts are connected by a set of links... [AL]
1496 * Add additionnal checkings on channel values in communication [AL]
1497 * New: MSG_task_get_source to see on which host a task was generated [HC]
1498 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1499 the number of tasks waiting to be received on channel and sent
1501 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1502 waits for the first task coming from a given host.. [AL]
1504 GRAS new functionnalities: [MQ]
1505 * Enhance the parsing macro to allow the size of multidimentional objects
1506 to be given thru annotations.
1507 * New example (and documentation): Matrix Multiplication a la RPC
1508 (as when I was young!) and fix a bunch of bugs found on the way.
1510 GRAS performance improvements: [MQ]
1512 * Reduce the amount of cbps creation/destruction by making it static to
1513 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1516 * Change libdata to a set so that we can search for stuff by ID (and thus
1517 reduce the insane amount of dict lookups)
1520 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1521 RL communication on top of SG since it's so uneffective.
1522 It may also be used for inter-thread communication in RL, one day.
1523 * Use gras_datadesc_copy() to exchange messages on top of SG
1525 - improve message exchange performance on top of SG
1526 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1527 * Don't exchange on the network the size of the used part of buffer,
1528 instead, specify the possible buffer size to read().
1530 - reduces the amount of read/write calls (one pair per exchange)
1531 - reduces the amount of exchanged data (the size)
1532 - allows to retrieve all arrived data on receiver side, if we don't need
1533 it right now (subsequent read will peek the buffer)
1534 - allows the receiver to proceed with the begining of the stream before
1535 everything is arrived
1536 - make it possible to build an iov transport (using readv/writev)
1538 - take care of the data with non-stable storage (like stacked data),
1540 * If possible, TCP send uses vector I/O (when writev() is here)
1541 - Don't use it for receive since we send data sizes and data on the
1542 same stream, so we wouldn't be able to chain large amount of chunks
1543 before having to flush the stuff to read the size.
1544 * Rework the transport plugin mecanism to simplify it and reduce the
1545 amount of pointer dereferencement when searching for the right function
1548 * I guess that now, we do almost as few system calls as possible while
1549 doing as few data copy as possible.
1551 To improve it further, we could try to send all the sizes first and then
1552 all the data (to use iov on receiving size), but it's only a partial
1553 solution: when you have 2 dimensional data, the sizes of the second
1554 dimension is data of the first dimension, so you need 3 streams.
1556 I'm not sure the potential performance gains justify the coding burden.
1558 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1560 SimGrid (3.00) stable; urgency=high
1563 * New! Give the possibility to hijack the surf parser and thus bypass
1564 MSG_create_environment and MSG_launch_application. Have a look at
1565 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1567 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1569 SimGrid (2.96) unstable; urgency=low
1574 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1575 This deprecates the xbt_error_t mecanisms.
1576 It modifies (simplifies) all XBT and GRAS API.
1577 MSG API keeps unchanged (exceptions raised by XBT are catched from
1578 within MSG and masked with existing error handling facilities)
1581 * New! Add a FATPIPE model. [AL]
1582 * New! Add a parallel task model. [AL]
1583 * New! Add automatically a loopback interface (in the default
1584 network model) if none was precised.
1587 * Bugfix: MSG_process_resume now works with the current running process.
1589 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1590 * Modification of MSG_task_get_compute_duration. Once a task has been
1591 processed, the value returned by this function is now equal to 0. [AL]
1592 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1593 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1594 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1595 of a m_task. For now, MSG_task_cancel only works with computation
1597 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1598 of the processor (in Mflop/s) regardless of the current load on the
1600 * API Change: use proper naming convention for MSG_getClock and
1601 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1603 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1604 This function changes the priority of a computation task. This priority
1605 doesn't affect the transfer rate. A priority of 2 will make a task
1606 receive two times more cpu power than the other ones. This function
1607 has been added to suit the needs of Nguyen The Loc and hasn't been that
1608 much tested yet. So if it fails, please report it and send me your code.
1610 * API Change: removed all functions and types that were marked "deprecated"
1611 since many months. Renamed MSG_global_init_args to MSG_global_init.
1613 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1615 SimGrid (2.95) unstable; urgency=low
1618 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1619 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1620 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1623 * Bugfix: Allow absolute paths for platform description files [MQ]
1624 * Bugfix: do free the variables after use. Leads to drastic performance
1626 * Implement max_duration (ie, timeouts) on resources [AL]
1629 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1631 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1632 the needed support (timeouts) [AL]
1635 * Implement measurement sockets. You can now get the bandwidth between two
1636 hosts thanks to AMOK (see below). [MQ]
1637 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1638 dynar over the network (yeah) [MQ]
1639 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1640 * simple/static token-ring example. [Alexandre Colucci and MQ]
1641 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1642 in place before [MQ]
1645 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1646 type parsing. "short" and "long" modifiers were also ignored; other
1647 modifier (such as reference level) are still ignored. [MQ]
1648 * Embeed the buffer size within the buffer itself on SG. [MQ]
1649 That way, send() are atomic and cannot get intermixed anymore (at least
1650 the ones which are less than 100k; bigger messages still have the issue)
1651 * Array size pushed by the field, not by the field type (or each
1652 and every long int will push stuff to the cbps) [MQ]
1653 * use select() to sleep since it allows to portably sleep less than one
1656 GRAS (minor cleanups)
1657 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1660 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1661 * type category 'ignored' killed as it was never used and were difficult
1663 * whether a type can cycle or not is now a flag, leaving room for more
1664 flags (as "ignored", if we feel the need one day ;)
1665 * Rename raw sockets to measurement sockets since "raw" has another
1666 meaning in networking community.
1669 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1670 and offers features not belonging to GRAS but that most applications
1671 need. One day, it may be a set of plugins loadable at runtime.
1672 * New module: bandwidth
1673 bandwidth measurement between arbitrary nodes running this module. [MQ]
1675 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1677 SimGrid (2.94) unstable; urgency=low
1679 The first beta release of SimGrid 3 !
1683 * Update the main page and the FAQ. Adding references to gforge.
1686 * Add a gras_os_getpid function.
1689 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1690 * Extend the logs so that they also print PID, hostname, date, ... if
1692 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1693 and kill the old version which were in testsuite/
1694 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1698 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1699 for users. Instead, each programming environment (GRAS, MSG,...) use it
1700 when needed to provide such a feature to users.
1701 Don't shortcut the mecanism or you will also shortcut the virtualization
1702 you need on the simulator.
1706 * Cleanups in configury with regard to compile optimization/warning flags.
1707 Also add -fno-loop-optimize to any powerpc since it's the optimization
1708 killing gcc (< 3.4.0).
1709 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1710 needed by MSG examples complications
1711 * Borrow configury beautifications from PHP
1714 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1715 warning (thanks loris for stumbling into this one).
1716 * Bugfix: stop loading private headers (gras_config.h) from the public
1720 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1721 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1722 RL would imply a DNS resolver, which is *hard* to do in a portable way
1723 (and therefore delayed).
1724 * Implement a real timer mecanism and use it in timing macros. This allows
1725 to avoid rounding errors and get a 0.000005 sec precision in timing
1726 macros. While I was at it, various cleanups:
1727 - allow to declare more than one timed section per file (fix a stupid bug)
1728 - move some private declaration to the right place
1729 - merge conditional execution and timing macros into emulation module
1730 - document the module
1731 - make sure the module cleanups its mess on gras_exit
1732 * Documentation improvements:
1733 - (new) how to compile applications using GRAS
1734 - (new) emulation support (timing macros)
1736 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1738 SimGrid (2.93) unstable; urgency=low
1740 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1743 - Use Paje properly where used. Still to be sanitized properly.
1744 - Portability fix: Add an implementation of the contexts using pthread
1748 - Add xbt_procname(): returns the name of the current process.
1749 Use it to show the current process's name in all logging.
1751 - fix detection of older flex version and the reaction, since we do
1752 depend on modern ones (we use lex_destroy)
1753 - Better separation of SG and RL in the libs: remove all simulation code
1754 from libgras. As a result, this lib is now only 200k when stripped.
1755 Some of the xbt modules may also be duplicated (two sets and such) and
1756 should be cleaned/killed before SG3.
1757 - Insist on using xlC on AIX because of weird problems involving gcc there.
1758 - Cleanup the make remote stuff. This is now done by scripts
1759 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1760 mainly for our private use, but we're working on changing them to user
1763 - Bugfix: flush the socket on close only if there is some *output*.
1764 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1766 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1769 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1771 SimGrid (2.92) unstable; urgency=low
1773 Alpha 3 on the path to SimGrid 3
1777 - New! First try of benchmarking macros.
1778 - New! First try so that gras_stub_generator generate deployment and
1779 remote compilation helpers.
1781 - Bugfix: Initialization fix in msg_test.
1785 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1787 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1789 - API Change: gras_os_sleep to take the amount of seconds as a double.
1790 Accepting an int was error prone since it was the only location where
1791 seconds were coded as such. It leaded to damn rounding errors.
1792 - Bugfix: Hard to belive that timers ever worked before this.
1794 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1796 SimGrid (2.91) unstable; urgency=low
1798 Alpha 2 on the path to SimGrid 3
1802 - Bug fix in the lmm_solver.
1804 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1805 through the function MSG_paje_output.
1806 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1807 - It is possible to bound the rate of a communication in MSG with
1808 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1809 to write it in the changelog).
1810 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1815 - Various cleanups to the autotools stuff
1816 - Begin to move Gras examples to examples/gras/
1817 - Let make distcheck work again (yeah!)
1819 - documentation overhauled using doxygen.
1820 gtk-doc-tools is dead in SimGrid now.
1821 - Automatically extract all existing logging categories, and add the list
1822 to the documentation (long standing one, to say the less)
1824 - Cleanup the known architecture table. Reorder the entries to group what
1825 should be, and use a more consistent naming scheme.
1826 (some of the test dataset are still to be regenerated)
1827 - New! Allow library to register globals on each process just as userdata
1829 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1830 lookup time (for now).
1831 Use it in msg and trp.
1832 This cleans a lot the internals and helps enforcing privacy of the
1833 headers between the gras components.
1834 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1835 - Bugfix: gras_os_time was delirious in RL.
1836 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1838 - Reenable GRAS now that it works.
1840 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1842 SimGrid (2.90) unstable; urgency=low
1844 Alpha 1 on the path to SimGrid 3
1846 * It is a long time since the last release of SimGrid. I'm sorry about
1847 that but as I had told you, I was rewriting a lot of things. I apologize
1848 to those who had been reporting bugs to me and that I had not answered.
1849 If your bug is still in the new version, please tell me. Here is a
1850 summary of the main changes.
1852 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1853 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1854 portability, speed, and a lot more but you'll figure it out later.
1855 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1856 I wanted to release the new MSG as soon as possible and I have
1857 broken GRAS, which is the reason why, for now, only MSG is fully
1858 functional. A laconic description of these projects is available
1859 in the documentation.
1861 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1862 kernel optimized for our needs (called SURF but only the developers
1863 should use it). Hence, MSG is now roughly 30 times faster and I think
1864 that by rewriting a little bit MSG, I could event speed it up a little
1865 bit more. Beside the gain in speed, it is also much easier to encode a
1866 new platform model with SURF than it was with SG. More to come...
1868 * REVOLUTION 3: I have tried to change a little as possible the API of
1869 MSG but a few things really had to disappear. The main differences
1870 with the previous version are :
1871 1) no more m_links_t and the corresponding functions. Platforms are
1872 directly read from a XML description and cannot be hard-coded
1873 anymore. The same format is used for application deployment
1874 description. The new format is described in the documentation.
1875 Have a look in tools/platform_generation. There is a tiny script
1876 that converts from the old platform format to the new one. Concerning
1877 the application deployment format, parsing the old one is tricky.
1878 I think most of you should however be able to convert your files. If
1879 it is really an issue, I can write a C code that does the conversion.
1881 2) the toolbox tbx does not exist anymore. We now have a library
1882 with much more data-structures but without the hash-tables (we have
1883 dictionaries that are much faster).
1885 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1887 *****************************************************************************
1888 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1889 * don't feel like converting the oldies. (Mt) *
1890 *****************************************************************************
1893 Version 2.90: "the long awaited one"
1894 - Finished rewriting and debugging MSG. Rewrote the documentation.
1895 - disable GRAS for now since it needs to be ported to the newest SG
1898 - Finish the port to windows (using mingw32 for cross-compile)
1901 - Main loop and datastructures of SURF. A cpu resource object is
1902 functional. Surf can thus be used to create cpu's with variable
1903 performance on which you can execute some actions.
1905 2004-11-15 Martin Quinson
1906 - Port to ARM. Simply added the alignment and size descriptions. Should
1907 work, but the ARM machines are so slow that I didn't had the opportunity
1908 to 'make check' over there yet.
1910 2004-11-15 Arnaud Legrand
1911 - Trace manager now written. It uses a heap structure and is therefore
1912 expected to be efficient. It may however be speeded up (particularly
1913 when many events occur at the same date) by using red and black
1914 trees. One day maybe...
1915 - Max-min linear system solver written. It uses a sparse matrix
1916 structure taking advantage of its expected use. Most operations are
1917 O(1) and free/calloc are called as few as possible. The computation of
1918 the minimum could however be improved by using a red and black tree
1921 2004-11-03 Arnaud Legrand
1922 - Rename every gras_* function that was in xbt/ to its xbt_
1924 - Add a heap and a doubly-linked list to xbt
1925 - Added a dichotomy to the dictionaries. make check works as well before
1926 so I assume that the patch is correct. I do not know however if things
1927 run effectively faster than before now. :)
1929 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1930 SimGrid, and the version number is bumped to 2.x
1932 2004-10-29 Martin Quinson
1933 - Introduction of the remote errors.
1934 They are the result of a RMI/RPC on the remote machine.
1935 ErrCodes being scalar values, you can't get the host on which those
1936 errors did happen. Extending the error mechanism as in Gnome is possible.
1937 No idea yet whether it is a good idea.
1939 2004-10-28 Martin Quinson
1940 - Interface revolution: the Starred Structure Eradication.
1941 I used to do typedef struct {} toto_t; and then handle *toto_t.
1942 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1943 - ???_t is a valid type (builded with typedef)
1944 - s_toto_t is a structure (access to fields with .)
1945 - s_toto is a structure needing 'struct' keyword to be used
1946 - e_toto_t is an enum
1947 - toto_t is an 'object' (struct*)
1949 typedef struct s_toto {} s_toto_t, *toto_t;
1950 typedef enum {} e_toto_t;
1951 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1952 s_toto_t) is private.
1954 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1955 so that it can set the variable to NULL. It was so for dicts and sets,
1956 it changed for dynars.
1958 - Fix a bunch of memleaks in dict_remove
1959 - Fix a bug in sg/server_socket opening: it failed all the time.
1961 2004-10-07 Martin Quinson
1962 - Speed up dynar lookup operation a bit.
1964 gras_dynar_get is dead.
1966 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1967 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1968 (which gives you the address of the stored data).
1970 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1971 copy of the data using an affectation to do the job and not a memcpy.
1973 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1974 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1976 It does not really speedup the dynar test because they are
1977 setting elements all the time (and look them seldom). But the dict does
1978 far more lookup than setting.
1980 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1982 2004-10-05 Martin Quinson
1983 - Allow to (en/dis)able the cycle detection at run time.
1985 Whether we should check for cycle or not is now a property of each
1986 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1987 datadesc_cycle_unset allow to remove this property when previously set.
1989 Note that the cycle detection is off by default since it impacts the
1990 performance. Watch the data you feed GRAS with ;)
1992 This property is hereditary. Any element embedded in a structure having it
1993 set have it set for the time of this data exchange.
1995 You should set it both on sender and receiver side. If you don't set it on
1996 sender side, it will enter an endless loop. If you forget on receiver
1997 side, the cycles won't be recreated after communication.
1999 - Header reorganization.
2000 Kill gras_private.h, each submodule must load the headers it needs.
2002 2004-10-04 Martin Quinson
2003 - Interface revolution: do not try to survive to malloc failure.
2005 Now, gras_malloc and friends call gras_abort() on failure.
2006 As a conclusion, malloc_error is not a valid error anymore, and all
2007 functions for which it was the only gras_error_t return value are
2008 changed. They now return void, or there result directly.
2009 This simplify the API a lot.
2011 2004-09-29 Martin Quinson
2012 - Re-enable raw sockets.
2013 Created by gras_socket_{client,server}_ext;
2014 Used with gras_raw_{send,recv}
2017 It should allow to kill the last bits of gras first version soon.
2019 This is not completely satisfactory yet (duplicate code with
2020 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2023 - Simplify transport plugin (internal) interface by not passing any
2024 argument to _server and _client, but embedding them in the socket
2027 2004-09-28 Martin Quinson
2028 - Finish the port to AIX.
2029 autoconf was my problem (segfault within the malloc replacement
2030 function. No idea why)
2032 2004-09-16 Martin Quinson
2033 - Fix some size_t madness on 64bit architectures.
2035 2004-09-08 Martin Quinson
2036 - Reduce the number of system headers loaded, overload some more system
2037 calls (such as malloc to cast the result of the system one, and work
2039 - Fix and reintroduce the config support
2041 2004-09-07 Martin Quinson
2042 - Source code reorganization to allow Arnaud to surf all over there.
2043 - Allow to document the logging categories.
2044 - Remove all uppercase from logging categories and useless cleanup in names.
2046 2004-08-18 Martin Quinson
2047 Version 0.6.2 (protocol not changed; API changed)
2048 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2049 gras_error_t), and NULL when not found. Functions expecting a msgtype
2050 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2051 hopefully usefull message.
2052 - Portability to prehistoric sparcs again
2054 2004-08-17 Martin Quinson
2055 Version 0.6.1 (protocol not changed; ABI not changed)
2056 - prealloc some buffers to speed things up
2058 2004-08-11 Martin Quinson
2059 Version 0.6 (protocol not changed; ABI expended)
2060 - The parsing macro can deal with the references, provided that you add
2061 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2063 2004-08-09 Martin Quinson
2064 Version 0.5 (protocol not changed; ABI changed)
2065 - Allow to off turn the cycle detection code in data exchange at
2066 compilation time. It should be at run time, but I'm short of time (and
2067 the config stuff is still broken). That way, we keep dict out of the
2068 critical path, which is good because the performance is poor:
2069 - search not dichotomial yet
2070 - dynar give no way to access their content and memcpy everytime
2071 - In composed data description (struct, ref and so on), stop foolness of
2072 keeping the subtype's ID, but store the type itself. This keeps sets out
2073 of the critical path, which is good since they rely on dynar and
2074 dictionnaries. The only loose of that is that we cannot detect the
2075 redeclaration of a structure/union with another content (but I'm not sure
2076 the code detected well this error before anyway). We still can detect
2077 the redefinition discrepancy for the other types.
2078 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2079 it breaks the code because of type-punning used all over the place).
2080 This breaks on all non-gcc architectures (for now).
2082 All those changes (plus the buffer of last time) allow me to gain 2 order
2083 of magnitude on cruel tests consisting of 800000 array of integers on two
2084 level of a hierarchical structure (200 secondes -> 4 secondes)
2087 - the selector of reference must now return the type it points to, not
2088 the ID of this type.
2090 2004-08-06 Martin Quinson
2091 Version 0.4 (protocol changed; ABI not changed)
2092 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2093 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2094 - Implement a buffer transport, and use it by default (it relies on tcp in
2095 real life and on sg in simulation).
2096 That's a bit hackish since I had a new field to the structure to store
2097 its data without interfering with the subtype ones. Inheritance
2098 is tricky in C. And that's a kind of reverse inheritance with one class
2099 derivating two classes. Or maybe a game with java interfaces. Anyway,
2100 that's damn hard in C (at least).
2101 Moreover, I got tired while trying to ensure plugin separation and
2102 genericity in SG mode. MSG wants me to do weird things, so let's go for
2103 cruel hacks (temporarily of course ;).
2104 See comment in transport_private.h:71
2105 - do not include all the _interface headers in private but in the files
2106 which really need them (to cut the compilation time when they are
2109 2004-07-26 Martin Quinson
2110 Version 0.3 (protocol not changed; ABI changed)
2111 - Major overhault of the datadesc interface to simplify it:
2112 - shorted the function names:
2113 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2114 - add a trivial way to push/pop integers into the cbps without malloc.
2115 This allows to make really generic sub_type description, which simply
2116 pop their size of the stack.
2117 - add a function gras_datadesc_ref_pop_arr() which does what users want
2118 most of the time: Declare a dynamic array (which pops its size of the
2119 stack) and declare a reference to it. Poor name, but anyway.
2120 - kill the post-send callback, add a post-receive one
2122 2004-07-23 Martin Quinson
2123 Version 0.2 (protocol changed; ABI changed)
2124 - add some testing for cpbs in the test cases, and fix some more bugs.
2125 This invalidate again the little64 data file, since I cannot regenerate
2127 - remove an awfull optimization in the logging stuff, allowing me to:
2128 - understand it again
2129 - learn gcc how to check that the argument match the provided format
2130 - fix all errors revealed by gcc after that
2131 - internal keys of dict are not \0 terminated. Deal with it properly in
2132 loggings instead of segfaulting when the user want to see the logs :-/
2134 2004-07-22 Martin Quinson
2135 - Fix some stupid bug preventing cbps (callback postit) from working
2137 2004-07-21 Martin Quinson
2138 - Some documentation cleanups
2139 - remove the useless last argument of msgtype_declare
2140 - rename the Virtu functions to fit into the 'os' namespace
2141 - move headers src/include -> src/include/gras/ and stop fooling with
2142 gras -> . symbolic link
2143 - make distcheck is now successful
2145 2004-07-19 Martin Quinson
2147 - Build shared library also
2148 - Install html doc to the right location
2149 - stop removing maintainer files in make clean
2150 - build tests only on make check
2152 2004-07-13 Martin Quinson
2154 - No major issue in previous version => change versionning schema
2155 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2156 corresponding dataset.
2158 2004-07-11 Martin Quinson
2160 - Get it working with any kind of structure (we can compute the padding
2161 bytes remotely for all the architectures I have access to)
2162 - Implement the structure parsing macro (still not quite robust/complete)
2163 - Improvement to the remote testing toysuite
2165 2004-07-10 Martin Quinson
2166 [autoconf mechanism]
2167 - get ride of a bunch of deprecated macros
2168 - actually run the test for two-compliment, not only compile it :-/
2169 - test whether the structures get packed (and bail out if yes. Damn.
2170 Alignment is a serious matter)
2171 - test whether the structures get compacted (but respecting the alignment
2172 constraints of each types)
2173 - test whether the array fields of structures can straddle alignment boundaries
2175 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2176 (took me 2 hours to find that bug, looking at the wrong place)
2177 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2178 before sending/receiving and are used to compute the offsets of fields
2179 - Given that padding size depend even on compiler options, keep track of
2180 alignment and aligned_size only for the current architecture. Not a big
2181 deal since we send structure fields one after the other (seems
2183 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2184 program, let it work on linux/gcc/little32. portability todo.
2186 2004-07-08 Martin Quinson
2187 - import and improve remote compilation support from FAST
2188 - make sure make check works on half a dozen of machines out there
2190 2004-07-07 Martin Quinson
2191 Let's say it's version 0.0.3 ;)
2192 - Implement conversions (yuhu!)
2193 - Let it work on solaris (beside conversion, of course)
2194 - Stupid me, using rand() to generate the conversion datatests in not wise.
2196 2004-07-06 Martin Quinson
2197 - Let make dist work, since I'm gonna need it to compile on remote hosts
2198 - Let Tests/datadesc_usage write the architecture on which the file was
2199 generated as first byte.
2200 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2201 architecture descriptions.
2202 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2203 on those architectures.
2204 - Optimization: send/recv array of scalar in one shoot
2206 2004-07-05 Martin Quinson
2207 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2209 - Plug a whole bunch of memleaks
2210 - each process now have to call gras_{init,exit}. One day, their log
2211 settings will be separated
2212 - Continue the code factorisation between SG, RL and common in Transport.
2214 2004-07-04 Martin Quinson
2216 - Redistribution between SG and RL.
2217 We wanna have to accept in SG, so move accepted related parts of RL in
2218 the common part. (more precisely, the dynar of all known sockets is no
2219 more a static in transport.c, but part of the process_data)
2221 [gras_stub_generator]
2222 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2224 2004-07-03 Martin Quinson
2225 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2226 them all up in one shot)
2228 - Fix the ugly case of reference to dynamic array.
2229 - New (semi-public) function gras_datadesc_size to allow the messaging
2230 layer to malloc the needed space for the buffer.
2232 - gras_socket_close now expect the socket to close (and not its address to
2233 put NULL in it after it). This is because the socket passed to handlers
2234 is one of their argument (=> not writable).
2236 - propagate the interface cleanup from last week in datadesc, ie remove a
2237 superfluous level of indirection. User pass adress of variable
2238 containing data (both when sending and receiving), and not of a variable
2239 being a pointer to the data. Let's say that I like it better ;)
2240 The price for that is constructs like "int msg=*(int*)payload" in
2241 handlers, but it's a fine price, IMHO.
2243 - Let it work in RL (yuhu)
2245 2004-06-21 Martin Quinson
2247 - porting SG plugin and SG select to new standards (works almost).
2248 - plug memleaks and fix bugs around.
2251 - cleanup the prototype of data recv and force users to specify when they
2252 want to handle references to objects. Test case working even for cycles.
2253 - plug memleaks. Valgrind is perfectly ok with this.
2255 2004-06-12 Martin Quinson
2257 - cleanup the separation between plugin and main code in plugin creation
2259 2004-06-11 Martin Quinson
2261 - Reput hook for raw sockets, needed for BW experiments
2262 - kill a few lines of dead code
2263 [Data description] Interface cleanup
2264 - gras_datadesc_by_name returns the searched type or NULL.
2265 That way, no variable is needed to use a type desc once, which makes
2267 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2268 two parameters were strings (field name, type name), leading to
2270 [Dicos] Interface cleanup
2271 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2272 This is consistant with the dynar API.
2274 2004-04-21 Martin Quinson
2276 - Porting to new standards.
2278 - interface cleanup.
2279 There is no bag anymore, no need to take extra provision to mask the
2280 pointers behind "ID".
2281 Better splitup of functions between files create/exchange/convert.
2282 This is still a bit artificial since convert and receive are so
2283 interleaved, but anyway.
2285 - add a queued message list to procdata (the ones not matching criteria
2287 - factorize some more code between SG and RL wrt procdata
2289 - use gras_exit in example to track memleaks
2290 - get rid of gs_example now that GS is properly integrated into gras
2291 - update run_test to integrate the lastest tests (datadesc)
2293 - rename WARNINGn macros to WARNn since it prooved error-prone
2295 2004-04-19 Martin Quinson
2297 - register init/exit functions within gras module mechanism
2298 - send/receive function.
2299 Convertion is not implemented, but short-cutted if not needed.
2300 struct/array elements are sent one by one (instead of block-wise), but
2301 nobody really cares (yet). Get a prototype before optimizing.
2302 - tests (using a file socket) for DD send/receive on:
2303 - base types: int, float
2304 - array: fixed size, string (ie ref to dynamic string)
2305 - structure: homogeneous, heterogeneous
2306 - chained list, graph with cycle
2307 Believe it or not, valgrind is not too unhappy with the results. The
2308 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2312 - Bugfix when using a filename explicitely (instead of '-')
2314 2004-04-09 Martin Quinson
2316 - factorize more code between RL and SG in socket creation
2317 - Complete the implementation and tests of:
2319 o file (only in RL, and mainly for debugging)
2321 I lost 3 days to design a portable address resolver, and then decided
2322 that the prototype mainly have to run on my box.
2323 Addressing portability too early may be like optimizing too early :-/
2325 - use gras_init in the Tests instead of the crappy parse_log_opt
2326 (the latter function is removed)
2327 [Conditional execution]
2328 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2329 [Code reorganisation]
2330 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2331 Build examples against the RL library, since there is no way to disable
2332 its creation for now.
2334 For information, the beginning of coding on GRAS was back in june
2335 2003. I guess that every line has been rewritten at least twice since